diff options
Diffstat (limited to 'drivers/gpu/drm/amd/pm')
33 files changed, 1909 insertions, 404 deletions
diff --git a/drivers/gpu/drm/amd/pm/amdgpu_pm.c b/drivers/gpu/drm/amd/pm/amdgpu_pm.c index 58c2246918fd..a57952b93e73 100644 --- a/drivers/gpu/drm/amd/pm/amdgpu_pm.c +++ b/drivers/gpu/drm/amd/pm/amdgpu_pm.c @@ -678,7 +678,12 @@ static ssize_t amdgpu_set_pp_table(struct device *dev, * clock labeled OD_MCLK * * - three <frequency, voltage> points labeled OD_VDDC_CURVE. - * They can be used to calibrate the sclk voltage curve. + * They can be used to calibrate the sclk voltage curve. This is + * available for Vega20 and NV1X. + * + * - voltage offset for the six anchor points of the v/f curve labeled + * OD_VDDC_CURVE. They can be used to calibrate the v/f curve. This + * is only availabe for some SMU13 ASICs. * * - voltage offset(in mV) applied on target voltage calculation. * This is available for Sienna Cichlid, Navy Flounder and Dimgrey @@ -719,12 +724,19 @@ static ssize_t amdgpu_set_pp_table(struct device *dev, * E.g., "p 2 0 800" would set the minimum core clock on core * 2 to 800Mhz. * - * For sclk voltage curve, enter the new values by writing a - * string that contains "vc point clock voltage" to the file. The - * points are indexed by 0, 1 and 2. E.g., "vc 0 300 600" will - * update point1 with clock set as 300Mhz and voltage as - * 600mV. "vc 2 1000 1000" will update point3 with clock set - * as 1000Mhz and voltage 1000mV. + * For sclk voltage curve, + * - For NV1X, enter the new values by writing a string that + * contains "vc point clock voltage" to the file. The points + * are indexed by 0, 1 and 2. E.g., "vc 0 300 600" will update + * point1 with clock set as 300Mhz and voltage as 600mV. "vc 2 + * 1000 1000" will update point3 with clock set as 1000Mhz and + * voltage 1000mV. + * - For SMU13 ASICs, enter the new values by writing a string that + * contains "vc anchor_point_index voltage_offset" to the file. + * There are total six anchor points defined on the v/f curve with + * index as 0 - 5. + * - "vc 0 10" will update the voltage offset for point1 as 10mv. + * - "vc 5 -10" will update the voltage offset for point6 as -10mv. * * To update the voltage offset applied for gfxclk/voltage calculation, * enter the new value by writing a string that contains "vo offset". @@ -871,13 +883,11 @@ static ssize_t amdgpu_get_pp_od_clk_voltage(struct device *dev, } if (ret == -ENOENT) { size = amdgpu_dpm_print_clock_levels(adev, OD_SCLK, buf); - if (size > 0) { - size += amdgpu_dpm_print_clock_levels(adev, OD_MCLK, buf + size); - size += amdgpu_dpm_print_clock_levels(adev, OD_VDDC_CURVE, buf + size); - size += amdgpu_dpm_print_clock_levels(adev, OD_VDDGFX_OFFSET, buf + size); - size += amdgpu_dpm_print_clock_levels(adev, OD_RANGE, buf + size); - size += amdgpu_dpm_print_clock_levels(adev, OD_CCLK, buf + size); - } + size += amdgpu_dpm_print_clock_levels(adev, OD_MCLK, buf + size); + size += amdgpu_dpm_print_clock_levels(adev, OD_VDDC_CURVE, buf + size); + size += amdgpu_dpm_print_clock_levels(adev, OD_VDDGFX_OFFSET, buf + size); + size += amdgpu_dpm_print_clock_levels(adev, OD_RANGE, buf + size); + size += amdgpu_dpm_print_clock_levels(adev, OD_CCLK, buf + size); } if (size == 0) @@ -3362,7 +3372,8 @@ static umode_t hwmon_attributes_visible(struct kobject *kobj, return 0; /* Skip crit temp on APU */ - if ((adev->flags & AMD_IS_APU) && (adev->family >= AMDGPU_FAMILY_CZ) && + if ((((adev->flags & AMD_IS_APU) && (adev->family >= AMDGPU_FAMILY_CZ)) || + (gc_ver == IP_VERSION(9, 4, 3))) && (attr == &sensor_dev_attr_temp1_crit.dev_attr.attr || attr == &sensor_dev_attr_temp1_crit_hyst.dev_attr.attr)) return 0; @@ -3395,9 +3406,10 @@ static umode_t hwmon_attributes_visible(struct kobject *kobj, attr == &sensor_dev_attr_pwm1_enable.dev_attr.attr)) /* can't manage state */ effective_mode &= ~S_IWUSR; - /* In the case of APUs, this is only implemented on Vangogh */ + /* not implemented yet for APUs other than GC 10.3.1 (vangogh) and 9.4.3 */ if (((adev->family == AMDGPU_FAMILY_SI) || - ((adev->flags & AMD_IS_APU) && (gc_ver != IP_VERSION(10, 3, 1)))) && + ((adev->flags & AMD_IS_APU) && (gc_ver != IP_VERSION(10, 3, 1)) && + (gc_ver != IP_VERSION(9, 4, 3)))) && (attr == &sensor_dev_attr_power1_cap_max.dev_attr.attr || attr == &sensor_dev_attr_power1_cap_min.dev_attr.attr || attr == &sensor_dev_attr_power1_cap.dev_attr.attr || @@ -3426,36 +3438,48 @@ static umode_t hwmon_attributes_visible(struct kobject *kobj, return 0; if ((adev->family == AMDGPU_FAMILY_SI || /* not implemented yet */ - adev->family == AMDGPU_FAMILY_KV) && /* not implemented yet */ + adev->family == AMDGPU_FAMILY_KV || /* not implemented yet */ + (gc_ver == IP_VERSION(9, 4, 3))) && (attr == &sensor_dev_attr_in0_input.dev_attr.attr || attr == &sensor_dev_attr_in0_label.dev_attr.attr)) return 0; - /* only APUs have vddnb */ - if (!(adev->flags & AMD_IS_APU) && + /* only APUs other than gc 9,4,3 have vddnb */ + if ((!(adev->flags & AMD_IS_APU) || (gc_ver == IP_VERSION(9, 4, 3))) && (attr == &sensor_dev_attr_in1_input.dev_attr.attr || attr == &sensor_dev_attr_in1_label.dev_attr.attr)) return 0; - /* no mclk on APUs */ - if ((adev->flags & AMD_IS_APU) && + /* no mclk on APUs other than gc 9,4,3*/ + if (((adev->flags & AMD_IS_APU) && (gc_ver != IP_VERSION(9, 4, 3))) && (attr == &sensor_dev_attr_freq2_input.dev_attr.attr || attr == &sensor_dev_attr_freq2_label.dev_attr.attr)) return 0; - /* only SOC15 dGPUs support hotspot and mem temperatures */ if (((adev->flags & AMD_IS_APU) || gc_ver < IP_VERSION(9, 0, 0)) && + (gc_ver != IP_VERSION(9, 4, 3)) && + (attr == &sensor_dev_attr_temp2_input.dev_attr.attr || + attr == &sensor_dev_attr_temp2_label.dev_attr.attr || + attr == &sensor_dev_attr_temp3_input.dev_attr.attr || + attr == &sensor_dev_attr_temp3_label.dev_attr.attr)) + return 0; + + /* hotspot temperature for gc 9,4,3*/ + if ((gc_ver == IP_VERSION(9, 4, 3)) && + (attr == &sensor_dev_attr_temp1_input.dev_attr.attr || + attr == &sensor_dev_attr_temp1_label.dev_attr.attr)) + return 0; + + /* only SOC15 dGPUs support hotspot and mem temperatures */ + if (((adev->flags & AMD_IS_APU) || gc_ver < IP_VERSION(9, 0, 0) || + (gc_ver == IP_VERSION(9, 4, 3))) && (attr == &sensor_dev_attr_temp2_crit.dev_attr.attr || attr == &sensor_dev_attr_temp2_crit_hyst.dev_attr.attr || attr == &sensor_dev_attr_temp3_crit.dev_attr.attr || attr == &sensor_dev_attr_temp3_crit_hyst.dev_attr.attr || attr == &sensor_dev_attr_temp1_emergency.dev_attr.attr || attr == &sensor_dev_attr_temp2_emergency.dev_attr.attr || - attr == &sensor_dev_attr_temp3_emergency.dev_attr.attr || - attr == &sensor_dev_attr_temp2_input.dev_attr.attr || - attr == &sensor_dev_attr_temp3_input.dev_attr.attr || - attr == &sensor_dev_attr_temp2_label.dev_attr.attr || - attr == &sensor_dev_attr_temp3_label.dev_attr.attr)) + attr == &sensor_dev_attr_temp3_emergency.dev_attr.attr)) return 0; /* only Vangogh has fast PPT limit and power labels */ diff --git a/drivers/gpu/drm/amd/pm/inc/smu_v13_0_0_pptable.h b/drivers/gpu/drm/amd/pm/inc/smu_v13_0_0_pptable.h index 566a0da59e53..1dc7a065a6d4 100644 --- a/drivers/gpu/drm/amd/pm/inc/smu_v13_0_0_pptable.h +++ b/drivers/gpu/drm/amd/pm/inc/smu_v13_0_0_pptable.h @@ -38,13 +38,12 @@ #define SMU_13_0_0_PP_THERMALCONTROLLER_NONE 0 #define SMU_13_0_0_PP_THERMALCONTROLLER_NAVI21 28 -#define SMU_13_0_0_PP_OVERDRIVE_VERSION 0x81 // OverDrive 8 Table Version 0.2 +#define SMU_13_0_0_PP_OVERDRIVE_VERSION 0x83 // OverDrive 8 Table Version 0.2 #define SMU_13_0_0_PP_POWERSAVINGCLOCK_VERSION 0x01 // Power Saving Clock Table Version 1.00 enum SMU_13_0_0_ODFEATURE_CAP { SMU_13_0_0_ODCAP_GFXCLK_LIMITS = 0, - SMU_13_0_0_ODCAP_GFXCLK_CURVE, SMU_13_0_0_ODCAP_UCLK_LIMITS, SMU_13_0_0_ODCAP_POWER_LIMIT, SMU_13_0_0_ODCAP_FAN_ACOUSTIC_LIMIT, @@ -59,13 +58,13 @@ enum SMU_13_0_0_ODFEATURE_CAP SMU_13_0_0_ODCAP_FAN_CURVE, SMU_13_0_0_ODCAP_AUTO_FAN_ACOUSTIC_LIMIT, SMU_13_0_0_ODCAP_POWER_MODE, + SMU_13_0_0_ODCAP_PER_ZONE_GFX_VOLTAGE_OFFSET, SMU_13_0_0_ODCAP_COUNT, }; enum SMU_13_0_0_ODFEATURE_ID { SMU_13_0_0_ODFEATURE_GFXCLK_LIMITS = 1 << SMU_13_0_0_ODCAP_GFXCLK_LIMITS, //GFXCLK Limit feature - SMU_13_0_0_ODFEATURE_GFXCLK_CURVE = 1 << SMU_13_0_0_ODCAP_GFXCLK_CURVE, //GFXCLK Curve feature SMU_13_0_0_ODFEATURE_UCLK_LIMITS = 1 << SMU_13_0_0_ODCAP_UCLK_LIMITS, //UCLK Limit feature SMU_13_0_0_ODFEATURE_POWER_LIMIT = 1 << SMU_13_0_0_ODCAP_POWER_LIMIT, //Power Limit feature SMU_13_0_0_ODFEATURE_FAN_ACOUSTIC_LIMIT = 1 << SMU_13_0_0_ODCAP_FAN_ACOUSTIC_LIMIT, //Fan Acoustic RPM feature @@ -80,6 +79,7 @@ enum SMU_13_0_0_ODFEATURE_ID SMU_13_0_0_ODFEATURE_FAN_CURVE = 1 << SMU_13_0_0_ODCAP_FAN_CURVE, //Fan Curve feature SMU_13_0_0_ODFEATURE_AUTO_FAN_ACOUSTIC_LIMIT = 1 << SMU_13_0_0_ODCAP_AUTO_FAN_ACOUSTIC_LIMIT, //Auto Fan Acoustic RPM feature SMU_13_0_0_ODFEATURE_POWER_MODE = 1 << SMU_13_0_0_ODCAP_POWER_MODE, //Optimized GPU Power Mode feature + SMU_13_0_0_ODFEATURE_PER_ZONE_GFX_VOLTAGE_OFFSET = 1 << SMU_13_0_0_ODCAP_PER_ZONE_GFX_VOLTAGE_OFFSET, //Perzone voltage offset feature SMU_13_0_0_ODFEATURE_COUNT = 16, }; @@ -89,10 +89,6 @@ enum SMU_13_0_0_ODSETTING_ID { SMU_13_0_0_ODSETTING_GFXCLKFMAX = 0, SMU_13_0_0_ODSETTING_GFXCLKFMIN, - SMU_13_0_0_ODSETTING_CUSTOM_GFX_VF_CURVE_A, - SMU_13_0_0_ODSETTING_CUSTOM_GFX_VF_CURVE_B, - SMU_13_0_0_ODSETTING_CUSTOM_GFX_VF_CURVE_C, - SMU_13_0_0_ODSETTING_CUSTOM_CURVE_VFT_FMIN, SMU_13_0_0_ODSETTING_UCLKFMIN, SMU_13_0_0_ODSETTING_UCLKFMAX, SMU_13_0_0_ODSETTING_POWERPERCENTAGE, @@ -117,6 +113,12 @@ enum SMU_13_0_0_ODSETTING_ID SMU_13_0_0_ODSETTING_FAN_CURVE_SPEED_5, SMU_13_0_0_ODSETTING_AUTO_FAN_ACOUSTIC_LIMIT, SMU_13_0_0_ODSETTING_POWER_MODE, + SMU_13_0_0_ODSETTING_PER_ZONE_GFX_VOLTAGE_OFFSET_POINT_1, + SMU_13_0_0_ODSETTING_PER_ZONE_GFX_VOLTAGE_OFFSET_POINT_2, + SMU_13_0_0_ODSETTING_PER_ZONE_GFX_VOLTAGE_OFFSET_POINT_3, + SMU_13_0_0_ODSETTING_PER_ZONE_GFX_VOLTAGE_OFFSET_POINT_4, + SMU_13_0_0_ODSETTING_PER_ZONE_GFX_VOLTAGE_OFFSET_POINT_5, + SMU_13_0_0_ODSETTING_PER_ZONE_GFX_VOLTAGE_OFFSET_POINT_6, SMU_13_0_0_ODSETTING_COUNT, }; #define SMU_13_0_0_MAX_ODSETTING 64 //Maximum Number of ODSettings diff --git a/drivers/gpu/drm/amd/pm/legacy-dpm/kv_dpm.c b/drivers/gpu/drm/amd/pm/legacy-dpm/kv_dpm.c index f5e08b60f66e..36c831b280ed 100644 --- a/drivers/gpu/drm/amd/pm/legacy-dpm/kv_dpm.c +++ b/drivers/gpu/drm/amd/pm/legacy-dpm/kv_dpm.c @@ -508,19 +508,19 @@ static int kv_enable_didt(struct amdgpu_device *adev, bool enable) pi->caps_db_ramping || pi->caps_td_ramping || pi->caps_tcp_ramping) { - amdgpu_gfx_rlc_enter_safe_mode(adev); + amdgpu_gfx_rlc_enter_safe_mode(adev, 0); if (enable) { ret = kv_program_pt_config_registers(adev, didt_config_kv); if (ret) { - amdgpu_gfx_rlc_exit_safe_mode(adev); + amdgpu_gfx_rlc_exit_safe_mode(adev, 0); return ret; } } kv_do_enable_didt(adev, enable); - amdgpu_gfx_rlc_exit_safe_mode(adev); + amdgpu_gfx_rlc_exit_safe_mode(adev, 0); } return 0; diff --git a/drivers/gpu/drm/amd/pm/legacy-dpm/si_dpm.c b/drivers/gpu/drm/amd/pm/legacy-dpm/si_dpm.c index d6d9e3b1b2c0..02e69ccff3ba 100644 --- a/drivers/gpu/drm/amd/pm/legacy-dpm/si_dpm.c +++ b/drivers/gpu/drm/amd/pm/legacy-dpm/si_dpm.c @@ -6925,23 +6925,6 @@ static int si_dpm_enable(struct amdgpu_device *adev) return 0; } -static int si_set_temperature_range(struct amdgpu_device *adev) -{ - int ret; - - ret = si_thermal_enable_alert(adev, false); - if (ret) - return ret; - ret = si_thermal_set_temperature_range(adev, R600_TEMP_RANGE_MIN, R600_TEMP_RANGE_MAX); - if (ret) - return ret; - ret = si_thermal_enable_alert(adev, true); - if (ret) - return ret; - - return ret; -} - static void si_dpm_disable(struct amdgpu_device *adev) { struct rv7xx_power_info *pi = rv770_get_pi(adev); @@ -7626,18 +7609,6 @@ static int si_dpm_process_interrupt(struct amdgpu_device *adev, static int si_dpm_late_init(void *handle) { - int ret; - struct amdgpu_device *adev = (struct amdgpu_device *)handle; - - if (!adev->pm.dpm_enabled) - return 0; - - ret = si_set_temperature_range(adev); - if (ret) - return ret; -#if 0 //TODO ? - si_dpm_powergate_uvd(adev, true); -#endif return 0; } diff --git a/drivers/gpu/drm/amd/pm/powerplay/hwmgr/smu7_powertune.c b/drivers/gpu/drm/amd/pm/powerplay/hwmgr/smu7_powertune.c index 32a5a00fd8ae..21be23ec3c79 100644 --- a/drivers/gpu/drm/amd/pm/powerplay/hwmgr/smu7_powertune.c +++ b/drivers/gpu/drm/amd/pm/powerplay/hwmgr/smu7_powertune.c @@ -973,7 +973,7 @@ int smu7_enable_didt_config(struct pp_hwmgr *hwmgr) PP_CAP(PHM_PlatformCaps_TDRamping) || PP_CAP(PHM_PlatformCaps_TCPRamping)) { - amdgpu_gfx_rlc_enter_safe_mode(adev); + amdgpu_gfx_rlc_enter_safe_mode(adev, 0); mutex_lock(&adev->grbm_idx_mutex); value = 0; value2 = cgs_read_register(hwmgr->device, mmGRBM_GFX_INDEX); @@ -1048,13 +1048,13 @@ int smu7_enable_didt_config(struct pp_hwmgr *hwmgr) } mutex_unlock(&adev->grbm_idx_mutex); - amdgpu_gfx_rlc_exit_safe_mode(adev); + amdgpu_gfx_rlc_exit_safe_mode(adev, 0); } return 0; error: mutex_unlock(&adev->grbm_idx_mutex); - amdgpu_gfx_rlc_exit_safe_mode(adev); + amdgpu_gfx_rlc_exit_safe_mode(adev, 0); return result; } @@ -1068,7 +1068,7 @@ int smu7_disable_didt_config(struct pp_hwmgr *hwmgr) PP_CAP(PHM_PlatformCaps_TDRamping) || PP_CAP(PHM_PlatformCaps_TCPRamping)) { - amdgpu_gfx_rlc_enter_safe_mode(adev); + amdgpu_gfx_rlc_enter_safe_mode(adev, 0); result = smu7_enable_didt(hwmgr, false); PP_ASSERT_WITH_CODE((result == 0), @@ -1081,12 +1081,12 @@ int smu7_disable_didt_config(struct pp_hwmgr *hwmgr) PP_ASSERT_WITH_CODE((0 == result), "Failed to disable DPM DIDT.", goto error); } - amdgpu_gfx_rlc_exit_safe_mode(adev); + amdgpu_gfx_rlc_exit_safe_mode(adev, 0); } return 0; error: - amdgpu_gfx_rlc_exit_safe_mode(adev); + amdgpu_gfx_rlc_exit_safe_mode(adev, 0); return result; } diff --git a/drivers/gpu/drm/amd/pm/powerplay/hwmgr/vega10_powertune.c b/drivers/gpu/drm/amd/pm/powerplay/hwmgr/vega10_powertune.c index 9757d47dd6b8..309a9d3bc1b7 100644 --- a/drivers/gpu/drm/amd/pm/powerplay/hwmgr/vega10_powertune.c +++ b/drivers/gpu/drm/amd/pm/powerplay/hwmgr/vega10_powertune.c @@ -915,7 +915,7 @@ static int vega10_enable_cac_driving_se_didt_config(struct pp_hwmgr *hwmgr) num_se = adev->gfx.config.max_shader_engines; - amdgpu_gfx_rlc_enter_safe_mode(adev); + amdgpu_gfx_rlc_enter_safe_mode(adev, 0); mutex_lock(&adev->grbm_idx_mutex); for (count = 0; count < num_se; count++) { @@ -940,7 +940,7 @@ static int vega10_enable_cac_driving_se_didt_config(struct pp_hwmgr *hwmgr) vega10_didt_set_mask(hwmgr, true); - amdgpu_gfx_rlc_exit_safe_mode(adev); + amdgpu_gfx_rlc_exit_safe_mode(adev, 0); return 0; } @@ -949,11 +949,11 @@ static int vega10_disable_cac_driving_se_didt_config(struct pp_hwmgr *hwmgr) { struct amdgpu_device *adev = hwmgr->adev; - amdgpu_gfx_rlc_enter_safe_mode(adev); + amdgpu_gfx_rlc_enter_safe_mode(adev, 0); vega10_didt_set_mask(hwmgr, false); - amdgpu_gfx_rlc_exit_safe_mode(adev); + amdgpu_gfx_rlc_exit_safe_mode(adev, 0); return 0; } @@ -966,7 +966,7 @@ static int vega10_enable_psm_gc_didt_config(struct pp_hwmgr *hwmgr) num_se = adev->gfx.config.max_shader_engines; - amdgpu_gfx_rlc_enter_safe_mode(adev); + amdgpu_gfx_rlc_enter_safe_mode(adev, 0); mutex_lock(&adev->grbm_idx_mutex); for (count = 0; count < num_se; count++) { @@ -985,7 +985,7 @@ static int vega10_enable_psm_gc_didt_config(struct pp_hwmgr *hwmgr) vega10_didt_set_mask(hwmgr, true); - amdgpu_gfx_rlc_exit_safe_mode(adev); + amdgpu_gfx_rlc_exit_safe_mode(adev, 0); vega10_program_gc_didt_config_registers(hwmgr, GCDiDtDroopCtrlConfig_vega10); if (PP_CAP(PHM_PlatformCaps_GCEDC)) @@ -1002,11 +1002,11 @@ static int vega10_disable_psm_gc_didt_config(struct pp_hwmgr *hwmgr) struct amdgpu_device *adev = hwmgr->adev; uint32_t data; - amdgpu_gfx_rlc_enter_safe_mode(adev); + amdgpu_gfx_rlc_enter_safe_mode(adev, 0); vega10_didt_set_mask(hwmgr, false); - amdgpu_gfx_rlc_exit_safe_mode(adev); + amdgpu_gfx_rlc_exit_safe_mode(adev, 0); if (PP_CAP(PHM_PlatformCaps_GCEDC)) { data = 0x00000000; @@ -1027,7 +1027,7 @@ static int vega10_enable_se_edc_config(struct pp_hwmgr *hwmgr) num_se = adev->gfx.config.max_shader_engines; - amdgpu_gfx_rlc_enter_safe_mode(adev); + amdgpu_gfx_rlc_enter_safe_mode(adev, 0); mutex_lock(&adev->grbm_idx_mutex); for (count = 0; count < num_se; count++) { @@ -1048,7 +1048,7 @@ static int vega10_enable_se_edc_config(struct pp_hwmgr *hwmgr) vega10_didt_set_mask(hwmgr, true); - amdgpu_gfx_rlc_exit_safe_mode(adev); + amdgpu_gfx_rlc_exit_safe_mode(adev, 0); return 