diff options
| -rw-r--r-- | drivers/gpu/drm/amd/amdgpu/amdgpu_device.c | 11 | ||||
| -rw-r--r-- | drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm.c | 11 | ||||
| -rw-r--r-- | drivers/gpu/drm/amd/display/dc/dml/dcn20/dcn20_fpu.c | 110 | ||||
| -rw-r--r-- | drivers/gpu/drm/amd/display/modules/info_packet/info_packet.c | 13 | ||||
| -rw-r--r-- | drivers/gpu/drm/amd/include/kgd_pp_interface.h | 80 | ||||
| -rw-r--r-- | drivers/gpu/drm/amd/pm/amdgpu_pm.c | 4 | ||||
| -rw-r--r-- | drivers/gpu/drm/amd/pm/swsmu/inc/pmfw_if/smu_v13_0_6_pmfw.h | 100 | ||||
| -rw-r--r-- | drivers/gpu/drm/amd/pm/swsmu/smu13/smu_v13_0_6_ppt.c | 153 | ||||
| -rw-r--r-- | drivers/gpu/drm/amd/pm/swsmu/smu_cmn.c | 3 |
9 files changed, 400 insertions, 85 deletions
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c index 8dee52ce26d0..93cf73d6fa11 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c @@ -2188,15 +2188,8 @@ static int amdgpu_device_parse_gpu_info_fw(struct amdgpu_device *adev) adev->firmware.gpu_info_fw = NULL; - if (adev->mman.discovery_bin) { - /* - * FIXME: The bounding box is still needed by Navi12, so - * temporarily read it from gpu_info firmware. Should be dropped - * when DAL no longer needs it. - */ - if (adev->asic_type != CHIP_NAVI12) - return 0; - } + if (adev->mman.discovery_bin) + return 0; switch (adev->asic_type) { default: diff --git a/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm.c b/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm.c index c8c00c2a5224..4e82ee4d74ac 100644 --- a/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm.c +++ b/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm.c @@ -6170,8 +6170,9 @@ create_stream_for_sink(struct amdgpu_dm_connector *aconnector, if (stream->signal == SIGNAL_TYPE_HDMI_TYPE_A) mod_build_hf_vsif_infopacket(stream, &stream->vsp_infopacket); - - if (stream->link->psr_settings.psr_feature_enabled || stream->link->replay_settings.replay_feature_enabled) { + else if (stream->signal == SIGNAL_TYPE_DISPLAY_PORT || + stream->signal == SIGNAL_TYPE_DISPLAY_PORT_MST || + stream->signal == SIGNAL_TYPE_EDP) { // // should decide stream support vsc sdp colorimetry capability // before building vsc info packet @@ -6187,8 +6188,9 @@ create_stream_for_sink(struct amdgpu_dm_connector *aconnector, if (stream->out_transfer_func->tf == TRANSFER_FUNCTION_GAMMA22) tf = TRANSFER_FUNC_GAMMA_22; mod_build_vsc_infopacket(stream, &stream->vsc_infopacket, stream->output_color_space, tf); - aconnector->psr_skip_count = AMDGPU_DM_PSR_ENTRY_DELAY; + if (stream->link->psr_settings.psr_feature_enabled) + aconnector->psr_skip_count = AMDGPU_DM_PSR_ENTRY_DELAY; } finish: dc_sink_release(sink); @@ -6914,8 +6916,7 @@ static int dm_encoder_helper_atomic_check(struct drm_encoder *encoder, if (IS_ERR(mst_state)) return PTR_ERR(mst_state); - if (!mst_state->pbn_div) - mst_state->pbn_div = dm_mst_get_pbn_divider(aconnector->mst_root->dc_link); + mst_state->pbn_div = dm_mst_get_pbn_divider(aconnector->mst_root->dc_link); if (!state->duplicated) { int max_bpc = conn_state->max_requested_bpc; diff --git a/drivers/gpu/drm/amd/display/dc/dml/dcn20/dcn20_fpu.c b/drivers/gpu/drm/amd/display/dc/dml/dcn20/dcn20_fpu.c index ec77b2b41ba3..d2271e308fa0 100644 --- a/drivers/gpu/drm/amd/display/dc/dml/dcn20/dcn20_fpu.c +++ b/drivers/gpu/drm/amd/display/dc/dml/dcn20/dcn20_fpu.c @@ -440,7 +440,115 @@ struct _vcs_dpi_soc_bounding_box_st dcn2_0_nv14_soc = { .use_urgent_burst_bw = 0 }; -struct _vcs_dpi_soc_bounding_box_st dcn2_0_nv12_soc = { 0 }; +struct _vcs_dpi_soc_bounding_box_st dcn2_0_nv12_soc = { + .clock_limits = { + { + .state = 0, + .dcfclk_mhz = 560.0, + .fabricclk_mhz = 560.0, + .dispclk_mhz = 513.0, + .dppclk_mhz = 513.0, + .phyclk_mhz = 540.0, + .socclk_mhz = 560.0, + .dscclk_mhz = 171.0, + .dram_speed_mts = 1069.0, + }, + { + .state = 1, + .dcfclk_mhz = 694.0, + .fabricclk_mhz = 694.0, + .dispclk_mhz = 642.0, + .dppclk_mhz = 642.0, + .phyclk_mhz = 600.0, + .socclk_mhz = 694.0, + .dscclk_mhz = 214.0, + .dram_speed_mts = 1324.0, + }, + { + .state = 2, + .dcfclk_mhz = 875.0, + .fabricclk_mhz = 875.0, + .dispclk_mhz = 734.0, + .dppclk_mhz = 734.0, + .phyclk_mhz = 810.0, + .socclk_mhz = 875.0, + .dscclk_mhz = 245.0, + .dram_speed_mts = 1670.0, + }, + { + .state = 3, + .dcfclk_mhz = 1000.0, + .fabricclk_mhz = 1000.0, + .dispclk_mhz = 1100.0, + .dppclk_mhz = 1100.0, + .phyclk_mhz = 810.0, + .socclk_mhz = 1000.0, + .dscclk_mhz = 367.0, + .dram_speed_mts = 2000.0, + }, + { + .state = 4, + .dcfclk_mhz = 1200.0, + .fabricclk_mhz = 1200.0, + .dispclk_mhz = 1284.0, + .dppclk_mhz = 1284.0, + .phyclk_mhz = 810.0, + .socclk_mhz = 1200.0, + .dscclk_mhz = 428.0, + .dram_speed_mts = 2000.0, + }, + { + .state = 5, + .dcfclk_mhz = 1200.0, + .fabricclk_mhz = 1200.0, + .dispclk_mhz = 1284.0, + .dppclk_mhz = 1284.0, + .phyclk_mhz = 810.0, + .socclk_mhz = 1200.0, + .dscclk_mhz = 428.0, + .dram_speed_mts = 2000.0, + }, + }, + + .num_states = 5, + .sr_exit_time_us = 1.9, + .sr_enter_plus_exit_time_us = 4.4, + .urgent_latency_us = 3.0, + .urgent_latency_pixel_data_only_us = 4.0, + .urgent_latency_pixel_mixed_with_vm_data_us = 4.0, + .urgent_latency_vm_data_only_us = 4.0, + .urgent_out_of_order_return_per_channel_pixel_only_bytes = 4096, + .urgent_out_of_order_return_per_channel_pixel_and_vm_bytes = 4096, + .urgent_out_of_order_return_per_channel_vm_only_bytes = 4096, + .pct_ideal_dram_sdp_bw_after_urgent_pixel_only = 40.0, + .pct_ideal_dram_sdp_bw_after_urgent_pixel_and_vm = 40.0, + .pct_ideal_dram_sdp_bw_after_urgent_vm_only = 40.0, + .max_avg_sdp_bw_use_normal_percent = 40.0, + .max_avg_dram_bw_use_normal_percent = 40.0, + .writeback_latency_us = 12.0, + .ideal_dram_bw_after_urgent_percent = 40.0, + .max_request_size_bytes = 256, + .dram_channel_width_bytes = 16, + .fabric_datapath_to_dcn_data_return_bytes = 64, + .dcn_downspread_percent = 0.5, + .downspread_percent = 0.5, + .dram_page_open_time_ns = 50.0, + .dram_rw_turnaround_time_ns = 17.5, + .dram_return_buffer_per_channel_bytes = 8192, + .round_trip_ping_latency_dcfclk_cycles = 131, + .urgent_out_of_order_return_per_channel_bytes = 4096, + .channel_interleave_bytes = 256, + .num_banks = 8, + .num_chans = 16, + .vmm_page_size_bytes = 4096, + .dram_clock_change_latency_us = 45.0, + .writeback_dram_clock_change_latency_us = 23.0, + .return_bus_width_bytes = 64, + .dispclk_dppclk_vco_speed_mhz = 3850, + .xfc_bus_transport_time_us = 20, + .xfc_xbuf_latency_tolerance_us = 50, + .use_urgent_burst_bw = 0, +}; struct _vcs_dpi_ip_params_st dcn2_1_ip = { .odm_capable = 1, diff --git a/drivers/gpu/drm/amd/display/modules/info_packet/info_packet.c b/drivers/gpu/drm/amd/display/modules/info_packet/info_packet.c index 84f9b412a4f1..738ee763f24a 100644 --- a/drivers/gpu/drm/amd/display/modules/info_packet/info_packet.c +++ b/drivers/gpu/drm/amd/display/modules/info_packet/info_packet.c @@ -147,12 +147,15 @@ void mod_build_vsc_infopacket(const struct dc_stream_state *stream, } /* VSC packet set to 4 for PSR-SU, or 2 for PSR1 */ - if (stream->link->psr_settings.psr_version == DC_PSR_VERSION_SU_1) - vsc_packet_revision = vsc_packet_rev4; - else if (stream->link->replay_settings.config.replay_supported) + if (stream->link->psr_settings.psr_feature_enabled) { + if (stream->link->psr_settings.psr_version == DC_PSR_VERSION_SU_1) + vsc_packet_revision = vsc_packet_rev4; + else if (stream->link->psr_settings.psr_version == DC_PSR_VERSION_1) + vsc_packet_revision = vsc_packet_rev2; + } + + if (stream->link->replay_settings.config.replay_supported) vsc_packet_revision = vsc_packet_rev4; - else if (stream->link->psr_settings.psr_version == DC_PSR_VERSION_1) - vsc_packet_revision = vsc_packet_rev2; /* Update to revision 5 for extended colorimetry support */ if (stream->use_vsc_sdp_for_colorimetry) diff --git a/drivers/gpu/drm/amd/include/kgd_pp_interface.h b/drivers/gpu/drm/amd/include/kgd_pp_interface.h index 1c5049e894e3..c2ccf3724e37 100644 --- a/drivers/gpu/drm/amd/include/kgd_pp_interface.h +++ b/drivers/gpu/drm/amd/include/kgd_pp_interface.h @@ -318,6 +318,7 @@ enum pp_xgmi_plpd_mode { #define MAX_GFX_CLKS 8 #define MAX_CLKS 4 #define NUM_VCN 4 +#define NUM_JPEG_ENG 32 struct seq_file; enum amd_pp_clock_type; @@ -774,6 +775,85 @@ struct gpu_metrics_v1_4 { uint16_t padding; }; +struct gpu_metrics_v1_5 { + struct metrics_table_header common_header; + + /* Temperature (Celsius) */ + uint16_t temperature_hotspot; + uint16_t temperature_mem; + uint16_t temperature_vrsoc; + + /* Power (Watts) */ + uint16_t curr_socket_power; + + /* Utilization (%) */ + uint16_t average_gfx_activity; + uint16_t average_umc_activity; // memory controller + uint16_t vcn_activity[NUM_VCN]; + uint16_t jpeg_activity[NUM_JPEG_ENG]; + + /* Energy (15.259uJ (2^-16) units) */ + uint64_t energy_accumulator; + + /* Driver attached timestamp (in ns) */ + uint64_t system_clock_counter; + + /* Throttle status */ + uint32_t throttle_status; + + /* Clock Lock Status. Each bit corresponds to clock instance */ + uint32_t gfxclk_lock_status; + + /* Link width (number of lanes) and speed (in 0.1 GT/s) */ + uint16_t pcie_link_width; + uint16_t pcie_link_speed; + + /* XGMI bus width and bitrate (in Gbps) */ + uint16_t xgmi_link_width; + uint16_t xgmi_link_speed; + + /* Utilization Accumulated (%) */ + uint32_t gfx_activity_acc; + uint32_t mem_activity_acc; + + /*PCIE accumulated bandwidth (GB/sec) */ + uint64_t pcie_bandwidth_acc; + + /*PCIE instantaneous bandwidth (GB/sec) */ + uint64_t pcie_bandwidth_inst; + + /* PCIE L0 to recovery state transition accumulated count */ + uint64_t pcie_l0_to_recov_count_acc; + + /* PCIE replay accumulated count */ + uint64_t pcie_replay_count_acc; + + /* PCIE replay rollover accumulated count */ + uint64_t pcie_replay_rover_count_acc; + + /* PCIE NAK sent accumulated count */ + uint32_t pcie_nak_sent_count_acc; + + /* PCIE NAK received accumulated count */ + uint32_t pcie_nak_rcvd_count_acc; + + /* XGMI accumulated data transfer size(KiloBytes) */ + uint64_t xgmi_read_data_acc[NUM_XGMI_LINKS]; + uint64_t xgmi_write_data_acc[NUM_XGMI_LINKS]; + + /* PMFW attached timestamp (10ns resolution) */ + uint64_t firmware_timestamp; + + /* Current clocks (Mhz) */ + uint16_t current_gfxclk[MAX_GFX_CLKS]; + uint16_t current_socclk[MAX_CLKS]; + uint16_t current_vclk0[MAX_CLKS]; + uint16_t