From a29997b7ac1f5c816b543e0c56aa2b5b56baac24 Mon Sep 17 00:00:00 2001 From: Dillon Varone Date: Wed, 13 Nov 2024 16:44:15 -0500 Subject: [PATCH 01/16] drm/amd/display: Limit VTotal range to max hw cap minus fp [WHY & HOW] Hardware does not support the VTotal to be between fp2 lines of the maximum possible VTotal, so add a capability flag to track it and apply where necessary. Cc: Mario Limonciello Cc: Alex Deucher Cc: stable@vger.kernel.org Reviewed-by: Jun Lei Reviewed-by: Anthony Koo Signed-off-by: Dillon Varone Signed-off-by: Alex Hung Tested-by: Daniel Wheeler Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/display/dc/dc.h | 1 + .../dc/dml2/dml21/dml21_translation_helper.c | 27 +++++++++++++++++-- .../dc/resource/dcn30/dcn30_resource.c | 1 + .../dc/resource/dcn302/dcn302_resource.c | 1 + .../dc/resource/dcn303/dcn303_resource.c | 1 + .../dc/resource/dcn32/dcn32_resource.c | 1 + .../dc/resource/dcn321/dcn321_resource.c | 1 + .../dc/resource/dcn35/dcn35_resource.c | 1 + .../dc/resource/dcn351/dcn351_resource.c | 1 + .../dc/resource/dcn401/dcn401_resource.c | 1 + .../amd/display/modules/freesync/freesync.c | 13 ++++++++- 11 files changed, 46 insertions(+), 3 deletions(-) diff --git a/drivers/gpu/drm/amd/display/dc/dc.h b/drivers/gpu/drm/amd/display/dc/dc.h index 49a31598bb88..1dda5fc2e212 100644 --- a/drivers/gpu/drm/amd/display/dc/dc.h +++ b/drivers/gpu/drm/amd/display/dc/dc.h @@ -290,6 +290,7 @@ struct dc_caps { uint16_t subvp_vertical_int_margin_us; bool seamless_odm; uint32_t max_v_total; + bool vtotal_limited_by_fp2; uint32_t max_disp_clock_khz_at_vmin; uint8_t subvp_drr_vblank_start_margin_us; bool cursor_not_scaled; diff --git a/drivers/gpu/drm/amd/display/dc/dml2/dml21/dml21_translation_helper.c b/drivers/gpu/drm/amd/display/dc/dml2/dml21/dml21_translation_helper.c index 138b4b1e42ed..f66493528f42 100644 --- a/drivers/gpu/drm/amd/display/dc/dml2/dml21/dml21_translation_helper.c +++ b/drivers/gpu/drm/amd/display/dc/dml2/dml21/dml21_translation_helper.c @@ -339,11 +339,22 @@ void dml21_apply_soc_bb_overrides(struct dml2_initialize_instance_in_out *dml_in // } } +static unsigned int calc_max_hardware_v_total(const struct dc_stream_state *stream) +{ + unsigned int max_hw_v_total = stream->ctx->dc->caps.max_v_total; + + if (stream->ctx->dc->caps.vtotal_limited_by_fp2) { + max_hw_v_total -= stream->timing.v_front_porch + 1; + } + + return max_hw_v_total; +} + static void populate_dml21_timing_config_from_stream_state(struct dml2_timing_cfg *timing, struct dc_stream_state *stream, struct dml2_context *dml_ctx) { - unsigned int hblank_start, vblank_start; + unsigned int hblank_start, vblank_start, min_hardware_refresh_in_uhz; timing->h_active = stream->timing.h_addressable + stream->timing.h_border_left + stream->timing.h_border_right; timing->v_active = stream->timing.v_addressable + stream->timing.v_border_bottom + stream->timing.v_border_top; @@ -371,11 +382,23 @@ static void populate_dml21_timing_config_from_stream_state(struct dml2_timing_cf - stream->timing.v_border_top - stream->timing.v_border_bottom; timing->drr_config.enabled = stream->ignore_msa_timing_param; - timing->drr_config.min_refresh_uhz = stream->timing.min_refresh_in_uhz; timing->drr_config.drr_active_variable = stream->vrr_active_variable; timing->drr_config.drr_active_fixed = stream->vrr_active_fixed; timing->drr_config.disallowed = !stream->allow_freesync; + /* limit min refresh rate to DC cap */ + min_hardware_refresh_in_uhz = stream->timing.min_refresh_in_uhz; + if (stream->ctx->dc->caps.max_v_total != 0) { + min_hardware_refresh_in_uhz = div64_u64((stream->timing.pix_clk_100hz * 100000000ULL), + (stream->timing.h_total * (long long)calc_max_hardware_v_total(stream))); + } + + if (stream->timing.min_refresh_in_uhz > min_hardware_refresh_in_uhz) { + timing->drr_config.min_refresh_uhz = stream->timing.min_refresh_in_uhz; + } else { + timing->drr_config.min_refresh_uhz = min_hardware_refresh_in_uhz; + } + if (dml_ctx->config.callbacks.get_max_flickerless_instant_vtotal_increase && stream->ctx->dc->config.enable_fpo_flicker_detection == 1) timing->drr_config.max_instant_vtotal_delta = dml_ctx->config.callbacks.get_max_flickerless_instant_vtotal_increase(stream, false); diff --git a/drivers/gpu/drm/amd/display/dc/resource/dcn30/dcn30_resource.c b/drivers/gpu/drm/amd/display/dc/resource/dcn30/dcn30_resource.c index cd31e4f16c14..bfd0eccbed28 100644 --- a/drivers/gpu/drm/amd/display/dc/resource/dcn30/dcn30_resource.c +++ b/drivers/gpu/drm/amd/display/dc/resource/dcn30/dcn30_resource.c @@ -2353,6 +2353,7 @@ static bool dcn30_resource_construct( dc->caps.dp_hdmi21_pcon_support = true; dc->caps.max_v_total = (1 << 15) - 1; + dc->caps.vtotal_limited_by_fp2 = true; /* read VBIOS LTTPR caps */ { diff --git a/drivers/gpu/drm/amd/display/dc/resource/dcn302/dcn302_resource.c b/drivers/gpu/drm/amd/display/dc/resource/dcn302/dcn302_resource.c index 02af8b8f4d27..7baefc910a3d 100644 --- a/drivers/gpu/drm/amd/display/dc/resource/dcn302/dcn302_resource.c +++ b/drivers/gpu/drm/amd/display/dc/resource/dcn302/dcn302_resource.c @@ -1233,6 +1233,7 @@ static bool dcn302_resource_construct( dc->caps.extended_aux_timeout_support = true; dc->caps.dmcub_support = true; dc->caps.max_v_total = (1 << 15) - 1; + dc->caps.vtotal_limited_by_fp2 = true; /* Color pipeline capabilities */ dc->caps.color.dpp.dcn_arch = 1; diff --git a/drivers/gpu/drm/amd/display/dc/resource/dcn303/dcn303_resource.c b/drivers/gpu/drm/amd/display/dc/resource/dcn303/dcn303_resource.c index 7002a8dd358a..8a57d46ad15f 100644 --- a/drivers/gpu/drm/amd/display/dc/resource/dcn303/dcn303_resource.c +++ b/drivers/gpu/drm/amd/display/dc/resource/dcn303/dcn303_resource.c @@ -1178,6 +1178,7 @@ static bool dcn303_resource_construct( dc->caps.extended_aux_timeout_support = true; dc->caps.dmcub_support = true; dc->caps.max_v_total = (1 << 15) - 1; + dc->caps.vtotal_limited_by_fp2 = true; /* Color pipeline capabilities */ dc->caps.color.dpp.dcn_arch = 1; diff --git a/drivers/gpu/drm/amd/display/dc/resource/dcn32/dcn32_resource.c b/drivers/gpu/drm/amd/display/dc/resource/dcn32/dcn32_resource.c index 01d1a11d5545..984d23bcbc27 100644 --- a/drivers/gpu/drm/amd/display/dc/resource/dcn32/dcn32_resource.c +++ b/drivers/gpu/drm/amd/display/dc/resource/dcn32/dcn32_resource.c @@ -2189,6 +2189,7 @@ static bool dcn32_resource_construct( dc->caps.dmcub_support = true; dc->caps.seamless_odm = true; dc->caps.max_v_total = (1 << 15) - 1; + dc->caps.vtotal_limited_by_fp2 = true; /* Color pipeline capabilities */ dc->caps.color.dpp.dcn_arch = 1; diff --git a/drivers/gpu/drm/amd/display/dc/resource/dcn321/dcn321_resource.c b/drivers/gpu/drm/amd/display/dc/resource/dcn321/dcn321_resource.c index 5cb74fd9cb7d..06b9479c8bd3 100644 --- a/drivers/gpu/drm/amd/display/dc/resource/dcn321/dcn321_resource.c +++ b/drivers/gpu/drm/amd/display/dc/resource/dcn321/dcn321_resource.c @@ -1742,6 +1742,7 @@ static bool dcn321_resource_construct( dc->caps.extended_aux_timeout_support = true; dc->caps.dmcub_support = true; dc->caps.max_v_total = (1 << 15) - 1; + dc->caps.vtotal_limited_by_fp2 = true; /* Color pipeline capabilities */ dc->caps.color.dpp.dcn_arch = 1; diff --git a/drivers/gpu/drm/amd/display/dc/resource/dcn35/dcn35_resource.c b/drivers/gpu/drm/amd/display/dc/resource/dcn35/dcn35_resource.c index 09a5a9474903..89e2adcf2a28 100644 --- a/drivers/gpu/drm/amd/display/dc/resource/dcn35/dcn35_resource.c +++ b/drivers/gpu/drm/amd/display/dc/resource/dcn35/dcn35_resource.c @@ -1850,6 +1850,7 @@ static bool dcn35_resource_construct( dc->caps.zstate_support = true; dc->caps.ips_support = true; dc->caps.max_v_total = (1 << 15) - 1; + dc->caps.vtotal_limited_by_fp2 = true; /* Color pipeline capabilities */ dc->caps.color.dpp.dcn_arch = 1; diff --git a/drivers/gpu/drm/amd/display/dc/resource/dcn351/dcn351_resource.c b/drivers/gpu/drm/amd/display/dc/resource/dcn351/dcn351_resource.c index fe382f9b6ff2..263a37c1cd3a 100644 --- a/drivers/gpu/drm/amd/display/dc/resource/dcn351/dcn351_resource.c +++ b/drivers/gpu/drm/amd/display/dc/resource/dcn351/dcn351_resource.c @@ -1829,6 +1829,7 @@ static bool dcn351_resource_construct( dc->caps.zstate_support = true; dc->caps.ips_support = true; dc->caps.max_v_total = (1 << 15) - 1; + dc->caps.vtotal_limited_by_fp2 = true; /* Color pipeline capabilities */ dc->caps.color.dpp.dcn_arch = 1; diff --git a/drivers/gpu/drm/amd/display/dc/resource/dcn401/dcn401_resource.c b/drivers/gpu/drm/amd/display/dc/resource/dcn401/dcn401_resource.c index db93bac247c0..2a3dabfe3cea 100644 --- a/drivers/gpu/drm/amd/display/dc/resource/dcn401/dcn401_resource.c +++ b/drivers/gpu/drm/amd/display/dc/resource/dcn401/dcn401_resource.c @@ -1864,6 +1864,7 @@ static bool dcn401_resource_construct( dc->caps.extended_aux_timeout_support = true; dc->caps.dmcub_support = true; dc->caps.max_v_total = (1 << 15) - 1; + dc->caps.vtotal_limited_by_fp2 = true; if (ASICREV_IS_GC_12_0_1_A0(dc->ctx->asic_id.hw_internal_rev)) dc->caps.dcc_plane_width_limit = 7680; diff --git a/drivers/gpu/drm/amd/display/modules/freesync/freesync.c b/drivers/gpu/drm/amd/display/modules/freesync/freesync.c index f980a84dceef..2b3964529539 100644 --- a/drivers/gpu/drm/amd/display/modules/freesync/freesync.c +++ b/drivers/gpu/drm/amd/display/modules/freesync/freesync.c @@ -122,6 +122,17 @@ static unsigned int calc_duration_in_us_from_v_total( return duration_in_us; } +static unsigned int calc_max_hardware_v_total(const struct dc_stream_state *stream) +{ + unsigned int max_hw_v_total = stream->ctx->dc->caps.max_v_total; + + if (stream->ctx->dc->caps.vtotal_limited_by_fp2) { + max_hw_v_total -= stream->timing.v_front_porch + 1; + } + + return max_hw_v_total; +} + unsigned int mod_freesync_calc_v_total_from_refresh( const struct dc_stream_state *stream, unsigned int refresh_in_uhz) @@ -1016,7 +1027,7 @@ void mod_freesync_build_vrr_params(struct mod_freesync *mod_freesync, if (stream->ctx->dc->caps.max_v_total != 0 && stream->timing.h_total != 0) { min_hardware_refresh_in_uhz = div64_u64((stream->timing.pix_clk_100hz * 100000000ULL), - (stream->timing.h_total * (long long)stream->ctx->dc->caps.max_v_total)); + (stream->timing.h_total * (long long)calc_max_hardware_v_total(stream))); } /* Limit minimum refresh rate to what can be supported by hardware */ min_refresh_in_uhz = min_hardware_refresh_in_uhz > in_config->min_refresh_in_uhz ? -- 2.50.1 From 