0; } @@ -1057,11 +1057,11 @@ static int vega10_disable_se_edc_config(struct pp_hwmgr *hwmgr) { struct amdgpu_device *adev = hwmgr->adev; - amdgpu_gfx_rlc_enter_safe_mode(adev); + amdgpu_gfx_rlc_enter_safe_mode(adev, 0); vega10_didt_set_mask(hwmgr, false); - amdgpu_gfx_rlc_exit_safe_mode(adev); + amdgpu_gfx_rlc_exit_safe_mode(adev, 0); return 0; } @@ -1075,7 +1075,7 @@ static int vega10_enable_psm_gc_edc_config(struct pp_hwmgr *hwmgr) num_se = adev->gfx.config.max_shader_engines; - amdgpu_gfx_rlc_enter_safe_mode(adev); + amdgpu_gfx_rlc_enter_safe_mode(adev, 0); vega10_program_gc_didt_config_registers(hwmgr, AvfsPSMResetConfig_vega10); @@ -1096,7 +1096,7 @@ static int vega10_enable_psm_gc_edc_config(struct pp_hwmgr *hwmgr) vega10_didt_set_mask(hwmgr, true); - amdgpu_gfx_rlc_exit_safe_mode(adev); + amdgpu_gfx_rlc_exit_safe_mode(adev, 0); vega10_program_gc_didt_config_registers(hwmgr, PSMGCEDCDroopCtrlConfig_vega10); @@ -1116,11 +1116,11 @@ static int vega10_disable_psm_gc_edc_config(struct pp_hwmgr *hwmgr) struct amdgpu_device *adev = hwmgr->adev; uint32_t data; - amdgpu_gfx_rlc_enter_safe_mode(adev); + amdgpu_gfx_rlc_enter_safe_mode(adev, 0); vega10_didt_set_mask(hwmgr, false); - amdgpu_gfx_rlc_exit_safe_mode(adev); + amdgpu_gfx_rlc_exit_safe_mode(adev, 0); if (PP_CAP(PHM_PlatformCaps_GCEDC)) { data = 0x00000000; @@ -1138,7 +1138,7 @@ static int vega10_enable_se_edc_force_stall_config(struct pp_hwmgr *hwmgr) struct amdgpu_device *adev = hwmgr->adev; int result; - amdgpu_gfx_rlc_enter_safe_mode(adev); + amdgpu_gfx_rlc_enter_safe_mode(adev, 0); mutex_lock(&adev->grbm_idx_mutex); WREG32_SOC15(GC, 0, mmGRBM_GFX_INDEX, 0xE0000000); @@ -1151,7 +1151,7 @@ static int vega10_enable_se_edc_force_stall_config(struct pp_hwmgr *hwmgr) vega10_didt_set_mask(hwmgr, false); - amdgpu_gfx_rlc_exit_safe_mode(adev); + amdgpu_gfx_rlc_exit_safe_mode(adev, 0); return 0; } diff --git a/drivers/gpu/drm/amd/pm/powerplay/inc/hwmgr.h b/drivers/gpu/drm/amd/pm/powerplay/inc/hwmgr.h index 5ce433e2c16a..f1580a26a850 100644 --- a/drivers/gpu/drm/amd/pm/powerplay/inc/hwmgr.h +++ b/drivers/gpu/drm/amd/pm/powerplay/inc/hwmgr.h @@ -359,7 +359,7 @@ struct pp_hwmgr_func { int (*set_ppfeature_status)(struct pp_hwmgr *hwmgr, uint64_t ppfeature_masks); int (*set_mp1_state)(struct pp_hwmgr *hwmgr, enum pp_mp1_state mp1_state); int (*asic_reset)(struct pp_hwmgr *hwmgr, enum SMU_ASIC_RESET_MODE mode); - int (*smu_i2c_bus_access)(struct pp_hwmgr *hwmgr, bool aquire); + int (*smu_i2c_bus_access)(struct pp_hwmgr *hwmgr, bool acquire); int (*set_df_cstate)(struct pp_hwmgr *hwmgr, enum pp_df_cstate state); int (*set_xgmi_pstate)(struct pp_hwmgr *hwmgr, uint32_t pstate); int (*disable_power_features_for_compute_performance)(struct pp_hwmgr *hwmgr, diff --git a/drivers/gpu/drm/amd/pm/swsmu/amdgpu_smu.c b/drivers/gpu/drm/amd/pm/swsmu/amdgpu_smu.c index 5633c5797e85..4dea79a0c5b5 100644 --- a/drivers/gpu/drm/amd/pm/swsmu/amdgpu_smu.c +++ b/drivers/gpu/drm/amd/pm/swsmu/amdgpu_smu.c @@ -733,6 +733,24 @@ static int smu_late_init(void *handle) return ret; } + /* + * Explicitly notify PMFW the power mode the system in. Since + * the PMFW may boot the ASIC with a different mode. + * For those supporting ACDC switch via gpio, PMFW will + * handle the switch automatically. Driver involvement + * is unnecessary. + */ + if (!smu->dc_controlled_by_gpio) { + ret = smu_set_power_source(smu, + adev->pm.ac_power ? SMU_POWER_SOURCE_AC : + SMU_POWER_SOURCE_DC); + if (ret) { + dev_err(adev->dev, "Failed to switch to %s mode!\n", + adev->pm.ac_power ? "AC" : "DC"); + return ret; + } + } + if ((adev->ip_versions[MP1_HWIP][0] == IP_VERSION(13, 0, 1)) || (adev->ip_versions[MP1_HWIP][0] == IP_VERSION(13, 0, 3))) return 0; @@ -804,11 +822,20 @@ static int smu_init_fb_allocations(struct smu_context *smu) } } + driver_table->domain = AMDGPU_GEM_DOMAIN_VRAM | AMDGPU_GEM_DOMAIN_GTT; /* VRAM allocation for driver table */ for (i = 0; i < SMU_TABLE_COUNT; i++) { if (tables[i].size == 0) continue; + /* If one of the tables has VRAM domain restriction, keep it in + * VRAM + */ + if ((tables[i].domain & + (AMDGPU_GEM_DOMAIN_VRAM | AMDGPU_GEM_DOMAIN_GTT)) == + AMDGPU_GEM_DOMAIN_VRAM) + driver_table->domain = AMDGPU_GEM_DOMAIN_VRAM; + if (i == SMU_TABLE_PMSTATUSLOG) continue; @@ -818,7 +845,6 @@ static int smu_init_fb_allocations(struct smu_context *smu) driver_table->size = max_table_size; driver_table->align = PAGE_SIZE; - driver_table->domain = AMDGPU_GEM_DOMAIN_VRAM; ret = amdgpu_bo_create_kernel(adev, driver_table->size, diff --git a/drivers/gpu/drm/amd/pm/swsmu/inc/pmfw_if/smu13_driver_if_aldebaran.h b/drivers/gpu/drm/amd/pm/swsmu/inc/pmfw_if/smu13_driver_if_aldebaran.h index 90200f31ff52..cddf45eebee8 100644 --- a/drivers/gpu/drm/amd/pm/swsmu/inc/pmfw_if/smu13_driver_if_aldebaran.h +++ b/drivers/gpu/drm/amd/pm/swsmu/inc/pmfw_if/smu13_driver_if_aldebaran.h @@ -24,6 +24,8 @@ #ifndef SMU13_DRIVER_IF_ALDEBARAN_H #define SMU13_DRIVER_IF_ALDEBARAN_H +#define SMU13_DRIVER_IF_VERSION_ALDE 0x08 + #define NUM_VCLK_DPM_LEVELS 8 #define NUM_DCLK_DPM_LEVELS 8 #define NUM_SOCCLK_DPM_LEVELS 8 diff --git a/drivers/gpu/drm/amd/pm/swsmu/inc/pmfw_if/smu13_driver_if_v13_0_0.h b/drivers/gpu/drm/amd/pm/swsmu/inc/pmfw_if/smu13_driver_if_v13_0_0.h index b686fb68a6e7..9dd1ed5b8940 100644 --- a/drivers/gpu/drm/amd/pm/swsmu/inc/pmfw_if/smu13_driver_if_v13_0_0.h +++ b/drivers/gpu/drm/amd/pm/swsmu/inc/pmfw_if/smu13_driver_if_v13_0_0.h @@ -24,8 +24,10 @@ #ifndef SMU13_DRIVER_IF_V13_0_0_H #define SMU13_DRIVER_IF_V13_0_0_H +#define SMU13_0_0_DRIVER_IF_VERSION 0x3D + //Increment this version if SkuTable_t or BoardTable_t change -#define PPTABLE_VERSION 0x26 +#define PPTABLE_VERSION 0x2B #define NUM_GFXCLK_DPM_LEVELS 16 #define NUM_SOCCLK_DPM_LEVELS 8 @@ -94,7 +96,7 @@ #define FEATURE_ATHUB_MMHUB_PG_BIT 48 #define FEATURE_SOC_PCC_BIT 49 #define FEATURE_EDC_PWRBRK_BIT 50 -#define FEATURE_SPARE_51_BIT 51 +#define FEATURE_BOMXCO_SVI3_PROG_BIT 51 #define FEATURE_SPARE_52_BIT 52 #define FEATURE_SPARE_53_BIT 53 #define FEATURE_SPARE_54_BIT 54 @@ -310,6 +312,7 @@ typedef enum { I2C_CONTROLLER_PROTOCOL_VR_IR35217, I2C_CONTROLLER_PROTOCOL_TMP_MAX31875, I2C_CONTROLLER_PROTOCOL_INA3221, + I2C_CONTROLLER_PROTOCOL_TMP_MAX6604, I2C_CONTROLLER_PROTOCOL_COUNT, } I2cControllerProtocol_e; @@ -568,6 +571,7 @@ typedef enum { } POWER_SOURCE_e; typedef enum { + MEM_VENDOR_PLACEHOLDER0, MEM_VENDOR_SAMSUNG, MEM_VENDOR_INFINEON, MEM_VENDOR_ELPIDA, @@ -577,7 +581,6 @@ typedef enum { MEM_VENDOR_MOSEL, MEM_VENDOR_WINBOND, MEM_VENDOR_ESMT, - MEM_VENDOR_PLACEHOLDER0, MEM_VENDOR_PLACEHOLDER1, MEM_VENDOR_PLACEHOLDER2, MEM_VENDOR_PLACEHOLDER3, @@ -665,7 +668,14 @@ typedef enum { #define PP_NUM_RTAVFS_PWL_ZONES 5 - +#define PP_OD_FEATURE_GFX_VF_CURVE_BIT 0 +#define PP_OD_FEATURE_PPT_BIT 2 +#define PP_OD_FEATURE_FAN_CURVE_BIT 3 +#define PP_OD_FEATURE_GFXCLK_BIT 7 +#define PP_OD_FEATURE_UCLK_BIT 8 +#define PP_OD_FEATURE_ZERO_FAN_BIT 9 +#define PP_OD_FEATURE_TEMPERATURE_BIT 10 +#define PP_OD_FEATURE_COUNT 13 // VBIOS or PPLIB configures telemetry slope and offset. Only slope expected to be set for SVI3 // Slope Q1.7, Offset Q1.2 @@ -687,10 +697,8 @@ typedef struct { //Voltage control int16_t VoltageOffsetPerZoneBoundary[PP_NUM_OD_VF_CURVE_POINTS]; - uint16_t VddGfxVmax; // in mV - uint8_t IdlePwrSavingFeaturesCtrl; - uint8_t RuntimePwrSavingFeaturesCtrl; + uint32_t Reserved; //Frequency changes int16_t GfxclkFmin; // MHz @@ -727,10 +735,9 @@ typedef struct { uint32_t FeatureCtrlMask; int16_t VoltageOffsetPerZoneBoundary; - uint16_t VddGfxVmax; // in mV + uint16_t Reserved1; - uint8_t IdlePwrSavingFeaturesCtrl; - uint8_t RuntimePwrSavingFeaturesCtrl; + uint16_t Reserved2; int16_t GfxclkFmin; // MHz int16_t GfxclkFmax; // MHz @@ -806,6 +813,9 @@ typedef enum { #define INVALID_BOARD_GPIO 0xFF +#define MARKETING_BASE_CLOCKS 0 +#define MARKETING_GAME_CLOCKS 1 +#define MARKETING_BOOST_CLOCKS 2 typedef struct { //PLL 0 @@ -1096,10 +1106,15 @@ typedef struct { uint16_t DcsExitHysteresis; //The min amount of time power credit accumulator should have a value > 0 before SMU exits the DCS throttling phase. uint16_t DcsTimeout; //This is the amount of time SMU FW waits for RLC to put GFX into GFXOFF before reverting to the fallback mechanism of throttling GFXCLK to Fmin. + uint8_t FoptEnabled; + uint8_t DcsSpare2[3]; + uint32_t DcsFoptM; //Tuning paramters to shift Fopt calculation + uint32_t DcsFoptB; //Tuning paramters to shift Fopt calculation - uint32_t DcsSpare[16]; + uint32_t DcsSpare[11]; // UCLK section + uint16_t ShadowFreqTableUclk[NUM_UCLK_DPM_LEVELS]; // In MHz uint8_t UseStrobeModeOptimizations; //Set to indicate that FW should use strobe mode optimizations uint8_t PaddingMem[3]; @@ -1245,8 +1260,13 @@ typedef struct { QuadraticInt_t qFeffCoeffBaseClock[POWER_SOURCE_COUNT]; QuadraticInt_t qFeffCoeffBoostClock[POWER_SOURCE_COUNT]; + uint16_t TemperatureLimit_Hynix; // In degrees Celsius. Memory temperature limit associated with Hynix + uint16_t TemperatureLimit_Micron; // In degrees Celsius. Memory temperature limit associated with Micron + uint16_t TemperatureFwCtfLimit_Hynix; + uint16_t TemperatureFwCtfLimit_Micron; + // SECTION: Sku Reserved - uint32_t Spare[43]; + uint32_t Spare[41]; // Padding for MMHUB - do not modify this uint32_t MmHubPadding[8]; @@ -1318,8 +1338,9 @@ typedef struct { // UCLK Spread Spectrum uint8_t UclkSpreadPercent[MEM_VENDOR_COUNT]; + uint8_t GfxclkSpreadEnable; + // FCLK Spread Spectrum - uint8_t FclkSpreadPadding; uint8_t FclkSpreadPercent; // Q4.4 uint16_t FclkSpreadFreq; // kHz @@ -1444,6 +1465,8 @@ typedef struct { uint8_t ThrottlingPercentage[THROTTLER_COUNT]; + uint8_t VmaxThrottlingPercentage; + uint8_t Padding1[3]; //metrics for D3hot entry/exit and driver ARM msgs uint32_t D3HotEntryCountPerMode[D3HOT_SEQUENCE_COUNT]; @@ -1463,7 +1486,7 @@ typedef struct { typedef struct { SmuMetrics_t SmuMetrics; - uint32_t Spare[30]; + uint32_t Spare[29]; // Padding - ignore uint32_t MmHubPadding[8]; // SMU internal use diff --git a/drivers/gpu/drm/amd/pm/swsmu/inc/pmfw_if/smu13_driver_if_v13_0_4.h b/drivers/gpu/drm/amd/pm/swsmu/inc/pmfw_if/smu13_driver_if_v13_0_4.h index 2162ecd1057d..fee9293b3f97 100644 --- a/drivers/gpu/drm/amd/pm/swsmu/inc/pmfw_if/smu13_driver_if_v13_0_4.h +++ b/drivers/gpu/drm/amd/pm/swsmu/inc/pmfw_if/smu13_driver_if_v13_0_4.h @@ -27,7 +27,7 @@ // *** IMPORTANT *** // SMU TEAM: Always increment the interface version if // any structure is changed in this file -#define PMFW_DRIVER_IF_VERSION 8 +#define SMU13_0_4_DRIVER_IF_VERSION 8 typedef struct { int32_t value; diff --git a/drivers/gpu/drm/amd/pm/swsmu/inc/pmfw_if/smu13_driver_if_v13_0_5.h b/drivers/gpu/drm/amd/pm/swsmu/inc/pmfw_if/smu13_driver_if_v13_0_5.h index aa971412b434..7589faa0232d 100644 --- a/drivers/gpu/drm/amd/pm/swsmu/inc/pmfw_if/smu13_driver_if_v13_0_5.h +++ b/drivers/gpu/drm/amd/pm/swsmu/inc/pmfw_if/smu13_driver_if_v13_0_5.h @@ -23,7 +23,7 @@ #ifndef __SMU13_DRIVER_IF_V13_0_5_H__ #define __SMU13_DRIVER_IF_V13_0_5_H__ -#define PMFW_DRIVER_IF_VERSION 4 +#define SMU13_0_5_DRIVER_IF_VERSION 4 // Throttler Status Bitmask #define THROTTLER_STATUS_BIT_SPL 0 diff --git a/drivers/gpu/drm/amd/pm/swsmu/inc/pmfw_if/smu13_driver_if_v13_0_6.h b/drivers/gpu/drm/amd/pm/swsmu/inc/pmfw_if/smu13_driver_if_v13_0_6.h index be596777cd2c..ca4a5e99ccd1 100644 --- a/drivers/gpu/drm/amd/pm/swsmu/inc/pmfw_if/smu13_driver_if_v13_0_6.h +++ b/drivers/gpu/drm/amd/pm/swsmu/inc/pmfw_if/smu13_driver_if_v13_0_6.h @@ -26,7 +26,7 @@ // *** IMPORTANT *** // PMFW TEAM: Always increment the interface version if // anything is changed in this file -#define SMU13_0_6_DRIVER_IF_VERSION 0x08042022 +#define SMU13_0_6_DRIVER_IF_VERSION 0x08042024 //I2C Interface #define NUM_I2C_CONTROLLERS 8 @@ -106,7 +106,7 @@ typedef enum { } UCLK_DPM_MODE_e; typedef struct { - //0-26 SOC, 27-29 SOCIO + //0-23 SOC, 24-26 SOCIO, 27-29 SOC uint16_t avgPsmCount[30]; uint16_t minPsmCount[30]; float avgPsmVoltage[30]; @@ -121,6 +121,17 @@ typedef struct { float minPsmVoltage[30]; } AvfsDebugTableXcd_t; +// Defines used for IH-based thermal interrupts to GFX driver - A/X only +#define IH_INTERRUPT_ID_TO_DRIVER 0xFE +#define IH_INTERRUPT_CONTEXT_ID_THERMAL_THROTTLING 0x7 + +//thermal over-temp mask defines for IH interrupt to host +#define THROTTLER_PROCHOT_BIT 0 +#define THROTTLER_PPT_BIT 1 +#define THROTTLER_THERMAL_SOCKET_BIT 2//AID, XCD, CCD throttling +#define THROTTLER_THERMAL_VR_BIT 3//VRHOT +#define THROTTLER_THERMAL_HBM_BIT 4 + // These defines are used with the following messages: // SMC_MSG_TransferTableDram2Smu // SMC_MSG_TransferTableSmu2Dram diff --git a/drivers/gpu/drm/amd/pm/swsmu/inc/pmfw_if/smu13_driver_if_v13_0_7.h b/drivers/gpu/drm/amd/pm/swsmu/inc/pmfw_if/smu13_driver_if_v13_0_7.h index 4c46a0392451..62b7c0daff68 100644 --- a/drivers/gpu/drm/amd/pm/swsmu/inc/pmfw_if/smu13_driver_if_v13_0_7.h +++ b/drivers/gpu/drm/amd/pm/swsmu/inc/pmfw_if/smu13_driver_if_v13_0_7.h @@ -25,7 +25,7 @@ // *** IMPORTANT *** // PMFW TEAM: Always increment the interface version on any change to this file -#define SMU13_DRIVER_IF_VERSION 0x35 +#define SMU13_0_7_DRIVER_IF_VERSION 0x35 //Increment this version if SkuTable_t or BoardTable_t change #define PPTABLE_VERSION 0x27 @@ -683,18 +683,12 @@ typedef struct { #define PP_OD_FEATURE_GFX_VF_CURVE_BIT 0 -#define PP_OD_FEATURE_VMAX_BIT 1 #define PP_OD_FEATURE_PPT_BIT 2 #define PP_OD_FEATURE_FAN_CURVE_BIT 3 -#define PP_OD_FEATURE_FREQ_DETER_BIT 4 -#define PP_OD_FEATURE_FULL_CTRL_BIT 5 -#define PP_OD_FEATURE_TDC_BIT 6 #define PP_OD_FEATURE_GFXCLK_BIT 7 #define PP_OD_FEATURE_UCLK_BIT 8 #define PP_OD_FEATURE_ZERO_FAN_BIT 9 #define PP_OD_FEATURE_TEMPERATURE_BIT 10 -#define PP_OD_FEATURE_POWER_FEATURE_CTRL_BIT 11 -#define PP_OD_FEATURE_ASIC_TDC_BIT 12 #define PP_OD_FEATURE_COUNT 13 typedef enum { @@ -713,10 +707,8 @@ typedef struct { //Voltage control int16_t VoltageOffsetPerZoneBoundary[PP_NUM_OD_VF_CURVE_POINTS]; - uint16_t VddGfxVmax; // in mV - uint8_t IdlePwrSavingFeaturesCtrl; - uint8_t RuntimePwrSavingFeaturesCtrl; + uint32_t Reserved; //Frequency changes int16_t GfxclkFmin; // MHz @@ -741,12 +733,7 @@ typedef struct { uint8_t MaxOpTemp; uint8_t Padding[4]; - uint16_t GfxVoltageFullCtrlMode; - uint16_t GfxclkFullCtrlMode; - uint16_t UclkFullCtrlMode; - int16_t AsicTdc; - - uint32_t Spare[10]; + uint32_t Spare[12]; uint32_t MmHubPadding[8]; // SMU internal use. Adding here instead of external as a workaround } OverDriveTable_t; @@ -759,10 +746,9 @@ typedef struct { uint32_t FeatureCtrlMask; int16_t VoltageOffsetPerZoneBoundary; - uint16_t VddGfxVmax; // in mV + uint16_t Reserved1; - uint8_t IdlePwrSavingFeaturesCtrl; - uint8_t RuntimePwrSavingFeaturesCtrl; + uint16_t Reserved2; int16_t GfxclkFmin; // MHz int16_t GfxclkFmax; // MHz @@ -785,12 +771,7 @@ typedef struct { uint8_t MaxOpTemp; uint8_t Padding[4]; - uint16_t GfxVoltageFullCtrlMode; - uint16_t GfxclkFullCtrlMode; - uint16_t UclkFullCtrlMode; - int16_t AsicTdc; - - uint32_t Spare[10]; + uint32_t Spare[12]; } OverDriveLimits_t; diff --git a/drivers/gpu/drm/amd/pm/swsmu/inc/pmfw_if/smu13_driver_if_yellow_carp.h b/drivers/gpu/drm/amd/pm/swsmu/inc/pmfw_if/smu13_driver_if_yellow_carp.h index 25540cb28208..7417634827ad 100644 --- a/drivers/gpu/drm/amd/pm/swsmu/inc/pmfw_if/smu13_driver_if_yellow_carp.h +++ b/drivers/gpu/drm/amd/pm/swsmu/inc/pmfw_if/smu13_driver_if_yellow_carp.h @@ -26,7 +26,7 @@ // *** IMPORTANT *** // SMU TEAM: Always increment the interface version if // any structure is changed in this file -#define SMU13_DRIVER_IF_VERSION 4 +#define SMU13_YELLOW_CARP_DRIVER_IF_VERSION 4 typedef struct { int32_t value; diff --git a/drivers/gpu/drm/amd/pm/swsmu/inc/pmfw_if/smu_v13_0_6_pmfw.h b/drivers/gpu/drm/amd/pm/swsmu/inc/pmfw_if/smu_v13_0_6_pmfw.h index bdccbb4a6276..252aef190c5c 100644 --- a/drivers/gpu/drm/amd/pm/swsmu/inc/pmfw_if/smu_v13_0_6_pmfw.h +++ b/drivers/gpu/drm/amd/pm/swsmu/inc/pmfw_if/smu_v13_0_6_pmfw.h @@ -123,9 +123,9 @@ typedef enum { VOLTAGE_GUARDBAND_COUNT } GFX_GUARDBAND_e; -#define SMU_METRICS_TABLE_VERSION 0x1 +#define SMU_METRICS_TABLE_VERSION 0x5 -typedef struct { +typedef struct __attribute__((packed, aligned(4))) { uint32_t AccumulationCounter; //TEMPERATURE @@ -198,11 +198,20 @@ typedef struct { uint32_t SocketThmResidencyAcc; uint32_t VrThmResidencyAcc; uint32_t HbmThmResidencyAcc; + uint32_t spare; + + // New Items at end to maintain driver compatibility + uint32_t