current_dclk0[MAX_CLKS]; + uint16_t current_uclk; + + uint16_t padding; +}; + /* * gpu_metrics_v2_0 is not recommended as it's not naturally aligned. * Use gpu_metrics_v2_1 or later instead. diff --git a/drivers/gpu/drm/amd/pm/amdgpu_pm.c b/drivers/gpu/drm/amd/pm/amdgpu_pm.c index 49028dde0f87..20c53eefd680 100644 --- a/drivers/gpu/drm/amd/pm/amdgpu_pm.c +++ b/drivers/gpu/drm/amd/pm/amdgpu_pm.c @@ -2128,7 +2128,9 @@ static int default_attr_update(struct amdgpu_device *adev, struct amdgpu_device_ if (amdgpu_dpm_is_overdrive_supported(adev)) *states = ATTR_STATE_SUPPORTED; } else if (DEVICE_ATTR_IS(mem_busy_percent)) { - if (adev->flags & AMD_IS_APU || gc_ver == IP_VERSION(9, 0, 1)) + if ((adev->flags & AMD_IS_APU && + gc_ver != IP_VERSION(9, 4, 3)) || + gc_ver == IP_VERSION(9, 0, 1)) *states = ATTR_STATE_UNSUPPORTED; } else if (DEVICE_ATTR_IS(pcie_bw)) { /* PCIe Perf counters won't work on APU nodes */ 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 fef2d290f3f2..7b812b9994d7 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 @@ -123,7 +123,7 @@ typedef enum { VOLTAGE_GUARDBAND_COUNT } GFX_GUARDBAND_e; -#define SMU_METRICS_TABLE_VERSION 0x9 +#define SMU_METRICS_TABLE_VERSION 0xB typedef struct __attribute__((packed, aligned(4))) { uint32_t AccumulationCounter; @@ -219,7 +219,103 @@ typedef struct __attribute__((packed, aligned(4))) { uint32_t PCIenReplayARolloverCountAcc; // The Pcie counter itself is accumulated uint32_t PCIeNAKSentCountAcc; // The Pcie counter itself is accumulated uint32_t PCIeNAKReceivedCountAcc; // The Pcie counter itself is accumulated -} MetricsTable_t; + + // VCN/JPEG ACTIVITY + uint32_t VcnBusy[4]; + uint32_t JpegBusy[32]; +} MetricsTableX_t; + +typedef struct __attribute__((packed, aligned(4))) { + uint32_t AccumulationCounter; + + //TEMPERATURE + uint32_t MaxSocketTemperature; + uint32_t MaxVrTemperature; + uint32_t MaxHbmTemperature; + uint64_t MaxSocketTemperatureAcc; + uint64_t MaxVrTemperatureAcc; + uint64_t MaxHbmTemperatureAcc; + + //POWER + uint32_t SocketPowerLimit; + uint32_t MaxSocketPowerLimit; + uint32_t SocketPower; + + //ENERGY + uint64_t Timestamp; + uint64_t SocketEnergyAcc; + uint64_t CcdEnergyAcc; + uint64_t XcdEnergyAcc; + uint64_t AidEnergyAcc; + uint64_t HbmEnergyAcc; + + //FREQUENCY + uint32_t CclkFrequencyLimit; + uint32_t GfxclkFrequencyLimit; + uint32_t FclkFrequency; + uint32_t UclkFrequency; + uint32_t SocclkFrequency[4]; + uint32_t VclkFrequency[4]; + uint32_t DclkFrequency[4]; + uint32_t LclkFrequency[4]; + uint64_t GfxclkFrequencyAcc[8]; + uint64_t CclkFrequencyAcc[96]; + + //FREQUENCY RANGE + uint32_t MaxCclkFrequency; + uint32_t MinCclkFrequency; + uint32_t MaxGfxclkFrequency; + uint32_t MinGfxclkFrequency; + uint32_t FclkFrequencyTable[4]; + uint32_t UclkFrequencyTable[4]; + uint32_t SocclkFrequencyTable[4]; + uint32_t VclkFrequencyTable[4]; + uint32_t DclkFrequencyTable[4]; + uint32_t LclkFrequencyTable[4]; + uint32_t MaxLclkDpmRange; + uint32_t MinLclkDpmRange; + + //XGMI + uint32_t XgmiWidth; + uint32_t XgmiBitrate; + uint64_t XgmiReadBandwidthAcc[8]; + uint64_t XgmiWriteBandwidthAcc[8]; + + //ACTIVITY + uint32_t SocketC0Residency; + uint32_t SocketGfxBusy; + uint32_t DramBandwidthUtilization; + uint64_t SocketC0ResidencyAcc; + uint64_t SocketGfxBusyAcc; + uint64_t DramBandwidthAcc; + uint32_t MaxDramBandwidth; + uint64_t DramBandwidthUtilizationAcc; + uint64_t PcieBandwidthAcc[4]; + + //THROTTLERS + uint32_t