0c0a19430bfdfedab437e77b9262e8e62ced384e Mon Sep 17 00:00:00 2001 From: Chris Park Date: Fri, 15 Nov 2024 15:44:52 -0500 Subject: [PATCH 02/16] drm/amd/display: Add hblank borrowing support [WHY] Some DSC timing failed at bandwidth validation due to hactive can't be evenly divided on each ODM segment. [HOW] Borrow from hblank to increase hactive to support these timing. Cc: Mario Limonciello Cc: Alex Deucher Cc: stable@vger.kernel.org Reviewed-by: Wenjing Liu Signed-off-by: Chris Park Signed-off-by: Alex Hung Tested-by: Daniel Wheeler Signed-off-by: Alex Deucher --- .../gpu/drm/amd/display/dc/core/dc_resource.c | 42 ++++++++++++++++++- drivers/gpu/drm/amd/display/dc/dc.h | 1 + .../gpu/drm/amd/display/dc/dc_spl_translate.c | 2 +- .../dc/dml2/dml21/dml21_translation_helper.c | 21 +++++++++- .../amd/display/dc/hwss/dcn32/dcn32_hwseq.c | 3 +- .../amd/display/dc/hwss/dcn401/dcn401_hwseq.c | 7 +++- .../gpu/drm/amd/display/dc/inc/core_types.h | 2 + .../gpu/drm/amd/display/dc/link/link_dpms.c | 3 +- .../dc/resource/dcn32/dcn32_resource.c | 1 + 9 files changed, 75 insertions(+), 7 deletions(-) diff --git a/drivers/gpu/drm/amd/display/dc/core/dc_resource.c b/drivers/gpu/drm/amd/display/dc/core/dc_resource.c index 619fad17de55..626f75b6ad00 100644 --- a/drivers/gpu/drm/amd/display/dc/core/dc_resource.c +++ b/drivers/gpu/drm/amd/display/dc/core/dc_resource.c @@ -2094,7 +2094,8 @@ int resource_get_odm_slice_dst_width(struct pipe_ctx *otg_master, count = resource_get_odm_slice_count(otg_master); h_active = timing->h_addressable + timing->h_border_left + - timing->h_border_right; + timing->h_border_right + + otg_master->hblank_borrow; width = h_active / count; if (otg_master->stream_res.tg) @@ -4026,6 +4027,41 @@ fail: return res; } +/** + * decide_hblank_borrow - Decides the horizontal blanking borrow value for a given pipe context. + * @pipe_ctx: Pointer to the pipe context structure. + * + * This function calculates the horizontal blanking borrow value for a given pipe context based on the + * display stream compression (DSC) configuration. If the horizontal active pixels (hactive) are less + * than the total width of the DSC slices, it sets the hblank_borrow value to the difference. If the + * total horizontal timing minus the hblank_borrow value is less than 32, it resets the hblank_borrow + * value to 0. + */ +static void decide_hblank_borrow(struct pipe_ctx *pipe_ctx) +{ + uint32_t hactive; + uint32_t ceil_slice_width; + struct dc_stream_state *stream = NULL; + + if (!pipe_ctx) + return; + + stream = pipe_ctx->stream; + + if (stream->timing.flags.DSC) { + hactive = stream->timing.h_addressable + stream->timing.h_border_left + stream->timing.h_border_right; + + /* Assume if determined slices does not divide Hactive evenly, Hborrow is needed for padding*/ + if (hactive % stream->timing.dsc_cfg.num_slices_h != 0) { + ceil_slice_width = (hactive / stream->timing.dsc_cfg.num_slices_h) + 1; + pipe_ctx->hblank_borrow = ceil_slice_width * stream->timing.dsc_cfg.num_slices_h - hactive; + + if (stream->timing.h_total - hactive - pipe_ctx->hblank_borrow < 32) + pipe_ctx->hblank_borrow = 0; + } + } +} + /** * dc_validate_global_state() - Determine if hardware can support a given state * @@ -4064,6 +4100,10 @@ enum dc_status dc_validate_global_state( if (pipe_ctx->stream != stream) continue; + /* Decide whether hblank borrow is needed and save it in pipe_ctx */ + if (dc->debug.enable_hblank_borrow) + decide_hblank_borrow(pipe_ctx); + if (dc->res_pool->funcs->patch_unknown_plane_state && pipe_ctx->plane_state && pipe_ctx->plane_state->tiling_info.gfx9.swizzle == DC_SW_UNKNOWN) { diff --git a/drivers/gpu/drm/amd/display/dc/dc.h b/drivers/gpu/drm/amd/display/dc/dc.h index 1dda5fc2e212..e9b9126c0401 100644 --- a/drivers/gpu/drm/amd/display/dc/dc.h +++ b/drivers/gpu/drm/amd/display/dc/dc.h @@ -1069,6 +1069,7 @@ struct dc_debug_options { unsigned int scale_to_sharpness_policy; bool skip_full_updated_if_possible; unsigned int enable_oled_edp_power_up_opt; + bool enable_hblank_borrow; }; diff --git a/drivers/gpu/drm/amd/display/dc/dc_spl_translate.c b/drivers/gpu/drm/amd/display/dc/dc_spl_translate.c index c8d8e335fa37..0e310fd48b5c 100644 --- a/drivers/gpu/drm/amd/display/dc/dc_spl_translate.c +++ b/drivers/gpu/drm/amd/display/dc/dc_spl_translate.c @@ -120,7 +120,7 @@ void translate_SPL_in_params_from_pipe_ctx(struct pipe_ctx *pipe_ctx, struct spl spl_in->odm_slice_index = resource_get_odm_slice_index(pipe_ctx); // Make spl input basic out info output_size width point to stream h active spl_in->basic_out.output_size.width = - stream->timing.h_addressable + stream->timing.h_border_left + stream->timing.h_border_right; + stream->timing.h_addressable + stream->timing.h_border_left + stream->timing.h_border_right + pipe_ctx->hblank_borrow; // Make spl input basic out info output_size height point to v active spl_in->basic_out.output_size.height = stream->timing.v_addressable + stream->timing.v_border_bottom + stream->timing.v_border_top; diff --git a/drivers/gpu/drm/amd/display/dc/dml2/dml21/dml21_translation_helper.c b/drivers/gpu/drm/amd/display/dc/dml2/dml21/dml21_translation_helper.c index f66493528f42..c6a5a8614679 100644 --- a/drivers/gpu/drm/amd/display/dc/dml2/dml21/dml21_translation_helper.c +++ b/drivers/gpu/drm/amd/display/dc/dml2/dml21/dml21_translation_helper.c @@ -445,6 +445,21 @@ static void populate_dml21_timing_config_from_stream_state(struct dml2_timing_cf timing->vblank_nom = timing->v_total - timing->v_active; } +/** + * adjust_dml21_hblank_timing_config_from_pipe_ctx - Adjusts the horizontal blanking timing configuration + * based on the pipe context. + * @timing: Pointer to the dml2_timing_cfg structure to be adjusted. + * @pipe: Pointer to the pipe_ctx structure containing the horizontal blanking borrow value. + * + * This function modifies the horizontal active and blank end timings by adding and subtracting + * the horizontal blanking borrow value from the pipe context, respectively. + */ +static void adjust_dml21_hblank_timing_config_from_pipe_ctx(struct dml2_timing_cfg *timing, struct pipe_ctx *pipe) +{ + timing->h_active += pipe->hblank_borrow; + timing->h_blank_end -= pipe->hblank_borrow; +} + static void populate_dml21_output_config_from_stream_state(struct dml2_link_output_cfg *output, struct dc_stream_state *stream, const struct pipe_ctx *pipe) { @@ -732,6 +747,7 @@ static const struct scaler_data *get_scaler_data_for_plane( temp_pipe->plane_state = pipe->plane_state; temp_pipe->plane_res.scl_data.taps = pipe->plane_res.scl_data.taps; temp_pipe->stream_res = pipe->stream_res; + temp_pipe->hblank_borrow = pipe->hblank_borrow; dml_ctx->config.callbacks.build_scaling_params(temp_pipe); break; } @@ -996,6 +1012,7 @@ bool dml21_map_dc_state_into_dml_display_cfg(const struct dc *in_dc, struct dc_s ASSERT(disp_cfg_stream_location >= 0 && disp_cfg_stream_location <= __DML2_WRAPPER_MAX_STREAMS_PLANES__); populate_dml21_timing_config_from_stream_state(&dml_dispcfg->stream_descriptors[disp_cfg_stream_location].timing, context->streams[stream_index], dml_ctx); + adjust_dml21_hblank_timing_config_from_pipe_ctx(&dml_dispcfg->stream_descriptors[disp_cfg_stream_location].timing, &context->res_ctx.pipe_ctx[stream_index]); populate_dml21_output_config_from_stream_state(&dml_dispcfg->stream_descriptors[disp_cfg_stream_location].output, context->streams[stream_index], &context->res_ctx.pipe_ctx[stream_index]); populate_dml21_stream_overrides_from_stream_state(&dml_dispcfg->stream_descriptors[disp_cfg_stream_location], context->streams[stream_index]); @@ -1134,12 +1151,12 @@ void dml21_populate_pipe_ctx_dlg_params(struct dml2_context *dml_ctx, struct dc_ struct dc_crtc_timing *timing = &pipe_ctx->stream->timing; union dml2_global_sync_programming *global_sync = &stream_programming->global_sync; - hactive = timing->h_addressable + timing->h_border_left + timing->h_border_right; + hactive = timing->h_addressable + timing->h_border_left + timing->h_border_right + pipe_ctx->hblank_borrow; vactive = timing->v_addressable + timing->v_border_bottom + timing->v_border_top; hblank_start = pipe_ctx->stream->timing.h_total - pipe_ctx->stream->timing.h_front_porch; vblank_start = pipe_ctx->stream->timing.v_total - pipe_ctx->stream->timing.v_front_porch; - hblank_end = hblank_start - timing->h_addressable - timing->h_border_left - timing->h_border_right; + hblank_end = hblank_start - timing->h_addressable - timing->h_border_left - timing->h_border_right - pipe_ctx->hblank_borrow; vblank_end = vblank_start - timing->v_addressable - timing->v_border_top - timing->v_border_bottom; if (dml_ctx->config.svp_pstate.callbacks.get_pipe_subvp_type(context, pipe_ctx) == SUBVP_PHANTOM) { diff --git a/drivers/gpu/drm/amd/display/dc/hwss/dcn32/dcn32_hwseq.c b/drivers/gpu/drm/amd/display/dc/hwss/dcn32/dcn32_hwseq.c index d7f8b2dcaa6b..fa11f075d1f9 100644 --- a/drivers/gpu/drm/amd/display/dc/hwss/dcn32/dcn32_hwseq.c +++ b/drivers/gpu/drm/amd/display/dc/hwss/dcn32/dcn32_hwseq.c @@ -1049,7 +1049,8 @@ void dcn32_update_dsc_on_stream(struct pipe_ctx *pipe_ctx, bool enable) } /* Enable DSC hw block */ - dsc_cfg.pic_width = (stream->timing.h_addressable + stream->timing.h_border_left + stream->timing.h_border_right) / opp_cnt; + dsc_cfg.pic_width = (stream->timing.h_addressable + pipe_ctx->hblank_borrow + + stream->timing.h_border_left + stream->timing.h_border_right) / opp_cnt; dsc_cfg.pic_height = stream->timing.v_addressable + stream->timing.v_border_top + stream->timing.v_border_bottom; dsc_cfg.pixel_encoding = stream->timing.pixel_encoding; dsc_cfg.color_depth = stream->timing.display_color_depth; diff --git a/drivers/gpu/drm/amd/display/dc/hwss/dcn401/dcn401_hwseq.c b/drivers/gpu/drm/amd/display/dc/hwss/dcn401/dcn401_hwseq.c index 5de11e2837c0..307782592789 100644 --- a/drivers/gpu/drm/amd/display/dc/hwss/dcn401/dcn401_hwseq.c +++ b/drivers/gpu/drm/amd/display/dc/hwss/dcn401/dcn401_hwseq.c @@ -820,6 +820,7 @@ enum dc_status dcn401_enable_stream_timing( int opp_cnt = 1; int opp_inst[MAX_PIPES] = {0}; struct pipe_ctx *opp_heads[MAX_PIPES] = {0}; + struct dc_crtc_timing patched_crtc_timing = stream->timing; bool manual_mode; unsigned int tmds_div = PIXEL_RATE_DIV_NA; unsigned int unused_div = PIXEL_RATE_DIV_NA; @@ -874,9 +875,13 @@ enum dc_status dcn401_enable_stream_timing( if (dc->hwseq->funcs.PLAT_58856_wa && (!dc_is_dp_signal(stream->signal))) dc->hwseq->funcs.PLAT_58856_wa(context, pipe_ctx); + /* if we are borrowing from hblank, h_addressable needs to be adjusted */ + if (dc->debug.enable_hblank_borrow) + patched_crtc_timing.h_addressable = patched_crtc_timing.h_addressable + pipe_ctx->hblank_borrow; + pipe_ctx->stream_res.tg->funcs->program_timing( pipe_ctx->stream_res.tg, - &stream->timing, + &patched_crtc_timing, pipe_ctx->pipe_dlg_param.vready_offset, pipe_ctx->pipe_dlg_param.vstartup_start, pipe_ctx->pipe_dlg_param.vupdate_offset, diff --git a/drivers/gpu/drm/amd/display/dc/inc/core_types.h b/drivers/gpu/drm/amd/display/dc/inc/core_types.h index 3061dca47dd2..2edd5b38ce4f 100644 --- a/drivers/gpu/drm/amd/display/dc/inc/core_types.h +++ b/drivers/gpu/drm/amd/display/dc/inc/core_types.h @@ -478,6 +478,8 @@ struct pipe_ctx { /* subvp_index: only valid if the pipe is a SUBVP_MAIN*/ uint8_t subvp_index; struct pixel_rate_divider pixel_rate_divider; + /* pixels borrowed from hblank to hactive */ + uint8_t hblank_borrow; }; /* Data used for dynamic link encoder assignment. diff --git a/drivers/gpu/drm/amd/display/dc/link/link_dpms.c b/drivers/gpu/drm/amd/display/dc/link/link_dpms.c index 41cab9ad6885..5d66bfc7fe6e 100644 --- a/drivers/gpu/drm/amd/display/dc/link/link_dpms.c +++ b/drivers/gpu/drm/amd/display/dc/link/link_dpms.c @@ -808,7 +808,8 @@ void link_set_dsc_on_stream(struct pipe_ctx *pipe_ctx, bool enable) enum optc_dsc_mode optc_dsc_mode; /* Enable DSC hw block */ - dsc_cfg.pic_width = (stream->timing.h_addressable + stream->timing.h_border_left + stream->timing.h_border_right) / opp_cnt; + dsc_cfg.pic_width = (stream->timing.h_addressable + pipe_ctx->hblank_borrow + + stream->timing.h_border_left + stream->timing.h_border_right) / opp_cnt; dsc_cfg.pic_height = stream->timing.v_addressable + stream->timing.v_border_top + stream->timing.v_border_bottom; dsc_cfg.pixel_encoding = stream->timing.pixel_encoding; dsc_cfg.color_depth = stream->timing.display_color_depth; diff --git a/drivers/gpu/drm/amd/display/dc/resource/dcn32/dcn32_resource.c b/drivers/gpu/drm/amd/display/dc/resource/dcn32/dcn32_resource.c index 984d23bcbc27..12d247a7ec45 100644 --- a/drivers/gpu/drm/amd/display/dc/resource/dcn32/dcn32_resource.c +++ b/drivers/gpu/drm/amd/display/dc/resource/dcn32/dcn32_resource.c @@ -2804,6 +2804,7 @@ struct pipe_ctx *dcn32_acquire_free_pipe_as_secondary_opp_head( free_pipe->plane_res.xfm = pool->transforms[free_pipe_idx]; free_pipe->plane_res.dpp = pool->dpps[free_pipe_idx]; free_pipe->plane_res.mpcc_inst = pool->dpps[free_pipe_idx]->inst; + free_pipe->hblank_borrow = otg_master->hblank_borrow; if (free_pipe->stream->timing.flags.DSC == 1) { dcn20_acquire_dsc(free_pipe->stream->ctx->dc, &new_ctx->res_ctx, -- 2.50.1 From 274e3f4596446955bf17680fd4eb5489f5ecac00 Mon Sep 17 00:00:00 2001 From: Pratap Nirujogi Date: Fri, 29 Nov 2024 14:52:08 -0500 Subject: [PATCH 03/16] drm/amdgpu: Fix ISP hw init issue ISP hw_init is not called with the recent changes related to hw init levels. AMDGPU_INIT_LEVEL_DEFAULT is ignoring the ISP IP block as AMDGPU_IP_BLK_MASK_ALL is derived using incorrect max number of IP blocks. Update AMDGPU_IP_BLK_MASK_ALL to use AMDGPU_MAX_IP_NUM instead of (AMDGPU_MAX_IP_NUM - 1) to fix the issue. Reviewed-by: Mario Limonciello Fixes: 14f2fe34f5c6 ("drm/amdgpu: Add init levels") Signed-off-by: Pratap Nirujogi Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_device.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c index 9095c05e0269..c3ca217b8d86 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c @@ -145,7 +145,7 @@ const char *amdgpu_asic_name[] = { "LAST", }; -#define AMDGPU_IP_BLK_MASK_ALL GENMASK(AMDGPU_MAX_IP_NUM - 1, 0) +#define AMDGPU_IP_BLK_MASK_ALL GENMASK(AMDGPU_MAX_IP_NUM, 0) /* * Default init level where all blocks are expected to be initialized. This is * the level of initialization expected by default and also after a full reset -- 2.50.1 From f3bb57b66dc439dd129eb509a4965f1e1aeea2b8 Mon Sep 17 00:00:00 2001 From: Yiqing Yao Date: Tue, 26 Nov 2024 18:36:11 +0800 Subject: [PATCH 04/16] drm/amdgpu: fix sriov reinit late orders Use found block to call correct init/resume function on the block. Set status.hw for resume and init. Print re-init result again. Change to use dev_info. Use amdgpu_device_ip_get_ip_block to get target block instead of loop. Fixes: 502d76308d45 ("drm/amdgpu: validate resume before function call") Signed-off-by: Yiqing Yao Reviewed-by: Lijo Lazar Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_device.c | 43 ++++++++++------------ 1 file changed, 20 insertions(+), 23 deletions(-) diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c index c3ca217b8d86..913eb4f08ee6 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c @@ -3670,9 +3670,11 @@ static int amdgpu_device_ip_reinit_early_sriov(struct amdgpu_device *adev) continue; r = block->version->funcs->hw_init(&adev->ip_blocks[i]); - DRM_INFO("RE-INIT-early: %s %s\n", block->version->funcs->name, r?"failed":"succeeded"); - if (r) + if (r) { + dev_err(adev->dev, "RE-INIT-early: %s failed\n", + block->version->funcs->name); return r; + } block->status.hw = true; } } @@ -3682,7 +3684,8 @@ static int