GfxclkFrequency[8]; + + //PSNs + uint64_t PublicSerialNumber_AID[4]; + uint64_t PublicSerialNumber_XCD[8]; + uint64_t PublicSerialNumber_CCD[12]; } MetricsTable_t; -#define SMU_VF_METRICS_TABLE_VERSION 0x1 +#define SMU_VF_METRICS_TABLE_VERSION 0x3 -typedef struct { +typedef struct __attribute__((packed, aligned(4))) { uint32_t AccumulationCounter; uint32_t InstGfxclk_TargFreq; uint64_t AccGfxclk_TargFreq; diff --git a/drivers/gpu/drm/amd/pm/swsmu/inc/pmfw_if/smu_v13_0_6_ppsmc.h b/drivers/gpu/drm/amd/pm/swsmu/inc/pmfw_if/smu_v13_0_6_ppsmc.h index b838e8db395a..ae4f44c4b877 100644 --- a/drivers/gpu/drm/amd/pm/swsmu/inc/pmfw_if/smu_v13_0_6_ppsmc.h +++ b/drivers/gpu/drm/amd/pm/swsmu/inc/pmfw_if/smu_v13_0_6_ppsmc.h @@ -82,7 +82,8 @@ #define PPSMC_MSG_SetSoftMaxGfxClk 0x31 #define PPSMC_MSG_GetMinGfxDpmFreq 0x32 #define PPSMC_MSG_GetMaxGfxDpmFreq 0x33 -#define PPSMC_Message_Count 0x34 +#define PPSMC_MSG_PrepareForDriverUnload 0x34 +#define PPSMC_Message_Count 0x35 //PPSMC Reset Types for driver msg argument #define PPSMC_RESET_TYPE_DRIVER_MODE_1_RESET 0x1 diff --git a/drivers/gpu/drm/amd/pm/swsmu/inc/smu_v13_0.h b/drivers/gpu/drm/amd/pm/swsmu/inc/smu_v13_0.h index df3baaab0037..6a0ac0bbaace 100644 --- a/drivers/gpu/drm/amd/pm/swsmu/inc/smu_v13_0.h +++ b/drivers/gpu/drm/amd/pm/swsmu/inc/smu_v13_0.h @@ -25,17 +25,6 @@ #include "amdgpu_smu.h" -#define SMU13_DRIVER_IF_VERSION_INV 0xFFFFFFFF -#define SMU13_DRIVER_IF_VERSION_YELLOW_CARP 0x04 -#define SMU13_DRIVER_IF_VERSION_ALDE 0x08 -#define SMU13_DRIVER_IF_VERSION_SMU_V13_0_0_0 0x37 -#define SMU13_DRIVER_IF_VERSION_SMU_V13_0_4 0x08 -#define SMU13_DRIVER_IF_VERSION_SMU_V13_0_5 0x04 -#define SMU13_DRIVER_IF_VERSION_SMU_V13_0_0_10 0x32 -#define SMU13_DRIVER_IF_VERSION_SMU_V13_0_7 0x37 -#define SMU13_DRIVER_IF_VERSION_SMU_V13_0_10 0x1D -#define SMU13_DRIVER_IF_VERSION_SMU_V13_0_6 0x0 - #define SMU13_MODE1_RESET_WAIT_TIME_IN_MS 500 //500ms /* MP Apertures */ @@ -62,6 +51,8 @@ #define CTF_OFFSET_HOTSPOT 5 #define CTF_OFFSET_MEM 5 +#define SMU_13_VCLK_SHIFT 16 + extern const int pmfw_decoded_link_speed[5]; extern const int pmfw_decoded_link_width[7]; @@ -130,6 +121,7 @@ struct smu_13_0_power_context { uint32_t power_source; uint8_t in_power_limit_boost_mode; enum smu_13_0_power_state power_state; + atomic_t throttle_status; }; #if defined(SWSMU_CODE_LAYER_L2) || defined(SWSMU_CODE_LAYER_L3) diff --git a/drivers/gpu/drm/amd/pm/swsmu/inc/smu_v13_0_7_pptable.h b/drivers/gpu/drm/amd/pm/swsmu/inc/smu_v13_0_7_pptable.h index 478862ded0bd..eadbe0149cae 100644 --- a/drivers/gpu/drm/amd/pm/swsmu/inc/smu_v13_0_7_pptable.h +++ b/drivers/gpu/drm/amd/pm/swsmu/inc/smu_v13_0_7_pptable.h @@ -38,13 +38,12 @@ #define SMU_13_0_7_PP_THERMALCONTROLLER_NONE 0 #define SMU_13_0_7_PP_THERMALCONTROLLER_NAVI21 28 -#define SMU_13_0_7_PP_OVERDRIVE_VERSION 0x81 // OverDrive 8 Table Version 0.2 +#define SMU_13_0_7_PP_OVERDRIVE_VERSION 0x83 // OverDrive 8 Table Version 0.2 #define SMU_13_0_7_PP_POWERSAVINGCLOCK_VERSION 0x01 // Power Saving Clock Table Version 1.00 enum SMU_13_0_7_ODFEATURE_CAP { SMU_13_0_7_ODCAP_GFXCLK_LIMITS = 0, - SMU_13_0_7_ODCAP_GFXCLK_CURVE, SMU_13_0_7_ODCAP_UCLK_LIMITS, SMU_13_0_7_ODCAP_POWER_LIMIT, SMU_13_0_7_ODCAP_FAN_ACOUSTIC_LIMIT, @@ -59,13 +58,13 @@ enum SMU_13_0_7_ODFEATURE_CAP SMU_13_0_7_ODCAP_FAN_CURVE, SMU_13_0_7_ODCAP_AUTO_FAN_ACOUSTIC_LIMIT, SMU_13_0_7_ODCAP_POWER_MODE, + SMU_13_0_7_ODCAP_PER_ZONE_GFX_VOLTAGE_OFFSET, SMU_13_0_7_ODCAP_COUNT, }; enum SMU_13_0_7_ODFEATURE_ID { SMU_13_0_7_ODFEATURE_GFXCLK_LIMITS = 1 << SMU_13_0_7_ODCAP_GFXCLK_LIMITS, //GFXCLK Limit feature - SMU_13_0_7_ODFEATURE_GFXCLK_CURVE = 1 << SMU_13_0_7_ODCAP_GFXCLK_CURVE, //GFXCLK Curve feature SMU_13_0_7_ODFEATURE_UCLK_LIMITS = 1 << SMU_13_0_7_ODCAP_UCLK_LIMITS, //UCLK Limit feature SMU_13_0_7_ODFEATURE_POWER_LIMIT = 1 << SMU_13_0_7_ODCAP_POWER_LIMIT, //Power Limit feature SMU_13_0_7_ODFEATURE_FAN_ACOUSTIC_LIMIT = 1 << SMU_13_0_7_ODCAP_FAN_ACOUSTIC_LIMIT, //Fan Acoustic RPM feature @@ -80,6 +79,7 @@ enum SMU_13_0_7_ODFEATURE_ID SMU_13_0_7_ODFEATURE_FAN_CURVE = 1 << SMU_13_0_7_ODCAP_FAN_CURVE, //Fan Curve feature SMU_13_0_7_ODFEATURE_AUTO_FAN_ACOUSTIC_LIMIT = 1 << SMU_13_0_7_ODCAP_AUTO_FAN_ACOUSTIC_LIMIT, //Auto Fan Acoustic RPM feature SMU_13_0_7_ODFEATURE_POWER_MODE = 1 << SMU_13_0_7_ODCAP_POWER_MODE, //Optimized GPU Power Mode feature + SMU_13_0_7_ODFEATURE_PER_ZONE_GFX_VOLTAGE_OFFSET = 1 << SMU_13_0_7_ODCAP_PER_ZONE_GFX_VOLTAGE_OFFSET, //Perzone voltage offset feature SMU_13_0_7_ODFEATURE_COUNT = 16, }; @@ -89,10 +89,6 @@ enum SMU_13_0_7_ODSETTING_ID { SMU_13_0_7_ODSETTING_GFXCLKFMAX = 0, SMU_13_0_7_ODSETTING_GFXCLKFMIN, - SMU_13_0_7_ODSETTING_CUSTOM_GFX_VF_CURVE_A, - SMU_13_0_7_ODSETTING_CUSTOM_GFX_VF_CURVE_B, - SMU_13_0_7_ODSETTING_CUSTOM_GFX_VF_CURVE_C, - SMU_13_0_7_ODSETTING_CUSTOM_CURVE_VFT_FMIN, SMU_13_0_7_ODSETTING_UCLKFMIN, SMU_13_0_7_ODSETTING_UCLKFMAX, SMU_13_0_7_ODSETTING_POWERPERCENTAGE, @@ -117,6 +113,12 @@ enum SMU_13_0_7_ODSETTING_ID SMU_13_0_7_ODSETTING_FAN_CURVE_SPEED_5, SMU_13_0_7_ODSETTING_AUTO_FAN_ACOUSTIC_LIMIT, SMU_13_0_7_ODSETTING_POWER_MODE, + SMU_13_0_7_ODSETTING_PER_ZONE_GFX_VOLTAGE_OFFSET_POINT_1, + SMU_13_0_7_ODSETTING_PER_ZONE_GFX_VOLTAGE_OFFSET_POINT_2, + SMU_13_0_7_ODSETTING_PER_ZONE_GFX_VOLTAGE_OFFSET_POINT_3, + SMU_13_0_7_ODSETTING_PER_ZONE_GFX_VOLTAGE_OFFSET_POINT_4, + SMU_13_0_7_ODSETTING_PER_ZONE_GFX_VOLTAGE_OFFSET_POINT_5, + SMU_13_0_7_ODSETTING_PER_ZONE_GFX_VOLTAGE_OFFSET_POINT_6, SMU_13_0_7_ODSETTING_COUNT, }; #define SMU_13_0_7_MAX_ODSETTING 64 //Maximum Number of ODSettings 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 c4000518dc56..275f708db636 100644 --- a/drivers/gpu/drm/amd/pm/swsmu/smu11/navi10_ppt.c +++ b/drivers/gpu/drm/amd/pm/swsmu/smu11/navi10_ppt.c @@ -3413,26 +3413,8 @@ static int navi10_post_smu_init(struct smu_context *smu) return 0; ret = navi10_run_umc_cdr_workaround(smu); - if (ret) { + if (ret) dev_err(adev->dev, "Failed to apply umc cdr workaround!\n"); - return ret; - } - - if (!smu->dc_controlled_by_gpio) { - /* - * For Navi1X, manually switch it to AC mode as PMFW - * may boot it with DC mode. - */ - ret = smu_v11_0_set_power_source(smu, - adev->pm.ac_power ? - SMU_POWER_SOURCE_AC : - SMU_POWER_SOURCE_DC); - if (ret) { - dev_err(adev->dev, "Failed to switch to %s mode!\n", - adev->pm.ac_power ? "AC" : "DC"); - return ret; - } - } return ret; } 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 75f18681e984..f7ed3e655e39 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 @@ -431,7 +431,13 @@ static int sienna_cichlid_append_powerplay_table(struct smu_context *smu) { struct atom_smc_dpm_info_v4_9 *smc_dpm_table; int index, ret; - I2cControllerConfig_t *table_member; + PPTable_beige_goby_t *ppt_beige_goby; + PPTable_t *ppt; + + if (smu->adev->ip_versions[MP1_HWIP][0] == IP_VERSION(11, 0, 13)) + ppt_beige_goby = smu->smu_table.driver_pptable; + else + ppt = smu->smu_table.driver_pptable; index = get_index_into_master_table(atom_master_list_of_data_tables_v2_1, smc_dpm_info); @@ -440,9 +446,13 @@ static int sienna_cichlid_append_powerplay_table(struct smu_context *smu) (uint8_t **)&smc_dpm_table); if (ret) return ret; - GET_PPTABLE_MEMBER(I2cControllers, &table_member); - memcpy(table_member, smc_dpm_table->I2cControllers, - sizeof(*smc_dpm_table) - sizeof(smc_dpm_table->table_header)); + + if (smu->adev->ip_versions[MP1_HWIP][0] == IP_VERSION(11, 0, 13)) + smu_memcpy_trailing(ppt_beige_goby, I2cControllers, BoardReserved, + smc_dpm_table, I2cControllers); + else + smu_memcpy_trailing(ppt, I2cControllers, BoardReserved, + smc_dpm_table, I2cControllers); return 0; } @@ -2067,33 +2077,94 @@ static int sienna_cichlid_display_disable_memory_clock_switch(struct smu_context return ret; } +static void sienna_cichlid_get_override_pcie_settings(struct smu_context *smu, + uint32_t *gen_speed_override, + uint32_t *lane_width_override) +{ + struct amdgpu_device *adev = smu->adev; + + *gen_speed_override = 0xff; + *lane_width_override = 0xff; + + switch (adev->pdev->device) { + case 0x73A0: + case 0x73A1: + case 0x73A2: + case 0x73A3: + case 0x73AB: + case 0x73AE: + /* Bit 7:0: PCIE lane width, 1 to 7 corresponds is x1 to x32 */ + *lane_width_override = 6; + break; + case 0x73E0: + case 0x73E1: + case 0x73E3: + *lane_width_override = 4; + break; + case 0x7420: + case 0x7421: + case 0x7422: + case 0x7423: + case 0x7424: + *lane_width_override = 3; + break; + default: + break; + } +} + +#define MAX(a, b) ((a) > (b) ? (a) : (b)) + static int sienna_cichlid_update_pcie_parameters(struct smu_context *smu, uint32_t pcie_gen_cap, uint32_t pcie_width_cap) { struct smu_11_0_dpm_context *dpm_context = smu->smu_dpm.dpm_context; - - uint32_t smu_pcie_arg; + struct smu_11_0_pcie_table *pcie_table = &dpm_context->dpm_tables.pcie_table; + uint32_t gen_speed_override, lane_width_override; uint8_t *table_member1, *table_member2; + uint32_t min_gen_speed, max_gen_speed; + uint32_t min_lane_width, max_lane_width; + uint32_t smu_pcie_arg; int ret, i; GET_PPTABLE_MEMBER(PcieGenSpeed, &table_member1); GET_PPTABLE_MEMBER(PcieLaneCount, &table_member2); - /* lclk dpm table setup */ - for (i = 0; i < MAX_PCIE_CONF; i++) { - dpm_context->dpm_tables.pcie_table.pcie_gen[i] = table_member1[i]; - dpm_context->dpm_tables.pcie_table.pcie_lane[i] = table_member2[i]; + sienna_cichlid_get_override_pcie_settings(smu, + &gen_speed_override, + &lane_width_override); + + /* PCIE gen speed override */ + if (gen_speed_override != 0xff) { + min_gen_speed = MIN(pcie_gen_cap, gen_speed_override); + max_gen_speed = MIN(pcie_gen_cap, gen_speed_override); + } else { + min_gen_speed = MAX(0, table_member1[0]); + max_gen_speed = MIN(pcie_gen_cap, table_member1[1]); + min_gen_speed = min_gen_speed > max_gen_speed ? + max_gen_speed : min_gen_speed; + } + pcie_table->pcie_gen[0] = min_gen_speed; + pcie_table->pcie_gen[1] = max_gen_speed; + + /* PCIE lane width override */ + if (lane_width_override != 0xff) { + min_lane_width = MIN(pcie_width_cap, lane_width_override); + max_lane_width = MIN(pcie_width_cap, lane_width_override); + } else { + min_lane_width = MAX(1, table_member2[0]); + max_lane_width = MIN(pcie_width_cap, table_member2[1]); + min_lane_width = min_lane_width > max_lane_width ? + max_lane_width : min_lane_width; } + pcie_table->pcie_lane[0] = min_lane_width; + pcie_table->pcie_lane[1] = max_lane_width; for (i = 0; i < NUM_LINK_LEVELS; i++) { - smu_pcie_arg = (i << 16) | - ((table_member1[i] <= pcie_gen_cap) ? - (table_member1[i] << 8) : - (pcie_gen_cap << 8)) | - ((table_member2[i] <= pcie_width_cap) ? - table_member2[i] : - pcie_width_cap); + smu_pcie_arg = (i << 16 | + pcie_table->pcie_gen[i] << 8 | + pcie_table->pcie_lane[i]); ret = smu_cmn_send_smc_msg_with_param(smu, SMU_MSG_OverridePcieParameters, @@ -2101,11 +2172,6 @@ static int sienna_cichlid_update_pcie_parameters(struct smu_context *smu, NULL); if (ret) return ret; - - if (table_member1[i] > pcie_gen_cap) - dpm_context->dpm_tables.pcie_table.pcie_gen[i] = pcie_gen_cap; - if (table_member2[i] > pcie_width_cap) - dpm_context->dpm_tables.pcie_table.pcie_lane[i] = pcie_width_cap; } return 0; 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 7433dcaa16e0..067b4e0b026c 100644 --- a/drivers/gpu/drm/amd/pm/swsmu/smu11/vangogh_ppt.c +++ b/drivers/gpu/drm/amd/pm/swsmu/smu11/vangogh_ppt.c @@ -582,7 +582,7 @@ static int vangogh_print_legacy_clk_levels(struct smu_context *smu, DpmClocks_t *clk_table = smu->smu_table.clocks_table; SmuMetrics_legacy_t metrics; struct smu_dpm_context *smu_dpm_ctx = &(smu->smu_dpm); - int i, size = 0, ret = 0; + int i, idx, size = 0, ret = 0; uint32_t cur_value = 0, value = 0, count = 0; bool cur_value_match_level = false; @@ -656,7 +656,8 @@ static int vangogh_print_legacy_clk_levels(struct smu_context *smu, case SMU_MCLK: case SMU_FCLK: for (i = 0; i < count; i++) { - ret = vangogh_get_dpm_clk_limited(smu, clk_type, i, &value); + idx = (clk_type == SMU_FCLK || clk_type == SMU_MCLK) ? (count - i - 1) : i; + ret = vangogh_get_dpm_clk_limited(smu, clk_type, idx, &value); if (ret) return ret; if (!value) @@ -683,7 +684,7 @@ static int vangogh_print_clk_levels(struct smu_context *smu, DpmClocks_t *clk_table = smu->smu_table.clocks_table; SmuMetrics_t metrics; struct smu_dpm_context *smu_dpm_ctx = &(smu->smu_dpm); - int i, size = 0, ret = 0; + int i, idx, size = 0, ret = 0; uint32_t cur_value = 0, value = 0, count = 0; bool cur_value_match_level = false; uint32_t min, max; @@ -765,7 +766,8 @@ static int vangogh_print_clk_levels(struct smu_context *smu, case SMU_MCLK: case SMU_FCLK: for (i = 0; i < count; i++) { - ret = vangogh_get_dpm_clk_limited(smu, clk_type, i, &value); + idx = (clk_type == SMU_FCLK || clk_type == SMU_MCLK) ? (count - i - 1) : i; + ret = vangogh_get_dpm_clk_limited(smu, clk_type, idx, &value); if (ret) return ret; if (!value) 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 5cdc07165480..8a8ba25c9ad7 100644 --- a/drivers/gpu/drm/amd/pm/swsmu/smu12/renoir_ppt.c +++ b/drivers/gpu/drm/amd/pm/swsmu/smu12/renoir_ppt.c @@ -494,7 +494,7 @@ static int renoir_set_fine_grain_gfx_freq_parameters(struct smu_context *smu) static int renoir_print_clk_levels(struct smu_context *smu, enum smu_clk_type clk_type, char *buf) { - int i, size = 0, ret = 0; + int i, idx, size = 0, ret = 0; uint32_t cur_value = 0, value = 0, count = 0, min = 0, max = 0; SmuMetrics_t metrics; struct smu_dpm_context *smu_dpm_ctx = &(smu->smu_dpm); @@ -594,7 +594,8 @@ static int renoir_print_clk_levels(struct smu_context *smu, case SMU_VCLK: case SMU_DCLK: for (i = 0; i < count; i++) { - ret = renoir_get_dpm_clk_limited(smu, clk_type, i, &value); + idx = (clk_type == SMU_FCLK || clk_type == SMU_MCLK) ? (count - i - 1) : i; + ret = renoir_get_dpm_clk_limited(smu, clk_type, idx, &value); if (ret) return ret; if (!value) diff --git a/drivers/gpu/drm/amd/pm/swsmu/smu13/aldebaran_ppt.c b/drivers/gpu/drm/amd/pm/swsmu/smu13/aldebaran_ppt.c index d30ec3005ea1..e80f122d8aec 100644 --- a/drivers/gpu/drm/amd/pm/swsmu/smu13/aldebaran_ppt.c +++ b/drivers/gpu/drm/amd/pm/swsmu/smu13/aldebaran_ppt.c @@ -2147,5 +2147,6 @@ void aldebaran_set_ppt_funcs(struct smu_context *smu) smu->clock_map = aldebaran_clk_map; smu->feature_map = aldebaran_feature_mask_map; smu->table_map = aldebaran_table_map; + smu->smc_driver_if_version = SMU13_DRIVER_IF_VERSION_ALDE; smu_v13_0_set_smu_mailbox_registers(smu); } diff --git a/drivers/gpu/drm/amd/pm/swsmu/smu13/smu_v13_0.c b/drivers/gpu/drm/amd/pm/swsmu/smu13/smu_v13_0.c index 393c6a7b9609..e52c563f0dac 100644 --- a/drivers/gpu/drm/amd/pm/swsmu/smu13/smu_v13_0.c +++ b/drivers/gpu/drm/amd/pm/swsmu/smu13/smu_v13_0.c @@ -269,45 +269,10 @@ int smu_v13_0_check_fw_version(struct smu_context *smu) smu_major = (smu_version >> 16) & 0xff; smu_minor = (smu_version >> 8) & 0xff; smu_debug = (smu_version >> 0) & 0xff; - if (smu->is_apu) + if (smu->is_apu || + adev->ip_versions[MP1_HWIP][0] == IP_VERSION(13, 0, 6)) adev->pm.fw_version = smu_version; - switch (adev->ip_versions[MP1_HWIP][0]) { - case IP_VERSION(13, 0, 2): - smu->smc_driver_if_version = SMU13_DRIVER_IF_VERSION_ALDE; - break; - case IP_VERSION(13, 0, 0): - smu->smc_driver_if_version = SMU13_DRIVER_IF_VERSION_SMU_V13_0_0_0; - break; - case IP_VERSION(13, 0, 10): - smu->smc_driver_if_version = SMU13_DRIVER_IF_VERSION_SMU_V13_0_0_10; - break; - case IP_VERSION(13, 0, 7): - smu->smc_driver_if_version = SMU13_DRIVER_IF_VERSION_SMU_V13_0_7; - break; - case IP_VERSION(13, 0, 1): - case IP_VERSION(13, 0, 3): - case IP_VERSION(13, 0, 8): - smu->smc_driver_if_version = SMU13_DRIVER_IF_VERSION_YELLOW_CARP; - break; - case IP_VERSION(13, 0, 4): - case IP_VERSION(13, 0, 11): - smu->smc_driver_if_version = SMU13_DRIVER_IF_VERSION_SMU_V13_0_4; - break; - case IP_VERSION(13, 0, 5): - smu->smc_driver_if_version = SMU13_DRIVER_IF_VERSION_SMU_V13_0_5; - break; - case IP_VERSION(13, 0, 6): - smu->smc_driver_if_version = SMU13_DRIVER_IF_VERSION_SMU_V13_0_6; - adev->pm.fw_version = smu_version; - break; - default: - dev_err(adev->dev, "smu unsupported IP version: 0x%x.\n", - adev->ip_versions[MP1_HWIP][0]); - smu->smc_driver_if_version = SMU13_DRIVER_IF_VERSION_INV; - break; - } - /* only for dGPU w/ SMU13*/ if (adev->pm.fw) dev_dbg(smu->adev->dev, "smu fw reported program %d, version = 0x%08x (%d.%d.