ProchotResidencyAcc; + uint32_t PptResidencyAcc; + uint32_t SocketThmResidencyAcc; + uint32_t VrThmResidencyAcc; + uint32_t HbmThmResidencyAcc; + uint32_t GfxLockXCDMak; + + // 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]; + + //XGMI Data tranfser size + uint64_t XgmiReadDataSizeAcc[8];//in KByte + uint64_t XgmiWriteDataSizeAcc[8];//in KByte + + // VCN/JPEG ACTIVITY + uint32_t VcnBusy[4]; + uint32_t JpegBusy[32]; +} MetricsTableA_t; #define SMU_VF_METRICS_TABLE_VERSION 0x3 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 900a2d9e6d85..b64e07b75937 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 @@ -245,6 +245,8 @@ struct PPTable_t { #define SMUQ10_TO_UINT(x) ((x) >> 10) #define SMUQ10_FRAC(x) ((x) & 0x3ff) #define SMUQ10_ROUND(x) ((SMUQ10_TO_UINT(x)) + ((SMUQ10_FRAC(x)) >= 0x200)) +#define GET_METRIC_FIELD(field) ((adev->flags & AMD_IS_APU) ?\ + (metrics_a->field) : (metrics_x->field)) struct smu_v13_0_6_dpm_map { enum smu_clk_type clk_type; @@ -327,7 +329,8 @@ static int smu_v13_0_6_tables_init(struct smu_context *smu) SMU_TABLE_INIT(tables, SMU_TABLE_PMSTATUSLOG, SMU13_TOOL_SIZE, PAGE_SIZE, AMDGPU_GEM_DOMAIN_VRAM); - SMU_TABLE_INIT(tables, SMU_TABLE_SMU_METRICS, sizeof(MetricsTable_t), + SMU_TABLE_INIT(tables, SMU_TABLE_SMU_METRICS, + max(sizeof(MetricsTableX_t), sizeof(MetricsTableA_t)), PAGE_SIZE, AMDGPU_GEM_DOMAIN_VRAM | AMDGPU_GEM_DOMAIN_GTT); @@ -335,12 +338,13 @@ static int smu_v13_0_6_tables_init(struct smu_context *smu) PAGE_SIZE, AMDGPU_GEM_DOMAIN_VRAM | AMDGPU_GEM_DOMAIN_GTT); - smu_table->metrics_table = kzalloc(sizeof(MetricsTable_t), GFP_KERNEL); + smu_table->metrics_table = kzalloc(max(sizeof(MetricsTableX_t), + sizeof(MetricsTableA_t)), GFP_KERNEL); if (!smu_table->metrics_table) return -ENOMEM; smu_table->metrics_time = 0; - smu_table->gpu_metrics_table_size = sizeof(struct gpu_metrics_v1_4); + smu_table->gpu_metrics_table_size = sizeof(struct gpu_metrics_v1_5); smu_table->gpu_metrics_table = kzalloc(smu_table->gpu_metrics_table_size, GFP_KERNEL); if (!smu_table->gpu_metrics_table) { @@ -431,9 +435,11 @@ static int smu_v13_0_6_get_metrics_table(struct smu_context *smu, static int smu_v13_0_6_setup_driver_pptable(struct smu_context *smu) { struct smu_table_context *smu_table = &smu->smu_table; - MetricsTable_t *metrics = (MetricsTable_t *)smu_table->metrics_table; + MetricsTableX_t *metrics_x = (MetricsTableX_t *)smu_table->metrics_table; + MetricsTableA_t *metrics_a = (MetricsTableA_t *)smu_table->metrics_table; struct PPTable_t *pptable = (struct PPTable_t *)smu_table->driver_pptable; + struct amdgpu_device *adev = smu->adev; int ret, i, retry = 100; /* Store one-time values in driver PPTable */ @@ -444,7 +450,7 @@ static int smu_v13_0_6_setup_driver_pptable(struct smu_context *smu) return ret; /* Ensure that metrics have been updated */ - if (metrics->AccumulationCounter) + if (GET_METRIC_FIELD(AccumulationCounter)) break; usleep_range(1000, 1100); @@ -454,29 +460,29 @@ static int smu_v13_0_6_setup_driver_pptable(struct smu_context *smu) return -ETIME; pptable->MaxSocketPowerLimit = - SMUQ10_ROUND(metrics->MaxSocketPowerLimit); + SMUQ10_ROUND(GET_METRIC_FIELD(MaxSocketPowerLimit)); pptable->MaxGfxclkFrequency = - SMUQ10_ROUND(metrics->MaxGfxclkFrequency); + SMUQ10_ROUND(GET_METRIC_FIELD(MaxGfxclkFrequency)); pptable->MinGfxclkFrequency = - SMUQ10_ROUND(metrics->MinGfxclkFrequency); + SMUQ10_ROUND(GET_METRIC_FIELD(MinGfxclkFrequency)); for (i = 0; i < 4; ++i) { pptable->FclkFrequencyTable[i] = - SMUQ10_ROUND(metrics->FclkFrequencyTable[i]); + SMUQ10_ROUND(GET_METRIC_FIELD(FclkFrequencyTable)[i]); pptable->UclkFrequencyTable[i] = - SMUQ10_ROUND(metrics->UclkFrequencyTable[i]); + SMUQ10_ROUND(GET_METRIC_FIELD(UclkFrequencyTable)[i]); pptable->SocclkFrequencyTable[i] = SMUQ10_ROUND( - metrics->SocclkFrequencyTable[i]); + GET_METRIC_FIELD(SocclkFrequencyTable)[i]); pptable->VclkFrequencyTable[i] = - SMUQ10_ROUND(metrics->VclkFrequencyTable[i]); + SMUQ10_ROUND(GET_METRIC_FIELD(VclkFrequencyTable)[i]); pptable->DclkFrequencyTable[i] = - SMUQ10_ROUND(metrics->DclkFrequencyTable[i]); + SMUQ10_ROUND(GET_METRIC_FIELD(DclkFrequencyTable)[i]); pptable->LclkFrequencyTable[i] = - SMUQ10_ROUND(metrics->LclkFrequencyTable[i]); + SMUQ10_ROUND(GET_METRIC_FIELD(LclkFrequencyTable)[i]); } /* use AID0 serial number by default */ - pptable->PublicSerialNumber_AID = metrics->PublicSerialNumber_AID[0]; + pptable->PublicSerialNumber_AID = GET_METRIC_FIELD(PublicSerialNumber_AID)[0]; pptable->Init = true; } @@ -778,7 +784,8 @@ static int smu_v13_0_6_get_smu_metrics_data(struct smu_context *smu, uint32_t *value) { struct smu_table_context *smu_table = &smu->smu_table; - MetricsTable_t *metrics = (MetricsTable_t *)smu_table->metrics_table; + MetricsTableX_t *metrics_x = (MetricsTableX_t *)smu_table->metrics_table; + MetricsTableA_t *metrics_a = (MetricsTableA_t *)smu_table->metrics_table; struct amdgpu_device *adev = smu->adev; int ret = 0; int xcc_id; @@ -793,50 +800,50 @@ static int smu_v13_0_6_get_smu_metrics_data(struct smu_context *smu, case METRICS_AVERAGE_GFXCLK: if (smu->smc_fw_version >= 0x552F00) { xcc_id = GET_INST(GC, 0); - *value = SMUQ10_ROUND(metrics->GfxclkFrequency[xcc_id]); + *value = SMUQ10_ROUND(GET_METRIC_FIELD(GfxclkFrequency)[xcc_id]); } else { *value = 0; } break; case METRICS_CURR_SOCCLK: case METRICS_AVERAGE_SOCCLK: - *value = SMUQ10_ROUND(metrics->SocclkFrequency[0]); + *value = SMUQ10_ROUND(GET_METRIC_FIELD(SocclkFrequency)[0]); break; case METRICS_CURR_UCLK: case METRICS_AVERAGE_UCLK: - *value = SMUQ10_ROUND(metrics->UclkFrequency); + *value = SMUQ10_ROUND(GET_METRIC_FIELD(UclkFrequency)); break; case METRICS_CURR_VCLK: - *value = SMUQ10_ROUND(metrics->VclkFrequency[0]); + *value = SMUQ10_ROUND(GET_METRIC_FIELD(VclkFrequency)[0]); break; case METRICS_CURR_DCLK: - *value = SMUQ10_ROUND(metrics->DclkFrequency[0]); + *value = SMUQ10_ROUND(GET_METRIC_FIELD(DclkFrequency)[0]); break; case METRICS_CURR_FCLK: - *value = SMUQ10_ROUND(metrics->FclkFrequency); + *value = SMUQ10_ROUND(GET_METRIC_FIELD(FclkFrequency)); break; case METRICS_AVERAGE_GFXACTIVITY: - *value = SMUQ10_ROUND(metrics->SocketGfxBusy); + *value = SMUQ10_ROUND(GET_METRIC_FIELD(SocketGfxBusy)); break; case METRICS_AVERAGE_MEMACTIVITY: - *value = SMUQ10_ROUND(metrics->DramBandwidthUtilization); + *value = SMUQ10_ROUND(GET_METRIC_FIELD(DramBandwidthUtilization)); break; case METRICS_CURR_SOCKETPOWER: - *value = SMUQ10_ROUND(metrics->SocketPower) << 8; + *value = SMUQ10_ROUND(GET_METRIC_FIELD(SocketPower)) << 8; break; case METRICS_TEMPERATURE_HOTSPOT: - *value = SMUQ10_ROUND(metrics->MaxSocketTemperature) * + *value = SMUQ10_ROUND(GET_METRIC_FIELD(MaxSocketTemperature)) * SMU_TEMPERATURE_UNITS_PER_CENTIGRADES; break; case METRICS_TEMPERATURE_MEM: - *value = SMUQ10_ROUND(metrics->MaxHbmTemperature) * + *value = SMUQ10_ROUND(GET_METRIC_FIELD(MaxHbmTemperature)) * SMU_TEMPERATURE_UNITS_PER_CENTIGRADES; break; /* This is the max of all VRs and not just SOC VR. * No need to define another data type for the same. */ case METRICS_TEMPERATURE_VRSOC: - *value = SMUQ10_ROUND(metrics->MaxVrTemperature) * + *value = SMUQ10_ROUND(GET_METRIC_FIELD(MaxVrTemperature)) * SMU_TEMPERATURE_UNITS_PER_CENTIGRADES; break; default: @@ -2022,67 +2029,70 @@ static int smu_v13_0_6_get_current_pcie_link_speed(struct smu_context *smu) 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_4 *gpu_metrics = - (struct gpu_metrics_v1_4 *)smu_table->gpu_metrics_table; + struct gpu_metrics_v1_5 *gpu_metrics = + (struct gpu_metrics_v1_5 *)smu_table->gpu_metrics_table; struct amdgpu_device *adev = smu->adev; - int ret = 0, xcc_id, inst, i; - MetricsTable_t *metrics; + int ret = 0, xcc_id, inst, i, j; + MetricsTableX_t *metrics_x; + MetricsTableA_t *metrics_a; u16 link_width_level; - metrics = kzalloc(sizeof(MetricsTable_t), GFP_KERNEL); - ret = smu_v13_0_6_get_metrics_table(smu, metrics, true); + metrics_x = kzalloc(max(sizeof(MetricsTableX_t), sizeof(MetricsTableA_t)), GFP_KERNEL); + ret = smu_v13_0_6_get_metrics_table(smu, metrics_x, true); if (ret) { - kfree(metrics); + kfree(metrics_x); return ret; } - smu_cmn_init_soft_gpu_metrics(gpu_metrics, 1, 4); + metrics_a = (MetricsTableA_t *)metrics_x; + + smu_cmn_init_soft_gpu_metrics(gpu_metrics, 1, 5); gpu_metrics->temperature_hotspot = - SMUQ10_ROUND(metrics->MaxSocketTemperature); + SMUQ10_ROUND(GET_METRIC_FIELD(MaxSocketTemperature)); /* Individual HBM stack temperature is not reported */ gpu_metrics->temperature_mem = - SMUQ10_ROUND(metrics->MaxHbmTemperature); + SMUQ10_ROUND(GET_METRIC_FIELD(MaxHbmTemperature)); /* Reports max temperature of all voltage rails */ gpu_metrics->temperature_vrsoc = - SMUQ10_ROUND(metrics->MaxVrTemperature); + SMUQ10_ROUND(GET_METRIC_FIELD(MaxVrTemperature)); gpu_metrics->average_gfx_activity = - SMUQ10_ROUND(metrics->SocketGfxBusy); + SMUQ10_ROUND(GET_METRIC_FIELD(SocketGfxBusy)); gpu_metrics->average_umc_activity = - SMUQ10_ROUND(metrics->DramBandwidthUtilization); + SMUQ10_ROUND(GET_METRIC_FIELD(DramBandwidthUtilization)); gpu_metrics->curr_socket_power = - SMUQ10_ROUND(metrics->SocketPower); + SMUQ10_ROUND(GET_METRIC_FIELD(SocketPower)); /* Energy counter reported in 15.259uJ (2^-16) units */ - gpu_metrics->energy_accumulator = metrics->SocketEnergyAcc; + gpu_metrics->energy_accumulator = GET_METRIC_FIELD(SocketEnergyAcc); for (i = 0; i < MAX_GFX_CLKS; i++) { xcc_id = GET_INST(GC, i); if (xcc_id >= 0) gpu_metrics->current_gfxclk[i] = - SMUQ10_ROUND(metrics->GfxclkFrequency[xcc_id]); + SMUQ10_ROUND(GET_METRIC_FIELD(GfxclkFrequency)[xcc_id]); if (i < MAX_CLKS) { gpu_metrics->current_socclk[i] = - SMUQ10_ROUND(metrics->SocclkFrequency[i]); + SMUQ10_ROUND(GET_METRIC_FIELD(SocclkFrequency)[i]); inst = GET_INST(VCN, i); if (inst >= 0) { gpu_metrics->current_vclk0[i] = - SMUQ10_ROUND(metrics->VclkFrequency[inst]); + SMUQ10_ROUND(GET_METRIC_FIELD(VclkFrequency)[inst]); gpu_metrics->current_dclk0[i] = - SMUQ10_ROUND(metrics->DclkFrequency[inst]); + SMUQ10_ROUND(GET_METRIC_FIELD(DclkFrequency)[inst]); } } } - gpu_metrics->current_uclk = SMUQ10_ROUND(metrics->UclkFrequency); + gpu_metrics->current_uclk = SMUQ10_ROUND(GET_METRIC_FIELD(UclkFrequency)); /* Throttle status is not reported through metrics now */ gpu_metrics->throttle_status = 0; /* Clock Lock Status. Each bit corresponds to each GFXCLK instance */ - gpu_metrics->gfxclk_lock_status = metrics->GfxLockXCDMak >> GET_INST(GC, 0); + gpu_metrics->gfxclk_lock_status = GET_METRIC_FIELD(GfxLockXCDMak) >> GET_INST(GC, 0); if (!