%d)\n", @@ -502,17 +467,26 @@ int smu_v13_0_init_smc_tables(struct smu_context *smu) ret = -ENOMEM; goto err3_out; } + + smu_table->user_overdrive_table = + kzalloc(tables[SMU_TABLE_OVERDRIVE].size, GFP_KERNEL); + if (!smu_table->user_overdrive_table) { + ret = -ENOMEM; + goto err4_out; + } } smu_table->combo_pptable = kzalloc(tables[SMU_TABLE_COMBO_PPTABLE].size, GFP_KERNEL); if (!smu_table->combo_pptable) { ret = -ENOMEM; - goto err4_out; + goto err5_out; } return 0; +err5_out: + kfree(smu_table->user_overdrive_table); err4_out: kfree(smu_table->boot_overdrive_table); err3_out: @@ -532,12 +506,14 @@ int smu_v13_0_fini_smc_tables(struct smu_context *smu) kfree(smu_table->gpu_metrics_table); kfree(smu_table->combo_pptable); + kfree(smu_table->user_overdrive_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->combo_pptable = NULL; + smu_table->user_overdrive_table = NULL; smu_table->boot_overdrive_table = NULL; smu_table->overdrive_table = NULL; smu_table->max_sustainable_clocks = NULL; @@ -573,11 +549,11 @@ int smu_v13_0_init_power(struct smu_context *smu) if (smu_power->power_context || smu_power->power_context_size != 0) return -EINVAL; - smu_power->power_context = kzalloc(sizeof(struct smu_13_0_dpm_context), + smu_power->power_context = kzalloc(sizeof(struct smu_13_0_power_context), GFP_KERNEL); if (!smu_power->power_context) return -ENOMEM; - smu_power->power_context_size = sizeof(struct smu_13_0_dpm_context); + smu_power->power_context_size = sizeof(struct smu_13_0_power_context); return 0; } diff --git a/drivers/gpu/drm/amd/pm/swsmu/smu13/smu_v13_0_0_ppt.c b/drivers/gpu/drm/amd/pm/swsmu/smu13/smu_v13_0_0_ppt.c index 09405ef1e3c8..a6083957ae51 100644 --- a/drivers/gpu/drm/amd/pm/swsmu/smu13/smu_v13_0_0_ppt.c +++ b/drivers/gpu/drm/amd/pm/swsmu/smu13/smu_v13_0_0_ppt.c @@ -46,7 +46,6 @@ #include "asic_reg/mp/mp_13_0_0_sh_mask.h" #include "smu_cmn.h" #include "amdgpu_ras.h" -#include "umc_v8_10.h" /* * DO NOT use these for err/warn/info/debug messages. @@ -237,6 +236,7 @@ static struct cmn2asic_mapping smu_v13_0_0_table_map[SMU_TABLE_COUNT] = { [SMU_TABLE_COMBO_PPTABLE] = {1, TABLE_COMBO_PPTABLE}, TAB_MAP(I2C_COMMANDS), TAB_MAP(ECCINFO), + TAB_MAP(OVERDRIVE), }; static struct cmn2asic_mapping smu_v13_0_0_pwr_src_map[SMU_POWER_SOURCE_COUNT] = { @@ -331,6 +331,11 @@ static int smu_v13_0_0_check_powerplay_table(struct smu_context *smu) struct smu_13_0_0_powerplay_table *powerplay_table = table_context->power_play_table; struct smu_baco_context *smu_baco = &smu->smu_baco; + PPTable_t *pptable = smu->smu_table.driver_pptable; + const OverDriveLimits_t * const overdrive_upperlimits = + &pptable->SkuTable.OverDriveLimitsBasicMax; + const OverDriveLimits_t * const overdrive_lowerlimits = + &pptable->SkuTable.OverDriveLimitsMin; if (powerplay_table->platform_caps & SMU_13_0_0_PP_PLATFORM_CAP_HARDWAREDC) smu->dc_controlled_by_gpio = true; @@ -342,6 +347,10 @@ static int smu_v13_0_0_check_powerplay_table(struct smu_context *smu) if (powerplay_table->platform_caps & SMU_13_0_0_PP_PLATFORM_CAP_MACO) smu_baco->maco_support = true; + if (!overdrive_lowerlimits->FeatureCtrlMask || + !overdrive_upperlimits->FeatureCtrlMask) + smu->od_enabled = false; + table_context->thermal_controller_type = powerplay_table->thermal_controller_type; @@ -461,7 +470,7 @@ static int smu_v13_0_0_tables_init(struct smu_context *smu) PAGE_SIZE, AMDGPU_GEM_DOMAIN_VRAM); SMU_TABLE_INIT(tables, SMU_TABLE_I2C_COMMANDS, sizeof(SwI2cRequest_t), PAGE_SIZE, AMDGPU_GEM_DOMAIN_VRAM); - SMU_TABLE_INIT(tables, SMU_TABLE_OVERDRIVE, sizeof(OverDriveTable_t), + SMU_TABLE_INIT(tables, SMU_TABLE_OVERDRIVE, sizeof(OverDriveTableExternal_t), PAGE_SIZE, AMDGPU_GEM_DOMAIN_VRAM); SMU_TABLE_INIT(tables, SMU_TABLE_PMSTATUSLOG, SMU13_TOOL_SIZE, PAGE_SIZE, AMDGPU_GEM_DOMAIN_VRAM); @@ -1022,17 +1031,119 @@ static int smu_v13_0_0_get_current_clk_freq_by_table(struct smu_context *smu, value); } +static bool smu_v13_0_0_is_od_feature_supported(struct smu_context *smu, + int od_feature_bit) +{ + PPTable_t *pptable = smu->smu_table.driver_pptable; + const OverDriveLimits_t * const overdrive_upperlimits = + &pptable->SkuTable.OverDriveLimitsBasicMax; + + return overdrive_upperlimits->FeatureCtrlMask & (1U << od_feature_bit); +} + +static void smu_v13_0_0_get_od_setting_limits(struct smu_context *smu, + int od_feature_bit, + bool lower_boundary, + int32_t *min, + int32_t *max) +{ + PPTable_t *pptable = smu->smu_table.driver_pptable; + const OverDriveLimits_t * const overdrive_upperlimits = + &pptable->SkuTable.OverDriveLimitsBasicMax; + const OverDriveLimits_t * const overdrive_lowerlimits = + &pptable->SkuTable.OverDriveLimitsMin; + int32_t od_min_setting, od_max_setting; + + switch (od_feature_bit) { + case PP_OD_FEATURE_GFXCLK_BIT: + if (lower_boundary) { + od_min_setting = overdrive_lowerlimits->GfxclkFmin; + od_max_setting = overdrive_upperlimits->GfxclkFmin; + } else { + od_min_setting = overdrive_lowerlimits->GfxclkFmax; + od_max_setting = overdrive_upperlimits->GfxclkFmax; + } + break; + case PP_OD_FEATURE_UCLK_BIT: + if (lower_boundary) { + od_min_setting = overdrive_lowerlimits->UclkFmin; + od_max_setting = overdrive_upperlimits->UclkFmin; + } else { + od_min_setting = overdrive_lowerlimits->UclkFmax; + od_max_setting = overdrive_upperlimits->UclkFmax; + } + break; + case PP_OD_FEATURE_GFX_VF_CURVE_BIT: + od_min_setting = overdrive_lowerlimits->VoltageOffsetPerZoneBoundary; + od_max_setting = overdrive_upperlimits->VoltageOffsetPerZoneBoundary; + break; + default: + break; + } + + if (min) + *min = od_min_setting; + if (max) + *max = od_max_setting; +} + +static void smu_v13_0_0_dump_od_table(struct smu_context *smu, + OverDriveTableExternal_t *od_table) +{ + struct amdgpu_device *adev = smu->adev; + + dev_dbg(adev->dev, "OD: Gfxclk: (%d, %d)\n", od_table->OverDriveTable.GfxclkFmin, + od_table->OverDriveTable.GfxclkFmax); + dev_dbg(adev->dev, "OD: Uclk: (%d, %d)\n", od_table->OverDriveTable.UclkFmin, + od_table->OverDriveTable.UclkFmax); +} + +static int smu_v13_0_0_get_overdrive_table(struct smu_context *smu, + OverDriveTableExternal_t *od_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; +} + +static int smu_v13_0_0_upload_overdrive_table(struct smu_context *smu, + OverDriveTableExternal_t *od_table) +{ + int ret = 0; + + ret = smu_cmn_update_table(smu, + SMU_TABLE_OVERDRIVE, + 0, + (void *)od_table, + true); + if (ret) + dev_err(smu->adev->dev, "Failed to upload overdrive table!\n"); + + return ret; +} + static int smu_v13_0_0_print_clk_levels(struct smu_context *smu, enum smu_clk_type clk_type, char *buf) { struct smu_dpm_context *smu_dpm = &smu->smu_dpm; struct smu_13_0_dpm_context *dpm_context = smu_dpm->dpm_context; + OverDriveTableExternal_t *od_table = + (OverDriveTableExternal_t *)smu->smu_table.overdrive_table; struct smu_13_0_dpm_table *single_dpm_table; struct smu_13_0_pcie_table *pcie_table; const int link_width[] = {0, 1, 2, 4, 8, 12, 16}; uint32_t gen_speed, lane_width; int i, curr_freq, size = 0; + int32_t min_value, max_value; int ret = 0; smu_cmn_get_sysfs_buf(&buf, &size); @@ -1149,6 +1260,89 @@ static int smu_v13_0_0_print_clk_levels(struct smu_context *smu, "*" : ""); break; + case SMU_OD_SCLK: + if (!smu_v13_0_0_is_od_feature_supported(smu, + PP_OD_FEATURE_GFXCLK_BIT)) + break; + + size += sysfs_emit_at(buf, size, "OD_SCLK:\n"); + size += sysfs_emit_at(buf, size, "0: %uMhz\n1: %uMhz\n", + od_table->OverDriveTable.GfxclkFmin, + od_table->OverDriveTable.GfxclkFmax); + break; + + case SMU_OD_MCLK: + if (!smu_v13_0_0_is_od_feature_supported(smu, + PP_OD_FEATURE_UCLK_BIT)) + break; + + size += sysfs_emit_at(buf, size, "OD_MCLK:\n"); + size += sysfs_emit_at(buf, size, "0: %uMhz\n1: %uMHz\n", + od_table->OverDriveTable.UclkFmin, + od_table->OverDriveTable.UclkFmax); + break; + + case SMU_OD_VDDC_CURVE: + if (!smu_v13_0_0_is_od_feature_supported(smu, + PP_OD_FEATURE_GFX_VF_CURVE_BIT)) + break; + + size += sysfs_emit_at(buf, size, "OD_VDDC_CURVE:\n"); + for (i = 0; i < PP_NUM_OD_VF_CURVE_POINTS; i++) + size += sysfs_emit_at(buf, size, "%d: %dmv\n", + i, + od_table->OverDriveTable.VoltageOffsetPerZoneBoundary[i]); + break; + + case SMU_OD_RANGE: + if (!smu_v13_0_0_is_od_feature_supported(smu, PP_OD_FEATURE_GFXCLK_BIT) && + !smu_v13_0_0_is_od_feature_supported(smu, PP_OD_FEATURE_UCLK_BIT) && + !smu_v13_0_0_is_od_feature_supported(smu, PP_OD_FEATURE_GFX_VF_CURVE_BIT)) + break; + + size += sysfs_emit_at(buf, size, "%s:\n", "OD_RANGE"); + + if (smu_v13_0_0_is_od_feature_supported(smu, PP_OD_FEATURE_GFXCLK_BIT)) { + smu_v13_0_0_get_od_setting_limits(smu, + PP_OD_FEATURE_GFXCLK_BIT, + true, + &min_value, + NULL); + smu_v13_0_0_get_od_setting_limits(smu, + PP_OD_FEATURE_GFXCLK_BIT, + false, + NULL, + &max_value); + size += sysfs_emit_at(buf, size, "SCLK: %7uMhz %10uMhz\n", + min_value, max_value); + } + + if (smu_v13_0_0_is_od_feature_supported(smu, PP_OD_FEATURE_UCLK_BIT)) { + smu_v13_0_0_get_od_setting_limits(smu, + PP_OD_FEATURE_UCLK_BIT, + true, + &min_value, + NULL); + smu_v13_0_0_get_od_setting_limits(smu, + PP_OD_FEATURE_UCLK_BIT, + false, + NULL, + &max_value); + size += sysfs_emit_at(buf, size, "MCLK: %7uMhz %10uMhz\n", + min_value, max_value); + } + + if (smu_v13_0_0_is_od_feature_supported(smu, PP_OD_FEATURE_GFX_VF_CURVE_BIT)) { + smu_v13_0_0_get_od_setting_limits(smu, + PP_OD_FEATURE_GFX_VF_CURVE_BIT, + true, + &min_value, + &max_value); + size += sysfs_emit_at(buf, size, "VDDC_CURVE: %7dmv %10dmv\n", + min_value, max_value); + } + break; + default: break; } @@ -1156,6 +1350,222 @@ static int smu_v13_0_0_print_clk_levels(struct smu_context *smu, return size; } +static int smu_v13_0_0_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; + OverDriveTableExternal_t *od_table = + (OverDriveTableExternal_t *)table_context->overdrive_table; + struct amdgpu_device *adev = smu->adev; + uint32_t offset_of_featurectrlmask; + int32_t minimum, maximum; + uint32_t feature_ctrlmask; + int i, ret = 0; + + switch (type) { + case PP_OD_EDIT_SCLK_VDDC_TABLE: + if (!smu_v13_0_0_is_od_feature_supported(smu, PP_OD_FEATURE_GFXCLK_BIT)) { + dev_warn(adev->dev, "GFXCLK_LIMITS setting not supported!\n"); + return -ENOTSUPP; + } + + for (i = 0; i < size; i += 2) { + if (i + 2 > size) { + dev_info(adev->dev, "invalid number of input parameters %d\n", size); + return -EINVAL; + } + + switch (input[i]) { + case 0: + smu_v13_0_0_get_od_setting_limits(smu, + PP_OD_FEATURE_GFXCLK_BIT, + true, + &minimum, + &maximum); + if (input[i + 1] < minimum || + input[i + 1] > maximum) { + dev_info(adev->dev, "GfxclkFmin (%ld) must be within [%u, %u]!\n", + input[i + 1], minimum, maximum); + return -EINVAL; + } + + od_table->OverDriveTable.GfxclkFmin = input[i + 1]; + od_table->OverDriveTable.FeatureCtrlMask |= 1U << PP_OD_FEATURE_GFXCLK_BIT; + break; + + case 1: + smu_v13_0_0_get_od_setting_limits(smu, + PP_OD_FEATURE_GFXCLK_BIT, + false, + &minimum, + &maximum); + if (input[i + 1] < minimum || + input[i + 1] > maximum) { + dev_info(adev->dev, "GfxclkFmax (%ld) must be within [%u, %u]!\n", + input[i + 1], minimum, maximum); + return -EINVAL; + } + + od_table->OverDriveTable.GfxclkFmax = input[i + 1]; + od_table->OverDriveTable.FeatureCtrlMask |= 1U << PP_OD_FEATURE_GFXCLK_BIT; + break; + + default: + dev_info(adev->dev, "Invalid SCLK_VDDC_TABLE index: %ld\n", input[i]); + dev_info(adev->dev, "Supported indices: [0:min,1:max]\n"); + return -EINVAL; + } + } + + if (od_table->OverDriveTable.GfxclkFmin > od_table->OverDriveTable.GfxclkFmax) { + dev_err(adev->dev, + "Invalid setting: GfxclkFmin(%u) is bigger than GfxclkFmax(%u)\n", + (uint32_t)od_table->OverDriveTable.GfxclkFmin, + (uint32_t)od_table->OverDriveTable.GfxclkFmax); + return -EINVAL; + } + break; + + case PP_OD_EDIT_MCLK_VDDC_TABLE: + if (!smu_v13_0_0_is_od_feature_supported(smu, PP_OD_FEATURE_UCLK_BIT)) { + dev_warn(adev->dev, "UCLK_LIMITS setting not supported!\n"); + return -ENOTSUPP; + } + + for (i = 0; i < size; i += 2) { + if (i + 2 > size) { + dev_info(adev->dev, "invalid number of input parameters %d\n", size); + return -EINVAL; + } + + switch (input[i]) { + case 0: + smu_v13_0_0_get_od_setting_limits(smu, + PP_OD_FEATURE_UCLK_BIT, + true, + &minimum, + &maximum); + if (input[i + 1] < minimum || + input[i + 1] > maximum) { + dev_info(adev->dev, "UclkFmin (%ld) must be within [%u, %u]!\n", + input[i + 1], minimum, maximum); + return -EINVAL; + } + + od_table->OverDriveTable.UclkFmin = input[i + 1]; + od_table->OverDriveTable.FeatureCtrlMask |= 1U << PP_OD_FEATURE_UCLK_BIT; + break; + + case 1: + smu_v13_0_0_get_od_setting_limits(smu, + PP_OD_FEATURE_UCLK_BIT, + false, + &minimum, + &maximum); + if (input[i + 1] < minimum || + input[i + 1] > maximum) { + dev_info(adev->dev, "UclkFmax (%ld) must be within [%u, %u]!\n", + input[i + 1], minimum, maximum); + return -EINVAL; + } + + od_table->OverDriveTable.UclkFmax = input[i + 1]; + od_table->OverDriveTable.FeatureCtrlMask |= 1U << PP_OD_FEATURE_UCLK_BIT; + break; + + default: + dev_info(adev->dev, "Invalid MCLK_VDDC_TABLE index: %ld\n", input[i]); + dev_info(adev->dev, "Supported indices: [0:min,1:max]\n"); + return -EINVAL; + } + } + + if (od_table->OverDriveTable.UclkFmin > od_table->OverDriveTable.UclkFmax) { + dev_err(adev->dev, + "Invalid setting: UclkFmin(%u) is bigger than UclkFmax(%u)\n", + (uint32_t)od_table->OverDriveTable.UclkFmin, + (uint32_t)od_table->OverDriveTable.UclkFmax); + return -EINVAL; + } + break; + + case PP_OD_EDIT_VDDC_CURVE: + if (!smu_v13_0_0_is_od_feature_supported(smu, PP_OD_FEATURE_GFX_VF_CURVE_BIT)) { + dev_warn(adev->dev, "VF curve setting not supported!\n"); + return -ENOTSUPP; + } + + if (input[0] >= PP_NUM_OD_VF_CURVE_POINTS || + input[0] < 0) + return -EINVAL; + + smu_v13_0_0_get_od_setting_limits(smu, + PP_OD_FEATURE_GFX_VF_CURVE_BIT, + true, + &minimum, + &maximum); + if (input[1] < minimum || + input[1] > maximum) { + dev_info(adev->dev, "Voltage offset (%ld) must be within [%d, %d]!\n", + input[1], minimum, maximum); + return -EINVAL; + } + + od_table->OverDriveTable.VoltageOffsetPerZoneBoundary[input[0]] = input[1]; + od_table->OverDriveTable.FeatureCtrlMask |= 1U << PP_OD_FEATURE_GFX_VF_CURVE_BIT; + break; + + case PP_OD_RESTORE_DEFAULT_TABLE: + feature_ctrlmask = od_table->OverDriveTable.FeatureCtrlMask; + memcpy(od_table, + table_context->boot_overdrive_table, + sizeof(OverDriveTableExternal_t)); + od_table->OverDriveTable.FeatureCtrlMask = feature_ctrlmask; + fallthrough; + + case PP_OD_COMMIT_DPM_TABLE: + /* + * The member below instructs PMFW the settings focused in + * this single operation. + * `uint32_t FeatureCtrlMask;` + * It does not contain actual informations about user's custom + * settings. Thus we do not cache it. + */ + offset_of_featurectrlmask = offsetof(OverDriveTable_t, FeatureCtrlMask); + if (memcmp((u8 *)od_table + offset_of_featurectrlmask, + table_context->user_overdrive_table + offset_of_featurectrlmask, + sizeof(OverDriveTableExternal_t) - offset_of_featurectrlmask)) { + smu_v13_0_0_dump_od_table(smu, od_table); + + ret = smu_v13_0_0_upload_overdrive_table(smu, od_table); + if (ret) { + dev_err(adev->dev, "Failed to upload overdrive table!\n"); + return ret; + } + + od_table->OverDriveTable.FeatureCtrlMask = 0; + memcpy(table_context->user_overdrive_table + offset_of_featurectrlmask, + (u8 *)od_table + offset_of_featurectrlmask, + sizeof(OverDriveTableExternal_t) - offset_of_featurectrlmask); + + if (!memcmp(table_context->user_overdrive_table, + table_context->boot_overdrive_table, + sizeof(OverDriveTableExternal_t))) + smu->user_dpm_profile.user_od = false; + else + smu->user_dpm_profile.user_od = true; + } + break; + + default: + return -ENOSYS; + } + + return ret; +} + static int smu_v13_0_0_force_clk_levels(struct smu_context *smu, enum smu_clk_type clk_type, uint32_t mask) @@ -1384,6 +1794,78 @@ static ssize_t smu_v13_0_0_get_gpu_metrics(struct smu_context *smu, return sizeof(struct gpu_metrics_v1_3); } +static int smu_v13_0_0_set_default_od_settings(struct smu_context *smu) +{ + OverDriveTableExternal_t *od_table = + (OverDriveTableExternal_t *)smu->smu_table.overdrive_table; + OverDriveTableExternal_t *boot_od_table = + (OverDriveTableExternal_t *)smu->smu_table.boot_overdrive_table; + OverDriveTableExternal_t *user_od_table = + (OverDriveTableExternal_t *)smu->smu_table.user_overdrive_table; + OverDriveTableExternal_t user_od_table_bak; + int ret = 0; + int i; + + ret = smu_v13_0_0_get_overdrive_table(smu, boot_od_table); + if (ret) + return ret; + + smu_v13_0_0_dump_od_table(smu, boot_od_table); + + memcpy(od_table, + boot_od_table, + sizeof(OverDriveTableExternal_t)); + + /* + * For S3/S4/Runpm resume, we need to setup those overdrive tables again, + * but we have to preserve user defined values in "user_od_table". + */ + if (!smu->adev->in_suspend) { + memcpy(user_od_table, + boot_od_table, + sizeof(OverDriveTableExternal_t)); + smu->user_dpm_profile.user_od = false; + } else if (smu->user_dpm_profile.user_od) { + memcpy(&user_od_table_bak, + user_od_table, + sizeof(OverDriveTableExternal_t)); + memcpy(user_od_table, + boot_od_table, + sizeof(OverDriveTableExternal_t)); + user_od_table->OverDriveTable.GfxclkFmin = + user_od_table_bak.OverDriveTable.GfxclkFmin; + user_od_table->OverDriveTable.GfxclkFmax = + user_od_table_bak.OverDriveTable.GfxclkFmax; + user_od_table->OverDriveTable.UclkFmin = + user_od_table_bak.OverDriveTable.UclkFmin; + user_od_table->OverDriveTable.UclkFmax = + user_od_table_bak.OverDriveTable.UclkFmax; + for (i = 0; i < PP_NUM_OD_VF_CURVE_POINTS; i++) + user_od_table->OverDriveTable.VoltageOffsetPerZoneBoundary[i] = + user_od_table_bak.OverDriveTable.VoltageOffsetPerZoneBoundary[i]; + } + + return 0; +} + +static int smu_v13_0_0_restore_user_od_settings(struct smu_context *smu) +{ + struct smu_table_context *table_context = &smu->smu_table; + OverDriveTableExternal_t *od_table = table_context->overdrive_table; + OverDriveTableExternal_t *user_od_table = table_context->user_overdrive_table; + int res; + + user_od_table->OverDriveTable.FeatureCtrlMask = 1U << PP_OD_FEATURE_GFXCLK_BIT | + 1U << PP_OD_FEATURE_UCLK_BIT | + 1U << PP_OD_FEATURE_GFX_VF_CURVE_BIT; + res = smu_v13_0_0_upload_overdrive_table(smu, user_od_table); + user_od_table->OverDriveTable.FeatureCtrlMask = 0; + if (res == 0) + memcpy(od_table, user_od_table, sizeof(OverDriveTableExternal_t)); + + return res; +} + static int smu_v13_0_0_populate_umd_state_clk(struct smu_context *smu) { struct smu_13_0_dpm_context *dpm_context = @@ -1696,10 +2178,39 @@ static int smu_v13_0_0_set_power_profile_mode(struct smu_context *smu, } } - /* conv PP_SMC_POWER_PROFILE* to WORKLOAD_PPLIB_*_BIT */ - workload_type = smu_cmn_to_asic_specific_index(smu, + if (smu->power_profile_mode == PP_SMC_POWER_PROFILE_COMPUTE && + (((smu->adev->pdev->device == 0x744C) && (smu->adev->pdev->revision == 0xC8)) || + ((smu->adev->pdev->device == 0x744C) && (smu->adev->pdev->revision == 0xCC)))) { + ret = smu_cmn_update_table(smu, + SMU_TABLE_ACTIVITY_MONITOR_COEFF, + WORKLOAD_PPLIB_COMPUTE_BIT, + (void *)(&activity_monitor_external), + false); + if (ret) { + dev_err(smu->adev->dev, "[%s] Failed to get activity monitor!", __func__); + return ret; + } + + ret = smu_cmn_update_table(smu, + SMU_TABLE_ACTIVITY_MONITOR_COEFF, + WORKLOAD_PPLIB_CUSTOM_BIT, + (void *)(&activity_monitor_external), + true); + if (ret) { + dev_err(smu->adev->dev, "[%s] Failed to set activity monitor!", __func__); + return ret; + } + + workload_type = smu_cmn_to_asic_specific_index(smu, + CMN2ASIC_MAPPING_WORKLOAD, + PP_SMC_POWER_PROFILE_CUSTOM); + } else { + /* conv PP_SMC_POWER_PROFILE* to WORKLOAD_PPLIB_*_BIT */ + workload_type = smu_cmn_to_asic_specific_index(smu, CMN2ASIC_MAPPING_WORKLOAD, smu->power_profile_mode); + } + if (workload_type < 0) return -EINVAL; @@ -2097,7 +2608,7 @@ static ssize_t smu_v13_0_0_get_ecc_info(struct smu_context *smu, ecc_table = (EccInfoTable_t *)smu_table->ecc_table; - for (i = 0; i < UMC_V8_10_TOTAL_CHANNEL_NUM(adev); i++) { + for (i = 0; i < ARRAY_SIZE(ecc_table->EccInfo); i++) { ecc_info_per_channel = &(eccinfo->ecc[i]); ecc_info_per_channel->ce_count_lo_chip = ecc_table->EccInfo[i].ce_count_lo_chip; @@ -2150,6 +2661,9 @@ static const struct pptable_funcs smu_v13_0_0_ppt_funcs = { .notify_memory_pool_location = smu_v13_0_notify_memory_pool_location, .get_gpu_metrics = smu_v13_0_0_get_gpu_metrics, .set_soft_freq_limited_range = smu_v13_0_set_soft_freq_limited_range, + .set_default_od_settings = smu_v13_0_0_set_default_od_settings, + .restore_user_od_settings = smu_v13_0_0_restore_user_od_settings, + .od_edit_dpm_table = smu_v13_0_0_od_edit_dpm_table, .init_pptable_microcode = smu_v13_0_init_pptable_microcode, .populate_umd_state_clk = smu_v13_0_0_populate_umd_state_clk, .set_performance_level = smu_v13_0_set_performance_level, @@ -2199,5 +2713,6 @@ void smu_v13_0_0_set_ppt_funcs(struct smu_context *smu) smu->table_map = smu_v13_0_0_table_map; smu->pwr_src_map = smu_v13_0_0_pwr_src_map; smu->workload_map = smu_v13_0_0_workload_map; + smu->smc_driver_if_version = SMU13_0_0_DRIVER_IF_VERSION; smu_v13_0_0_set_smu_mailbox_registers(smu); } diff --git a/drivers/gpu/drm/amd/pm/swsmu/smu13/smu_v13_0_4_ppt.c b/drivers/gpu/drm/amd/pm/swsmu/smu13/smu_v13_0_4_ppt.c index 8fa9a36c38b6..ef37dda9908f 100644 --- a/drivers/gpu/drm/amd/pm/swsmu/smu13/smu_v13_0_4_ppt.c +++ b/drivers/gpu/drm/amd/pm/swsmu/smu13/smu_v13_0_4_ppt.c @@ -54,6 +54,10 @@ #define FEATURE_MASK(feature) (1ULL << feature) +#define SMU_13_0_4_UMD_PSTATE_GFXCLK 938 +#define SMU_13_0_4_UMD_PSTATE_SOCCLK 938 +#define SMU_13_0_4_UMD_PSTATE_FCLK 1875 + #define SMC_DPM_FEATURE ( \ FEATURE_MASK(FEATURE_CCLK_DPM_BIT) | \ FEATURE_MASK(FEATURE_VCN_DPM_BIT) | \ @@ -478,7 +482,7 @@ static int smu_v13_0_4_get_dpm_level_count(struct smu_context *smu, static int smu_v13_0_4_print_clk_levels(struct smu_context *smu, enum smu_clk_type clk_type, char *buf) { - int i, size = 0, ret = 0; + int i, idx, size = 0, ret = 0; uint32_t cur_value = 0, value = 0, count = 0; uint32_t min, max; @@ -512,7 +516,8 @@ static int smu_v13_0_4_print_clk_levels(struct smu_context *smu, break; for (i = 0; i < count; i++) { - ret = smu_v13_0_4_get_dpm_freq_by_index(smu, clk_type, i, &value); + idx = (clk_type == SMU_FCLK || clk_type == SMU_MCLK) ? (count - i - 1) : i; + ret = smu_v13_0_4_get_dpm_freq_by_index(smu, clk_type, idx, &value); if (ret) break; @@ -830,6 +835,8 @@ static int smu_v13_0_4_set_soft_freq_limited_range(struct smu_context *smu, uint32_t max) { enum smu_message_type msg_set_min, msg_set_max; + uint32_t min_clk = min; + uint32_t max_clk = max; int ret = 0; if (!smu_v13_0_4_clk_dpm_is_enabled(smu, clk_type)) @@ -858,12 +865,17 @@ static int smu_v13_0_4_set_soft_freq_limited_range(struct smu_context *smu, return -EINVAL; } - ret = smu_cmn_send_smc_msg_with_param(smu, msg_set_min, min, NULL); + if (clk_type == SMU_VCLK) { + min_clk = min << SMU_13_VCLK_SHIFT; + max_clk = max << SMU_13_VCLK_SHIFT; + } + + ret = smu_cmn_send_smc_msg_with_param(smu, msg_set_min, min_clk, NULL); if (ret) return ret; return smu_cmn_send_smc_msg_with_param(smu, msg_set_max, - max, NULL); + max_clk, NULL); } static int smu_v13_0_4_force_clk_levels(struct smu_context *smu, @@ -900,6 +912,50 @@ static int smu_v13_0_4_force_clk_levels(struct smu_context *smu, return ret; } +static int smu_v13_0_4_get_dpm_profile_freq(struct smu_context *smu, + enum amd_dpm_forced_level level, + enum smu_clk_type clk_type, + uint32_t *min_clk, + uint32_t *max_clk) +{ + int ret = 0; + uint32_t clk_limit = 0; + + switch (clk_type) { + case SMU_GFXCLK: + case SMU_SCLK: + clk_limit = SMU_13_0_4_UMD_PSTATE_GFXCLK; + if (level == AMD_DPM_FORCED_LEVEL_PROFILE_PEAK) + smu_v13_0_4_get_dpm_ultimate_freq(smu, SMU_SCLK, NULL, &clk_limit); + else if (level == AMD_DPM_FORCED_LEVEL_PROFILE_MIN_SCLK) + smu_v13_0_4_get_dpm_ultimate_freq(smu, SMU_SCLK, &clk_limit, NULL); + break; + case SMU_SOCCLK: + clk_limit = SMU_13_0_4_UMD_PSTATE_SOCCLK; + if (level == AMD_DPM_FORCED_LEVEL_PROFILE_PEAK) + smu_v13_0_4_get_dpm_ultimate_freq(smu, SMU_SOCCLK, NULL, &clk_limit); + break; + case SMU_FCLK: + clk_limit = SMU_13_0_4_UMD_PSTATE_FCLK; + if (level == AMD_DPM_FORCED_LEVEL_PROFILE_PEAK) + smu_v13_0_4_get_dpm_ultimate_freq(smu, SMU_FCLK, NULL, &clk_limit); + else if (level == AMD_DPM_FORCED_LEVEL_PROFILE_MIN_MCLK) + smu_v13_0_4_get_dpm_ultimate_freq(smu, SMU_FCLK, &clk_limit, NULL); + break; + case SMU_VCLK: + smu_v13_0_4_get_dpm_ultimate_freq(smu, SMU_VCLK, NULL, &clk_limit); + break; + case SMU_DCLK: + smu_v13_0_4_get_dpm_ultimate_freq(smu, SMU_DCLK, NULL, &clk_limit); + break; + default: + ret = -EINVAL; + break; + } + *min_clk = *max_clk = clk_limit; + return ret; +} + static int smu_v13_0_4_set_performance_level(struct smu_context *smu, enum amd_dpm_forced_level level) { @@ -907,6 +963,8 @@ static int smu_v13_0_4_set_performance_level(struct smu_context *smu, uint32_t sclk_min = 0, sclk_max = 0; uint32_t fclk_min = 0, fclk_max = 0; uint32_t socclk_min = 0, socclk_max = 0; + uint32_t vclk_min = 0, vclk_max = 0; + uint32_t dclk_min = 0, dclk_max = 0; int ret = 0; switch (level) { @@ -914,28 +972,42 @@ static int smu_v13_0_4_set_performance_level(struct smu_context *smu, smu_v13_0_4_get_dpm_ultimate_freq(smu, SMU_SCLK, NULL, &sclk_max); smu_v13_0_4_get_dpm_ultimate_freq(smu, SMU_FCLK, NULL, &fclk_max); smu_v13_0_4_get_dpm_ultimate_freq(smu, SMU_SOCCLK, NULL, &socclk_max); + smu_v13_0_4_get_dpm_ultimate_freq(smu, SMU_VCLK, NULL, &vclk_max); + smu_v13_0_4_get_dpm_ultimate_freq(smu, SMU_DCLK, NULL, &dclk_max); sclk_min = sclk_max; fclk_min = fclk_max; socclk_min = socclk_max; + vclk_min = vclk_max; + dclk_min = dclk_max; break; case AMD_DPM_FORCED_LEVEL_LOW: smu_v13_0_4_get_dpm_ultimate_freq(smu, SMU_SCLK, &sclk_min, NULL); smu_v13_0_4_get_dpm_ultimate_freq(smu, SMU_FCLK, &fclk_min, NULL); smu_v13_0_4_get_dpm_ultimate_freq(smu, SMU_SOCCLK, &socclk_min, NULL); + smu_v13_0_4_get_dpm_ultimate_freq(smu, SMU_VCLK, &vclk_min, NULL); + smu_v13_0_4_get_dpm_ultimate_freq(smu, SMU_DCLK, &dclk_min, NULL); sclk_max = sclk_min; fclk_max = fclk_min; socclk_max = socclk_min; + vclk_max = vclk_min; + dclk_max = dclk_min; break; case AMD_DPM_FORCED_LEVEL_AUTO: smu_v13_0_4_get_dpm_ultimate_freq(smu, SMU_SCLK, &sclk_min, &sclk_max); smu_v13_0_4_get_dpm_ultimate_freq(smu, SMU_FCLK, &fclk_min, &fclk_max); smu_v13_0_4_get_dpm_ultimate_freq(smu, SMU_SOCCLK, &socclk_min, &socclk_max); + smu_v13_0_4_get_dpm_ultimate_freq(smu, SMU_VCLK, &vclk_min, &vclk_max); + smu_v13_0_4_get_dpm_ultimate_freq(smu, SMU_DCLK, &dclk_min, &dclk_max); break; case AMD_DPM_FORCED_LEVEL_PROFILE_STANDARD: case AMD_DPM_FORCED_LEVEL_PROFILE_MIN_SCLK: case AMD_DPM_FORCED_LEVEL_PROFILE_MIN_MCLK: case AMD_DPM_FORCED_LEVEL_PROFILE_PEAK: - /* Temporarily do nothing since the optimal clocks haven't been provided yet */ + smu_v13_0_4_get_dpm_profile_freq(smu, level, SMU_SCLK, &sclk_min, &sclk_max); + smu_v13_0_4_get_dpm_profile_freq(smu, level, SMU_FCLK, &fclk_min, &fclk_max); + smu_v13_0_4_get_dpm_profile_freq(smu, level, SMU_SOCCLK, &socclk_min, &socclk_max); + smu_v13_0_4_get_dpm_profile_freq(smu, level, SMU_VCLK, &vclk_min, &vclk_max); + smu_v13_0_4_get_dpm_profile_freq(smu, level, SMU_DCLK, &dclk_min, &dclk_max); break; case AMD_DPM_FORCED_LEVEL_MANUAL: case AMD_DPM_FORCED_LEVEL_PROFILE_EXIT: @@ -975,6 +1047,23 @@ static int smu_v13_0_4_set_performance_level(struct smu_context *smu, return ret; } + if (vclk_min && vclk_max) { + ret = smu_v13_0_4_set_soft_freq_limited_range(smu, + SMU_VCLK, + vclk_min, + vclk_max); + if (ret) + return ret; + } + + if (dclk_min && dclk_max) { + ret = smu_v13_0_4_set_soft_freq_limited_range(smu, + SMU_DCLK, + dclk_min, + dclk_max); + if (ret) + return ret; + } return ret; } @@ -1043,6 +1132,7 @@ void smu_v13_0_4_set_ppt_funcs(struct smu_context *smu) smu->message_map = smu_v13_0_4_message_map; smu->feature_map = smu_v13_0_4_feature_mask_map; smu->table_map = smu_v13_0_4_table_map; + smu->smc_driver_if_version = SMU13_0_4_DRIVER_IF_VERSION; smu->is_apu = true; if (adev->ip_versions[MP1_HWIP][0] == IP_VERSION(13, 0, 4)) diff --git a/drivers/gpu/drm/amd/pm/swsmu/smu13/smu_v13_0_5_ppt.c b/drivers/gpu/drm/amd/pm/swsmu/smu13/smu_v13_0_5_ppt.c index 66445964efbd..42f110602eb1 100644 --- a/drivers/gpu/drm/amd/pm/swsmu/smu13/smu_v13_0_5_ppt.c +++ b/drivers/gpu/drm/amd/pm/swsmu/smu13/smu_v13_0_5_ppt.c @@ -831,6 +831,8 @@ static int smu_v13_0_5_set_soft_freq_limited_range(struct smu_context *smu, uint32_t max) { enum smu_message_type msg_set_min, msg_set_max; + uint32_t min_clk = min; + uint32_t max_clk = max; int ret = 0; if (!smu_v13_0_5_clk_dpm_is_enabled(smu, clk_type)) @@ -851,11 +853,16 @@ static int smu_v13_0_5_set_soft_freq_limited_range(struct smu_context *smu, return -EINVAL; } - ret = smu_cmn_send_smc_msg_with_param(smu, msg_set_min, min, NULL); + if (clk_type == SMU_VCLK) { + min_clk = min << SMU_13_VCLK_SHIFT; + max_clk = max << SMU_13_VCLK_SHIFT; + } + + ret = smu_cmn_send_smc_msg_with_param(smu, msg_set_min, min_clk, NULL); if (ret) goto out; - ret = smu_cmn_send_smc_msg_with_param(smu, msg_set_max, max, NULL); + ret = smu_cmn_send_smc_msg_with_param(smu, msg_set_max, max_clk, NULL); if (ret) goto out; @@ -866,7 +873,7 @@ out: static int smu_v13_0_5_print_clk_levels(struct smu_context *smu, enum smu_clk_type clk_type, char *buf) { - int i, size = 0, ret = 0; + int i, idx, size = 0, ret = 0; uint32_t cur_value = 0, value = 0, count = 0; uint32_t min = 0, max = 0; @@ -898,7 +905,8 @@ static int smu_v13_0_5_print_clk_levels(struct smu_context *smu, goto print_clk_out; for (i = 0; i < count; i++) { - ret = smu_v13_0_5_get_dpm_freq_by_index(smu, clk_type, i, &value); + idx = (clk_type == SMU_MCLK) ? (count - i - 1) : i; + ret = smu_v13_0_5_get_dpm_freq_by_index(smu, clk_type, idx, &value); if (ret) goto print_clk_out; @@ -970,31 +978,79 @@ force_level_out: return ret; } +static int smu_v13_0_5_get_dpm_profile_freq(struct smu_context *smu, + enum amd_dpm_forced_level level, + enum smu_clk_type clk_type, + uint32_t *min_clk, + uint32_t *max_clk) +{ + int ret = 0; + uint32_t clk_limit = 0; + + switch (clk_type) { + case SMU_GFXCLK: + case SMU_SCLK: + clk_limit = SMU_13_0_5_UMD_PSTATE_GFXCLK; + if (level == AMD_DPM_FORCED_LEVEL_PROFILE_PEAK) + smu_v13_0_5_get_dpm_ultimate_freq(smu, SMU_SCLK, NULL, &clk_limit); + else if (level == AMD_DPM_FORCED_LEVEL_PROFILE_MIN_SCLK) + smu_v13_0_5_get_dpm_ultimate_freq(smu, SMU_SCLK, &clk_limit, NULL); + break; + case SMU_VCLK: + smu_v13_0_5_get_dpm_ultimate_freq(smu, SMU_VCLK, NULL, &clk_limit); + break; + case SMU_DCLK: + smu_v13_0_5_get_dpm_ultimate_freq(smu, SMU_DCLK, NULL, &clk_limit); + break; + default: + ret = -EINVAL; + break; + } + *min_clk = *max_clk = clk_limit; + return ret; +} + static int smu_v13_0_5_set_performance_level(struct smu_context *smu, enum amd_dpm_forced_level level) { struct amdgpu_device *adev = smu->adev; uint32_t sclk_min = 0, sclk_max = 0; + uint32_t vclk_min = 0, vclk_max = 0; + uint32_t dclk_min = 0, dclk_max = 0; int ret = 0; switch (level) { case AMD_DPM_FORCED_LEVEL_HIGH: smu_v13_0_5_get_dpm_ultimate_freq(smu, SMU_SCLK, NULL, &sclk_max); + smu_v13_0_5_get_dpm_ultimate_freq(smu, SMU_VCLK, NULL, &vclk_max); + smu_v13_0_5_get_dpm_ultimate_freq(smu, SMU_DCLK, NULL, &dclk_max); sclk_min = sclk_max; + vclk_min = vclk_max; + dclk_min = dclk_max; break; case AMD_DPM_FORCED_LEVEL_LOW: smu_v13_0_5_get_dpm_ultimate_freq(smu, SMU_SCLK, &sclk_min, NULL); + smu_v13_0_5_get_dpm_ultimate_freq(smu, SMU_VCLK, &vclk_min, NULL); + smu_v13_0_5_get_dpm_ultimate_freq(smu, SMU_DCLK, &dclk_min, NULL); sclk_max = sclk_min; + vclk_max = vclk_min; + dclk_max = dclk_min; break; case AMD_DPM_FORCED_LEVEL_AUTO: smu_v13_0_5_get_dpm_ultimate_freq(smu, SMU_SCLK, &sclk_min, &sclk_max); + smu_v13_0_5_get_dpm_ultimate_freq(smu, SMU_VCLK, &vclk_min, &vclk_max); + smu_v13_0_5_get_dpm_ultimate_freq(smu, SMU_DCLK, &dclk_min, &dclk_max); break; case AMD_DPM_FORCED_LEVEL_PROFILE_STANDARD: case AMD_DPM_FORCED_LEVEL_PROFILE_MIN_SCLK: - case AMD_DPM_FORCED_LEVEL_PROFILE_MIN_MCLK: case AMD_DPM_FORCED_LEVEL_PROFILE_PEAK: - /* Temporarily do nothing since the optimal clocks haven't been provided yet */ + smu_v13_0_5_get_dpm_profile_freq(smu, level, SMU_SCLK, &sclk_min, &sclk_max); + smu_v13_0_5_get_dpm_profile_freq(smu, level, SMU_VCLK, &vclk_min, &vclk_max); + smu_v13_0_5_get_dpm_profile_freq(smu, level, SMU_DCLK, &dclk_min, &dclk_max); break; + case AMD_DPM_FORCED_LEVEL_PROFILE_MIN_MCLK: + dev_err(adev->dev, "The performance level profile_min_mclk is not supported."); + return -EOPNOTSUPP; case AMD_DPM_FORCED_LEVEL_MANUAL: case AMD_DPM_FORCED_LEVEL_PROFILE_EXIT: return 0; @@ -1015,6 +1071,23 @@ static int smu_v13_0_5_set_performance_level(struct smu_context *smu, smu->gfx_actual_soft_max_freq = sclk_max; } + if (vclk_min && vclk_max) { + ret = smu_v13_0_5_set_soft_freq_limited_range(smu, + SMU_VCLK, + vclk_min, + vclk_max); + if (ret) + return ret; + } + + if (dclk_min && dclk_max) { + ret = smu_v13_0_5_set_soft_freq_limited_range(smu, + SMU_DCLK, + dclk_min, + dclk_max); + if (ret) + return ret; + } return ret; } @@ -1068,6 +1141,7 @@ void smu_v13_0_5_set_ppt_funcs(struct smu_context *smu) smu->feature_map = smu_v13_0_5_feature_mask_map; smu->table_map = smu_v13_0_5_table_map; smu->is_apu = true; + smu->smc_driver_if_version = SMU13_0_5_DRIVER_IF_VERSION; smu->param_reg = SOC15_REG_OFFSET(MP1, 0, mmMP1_C2PMSG_34); smu->msg_reg = SOC15_REG_OFFSET(MP1, 0, mmMP1_C2PMSG_2); smu->resp_reg = SOC15_REG_OFFSET(MP1, 0, mmMP1_C2PMSG_33); diff --git a/drivers/gpu/drm/amd/pm/swsmu/smu13/smu_v13_0_5_ppt.h b/drivers/gpu/drm/amd/pm/swsmu/smu13/smu_v13_0_5_ppt.h index 40bc0f8e6d61..263cd651855e 100644 --- a/drivers/gpu/drm/amd/pm/swsmu/smu13/smu_v13_0_5_ppt.h +++ b/drivers/gpu/drm/amd/pm/swsmu/smu13/smu_v13_0_5_ppt.h @@ -24,6 +24,6 @@ #define __SMU_V13_0_5_PPT_H__ extern void smu_v13_0_5_set_ppt_funcs(struct smu_context *smu); -#define SMU_13_0_5_UMD_PSTATE_GFXCLK 1100 +#define SMU_13_0_5_UMD_PSTATE_GFXCLK 700 #endif diff --git a/drivers/gpu/drm/amd/pm/swsmu/smu13/smu_v13_0_6_ppt.c b/drivers/gpu/drm/amd/pm/swsmu/smu13/smu_v13_0_6_ppt.c index ea8f3d6fb98b..a92ea4601ea4 100644 --- a/drivers/gpu/drm/amd/pm/swsmu/smu13/smu_v13_0_6_ppt.c +++ b/drivers/gpu/drm/amd/pm/swsmu/smu13/smu_v13_0_6_ppt.c @@ -80,7 +80,10 @@ /* possible frequency drift (1Mhz) */ #define EPSILON 1 -#define smnPCIE_ESM_CTRL 0x111003D0 +#define smnPCIE_ESM_CTRL 0x193D0 +#define smnPCIE_LC_LINK_WIDTH_CNTL 0x1ab40288 +#define PCIE_LC_LINK_WIDTH_CNTL__LC_LINK_WIDTH_RD_MASK 0x00000070L +#define PCIE_LC_LINK_WIDTH_CNTL__LC_LINK_WIDTH_RD__SHIFT 0x4 static const struct cmn2asic_msg_mapping smu_v13_0_6_message_map[SMU_MSG_MAX_COUNT] = { MSG_MAP(TestMessage, PPSMC_MSG_TestMessage, 0), @@ -122,6 +125,7 @@ static const struct cmn2asic_msg_mapping smu_v13_0_6_message_map[SMU_MSG_MAX_COU MSG_MAP(GetMaxGfxclkFrequency, PPSMC_MSG_GetMaxGfxDpmFreq, 0), MSG_MAP(SetSoftMinGfxclk, PPSMC_MSG_SetSoftMinGfxClk, 0), MSG_MAP(SetSoftMaxGfxClk, PPSMC_MSG_SetSoftMaxGfxClk, 0), + MSG_MAP(PrepareMp1ForUnload, PPSMC_MSG_PrepareForDriverUnload, 0), }; static const struct cmn2asic_mapping smu_v13_0_6_clk_map[SMU_CLK_COUNT] = { @@ -171,18 +175,12 @@ static const struct cmn2asic_mapping smu_v13_0_6_table_map[SMU_TABLE_COUNT] = { TAB_MAP(I2C_COMMANDS), }; -#define THROTTLER_PROCHOT_GFX_BIT 0 -#define THROTTLER_PPT_BIT 1 -#define THROTTLER_TEMP_SOC_BIT 2 -#define THROTTLER_TEMP_VR_GFX_BIT 3 -#define THROTTLER_TEMP_HBM_BIT 4 - static const uint8_t smu_v13_0_6_throttler_map[] = { [THROTTLER_PPT_BIT] = (SMU_THROTTLER_PPT0_BIT), - [THROTTLER_TEMP_SOC_BIT] = (SMU_THROTTLER_TEMP_GPU_BIT), - [THROTTLER_TEMP_HBM_BIT] = (SMU_THROTTLER_TEMP_MEM_BIT), - [THROTTLER_TEMP_VR_GFX_BIT] = (SMU_THROTTLER_TEMP_VR_GFX_BIT), - [THROTTLER_PROCHOT_GFX_BIT] = (SMU_THROTTLER_PROCHOT_GFX_BIT), + [THROTTLER_THERMAL_SOCKET_BIT] = (SMU_THROTTLER_TEMP_GPU_BIT), + [THROTTLER_THERMAL_HBM_BIT] = (SMU_THROTTLER_TEMP_MEM_BIT), + [THROTTLER_THERMAL_VR_BIT] = (SMU_THROTTLER_TEMP_VR_GFX_BIT), + [THROTTLER_PROCHOT_BIT] = (SMU_THROTTLER_PROCHOT_GFX_BIT), }; struct PPTable_t { @@ -197,10 +195,12 @@ struct PPTable_t { uint32_t LclkFrequencyTable[4]; uint32_t MaxLclkDpmRange; uint32_t MinLclkDpmRange; + uint64_t PublicSerialNumber_AID; bool Init; }; #define SMUQ10_TO_UINT(x) ((x) >> 10) +#define SMUQ16_TO_UINT(x) ((x) >> 16) struct smu_v13_0_6_dpm_map { enum smu_clk_type clk_type; @@ -220,10 +220,12 @@ static int smu_v13_0_6_tables_init(struct smu_context *smu) PAGE_SIZE, AMDGPU_GEM_DOMAIN_VRAM); SMU_TABLE_INIT(tables, SMU_TABLE_SMU_METRICS, sizeof(MetricsTable_t), - PAGE_SIZE, AMDGPU_GEM_DOMAIN_VRAM); + PAGE_SIZE, + AMDGPU_GEM_DOMAIN_VRAM | AMDGPU_GEM_DOMAIN_GTT); SMU_TABLE_INIT(tables, SMU_TABLE_I2C_COMMANDS, sizeof(SwI2cRequest_t), - PAGE_SIZE, AMDGPU_GEM_DOMAIN_VRAM); + PAGE_SIZE, + AMDGPU_GEM_DOMAIN_VRAM | AMDGPU_GEM_DOMAIN_GTT); smu_table->metrics_table = kzalloc(sizeof(MetricsTable_t), GFP_KERNEL); if (!smu_table->metrics_table) @@ -355,6 +357,9 @@ static int smu_v13_0_6_setup_driver_pptable(struct smu_context *smu) SMUQ10_TO_UINT(metrics->LclkFrequencyTable[i]); } + /* use AID0 serial number by default */ + pptable->PublicSerialNumber_AID = metrics->PublicSerialNumber_AID[0]; + pptable->Init = true; } @@ -385,7 +390,7 @@ static int smu_v13_0_6_get_dpm_ultimate_freq(struct smu_context *smu, break; case SMU_SOCCLK: if (pptable->Init) - clock_limit = pptable->UclkFrequencyTable[0]; + clock_limit = pptable->SocclkFrequencyTable[0]; break; case SMU_FCLK: if (pptable->Init) @@ -638,16 +643,14 @@ static int smu_v13_0_6_freqs_in_same_level(int32_t frequency1, return (abs(frequency1 - frequency2) <= EPSILON); } -static uint32_t smu_v13_0_6_get_throttler_status(struct smu_context *smu, - MetricsTable_t *metrics) +static uint32_t smu_v13_0_6_get_throttler_status(struct smu_context *smu) { + struct smu_power_context *smu_power = &smu->smu_power; + struct smu_13_0_power_context *power_context = smu_power->power_context; uint32_t throttler_status = 0; - throttler_status |= metrics->ProchotResidencyAcc > 0 ? 1U << THROTTLER_PROCHOT_GFX_BIT : 0; - throttler_status |= metrics->PptResidencyAcc > 0 ? 1U << THROTTLER_PPT_BIT : 0; - throttler_status |= metrics->SocketThmResidencyAcc > 0 ? 1U << THROTTLER_TEMP_SOC_BIT : 0; - throttler_status |= metrics->VrThmResidencyAcc > 0 ? 1U << THROTTLER_TEMP_VR_GFX_BIT : 0; - throttler_status |= metrics->HbmThmResidencyAcc > 0 ? 1U << THROTTLER_TEMP_HBM_BIT : 0; + throttler_status = atomic_read(&power_context->throttle_status); + dev_dbg(smu->adev->dev, "SMU Throttler status: %u", throttler_status); return throttler_status; } @@ -658,7 +661,10 @@ static int smu_v13_0_6_get_smu_metrics_data(struct smu_context *smu, { struct smu_table_context *smu_table = &smu->smu_table; MetricsTable_t *metrics = (MetricsTable_t *)smu_table->metrics_table; + struct amdgpu_device *adev = smu->adev; + uint32_t smu_version; int ret = 0; + int xcc_id; ret = smu_v13_0_6_get_metrics_table(smu, NULL, false); if (ret) @@ -668,7 +674,13 @@ static int smu_v13_0_6_get_smu_metrics_data(struct smu_context *smu, switch (member) { case METRICS_CURR_GFXCLK: case METRICS_AVERAGE_GFXCLK: - *value = 0; + smu_cmn_get_smc_version(smu, NULL, &smu_version); + if (smu_version >= 0x552F00) { + xcc_id = GET_INST(GC, 0); + *value = SMUQ10_TO_UINT(metrics->GfxclkFrequency[xcc_id]); + } else { + *value = 0; + } break; case METRICS_CURR_SOCCLK: case METRICS_AVERAGE_SOCCLK: @@ -708,9 +720,6 @@ static int smu_v13_0_6_get_smu_metrics_data(struct smu_context *smu, case METRICS_TEMPERATURE_VRSOC: *value = SMUQ10_TO_UINT(metrics->MaxVrTemperature); break; - case METRICS_THROTTLER_STATUS: - *value = smu_v13_0_6_get_throttler_status(smu, metrics); - break; default: *value = UINT_MAX; break; @@ -1246,21 +1255,6 @@ static int smu_v13_0_6_get_power_limit(struct smu_context *smu, uint32_t power_limit = 0; int ret; - if (!smu_cmn_feature_is_enabled(smu, SMU_FEATURE_PPT_BIT)) { - if (current_power_limit) - *current_power_limit = 0; - if (default_power_limit) - *default_power_limit = 0; - if (max_power_limit) - *max_power_limit = 0; - - dev_warn( - smu->adev->dev, - "PPT feature is not enabled, power values can't be fetched."); - - return 0; - } - ret = smu_cmn_send_smc_msg(smu, SMU_MSG_GetPptLimit, &power_limit); if (ret) { @@ -1287,16 +1281,147 @@ static int smu_v13_0_6_set_power_limit(struct smu_context *smu, return smu_v13_0_set_power_limit(smu, limit_type, limit); } -static int smu_v13_0_6_system_features_control(struct smu_context *smu, - bool enable) +static int smu_v13_0_6_irq_process(struct amdgpu_device *adev, + struct amdgpu_irq_src *source, + struct amdgpu_iv_entry *entry) { - int ret; + struct smu_context *smu = adev->powerplay.pp_handle; + struct smu_power_context *smu_power = &smu->smu_power; + struct smu_13_0_power_context *power_context = smu_power->power_context; + uint32_t client_id = entry->client_id; + uint32_t ctxid = entry->src_data[0]; + uint32_t src_id = entry->src_id; + uint32_t data; + + if (client_id == SOC15_IH_CLIENTID_MP1) { + if (src_id == IH_INTERRUPT_ID_TO_DRIVER) { + /* ACK SMUToHost interrupt */ + data = RREG32_SOC15(MP1, 0, regMP1_SMN_IH_SW_INT_CTRL); + data = REG_SET_FIELD(data, MP1_SMN_IH_SW_INT_CTRL, INT_ACK, 1); + WREG32_SOC15(MP1, 0, regMP1_SMN_IH_SW_INT_CTRL, data); + /* + * ctxid is used to distinguish different events for SMCToHost + * interrupt. + */ + switch (ctxid) { + case IH_INTERRUPT_CONTEXT_ID_THERMAL_THROTTLING: + /* + * Increment the throttle interrupt counter + */ + atomic64_inc(&smu->throttle_int_counter); + + if (!atomic_read(&adev->throttling_logging_enabled)) + return 0; + + /* This uses the new method which fixes the + * incorrect throttling status reporting + * through metrics table. For older FWs, + * it will be ignored. + */ + if (__ratelimit(&adev->throttling_logging_rs)) { + atomic_set( + &power_context->throttle_status, + entry->src_data[1]); + schedule_work(&smu->throttling_logging_work); + } + + break; + } + } + } + + return 0; +} + +static int smu_v13_0_6_set_irq_state(struct amdgpu_device *adev, + struct amdgpu_irq_src *source, + unsigned tyep, + enum amdgpu_interrupt_state state) +{ + uint32_t val = 0; + + switch (state) { + case AMDGPU_IRQ_STATE_DISABLE: + /* For MP1 SW irqs */ + val = RREG32_SOC15(MP1, 0, regMP1_SMN_IH_SW_INT_CTRL); + val = REG_SET_FIELD(val, MP1_SMN_IH_SW_INT_CTRL, INT_MASK, 1); + WREG32_SOC15(MP1, 0, regMP1_SMN_IH_SW_INT_CTRL, val); - /* Nothing to be done for APU */ - if (smu->adev->flags & AMD_IS_APU) + break; + case AMDGPU_IRQ_STATE_ENABLE: + /* For MP1 SW irqs */ + val = RREG32_SOC15(MP1, 0, regMP1_SMN_IH_SW_INT); + val = REG_SET_FIELD(val, MP1_SMN_IH_SW_INT, ID, 0xFE); + val = REG_SET_FIELD(val, MP1_SMN_IH_SW_INT, VALID, 0); + WREG32_SOC15(MP1, 0, regMP1_SMN_IH_SW_INT, val); + + val = RREG32_SOC15(MP1, 0, regMP1_SMN_IH_SW_INT_CTRL); + val = REG_SET_FIELD(val, MP1_SMN_IH_SW_INT_CTRL, INT_MASK, 0); + WREG32_SOC15(MP1, 0, regMP1_SMN_IH_SW_INT_CTRL, val); + + break; + default: + break; + } + + return 0; +} + +static const struct amdgpu_irq_src_funcs smu_v13_0_6_irq_funcs = +{ + .set = smu_v13_0_6_set_irq_state, + .process = smu_v13_0_6_irq_process, +}; + +static int smu_v13_0_6_register_irq_handler(struct smu_context *smu) +{ + struct amdgpu_device *adev = smu->adev; + struct amdgpu_irq_src *irq_src = &smu->irq_source; + int ret = 0; + + if (amdgpu_sriov_vf(adev)) return 0; - ret = smu_v13_0_system_features_control(smu, enable); + irq_src->num_types = 1; + irq_src->funcs = &smu_v13_0_6_irq_funcs; + + ret = amdgpu_irq_add_id(adev, SOC15_IH_CLIENTID_MP1, + IH_INTERRUPT_ID_TO_DRIVER, + irq_src); + if (ret) + return ret; + + return ret; +} + +static int smu_v13_0_6_notify_unload(struct smu_context *smu) +{ + uint32_t smu_version; + + smu_cmn_get_smc_version(smu, NULL, &smu_version); + if (smu_version <= 0x553500) + return 0; + + dev_dbg(smu->adev->dev, "Notify PMFW about driver unload"); + /* Ignore return, just intimate FW that driver is not going to be there */ + smu_cmn_send_smc_msg(smu, SMU_MSG_PrepareMp1ForUnload, NULL); + + return 0; +} + +static int smu_v13_0_6_system_features_control(struct smu_context *smu, + bool enable) +{ + struct amdgpu_device *adev = smu->adev; + int ret = 0; + + if (enable) { + if (!(adev->flags & AMD_IS_APU)) + ret = smu_v13_0_system_features_control(smu, enable); + } else { + /* Notify FW that the device is no longer driver managed */ + smu_v13_0_6_notify_unload(smu); + } return ret; } @@ -1737,19 +1862,11 @@ static void smu_v13_0_6_i2c_control_fini(struct smu_context *smu) static void smu_v13_0_6_get_unique_id(struct smu_context *smu) { struct amdgpu_device *adev = smu->adev; - //SmuMetrics_t *metrics = smu->smu_table.metrics_table; - uint32_t upper32 = 0, lower32 = 0; - int ret; - - ret = smu_cmn_get_metrics_table(smu, NULL, false); - if (ret) - goto out; - - //upper32 = metrics->PublicSerialNumUpper32; - //lower32 = metrics->PublicSerialNumLower32; + struct smu_table_context *smu_table = &smu->smu_table; + struct PPTable_t *pptable = + (struct PPTable_t *)smu_table->driver_pptable; -out: - adev->unique_id = ((uint64_t)upper32 << 32) | lower32; + adev->unique_id = pptable->PublicSerialNumber_AID; if (adev->serial[0] == '\0') sprintf(adev->serial, "%016llx", adev->unique_id); } @@ -1774,37 +1891,35 @@ static int smu_v13_0_6_allow_xgmi_power_down(struct smu_context *smu, bool en) en ? 0 : 1, NULL); } -static const struct throttling_logging_label { - uint32_t feature_mask; - const char *label; -} logging_label[] = { - { (1U << THROTTLER_TEMP_HBM_BIT), "HBM" }, - { (1U << THROTTLER_TEMP_SOC_BIT), "SOC" }, - { (1U << THROTTLER_TEMP_VR_GFX_BIT), "VR limit" }, +static const char *const throttling_logging_label[] = { + [THROTTLER_PROCHOT_BIT] = "Prochot", + [THROTTLER_PPT_BIT] = "PPT", + [THROTTLER_THERMAL_SOCKET_BIT] = "SOC", + [THROTTLER_THERMAL_VR_BIT] = "VR", + [THROTTLER_THERMAL_HBM_BIT] = "HBM" }; + static void smu_v13_0_6_log_thermal_throttling_event(struct smu_context *smu) { - int ret; int throttler_idx, throtting_events = 0, buf_idx = 0; struct amdgpu_device *adev = smu->adev; uint32_t throttler_status; char log_buf[256]; - ret = smu_v13_0_6_get_smu_metrics_data(smu, METRICS_THROTTLER_STATUS, - &throttler_status); - if (ret) + throttler_status = smu_v13_0_6_get_throttler_status(smu); + if (!throttler_status) return; memset(log_buf, 0, sizeof(log_buf)); - for (throttler_idx = 0; throttler_idx < ARRAY_SIZE(logging_label); + for (throttler_idx = 0; + throttler_idx < ARRAY_SIZE(throttling_logging_label); throttler_idx++) { - if (throttler_status & - logging_label[throttler_idx].feature_mask) { + if (throttler_status & (1U << throttler_idx)) { throtting_events++; - buf_idx += snprintf(log_buf + buf_idx, - sizeof(log_buf) - buf_idx, "%s%s", - throtting_events > 1 ? " and " : "", - logging_label[throttler_idx].label); + buf_idx += snprintf( + log_buf + buf_idx, sizeof(log_buf) - buf_idx, + "%s%s", throtting_events > 1 ? " and " : "", + throttling_logging_label[throttler_idx]); if (buf_idx >= sizeof(log_buf)) { dev_err(adev->dev, "buffer overflow!\n"); log_buf[sizeof(log_buf) - 1] = '\0'; @@ -1813,16 +1928,24 @@ static void smu_v13_0_6_log_thermal_throttling_event(struct smu_context *smu) } } - dev_warn( - adev->dev, - "WARN: GPU thermal throttling temperature reached, expect performance decrease. %s.\n", - log_buf); + dev_warn(adev->dev, + "WARN: GPU is throttled, expect performance decrease. %s.\n", + log_buf); kgd2kfd_smi_event_throttle( smu->adev->kfd.dev, smu_cmn_get_indep_throttler_status(throttler_status, smu_v13_0_6_throttler_map)); } +static int +smu_v13_0_6_get_current_pcie_link_width_level(struct smu_context *smu) +{ + struct amdgpu_device *adev = smu->adev; + + return REG_GET_FIELD(RREG32_PCIE(smnPCIE_LC_LINK_WIDTH_CNTL), + PCIE_LC_LINK_WIDTH_CNTL, LC_LINK_WIDTH_RD); +} + static int smu_v13_0_6_get_current_pcie_link_speed(struct smu_context *smu) { struct amdgpu_device *adev = smu->adev; @@ -1841,8 +1964,12 @@ static ssize_t smu_v13_0_6_get_gpu_metrics(struct smu_context *smu, void **table struct smu_table_context *smu_table = &smu->smu_table; struct gpu_metrics_v1_3 *gpu_metrics = (struct gpu_metrics_v1_3 *)smu_table->gpu_metrics_table; + struct amdgpu_device *adev = smu->adev; + int ret = 0, inst0, xcc0; MetricsTable_t *metrics; - int i, ret = 0; + + inst0 = adev->sdma.instance[0].aid_id; + xcc0 = GET_INST(GC, 0); metrics = kzalloc(sizeof(MetricsTable_t), GFP_KERNEL); ret = smu_v13_0_6_get_metrics_table(smu, metrics, true); @@ -1851,51 +1978,59 @@ static ssize_t smu_v13_0_6_get_gpu_metrics(struct smu_context *smu, void **table smu_cmn_init_soft_gpu_metrics(gpu_metrics, 1, 3); - /* TODO: Decide on how to fill in zero value fields */ - gpu_metrics->temperature_edge = 0; - gpu_metrics->temperature_hotspot = 0; - gpu_metrics->temperature_mem = 0; - gpu_metrics->temperature_vrgfx = 0; - gpu_metrics->temperature_vrsoc = 0; - gpu_metrics->temperature_vrmem = 0; - - gpu_metrics->average_gfx_activity = 0; - gpu_metrics->average_umc_activity = 0; - gpu_metrics->average_mm_activity = 0; - - gpu_metrics->average_socket_power = 0; - gpu_metrics->energy_accumulator = 0; - - gpu_metrics->average_gfxclk_frequency = 0; - gpu_metrics->average_socclk_frequency = 0; - gpu_metrics->average_uclk_frequency = 0; - gpu_metrics->average_vclk0_frequency = 0; - gpu_metrics->average_dclk0_frequency = 0; - - gpu_metrics->current_gfxclk = 0; - gpu_metrics->current_socclk = 0; - gpu_metrics->current_uclk = 0; - gpu_metrics->current_vclk0 = 0; - gpu_metrics->current_dclk0 = 0; - + gpu_metrics->temperature_hotspot = + SMUQ10_TO_UINT(metrics->MaxSocketTemperature); + /* Individual HBM stack temperature is not reported */ + gpu_metrics->temperature_mem = + SMUQ10_TO_UINT(metrics->MaxHbmTemperature); + /* Reports max temperature of all voltage rails */ + gpu_metrics->temperature_vrsoc = + SMUQ10_TO_UINT(metrics->MaxVrTemperature); + + gpu_metrics->average_gfx_activity = + SMUQ10_TO_UINT(metrics->SocketGfxBusy); + gpu_metrics->average_umc_activity = + SMUQ10_TO_UINT(metrics->DramBandwidthUtilization); + + gpu_metrics->average_socket_power = + SMUQ10_TO_UINT(metrics->SocketPower); + gpu_metrics->energy_accumulator = + SMUQ16_TO_UINT(metrics->SocketEnergyAcc); + + gpu_metrics->current_gfxclk = + SMUQ10_TO_UINT(metrics->GfxclkFrequency[xcc0]); + gpu_metrics->current_socclk = + SMUQ10_TO_UINT(metrics->SocclkFrequency[inst0]); + gpu_metrics->current_uclk = SMUQ10_TO_UINT(metrics->UclkFrequency); + gpu_metrics->current_vclk0 = + SMUQ10_TO_UINT(metrics->VclkFrequency[inst0]); + gpu_metrics->current_dclk0 = + SMUQ10_TO_UINT(metrics->DclkFrequency[inst0]); + + gpu_metrics->average_gfxclk_frequency = gpu_metrics->current_gfxclk; + gpu_metrics->average_socclk_frequency = gpu_metrics->current_socclk; + gpu_metrics->average_uclk_frequency = gpu_metrics->current_uclk; + gpu_metrics->average_vclk0_frequency = gpu_metrics->current_vclk0; + gpu_metrics->average_dclk0_frequency = gpu_metrics->current_dclk0; + + /* Throttle status is not reported through metrics now */ gpu_metrics->throttle_status = 0; - gpu_metrics->indep_throttle_status = smu_cmn_get_indep_throttler_status( - gpu_metrics->throttle_status, smu_v13_0_6_throttler_map); - gpu_metrics->current_fan_speed = 0; - - gpu_metrics->pcie_link_width = 0; - gpu_metrics->pcie_link_speed = smu_v13_0_6_get_current_pcie_link_speed(smu); + if (!