(adev->flags & AMD_IS_APU)) { link_width_level = smu_v13_0_6_get_current_pcie_link_width_level(smu); @@ -2094,38 +2104,57 @@ static ssize_t smu_v13_0_6_get_gpu_metrics(struct smu_context *smu, void **table gpu_metrics->pcie_link_speed = smu_v13_0_6_get_current_pcie_link_speed(smu); gpu_metrics->pcie_bandwidth_acc = - SMUQ10_ROUND(metrics->PcieBandwidthAcc[0]); + SMUQ10_ROUND(metrics_x->PcieBandwidthAcc[0]); gpu_metrics->pcie_bandwidth_inst = - SMUQ10_ROUND(metrics->PcieBandwidth[0]); + SMUQ10_ROUND(metrics_x->PcieBandwidth[0]); gpu_metrics->pcie_l0_to_recov_count_acc = - metrics->PCIeL0ToRecoveryCountAcc; + metrics_x->PCIeL0ToRecoveryCountAcc; gpu_metrics->pcie_replay_count_acc = - metrics->PCIenReplayAAcc; + metrics_x->PCIenReplayAAcc; gpu_metrics->pcie_replay_rover_count_acc = - metrics->PCIenReplayARolloverCountAcc; + metrics_x->PCIenReplayARolloverCountAcc; + gpu_metrics->pcie_nak_sent_count_acc = + metrics_x->PCIeNAKSentCountAcc; + gpu_metrics->pcie_nak_rcvd_count_acc = + metrics_x->PCIeNAKReceivedCountAcc; } gpu_metrics->system_clock_counter = ktime_get_boottime_ns(); gpu_metrics->gfx_activity_acc = - SMUQ10_ROUND(metrics->SocketGfxBusyAcc); + SMUQ10_ROUND(GET_METRIC_FIELD(SocketGfxBusyAcc)); gpu_metrics->mem_activity_acc = - SMUQ10_ROUND(metrics->DramBandwidthUtilizationAcc); + SMUQ10_ROUND(GET_METRIC_FIELD(DramBandwidthUtilizationAcc)); for (i = 0; i < NUM_XGMI_LINKS; i++) { gpu_metrics->xgmi_read_data_acc[i] = - SMUQ10_ROUND(metrics->XgmiReadDataSizeAcc[i]); + SMUQ10_ROUND(GET_METRIC_FIELD(XgmiReadDataSizeAcc)[i]); gpu_metrics->xgmi_write_data_acc[i] = - SMUQ10_ROUND(metrics->XgmiWriteDataSizeAcc[i]); + SMUQ10_ROUND(GET_METRIC_FIELD(XgmiWriteDataSizeAcc)[i]); + } + + for (i = 0; i < adev->jpeg.num_jpeg_inst; ++i) { + inst = GET_INST(JPEG, i); + for (j = 0; j < adev->jpeg.num_jpeg_rings; ++j) { + gpu_metrics->jpeg_activity[(i * adev->jpeg.num_jpeg_rings) + j] = + SMUQ10_ROUND(GET_METRIC_FIELD(JpegBusy) + [(inst * adev->jpeg.num_jpeg_rings) + j]); + } + } + + for (i = 0; i < adev->vcn.num_vcn_inst; ++i) { + inst = GET_INST(VCN, i); + gpu_metrics->vcn_activity[i] = + SMUQ10_ROUND(GET_METRIC_FIELD(VcnBusy)[inst]); } - gpu_metrics->xgmi_link_width = SMUQ10_ROUND(metrics->XgmiWidth); - gpu_metrics->xgmi_link_speed = SMUQ10_ROUND(metrics->XgmiBitrate); + gpu_metrics->xgmi_link_width = SMUQ10_ROUND(GET_METRIC_FIELD(XgmiWidth)); + gpu_metrics->xgmi_link_speed = SMUQ10_ROUND(GET_METRIC_FIELD(XgmiBitrate)); - gpu_metrics->firmware_timestamp = metrics->Timestamp; + gpu_metrics->firmware_timestamp = GET_METRIC_FIELD(Timestamp); *table = (void *)gpu_metrics; - kfree(metrics); + kfree(metrics_x); return sizeof(*gpu_metrics); } diff --git a/drivers/gpu/drm/amd/pm/swsmu/smu_cmn.c b/drivers/gpu/drm/amd/pm/swsmu/smu_cmn.c index 001a5cf09657..00cd615bbcdc 100644 --- a/drivers/gpu/drm/amd/pm/swsmu/smu_cmn.c +++ b/drivers/gpu/drm/amd/pm/swsmu/smu_cmn.c @@ -989,6 +989,9 @@ void smu_cmn_init_soft_gpu_metrics(void *table, uint8_t frev, uint8_t crev) case METRICS_VERSION(1, 4): structure_size = sizeof(struct gpu_metrics_v1_4); break; + case METRICS_VERSION(1, 5): + structure_size = sizeof(struct gpu_metrics_v1_5); + break; case METRICS_VERSION(2, 0): structure_size = sizeof(struct gpu_metrics_v2_0); break; |