(adev->flags & AMD_IS_APU)) { + gpu_metrics->pcie_link_width = + smu_v13_0_6_get_current_pcie_link_width_level(smu); + gpu_metrics->pcie_link_speed = + smu_v13_0_6_get_current_pcie_link_speed(smu); + } gpu_metrics->system_clock_counter = ktime_get_boottime_ns(); - gpu_metrics->gfx_activity_acc = 0; - gpu_metrics->mem_activity_acc = 0; + gpu_metrics->gfx_activity_acc = + SMUQ10_TO_UINT(metrics->SocketGfxBusyAcc); + gpu_metrics->mem_activity_acc = + SMUQ10_TO_UINT(metrics->DramBandwidthUtilizationAcc); - for (i = 0; i < NUM_HBM_INSTANCES; i++) - gpu_metrics->temperature_hbm[i] = 0; - - gpu_metrics->firmware_timestamp = 0; + gpu_metrics->firmware_timestamp = metrics->Timestamp; *table = (void *)gpu_metrics; kfree(metrics); @@ -1905,27 +2040,27 @@ static ssize_t smu_v13_0_6_get_gpu_metrics(struct smu_context *smu, void **table static int smu_v13_0_6_mode2_reset(struct smu_context *smu) { - u32 smu_version; int ret = 0, index; struct amdgpu_device *adev = smu->adev; int timeout = 10; - smu_cmn_get_smc_version(smu, NULL, &smu_version); - index = smu_cmn_to_asic_specific_index(smu, CMN2ASIC_MAPPING_MSG, SMU_MSG_GfxDeviceDriverReset); mutex_lock(&smu->message_lock); + ret = smu_cmn_send_msg_without_waiting(smu, (uint16_t)index, SMU_RESET_MODE_2); + /* This is similar to FLR, wait till max FLR timeout */ msleep(100); + dev_dbg(smu->adev->dev, "restore config space...\n"); /* Restore the config space saved during init */ amdgpu_device_load_pci_state(adev->pdev); dev_dbg(smu->adev->dev, "wait for reset ack\n"); - while (ret == -ETIME && timeout) { + do { ret = smu_cmn_wait_for_response(smu); /* Wait a bit more time for getting ACK */ if (ret == -ETIME) { @@ -1934,16 +2069,14 @@ static int smu_v13_0_6_mode2_reset(struct smu_context *smu) continue; } - if (ret != 1) { + if (ret) { dev_err(adev->dev, - "failed to send mode2 message \tparam: 0x%08x response %#x\n", + "failed to send mode2 message \tparam: 0x%08x error code %d\n", SMU_RESET_MODE_2, ret); goto out; } - } + } while (ret == -ETIME && timeout); - if (ret == 1) - ret = 0; out: mutex_unlock(&smu->message_lock); @@ -2032,11 +2165,9 @@ static const struct pptable_funcs smu_v13_0_6_ppt_funcs = { .feature_is_enabled = smu_cmn_feature_is_enabled, .set_power_limit = smu_v13_0_6_set_power_limit, .set_xgmi_pstate = smu_v13_0_set_xgmi_pstate, - /* TODO: Thermal limits unknown, skip these for now - .register_irq_handler = smu_v13_0_register_irq_handler, + .register_irq_handler = smu_v13_0_6_register_irq_handler, .enable_thermal_alert = smu_v13_0_enable_thermal_alert, .disable_thermal_alert = smu_v13_0_disable_thermal_alert, - */ .setup_pptable = smu_v13_0_6_setup_pptable, .baco_is_support = smu_v13_0_6_is_baco_supported, .get_dpm_ultimate_freq = smu_v13_0_6_get_dpm_ultimate_freq, @@ -2065,5 +2196,6 @@ void smu_v13_0_6_set_ppt_funcs(struct smu_context *smu) smu->clock_map = smu_v13_0_6_clk_map; smu->feature_map = smu_v13_0_6_feature_mask_map; smu->table_map = smu_v13_0_6_table_map; + smu->smc_driver_if_version = SMU13_0_6_DRIVER_IF_VERSION; smu_v13_0_set_smu_mailbox_registers(smu); } diff --git a/drivers/gpu/drm/amd/pm/swsmu/smu13/smu_v13_0_7_ppt.c b/drivers/gpu/drm/amd/pm/swsmu/smu13/smu_v13_0_7_ppt.c index 3d9ff46706fb..cda4e818aab7 100644 --- a/drivers/gpu/drm/amd/pm/swsmu/smu13/smu_v13_0_7_ppt.c +++ b/drivers/gpu/drm/amd/pm/swsmu/smu13/smu_v13_0_7_ppt.c @@ -125,6 +125,7 @@ static struct cmn2asic_msg_mapping smu_v13_0_7_message_map[SMU_MSG_MAX_COUNT] = MSG_MAP(ArmD3, PPSMC_MSG_ArmD3, 0), MSG_MAP(AllowGpo, PPSMC_MSG_SetGpoAllow, 0), MSG_MAP(GetPptLimit, PPSMC_MSG_GetPptLimit, 0), + MSG_MAP(NotifyPowerSource, PPSMC_MSG_NotifyPowerSource, 0), }; static struct cmn2asic_mapping smu_v13_0_7_clk_map[SMU_CLK_COUNT] = { @@ -205,6 +206,7 @@ static struct cmn2asic_mapping smu_v13_0_7_table_map[SMU_TABLE_COUNT] = { TAB_MAP(DRIVER_SMU_CONFIG), TAB_MAP(ACTIVITY_MONITOR_COEFF), [SMU_TABLE_COMBO_PPTABLE] = {1, TABLE_COMBO_PPTABLE}, + TAB_MAP(OVERDRIVE), }; static struct cmn2asic_mapping smu_v13_0_7_pwr_src_map[SMU_POWER_SOURCE_COUNT] = { @@ -321,6 +323,10 @@ static int smu_v13_0_7_check_powerplay_table(struct smu_context *smu) struct smu_baco_context *smu_baco = &smu->smu_baco; PPTable_t *smc_pptable = table_context->driver_pptable; BoardTable_t *BoardTable = &smc_pptable->BoardTable; + const OverDriveLimits_t * const overdrive_upperlimits = + &smc_pptable->SkuTable.OverDriveLimitsBasicMax; + const OverDriveLimits_t * const overdrive_lowerlimits = + &smc_pptable->SkuTable.OverDriveLimitsMin; if (powerplay_table->platform_caps & SMU_13_0_7_PP_PLATFORM_CAP_HARDWAREDC) smu->dc_controlled_by_gpio = true; @@ -332,6 +338,10 @@ static int smu_v13_0_7_check_powerplay_table(struct smu_context *smu) if (smu_baco->platform_support && (BoardTable->HsrEnabled || BoardTable->VddqOffEnabled)) smu_baco->maco_support = true; + if (!overdrive_lowerlimits->FeatureCtrlMask || + !overdrive_upperlimits->FeatureCtrlMask) + smu->od_enabled = false; + table_context->thermal_controller_type = powerplay_table->thermal_controller_type; @@ -478,7 +488,7 @@ static int smu_v13_0_7_tables_init(struct smu_context *smu) PAGE_SIZE, AMDGPU_GEM_DOMAIN_VRAM); SMU_TABLE_INIT(tables, SMU_TABLE_I2C_COMMANDS, sizeof(SwI2cRequest_t), PAGE_SIZE, AMDGPU_GEM_DOMAIN_VRAM); - SMU_TABLE_INIT(tables, SMU_TABLE_OVERDRIVE, sizeof(OverDriveTable_t), + SMU_TABLE_INIT(tables, SMU_TABLE_OVERDRIVE, sizeof(OverDriveTableExternal_t), PAGE_SIZE, AMDGPU_GEM_DOMAIN_VRAM); SMU_TABLE_INIT(tables, SMU_TABLE_PMSTATUSLOG, SMU13_TOOL_SIZE, PAGE_SIZE, AMDGPU_GEM_DOMAIN_VRAM); @@ -1012,16 +1022,118 @@ static int smu_v13_0_7_get_current_clk_freq_by_table(struct smu_context *smu, value); } +static bool smu_v13_0_7_is_od_feature_supported(struct smu_context *smu, + int od_feature_bit) +{ + PPTable_t *pptable = smu->smu_table.driver_pptable; + const OverDriveLimits_t * const overdrive_upperlimits = + &pptable->SkuTable.OverDriveLimitsBasicMax; + + return overdrive_upperlimits->FeatureCtrlMask & (1U << od_feature_bit); +} + +static void smu_v13_0_7_get_od_setting_limits(struct smu_context *smu, + int od_feature_bit, + bool lower_boundary, + int32_t *min, + int32_t *max) +{ + PPTable_t *pptable = smu->smu_table.driver_pptable; + const OverDriveLimits_t * const overdrive_upperlimits = + &pptable->SkuTable.OverDriveLimitsBasicMax; + const OverDriveLimits_t * const overdrive_lowerlimits = + &pptable->SkuTable.OverDriveLimitsMin; + int32_t od_min_setting, od_max_setting; + + switch (od_feature_bit) { + case PP_OD_FEATURE_GFXCLK_BIT: + if (lower_boundary) { + od_min_setting = overdrive_lowerlimits->GfxclkFmin; + od_max_setting = overdrive_upperlimits->GfxclkFmin; + } else { + od_min_setting = overdrive_lowerlimits->GfxclkFmax; + od_max_setting = overdrive_upperlimits->GfxclkFmax; + } + break; + case PP_OD_FEATURE_UCLK_BIT: + if (lower_boundary) { + od_min_setting = overdrive_lowerlimits->UclkFmin; + od_max_setting = overdrive_upperlimits->UclkFmin; + } else { + od_min_setting = overdrive_lowerlimits->UclkFmax; + od_max_setting = overdrive_upperlimits->UclkFmax; + } + break; + case PP_OD_FEATURE_GFX_VF_CURVE_BIT: + od_min_setting = overdrive_lowerlimits->VoltageOffsetPerZoneBoundary; + od_max_setting = overdrive_upperlimits->VoltageOffsetPerZoneBoundary; + break; + default: + break; + } + + if (min) + *min = od_min_setting; + if (max) + *max = od_max_setting; +} + +static void smu_v13_0_7_dump_od_table(struct smu_context *smu, + OverDriveTableExternal_t *od_table) +{ + struct amdgpu_device *adev = smu->adev; + + dev_dbg(adev->dev, "OD: Gfxclk: (%d, %d)\n", od_table->OverDriveTable.GfxclkFmin, + od_table->OverDriveTable.GfxclkFmax); + dev_dbg(adev->dev, "OD: Uclk: (%d, %d)\n", od_table->OverDriveTable.UclkFmin, + od_table->OverDriveTable.UclkFmax); +} + +static int smu_v13_0_7_get_overdrive_table(struct smu_context *smu, + OverDriveTableExternal_t *od_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; +} + +static int smu_v13_0_7_upload_overdrive_table(struct smu_context *smu, + OverDriveTableExternal_t *od_table) +{ + int ret = 0; + + ret = smu_cmn_update_table(smu, + SMU_TABLE_OVERDRIVE, + 0, + (void *)od_table, + true); + if (ret) + dev_err(smu->adev->dev, "Failed to upload overdrive table!\n"); + + return ret; +} + static int smu_v13_0_7_print_clk_levels(struct smu_context *smu, enum smu_clk_type clk_type, char *buf) { struct smu_dpm_context *smu_dpm = &smu->smu_dpm; struct smu_13_0_dpm_context *dpm_context = smu_dpm->dpm_context; + OverDriveTableExternal_t *od_table = + (OverDriveTableExternal_t *)smu->smu_table.overdrive_table; struct smu_13_0_dpm_table *single_dpm_table; struct smu_13_0_pcie_table *pcie_table; uint32_t gen_speed, lane_width; int i, curr_freq, size = 0; + int32_t min_value, max_value; int ret = 0; smu_cmn_get_sysfs_buf(&buf, &size); @@ -1138,6 +1250,89 @@ static int smu_v13_0_7_print_clk_levels(struct smu_context *smu, "*" : ""); break; + case SMU_OD_SCLK: + if (!smu_v13_0_7_is_od_feature_supported(smu, + PP_OD_FEATURE_GFXCLK_BIT)) + break; + + size += sysfs_emit_at(buf, size, "OD_SCLK:\n"); + size += sysfs_emit_at(buf, size, "0: %uMhz\n1: %uMhz\n", + od_table->OverDriveTable.GfxclkFmin, + od_table->OverDriveTable.GfxclkFmax); + break; + + case SMU_OD_MCLK: + if (!smu_v13_0_7_is_od_feature_supported(smu, + PP_OD_FEATURE_UCLK_BIT)) + break; + + size += sysfs_emit_at(buf, size, "OD_MCLK:\n"); + size += sysfs_emit_at(buf, size, "0: %uMhz\n1: %uMHz\n", + od_table->OverDriveTable.UclkFmin, + od_table->OverDriveTable.UclkFmax); + break; + + case SMU_OD_VDDC_CURVE: + if (!smu_v13_0_7_is_od_feature_supported(smu, + PP_OD_FEATURE_GFX_VF_CURVE_BIT)) + break; + + size += sysfs_emit_at(buf, size, "OD_VDDC_CURVE:\n"); + for (i = 0; i < PP_NUM_OD_VF_CURVE_POINTS; i++) + size += sysfs_emit_at(buf, size, "%d: %dmv\n", + i, + od_table->OverDriveTable.VoltageOffsetPerZoneBoundary[i]); + break; + + case SMU_OD_RANGE: + if (!smu_v13_0_7_is_od_feature_supported(smu, PP_OD_FEATURE_GFXCLK_BIT) && + !smu_v13_0_7_is_od_feature_supported(smu, PP_OD_FEATURE_UCLK_BIT) && + !smu_v13_0_7_is_od_feature_supported(smu, PP_OD_FEATURE_GFX_VF_CURVE_BIT)) + break; + + size += sysfs_emit_at(buf, size, "%s:\n", "OD_RANGE"); + + if (smu_v13_0_7_is_od_feature_supported(smu, PP_OD_FEATURE_GFXCLK_BIT)) { + smu_v13_0_7_get_od_setting_limits(smu, + PP_OD_FEATURE_GFXCLK_BIT, + true, + &min_value, + NULL); + smu_v13_0_7_get_od_setting_limits(smu, + PP_OD_FEATURE_GFXCLK_BIT, + false, + NULL, + &max_value); + size += sysfs_emit_at(buf, size, "SCLK: %7uMhz %10uMhz\n", + min_value, max_value); + } + + if (smu_v13_0_7_is_od_feature_supported(smu, PP_OD_FEATURE_UCLK_BIT)) { + smu_v13_0_7_get_od_setting_limits(smu, + PP_OD_FEATURE_UCLK_BIT, + true, + &min_value, + NULL); + smu_v13_0_7_get_od_setting_limits(smu, + PP_OD_FEATURE_UCLK_BIT, + false, + NULL, + &max_value); + size += sysfs_emit_at(buf, size, "MCLK: %7uMhz %10uMhz\n", + min_value, max_value); + } + + if (smu_v13_0_7_is_od_feature_supported(smu, PP_OD_FEATURE_GFX_VF_CURVE_BIT)) { + smu_v13_0_7_get_od_setting_limits(smu, + PP_OD_FEATURE_GFX_VF_CURVE_BIT, + true, + &min_value, + &max_value); + size += sysfs_emit_at(buf, size, "VDDC_CURVE: %7dmv %10dmv\n", + min_value, max_value); + } + break; + default: break; } @@ -1145,6 +1340,222 @@ static int smu_v13_0_7_print_clk_levels(struct smu_context *smu, return size; } +static int smu_v13_0_7_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; + OverDriveTableExternal_t *od_table = + (OverDriveTableExternal_t *)table_context->overdrive_table; + struct amdgpu_device *adev = smu->adev; + uint32_t offset_of_featurectrlmask; + int32_t minimum, maximum; + uint32_t feature_ctrlmask; + int i, ret = 0; + + switch (type) { + case PP_OD_EDIT_SCLK_VDDC_TABLE: + if (!smu_v13_0_7_is_od_feature_supported(smu, PP_OD_FEATURE_GFXCLK_BIT)) { + dev_warn(adev->dev, "GFXCLK_LIMITS setting not supported!\n"); + return -ENOTSUPP; + } + + for (i = 0; i < size; i += 2) { + if (i + 2 > size) { + dev_info(adev->dev, "invalid number of input parameters %d\n", size); + return -EINVAL; + } + + switch (input[i]) { + case 0: + smu_v13_0_7_get_od_setting_limits(smu, + PP_OD_FEATURE_GFXCLK_BIT, + true, + &minimum, + &maximum); + if (input[i + 1] < minimum || + input[i + 1] > maximum) { + dev_info(adev->dev, "GfxclkFmin (%ld) must be within [%u, %u]!\n", + input[i + 1], minimum, maximum); + return -EINVAL; + } + + od_table->OverDriveTable.GfxclkFmin = input[i + 1]; + od_table->OverDriveTable.FeatureCtrlMask |= 1U << PP_OD_FEATURE_GFXCLK_BIT; + break; + + case 1: + smu_v13_0_7_get_od_setting_limits(smu, + PP_OD_FEATURE_GFXCLK_BIT, + false, + &minimum, + &maximum); + if (input[i + 1] < minimum || + input[i + 1] > maximum) { + dev_info(adev->dev, "GfxclkFmax (%ld) must be within [%u, %u]!\n", + input[i + 1], minimum, maximum); + return -EINVAL; + } + + od_table->OverDriveTable.GfxclkFmax = input[i + 1]; + od_table->OverDriveTable.FeatureCtrlMask |= 1U << PP_OD_FEATURE_GFXCLK_BIT; + break; + + default: + dev_info(adev->dev, "Invalid SCLK_VDDC_TABLE index: %ld\n", input[i]); + dev_info(adev->dev, "Supported indices: [0:min,1:max]\n"); + return -EINVAL; + } + } + + if (od_table->OverDriveTable.GfxclkFmin > od_table->OverDriveTable.GfxclkFmax) { + dev_err(adev->dev, + "Invalid setting: GfxclkFmin(%u) is bigger than GfxclkFmax(%u)\n", + (uint32_t)od_table->OverDriveTable.GfxclkFmin, + (uint32_t)od_table->OverDriveTable.GfxclkFmax); + return -EINVAL; + } + break; + + case PP_OD_EDIT_MCLK_VDDC_TABLE: + if (!smu_v13_0_7_is_od_feature_supported(smu, PP_OD_FEATURE_UCLK_BIT)) { + dev_warn(adev->dev, "UCLK_LIMITS setting not supported!\n"); + return -ENOTSUPP; + } + + for (i = 0; i < size; i += 2) { + if (i + 2 > size) { + dev_info(adev->dev, "invalid number of input parameters %d\n", size); + return -EINVAL; + } + + switch (input[i]) { + case 0: + smu_v13_0_7_get_od_setting_limits(smu, + PP_OD_FEATURE_UCLK_BIT, + true, + &minimum, + &maximum); + if (input[i + 1] < minimum || + input[i + 1] > maximum) { + dev_info(adev->dev, "UclkFmin (%ld) must be within [%u, %u]!\n", + input[i + 1], minimum, maximum); + return -EINVAL; + } + + od_table->OverDriveTable.UclkFmin = input[i + 1]; + od_table->OverDriveTable.FeatureCtrlMask |= 1U << PP_OD_FEATURE_UCLK_BIT; + break; + + case 1: + smu_v13_0_7_get_od_setting_limits(smu, + PP_OD_FEATURE_UCLK_BIT, + false, + &minimum, + &maximum); + if (input[i + 1] < minimum || + input[i + 1] > maximum) { + dev_info(adev->dev, "UclkFmax (%ld) must be within [%u, %u]!\n", + input[i + 1], minimum, maximum); + return -EINVAL; + } + + od_table->OverDriveTable.UclkFmax = input[i + 1]; + od_table->OverDriveTable.FeatureCtrlMask |= 1U << PP_OD_FEATURE_UCLK_BIT; + break; + + default: + dev_info(adev->dev, "Invalid MCLK_VDDC_TABLE index: %ld\n", input[i]); + dev_info(adev->dev, "Supported indices: [0:min,1:max]\n"); + return -EINVAL; + } + } + + if (od_table->OverDriveTable.UclkFmin > od_table->OverDriveTable.UclkFmax) { + dev_err(adev->dev, + "Invalid setting: UclkFmin(%u) is bigger than UclkFmax(%u)\n", + (uint32_t)od_table->OverDriveTable.UclkFmin, + (uint32_t)od_table->OverDriveTable.UclkFmax); + return -EINVAL; + } + break; + + case PP_OD_EDIT_VDDC_CURVE: + if (!smu_v13_0_7_is_od_feature_supported(smu, PP_OD_FEATURE_GFX_VF_CURVE_BIT)) { + dev_warn(adev->dev, "VF curve setting not supported!\n"); + return -ENOTSUPP; + } + + if (input[0] >= PP_NUM_OD_VF_CURVE_POINTS || + input[0] < 0) + return -EINVAL; + + smu_v13_0_7_get_od_setting_limits(smu, + PP_OD_FEATURE_GFX_VF_CURVE_BIT, + true, + &minimum, + &maximum); + if (input[1] < minimum || + input[1] > maximum) { + dev_info(adev->dev, "Voltage offset (%ld) must be within [%d, %d]!\n", + input[1], minimum, maximum); + return -EINVAL; + } + + od_table->OverDriveTable.VoltageOffsetPerZoneBoundary[input[0]] = input[1]; + od_table->OverDriveTable.FeatureCtrlMask |= 1U << PP_OD_FEATURE_GFX_VF_CURVE_BIT; + break; + + case PP_OD_RESTORE_DEFAULT_TABLE: + feature_ctrlmask = od_table->OverDriveTable.FeatureCtrlMask; + memcpy(od_table, + table_context->boot_overdrive_table, + sizeof(OverDriveTableExternal_t)); + od_table->OverDriveTable.FeatureCtrlMask = feature_ctrlmask; + fallthrough; + + case PP_OD_COMMIT_DPM_TABLE: + /* + * The member below instructs PMFW the settings focused in + * this single operation. + * `uint32_t FeatureCtrlMask;` + * It does not contain actual informations about user's custom + * settings. Thus we do not cache it. + */ + offset_of_featurectrlmask = offsetof(OverDriveTable_t, FeatureCtrlMask); + if (memcmp((u8 *)od_table + offset_of_featurectrlmask, + table_context->user_overdrive_table + offset_of_featurectrlmask, + sizeof(OverDriveTableExternal_t) - offset_of_featurectrlmask)) { + smu_v13_0_7_dump_od_table(smu, od_table); + + ret = smu_v13_0_7_upload_overdrive_table(smu, od_table); + if (ret) { + dev_err(adev->dev, "Failed to upload overdrive table!\n"); + return ret; + } + + od_table->OverDriveTable.FeatureCtrlMask = 0; + memcpy(table_context->user_overdrive_table + offset_of_featurectrlmask, + (u8 *)od_table + offset_of_featurectrlmask, + sizeof(OverDriveTableExternal_t) - offset_of_featurectrlmask); + + if (!memcmp(table_context->user_overdrive_table, + table_context->boot_overdrive_table, + sizeof(OverDriveTableExternal_t))) + smu->user_dpm_profile.user_od = false; + else + smu->user_dpm_profile.user_od = true; + } + break; + + default: + return -ENOSYS; + } + + return ret; +} + static int smu_v13_0_7_force_clk_levels(struct smu_context *smu, enum smu_clk_type clk_type, uint32_t mask) @@ -1370,6 +1781,78 @@ static ssize_t smu_v13_0_7_get_gpu_metrics(struct smu_context *smu, return sizeof(struct gpu_metrics_v1_3); } +static int smu_v13_0_7_set_default_od_settings(struct smu_context *smu) +{ + OverDriveTableExternal_t *od_table = + (OverDriveTableExternal_t *)smu->smu_table.overdrive_table; + OverDriveTableExternal_t *boot_od_table = + (OverDriveTableExternal_t *)smu->smu_table.boot_overdrive_table; + OverDriveTableExternal_t *user_od_table = + (OverDriveTableExternal_t *)smu->smu_table.user_overdrive_table; + OverDriveTableExternal_t user_od_table_bak; + int ret = 0; + int i; + + ret = smu_v13_0_7_get_overdrive_table(smu, boot_od_table); + if (ret) + return ret; + + smu_v13_0_7_dump_od_table(smu, boot_od_table); + + memcpy(od_table, + boot_od_table, + sizeof(OverDriveTableExternal_t)); + + /* + * For S3/S4/Runpm resume, we need to setup those overdrive tables again, + * but we have to preserve user defined values in "user_od_table". + */ + if (!smu->adev->in_suspend) { + memcpy(user_od_table, + boot_od_table, + sizeof(OverDriveTableExternal_t)); + smu->user_dpm_profile.user_od = false; + } else if (smu->user_dpm_profile.user_od) { + memcpy(&user_od_table_bak, + user_od_table, + sizeof(OverDriveTableExternal_t)); + memcpy(user_od_table, + boot_od_table, + sizeof(OverDriveTableExternal_t)); + user_od_table->OverDriveTable.GfxclkFmin = + user_od_table_bak.OverDriveTable.GfxclkFmin; + user_od_table->OverDriveTable.GfxclkFmax = + user_od_table_bak.OverDriveTable.GfxclkFmax; + user_od_table->OverDriveTable.UclkFmin = + user_od_table_bak.OverDriveTable.UclkFmin; + user_od_table->OverDriveTable.UclkFmax = + user_od_table_bak.OverDriveTable.UclkFmax; + for (i = 0; i < PP_NUM_OD_VF_CURVE_POINTS; i++) + user_od_table->OverDriveTable.VoltageOffsetPerZoneBoundary[i] = + user_od_table_bak.OverDriveTable.VoltageOffsetPerZoneBoundary[i]; + } + + return 0; +} + +static int smu_v13_0_7_restore_user_od_settings(struct smu_context *smu) +{ + struct smu_table_context *table_context = &smu->smu_table; + OverDriveTableExternal_t *od_table = table_context->overdrive_table; + OverDriveTableExternal_t *user_od_table = table_context->user_overdrive_table; + int res; + + user_od_table->OverDriveTable.FeatureCtrlMask = 1U << PP_OD_FEATURE_GFXCLK_BIT | + 1U << PP_OD_FEATURE_UCLK_BIT | + 1U << PP_OD_FEATURE_GFX_VF_CURVE_BIT; + res = smu_v13_0_7_upload_overdrive_table(smu, user_od_table); + user_od_table->OverDriveTable.FeatureCtrlMask = 0; + if (res == 0) + memcpy(od_table, user_od_table, sizeof(OverDriveTableExternal_t)); + + return res; +} + static int smu_v13_0_7_populate_umd_state_clk(struct smu_context *smu) { struct smu_13_0_dpm_context *dpm_context = @@ -1759,6 +2242,9 @@ static const struct pptable_funcs smu_v13_0_7_ppt_funcs = { .notify_memory_pool_location = smu_v13_0_notify_memory_pool_location, .get_gpu_metrics = smu_v13_0_7_get_gpu_metrics, .set_soft_freq_limited_range = smu_v13_0_set_soft_freq_limited_range, + .set_default_od_settings = smu_v13_0_7_set_default_od_settings, + .restore_user_od_settings = smu_v13_0_7_restore_user_od_settings, + .od_edit_dpm_table = smu_v13_0_7_od_edit_dpm_table, .set_performance_level = smu_v13_0_set_performance_level, .gfx_off_control = smu_v13_0_gfx_off_control, .get_fan_speed_pwm = smu_v13_0_7_get_fan_speed_pwm, @@ -1770,6 +2256,7 @@ static const struct pptable_funcs smu_v13_0_7_ppt_funcs = { .enable_mgpu_fan_boost = smu_v13_0_7_enable_mgpu_fan_boost, .get_power_limit = smu_v13_0_7_get_power_limit, .set_power_limit = smu_v13_0_set_power_limit, + .set_power_source = smu_v13_0_set_power_source, .get_power_profile_mode = smu_v13_0_7_get_power_profile_mode, .set_power_profile_mode = smu_v13_0_7_set_power_profile_mode, .set_tool_table_location = smu_v13_0_set_tool_table_location, @@ -1796,5 +2283,6 @@ void smu_v13_0_7_set_ppt_funcs(struct smu_context *smu) smu->table_map = smu_v13_0_7_table_map; smu->pwr_src_map = smu_v13_0_7_pwr_src_map; smu->workload_map = smu_v13_0_7_workload_map; + smu->smc_driver_if_version = SMU13_0_7_DRIVER_IF_VERSION; smu_v13_0_set_smu_mailbox_registers(smu); } diff --git a/drivers/gpu/drm/amd/pm/swsmu/smu13/yellow_carp_ppt.c b/drivers/gpu/drm/amd/pm/swsmu/smu13/yellow_carp_ppt.c index 04e56b0b3033..a1be2029ba4a 100644 --- a/drivers/gpu/drm/amd/pm/swsmu/smu13/yellow_carp_ppt.c +++ b/drivers/gpu/drm/amd/pm/swsmu/smu13/yellow_carp_ppt.c @@ -47,6 +47,14 @@ #define SMUIO_GFX_MISC_CNTL__PWR_GFXOFF_STATUS_MASK 0x00000006L #define SMUIO_GFX_MISC_CNTL__PWR_GFXOFF_STATUS__SHIFT 0x1L +#define SMU_13_0_8_UMD_PSTATE_GFXCLK 533 +#define SMU_13_0_8_UMD_PSTATE_SOCCLK 533 +#define SMU_13_0_8_UMD_PSTATE_FCLK 800 + +#define SMU_13_0_1_UMD_PSTATE_GFXCLK 700 +#define SMU_13_0_1_UMD_PSTATE_SOCCLK 678 +#define SMU_13_0_1_UMD_PSTATE_FCLK 1800 + #define FEATURE_MASK(feature) (1ULL << feature) #define SMC_DPM_FEATURE ( \ FEATURE_MASK(FEATURE_CCLK_DPM_BIT) | \ @@ -957,6 +965,9 @@ static int yellow_carp_set_soft_freq_limited_range(struct smu_context *smu, uint32_t max) { enum smu_message_type msg_set_min, msg_set_max; + uint32_t min_clk = min; + uint32_t max_clk = max; + int ret = 0; if (!yellow_carp_clk_dpm_is_enabled(smu, clk_type)) @@ -985,11 +996,17 @@ static int yellow_carp_set_soft_freq_limited_range(struct smu_context *smu, return -EINVAL; } - ret = smu_cmn_send_smc_msg_with_param(smu, msg_set_min, min, NULL); + if (clk_type == SMU_VCLK) { + min_clk = min << SMU_13_VCLK_SHIFT; + max_clk = max << SMU_13_VCLK_SHIFT; + } + + ret = smu_cmn_send_smc_msg_with_param(smu, msg_set_min, min_clk, NULL); + if (ret) goto out; - ret = smu_cmn_send_smc_msg_with_param(smu, msg_set_max, max, NULL); + ret = smu_cmn_send_smc_msg_with_param(smu, msg_set_max, max_clk, NULL); if (ret) goto out; @@ -997,12 +1014,49 @@ out: return ret; } +static uint32_t yellow_carp_get_umd_pstate_clk_default(struct smu_context *smu, + enum smu_clk_type clk_type) +{ + uint32_t clk_limit = 0; + struct amdgpu_device *adev = smu->adev; + + switch (clk_type) { + case SMU_GFXCLK: + case SMU_SCLK: + if ((adev->ip_versions[MP1_HWIP][0]) == IP_VERSION(13, 0, 8)) + clk_limit = SMU_13_0_8_UMD_PSTATE_GFXCLK; + if ((adev->ip_versions[MP1_HWIP][0]) == IP_VERSION(13, 0, 1) || + (adev->ip_versions[MP1_HWIP][0]) == IP_VERSION(13, 0, 3)) + clk_limit = SMU_13_0_1_UMD_PSTATE_GFXCLK; + break; + case SMU_SOCCLK: + if ((adev->ip_versions[MP1_HWIP][0]) == IP_VERSION(13, 0, 8)) + clk_limit = SMU_13_0_8_UMD_PSTATE_SOCCLK; + if ((adev->ip_versions[MP1_HWIP][0]) == IP_VERSION(13, 0, 1) || + (adev->ip_versions[MP1_HWIP][0]) == IP_VERSION(13, 0, 3)) + clk_limit = SMU_13_0_1_UMD_PSTATE_SOCCLK; + break; + case SMU_FCLK: + if ((adev->ip_versions[MP1_HWIP][0]) == IP_VERSION(13, 0, 8)) + clk_limit = SMU_13_0_8_UMD_PSTATE_FCLK; + if ((adev->ip_versions[MP1_HWIP][0]) == IP_VERSION(13, 0, 1) || + (adev->ip_versions[MP1_HWIP][0]) == IP_VERSION(13, 0, 3)) + clk_limit = SMU_13_0_1_UMD_PSTATE_FCLK; + break; + default: + break; + } + + return clk_limit; +} + static int yellow_carp_print_clk_levels(struct smu_context *smu, enum smu_clk_type clk_type, char *buf) { - int i, size = 0, ret = 0; + int i, idx, size = 0, ret = 0; uint32_t cur_value = 0, value = 0, count = 0; uint32_t min, max; + uint32_t clk_limit = 0; smu_cmn_get_sysfs_buf(&buf, &size); @@ -1033,7 +1087,8 @@ static int yellow_carp_print_clk_levels(struct smu_context *smu, goto print_clk_out; for (i = 0; i < count; i++) { - ret = yellow_carp_get_dpm_freq_by_index(smu, clk_type, i, &value); + idx = (clk_type == SMU_FCLK || clk_type == SMU_MCLK) ? (count - i - 1) : i; + ret = yellow_carp_get_dpm_freq_by_index(smu, clk_type, idx, &value); if (ret) goto print_clk_out; @@ -1043,6 +1098,7 @@ static int yellow_carp_print_clk_levels(struct smu_context *smu, break; case SMU_GFXCLK: case SMU_SCLK: + clk_limit = yellow_carp_get_umd_pstate_clk_default(smu, clk_type); ret = yellow_carp_get_current_clk_freq(smu, clk_type, &cur_value); if (ret) goto print_clk_out; @@ -1057,7 +1113,7 @@ static int yellow_carp_print_clk_levels(struct smu_context *smu, size += sysfs_emit_at(buf, size, "0: %uMhz %s\n", min, i == 0 ? "*" : ""); size += sysfs_emit_at(buf, size, "1: %uMhz %s\n", - i == 1 ? cur_value : YELLOW_CARP_UMD_PSTATE_GFXCLK, + i == 1 ? cur_value : clk_limit, i == 1 ? "*" : ""); size += sysfs_emit_at(buf, size, "2: %uMhz %s\n", max, i == 2 ? "*" : ""); @@ -1106,6 +1162,49 @@ force_level_out: return ret; } +static int yellow_carp_get_dpm_profile_freq(struct smu_context *smu, + enum amd_dpm_forced_level level, + enum smu_clk_type clk_type, + uint32_t *min_clk, + uint32_t *max_clk) +{ + int ret = 0; + uint32_t clk_limit = 0; + + clk_limit = yellow_carp_get_umd_pstate_clk_default(smu, clk_type); + + switch (clk_type) { + case SMU_GFXCLK: + case SMU_SCLK: + if (level == AMD_DPM_FORCED_LEVEL_PROFILE_PEAK) + yellow_carp_get_dpm_ultimate_freq(smu, SMU_SCLK, NULL, &clk_limit); + else if (level == AMD_DPM_FORCED_LEVEL_PROFILE_MIN_SCLK) + yellow_carp_get_dpm_ultimate_freq(smu, SMU_SCLK, &clk_limit, NULL); + break; + case SMU_SOCCLK: + if (level == AMD_DPM_FORCED_LEVEL_PROFILE_PEAK) + yellow_carp_get_dpm_ultimate_freq(smu, SMU_SOCCLK, NULL, &clk_limit); + break; + case SMU_FCLK: + if (level == AMD_DPM_FORCED_LEVEL_PROFILE_PEAK) + yellow_carp_get_dpm_ultimate_freq(smu, SMU_FCLK, NULL, &clk_limit); + else if (level == AMD_DPM_FORCED_LEVEL_PROFILE_MIN_MCLK) + yellow_carp_get_dpm_ultimate_freq(smu, SMU_FCLK, &clk_limit, NULL); + break; + case SMU_VCLK: + yellow_carp_get_dpm_ultimate_freq(smu, SMU_VCLK, NULL, &clk_limit); + break; + case SMU_DCLK: + yellow_carp_get_dpm_ultimate_freq(smu, SMU_DCLK, NULL, &clk_limit); + break; + default: + ret = -EINVAL; + break; + } + *min_clk = *max_clk = clk_limit; + return ret; +} + static int yellow_carp_set_performance_level(struct smu_context *smu, enum amd_dpm_forced_level level) { @@ -1113,6 +1212,9 @@ static int yellow_carp_set_performance_level(struct smu_context *smu, uint32_t sclk_min = 0, sclk_max = 0; uint32_t fclk_min = 0, fclk_max = 0; uint32_t socclk_min = 0, socclk_max = 0; + uint32_t vclk_min = 0, vclk_max = 0; + uint32_t dclk_min = 0, dclk_max = 0; + int ret = 0; switch (level) { @@ -1120,28 +1222,42 @@ static int yellow_carp_set_performance_level(struct smu_context *smu, yellow_carp_get_dpm_ultimate_freq(smu, SMU_SCLK, NULL, &sclk_max); yellow_carp_get_dpm_ultimate_freq(smu, SMU_FCLK, NULL, &fclk_max); yellow_carp_get_dpm_ultimate_freq(smu, SMU_SOCCLK, NULL, &socclk_max); + yellow_carp_get_dpm_ultimate_freq(smu, SMU_VCLK, NULL, &vclk_max); + yellow_carp_get_dpm_ultimate_freq(smu, SMU_DCLK, NULL, &dclk_max); sclk_min = sclk_max; fclk_min = fclk_max; socclk_min = socclk_max; + vclk_min = vclk_max; + dclk_min = dclk_max; break; case AMD_DPM_FORCED_LEVEL_LOW: yellow_carp_get_dpm_ultimate_freq(smu, SMU_SCLK, &sclk_min, NULL); yellow_carp_get_dpm_ultimate_freq(smu, SMU_FCLK, &fclk_min, NULL); yellow_carp_get_dpm_ultimate_freq(smu, SMU_SOCCLK, &socclk_min, NULL); + yellow_carp_get_dpm_ultimate_freq(smu, SMU_VCLK, &vclk_min, NULL); + yellow_carp_get_dpm_ultimate_freq(smu, SMU_DCLK, &dclk_min, NULL); sclk_max = sclk_min; fclk_max = fclk_min; socclk_max = socclk_min; + vclk_max = vclk_min; + dclk_max = dclk_min; break; case AMD_DPM_FORCED_LEVEL_AUTO: yellow_carp_get_dpm_ultimate_freq(smu, SMU_SCLK, &sclk_min, &sclk_max); yellow_carp_get_dpm_ultimate_freq(smu, SMU_FCLK, &fclk_min, &fclk_max); yellow_carp_get_dpm_ultimate_freq(smu, SMU_SOCCLK, &socclk_min, &socclk_max); + yellow_carp_get_dpm_ultimate_freq(smu, SMU_VCLK, &vclk_min, &vclk_max); + yellow_carp_get_dpm_ultimate_freq(smu, SMU_DCLK, &dclk_min, &dclk_max); break; case AMD_DPM_FORCED_LEVEL_PROFILE_STANDARD: case AMD_DPM_FORCED_LEVEL_PROFILE_MIN_SCLK: case AMD_DPM_FORCED_LEVEL_PROFILE_MIN_MCLK: case AMD_DPM_FORCED_LEVEL_PROFILE_PEAK: - /* Temporarily do nothing since the optimal clocks haven't been provided yet */ + yellow_carp_get_dpm_profile_freq(smu, level, SMU_SCLK, &sclk_min, &sclk_max); + yellow_carp_get_dpm_profile_freq(smu, level, SMU_FCLK, &fclk_min, &fclk_max); + yellow_carp_get_dpm_profile_freq(smu, level, SMU_SOCCLK, &socclk_min, &socclk_max); + yellow_carp_get_dpm_profile_freq(smu, level, SMU_VCLK, &vclk_min, &vclk_max); + yellow_carp_get_dpm_profile_freq(smu, level, SMU_DCLK, &dclk_min, &dclk_max); break; case AMD_DPM_FORCED_LEVEL_MANUAL: case AMD_DPM_FORCED_LEVEL_PROFILE_EXIT: @@ -1181,6 +1297,24 @@ static int yellow_carp_set_performance_level(struct smu_context *smu, return ret; } + if (vclk_min && vclk_max) { + ret = yellow_carp_set_soft_freq_limited_range(smu, + SMU_VCLK, + vclk_min, + vclk_max); + if (ret) + return ret; + } + + if (dclk_min && dclk_max) { + ret = yellow_carp_set_soft_freq_limited_range(smu, + SMU_DCLK, + dclk_min, + dclk_max); + if (ret) + return ret; + } + return ret; } @@ -1234,5 +1368,6 @@ void yellow_carp_set_ppt_funcs(struct smu_context *smu) smu->feature_map = yellow_carp_feature_mask_map; smu->table_map = yellow_carp_table_map; smu->is_apu = true; + smu->smc_driver_if_version = SMU13_YELLOW_CARP_DRIVER_IF_VERSION; smu_v13_0_set_smu_mailbox_registers(smu); } diff --git a/drivers/gpu/drm/amd/pm/swsmu/smu13/yellow_carp_ppt.h b/drivers/gpu/drm/amd/pm/swsmu/smu13/yellow_carp_ppt.h index a9205a8ea3ad..b3ad8352c68a 100644 --- a/drivers/gpu/drm/amd/pm/swsmu/smu13/yellow_carp_ppt.h +++ b/drivers/gpu/drm/amd/pm/swsmu/smu13/yellow_carp_ppt.h @@ -24,6 +24,5 @@ #define __YELLOW_CARP_PPT_H__ extern void yellow_carp_set_ppt_funcs(struct smu_context *smu); -#define YELLOW_CARP_UMD_PSTATE_GFXCLK 1100 #endif |