amdgpu_device_ip_reinit_early_sriov(struct amdgpu_device *adev) static int amdgpu_device_ip_reinit_late_sriov(struct amdgpu_device *adev) { - int i, r; + struct amdgpu_ip_block *block; + int i, r = 0; static enum amd_ip_block_type ip_order[] = { AMD_IP_BLOCK_TYPE_SMC, @@ -3697,34 +3700,28 @@ static int amdgpu_device_ip_reinit_late_sriov(struct amdgpu_device *adev) }; for (i = 0; i < ARRAY_SIZE(ip_order); i++) { - int j; - struct amdgpu_ip_block *block; - - for (j = 0; j < adev->num_ip_blocks; j++) { - block = &adev->ip_blocks[j]; + block = amdgpu_device_ip_get_ip_block(adev, ip_order[i]); - if (block->version->type != ip_order[i] || - !block->status.valid || - block->status.hw) - continue; + if (!block) + continue; + if (block->status.valid && !block->status.hw) { if (block->version->type == AMD_IP_BLOCK_TYPE_SMC) { - r = amdgpu_ip_block_resume(&adev->ip_blocks[i]); - if (r) - return r; + r = amdgpu_ip_block_resume(block); } else { - r = block->version->funcs->hw_init(&adev->ip_blocks[i]); - if (r) { - DRM_ERROR("hw_init of IP block <%s> failed %d\n", - adev->ip_blocks[i].version->funcs->name, r); - return r; - } - block->status.hw = true; + r = block->version->funcs->hw_init(block); + } + + if (r) { + dev_err(adev->dev, "RE-INIT-late: %s failed\n", + block->version->funcs->name); + break; } + block->status.hw = true; } } - return 0; + return r; } /** -- 2.50.1 From c3d06a3b6acd6b8c9595d677d049555f475703df Mon Sep 17 00:00:00 2001 From: Alex Deucher Date: Sat, 16 Nov 2024 09:22:14 -0500 Subject: [PATCH 05/16] Revert "drm/amd/pm: correct the workload setting" This reverts commit 74e1006430a5377228e49310f6d915628609929e. This causes a regression in the workload selection. A more extensive fix is being worked on. For now, revert. This came back after a merge in 6.13-rc1, so revert again. Link: https://gitlab.freedesktop.org/drm/amd/-/issues/3618 Fixes: 74e1006430a5 ("drm/amd/pm: correct the workload setting") Signed-off-by: Alex Deucher (cherry picked from commit 44f392fbf628a7ff2d8bb8e83ca1851261f81a6f) --- drivers/gpu/drm/amd/pm/swsmu/amdgpu_smu.c | 49 ++++++------------- drivers/gpu/drm/amd/pm/swsmu/inc/amdgpu_smu.h | 4 +- .../gpu/drm/amd/pm/swsmu/smu11/arcturus_ppt.c | 5 +- .../gpu/drm/amd/pm/swsmu/smu11/navi10_ppt.c | 5 +- .../amd/pm/swsmu/smu11/sienna_cichlid_ppt.c | 5 +- .../gpu/drm/amd/pm/swsmu/smu11/vangogh_ppt.c | 4 +- .../gpu/drm/amd/pm/swsmu/smu12/renoir_ppt.c | 4 +- .../drm/amd/pm/swsmu/smu13/smu_v13_0_0_ppt.c | 20 ++------ .../drm/amd/pm/swsmu/smu13/smu_v13_0_7_ppt.c | 5 +- .../drm/amd/pm/swsmu/smu14/smu_v14_0_2_ppt.c | 9 ++-- drivers/gpu/drm/amd/pm/swsmu/smu_cmn.c | 8 --- drivers/gpu/drm/amd/pm/swsmu/smu_cmn.h | 2 - 12 files changed, 36 insertions(+), 84 deletions(-) diff --git a/drivers/gpu/drm/amd/pm/swsmu/amdgpu_smu.c b/drivers/gpu/drm/amd/pm/swsmu/amdgpu_smu.c index b8355293518f..563824f42b1d 100644 --- a/drivers/gpu/drm/amd/pm/swsmu/amdgpu_smu.c +++ b/drivers/gpu/drm/amd/pm/swsmu/amdgpu_smu.c @@ -1261,33 +1261,26 @@ static int smu_sw_init(struct amdgpu_ip_block *ip_block) smu->watermarks_bitmap = 0; smu->power_profile_mode = PP_SMC_POWER_PROFILE_BOOTUP_DEFAULT; smu->default_power_profile_mode = PP_SMC_POWER_PROFILE_BOOTUP_DEFAULT; - smu->user_dpm_profile.user_workload_mask = 0; atomic_set(&smu->smu_power.power_gate.vcn_gated, 1); atomic_set(&smu->smu_power.power_gate.jpeg_gated, 1); atomic_set(&smu->smu_power.power_gate.vpe_gated, 1); atomic_set(&smu->smu_power.power_gate.umsch_mm_gated, 1); - smu->workload_priority[PP_SMC_POWER_PROFILE_BOOTUP_DEFAULT] = 0; - smu->workload_priority[PP_SMC_POWER_PROFILE_FULLSCREEN3D] = 1; - smu->workload_priority[PP_SMC_POWER_PROFILE_POWERSAVING] = 2; - smu->workload_priority[PP_SMC_POWER_PROFILE_VIDEO] = 3; - smu->workload_priority[PP_SMC_POWER_PROFILE_VR] = 4; - smu->workload_priority[PP_SMC_POWER_PROFILE_COMPUTE] = 5; - smu->workload_priority[PP_SMC_POWER_PROFILE_CUSTOM] = 6; + smu->workload_prority[PP_SMC_POWER_PROFILE_BOOTUP_DEFAULT] = 0; + smu->workload_prority[PP_SMC_POWER_PROFILE_FULLSCREEN3D] = 1; + smu->workload_prority[PP_SMC_POWER_PROFILE_POWERSAVING] = 2; + smu->workload_prority[PP_SMC_POWER_PROFILE_VIDEO] = 3; + smu->workload_prority[PP_SMC_POWER_PROFILE_VR] = 4; + smu->workload_prority[PP_SMC_POWER_PROFILE_COMPUTE] = 5; + smu->workload_prority[PP_SMC_POWER_PROFILE_CUSTOM] = 6; if (smu->is_apu || - !smu_is_workload_profile_available(smu, PP_SMC_POWER_PROFILE_FULLSCREEN3D)) { - smu->driver_workload_mask = - 1 << smu->workload_priority[PP_SMC_POWER_PROFILE_BOOTUP_DEFAULT]; - } else { - smu->driver_workload_mask = - 1 << smu->workload_priority[PP_SMC_POWER_PROFILE_FULLSCREEN3D]; - smu->default_power_profile_mode = PP_SMC_POWER_PROFILE_FULLSCREEN3D; - } + !smu_is_workload_profile_available(smu, PP_SMC_POWER_PROFILE_FULLSCREEN3D)) + smu->workload_mask = 1 << smu->workload_prority[PP_SMC_POWER_PROFILE_BOOTUP_DEFAULT]; + else + smu->workload_mask = 1 << smu->workload_prority[PP_SMC_POWER_PROFILE_FULLSCREEN3D]; - smu->workload_mask = smu->driver_workload_mask | - smu->user_dpm_profile.user_workload_mask; smu->workload_setting[0] = PP_SMC_POWER_PROFILE_BOOTUP_DEFAULT; smu->workload_setting[1] = PP_SMC_POWER_PROFILE_FULLSCREEN3D; smu->workload_setting[2] = PP_SMC_POWER_PROFILE_POWERSAVING; @@ -2366,20 +2359,17 @@ static int smu_switch_power_profile(void *handle, return -EINVAL; if (!en) { - smu->driver_workload_mask &= ~(1 << smu->workload_priority[type]); + smu->workload_mask &= ~(1 << smu->workload_prority[type]); index = fls(smu->workload_mask); index = index > 0 && index <= WORKLOAD_POLICY_MAX ? index - 1 : 0; workload[0] = smu->workload_setting[index]; } else { - smu->driver_workload_mask |= (1 << smu->workload_priority[type]); + smu->workload_mask |= (1 << smu->workload_prority[type]); index = fls(smu->workload_mask); index = index <= WORKLOAD_POLICY_MAX ? index - 1 : 0; workload[0] = smu->workload_setting[index]; } - smu->workload_mask = smu->driver_workload_mask | - smu->user_dpm_profile.user_workload_mask; - if (smu_dpm_ctx->dpm_level != AMD_DPM_FORCED_LEVEL_MANUAL && smu_dpm_ctx->dpm_level != AMD_DPM_FORCED_LEVEL_PERF_DETERMINISM) smu_bump_power_profile_mode(smu, workload, 0); @@ -3074,23 +3064,12 @@ static int smu_set_power_profile_mode(void *handle, uint32_t param_size) { struct smu_context *smu = handle; - int ret; if (!smu->pm_enabled || !smu->adev->pm.dpm_enabled || !smu->ppt_funcs->set_power_profile_mode) return -EOPNOTSUPP; - if (smu->user_dpm_profile.user_workload_mask & - (1 << smu->workload_priority[param[param_size]])) - return 0; - - smu->user_dpm_profile.user_workload_mask = - (1 << smu->workload_priority[param[param_size]]); - smu->workload_mask = smu->user_dpm_profile.user_workload_mask | - smu->driver_workload_mask; - ret = smu_bump_power_profile_mode(smu, param, param_size); - - return ret; + return smu_bump_power_profile_mode(smu, param, param_size); } static int smu_get_fan_control_mode(void *handle, u32 *fan_mode) diff --git a/drivers/gpu/drm/amd/pm/swsmu/inc/amdgpu_smu.h b/drivers/gpu/drm/amd/pm/swsmu/inc/amdgpu_smu.h index d665c47f19b7..4ebcc1e53ea2 100644 --- a/drivers/gpu/drm/amd/pm/swsmu/inc/amdgpu_smu.h +++ b/drivers/gpu/drm/amd/pm/swsmu/inc/amdgpu_smu.h @@ -240,7 +240,6 @@ struct smu_user_dpm_profile { /* user clock state information */ uint32_t clk_mask[SMU_CLK_COUNT]; uint32_t clk_dependency; - uint32_t user_workload_mask; }; #define SMU_TABLE_INIT(tables, table_id, s, a, d) \ @@ -558,8 +557,7 @@ struct smu_context { bool disable_uclk_switch; uint32_t workload_mask; - uint32_t driver_workload_mask; - uint32_t workload_priority[WORKLOAD_POLICY_MAX]; + uint32_t workload_prority[WORKLOAD_POLICY_MAX]; uint32_t workload_setting[WORKLOAD_POLICY_MAX]; uint32_t power_profile_mode; uint32_t default_power_profile_mode; diff --git a/drivers/gpu/drm/amd/pm/swsmu/smu11/arcturus_ppt.c b/drivers/gpu/drm/amd/pm/swsmu/smu11/arcturus_ppt.c index 12125303bb79..a15754b1989f 100644 --- a/drivers/gpu/drm/amd/pm/swsmu/smu11/arcturus_ppt.c +++ b/drivers/gpu/drm/amd/pm/swsmu/smu11/arcturus_ppt.c @@ -1459,6 +1459,7 @@ static int arcturus_set_power_profile_mode(struct smu_context *smu, return -EINVAL; } + if ((profile_mode == PP_SMC_POWER_PROFILE_CUSTOM) && (smu->smc_fw_version >= 0x360d00)) { if (size != 10) @@ -1526,14 +1527,14 @@ static int arcturus_set_power_profile_mode(struct smu_context *smu, ret = smu_cmn_send_smc_msg_with_param(smu, SMU_MSG_SetWorkloadMask, - smu->workload_mask, + 1 << workload_type, NULL); if (ret) { dev_err(smu->adev->dev, "Fail to set workload type %d\n", workload_type); return ret; } - smu_cmn_assign_power_profile(smu); + smu->power_profile_mode = profile_mode; return 0; } 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 211635dabed8..faa8e7d9c3c6 100644 --- a/drivers/gpu/drm/amd/pm/swsmu/smu11/navi10_ppt.c +++ b/drivers/gpu/drm/amd/pm/swsmu/smu11/navi10_ppt.c @@ -2083,13 +2083,10 @@ static int navi10_set_power_profile_mode(struct smu_context *smu, long *input, u smu->power_profile_mode); if (workload_type < 0) return -EINVAL; - ret = smu_cmn_send_smc_msg_with_param(smu, SMU_MSG_SetWorkloadMask, - smu->workload_mask, NULL); + 1 << workload_type, NULL); if (ret) dev_err(smu->adev->dev, "[%s] Failed to set work load mask!", __func__); - else - smu_cmn_assign_power_profile(smu); 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 d0ed0d060a8a..a9cb28ce2133 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 @@ -1788,13 +1788,10 @@ static int sienna_cichlid_set_power_profile_mode(struct smu_context *smu, long * smu->power_profile_mode); if (workload_type < 0) return -EINVAL; - ret = smu_cmn_send_smc_msg_with_param(smu, SMU_MSG_SetWorkloadMask, - smu->workload_mask, NULL); + 1 << workload_type, NULL); if (ret) dev_err(smu->adev->dev, "[%s] Failed to set work load mask!", __func__); - else - smu_cmn_assign_power_profile(smu); return ret; } 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 f89c487dce72..cd3e9ba3eff4 100644 --- a/drivers/gpu/drm/amd/pm/swsmu/smu11/vangogh_ppt.c +++ b/drivers/gpu/drm/amd/pm/swsmu/smu11/vangogh_ppt.c @@ -1081,7 +1081,7 @@ static int vangogh_set_power_profile_mode(struct smu_context *smu, long *input, } ret = smu_cmn_send_smc_msg_with_param(smu, SMU_MSG_ActiveProcessNotify, - smu->workload_mask, + 1 << workload_type, NULL); if (ret) { dev_err_once(smu->adev->dev, "Fail to set workload type %d\n", @@ -1089,7 +1089,7 @@ static int vangogh_set_power_profile_mode(struct smu_context *smu, long *input, return ret; } - smu_cmn_assign_power_profile(smu); + smu->power_profile_mode = profile_mode; return 0; } diff --git a/drivers/gpu/drm/amd/pm/swsmu/smu12/renoir_ppt.c b/drivers/gpu/drm/amd/pm/swsmu/smu12/renoir_ppt.c index 75a9ea87f419..a34797f3576b 100644 --- a/drivers/gpu/drm/amd/pm/swsmu/smu12/renoir_ppt.c +++ b/drivers/gpu/drm/amd/pm/swsmu/smu12/renoir_ppt.c @@ -892,14 +892,14 @@ static int renoir_set_power_profile_mode(struct smu_context *smu, long *input, u } ret = smu_cmn_send_smc_msg_with_param(smu, SMU_MSG_ActiveProcessNotify, - smu->workload_mask, + 1 << workload_type, NULL); if (ret) { dev_err_once(smu->adev->dev, "Fail to set workload type %d\n", workload_type); return ret; } - smu_cmn_assign_power_profile(smu); + smu->power_profile_mode = profile_mode; 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 80c6b1e523aa..199bdd9720d3 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 @@ -2579,7 +2579,7 @@ static int smu_v13_0_0_set_power_profile_mode(struct smu_context *smu, DpmActivityMonitorCoeffInt_t *activity_monitor = &(activity_monitor_external.DpmActivityMonitorCoeffInt); int workload_type, ret = 0; - u32 workload_mask; + u32 workload_mask, selected_workload_mask; smu->power_profile_mode = input[size]; @@ -2646,7 +2646,7 @@ static int smu_v13_0_0_set_power_profile_mode(struct smu_context *smu, if (workload_type < 0) return -EINVAL; - workload_mask = 1 << workload_type; + selected_workload_mask = workload_mask = 1 << workload_type; /* Add optimizations for SMU13.0.0/10. Reuse the power saving profile */ if ((amdgpu_ip_version(smu->adev, MP1_HWIP, 0) == IP_VERSION(13, 0, 0) && @@ -2661,22 +2661,12 @@ static int smu_v13_0_0_set_power_profile_mode(struct smu_context *smu, workload_mask |= 1 << workload_type; } - smu->workload_mask |= workload_mask; ret = smu_cmn_send_smc_msg_with_param(smu, SMU_MSG_SetWorkloadMask, - smu->workload_mask, + workload_mask, NULL); - if (!ret) { - smu_cmn_assign_power_profile(smu); - if (smu->power_profile_mode == PP_SMC_POWER_PROFILE_POWERSAVING) { - workload_type = smu_cmn_to_asic_specific_index(smu, - CMN2ASIC_MAPPING_WORKLOAD, - PP_SMC_POWER_PROFILE_FULLSCREEN3D); - smu->power_profile_mode = smu->workload_mask & (1 << workload_type) - ? PP_SMC_POWER_PROFILE_FULLSCREEN3D - : PP_SMC_POWER_PROFILE_BOOTUP_DEFAULT; - } - } + if (!ret) + smu->workload_mask = selected_workload_mask; return ret; } 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 4fd0354bd312..34c1e0c7e1e4 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 @@ -2595,14 +2595,13 @@ static int smu_v13_0_7_set_power_profile_mode(struct smu_context *smu, long *inp smu->power_profile_mode); if (workload_type < 0) return -EINVAL; - ret = smu_cmn_send_smc_msg_with_param(smu, SMU_MSG_SetWorkloadMask, - smu->workload_mask, NULL); + 1 << workload_type, NULL); if (ret) dev_err(smu->adev->dev, "[%s] Failed to set work load mask!", __func__); else - smu_cmn_assign_power_profile(smu); + smu->workload_mask = (1 << workload_type); return ret; } diff --git a/drivers/gpu/drm/amd/pm/swsmu/smu14/smu_v14_0_2_ppt.c b/drivers/gpu/drm/amd/pm/swsmu/smu14/smu_v14_0_2_ppt.c index 687a0f5ac94f..98e01a06add8 100644 --- a/drivers/gpu/drm/amd/pm/swsmu/smu14/smu_v14_0_2_ppt.c +++ b/drivers/gpu/drm/amd/pm/swsmu/smu14/smu_v14_0_2_ppt.c @@ -1817,11 +1817,12 @@ static int smu_v14_0_2_set_power_profile_mode(struct smu_context *smu, if (workload_type < 0) return -EINVAL; - ret = smu_cmn_send_smc_msg_with_param(smu, SMU_MSG_SetWorkloadMask, - smu->workload_mask, NULL); - + ret = smu_cmn_send_smc_msg_with_param(smu, + SMU_MSG_SetWorkloadMask, + 1 << workload_type, + NULL); if (!ret) - smu_cmn_assign_power_profile(smu); + smu->workload_mask = 1 << workload_type; return ret; } diff --git a/drivers/gpu/drm/amd/pm/swsmu/smu_cmn.c b/drivers/gpu/drm/amd/pm/swsmu/smu_cmn.c index dbbd3759bff3..0bc6fb7d7223 100644 --- a/drivers/gpu/drm/amd/pm/swsmu/smu_cmn.c +++ b/drivers/gpu/drm/amd/pm/swsmu/smu_cmn.c @@ -1144,14 +1144,6 @@ int smu_cmn_set_mp1_state(struct smu_context *smu, return ret; } -void smu_cmn_assign_power_profile(struct smu_context *smu) -{ - uint32_t index; - index = fls(smu->workload_mask); - index = index > 0 && index <= WORKLOAD_POLICY_MAX ? index - 1 : 0; - smu->power_profile_mode = smu->workload_setting[index]; -} - bool smu_cmn_is_audio_func_enabled(struct amdgpu_device *adev) { struct pci_dev *p = NULL; diff --git a/drivers/gpu/drm/amd/pm/swsmu/smu_cmn.h b/drivers/gpu/drm/amd/pm/swsmu/smu_cmn.h index 8a801e389659..1de685defe85 100644 --- a/drivers/gpu/drm/amd/pm/swsmu/smu_cmn.h +++ b/drivers/gpu/drm/amd/pm/swsmu/smu_cmn.h @@ -130,8 +130,6 @@ void smu_cmn_init_soft_gpu_metrics(void *table, uint8_t frev, uint8_t crev); int smu_cmn_set_mp1_state(struct smu_context *smu, enum pp_mp1_state mp1_state); -void smu_cmn_assign_power_profile(struct smu_context *smu); - /* * Helper function to make sysfs_emit_at() happy. Align buf to * the current page boundary and record the offset. -- 2.50.1 From 1443dd3c67f6d1a8bd1f810e598e2f0c6f19205c Mon Sep 17 00:00:00 2001 From: Alex Deucher Date: Sat, 16 Nov 2024 08:20:59 -0500 Subject: [PATCH 06/16] drm/amd/pm: fix and simplify workload handling smu->workload_mask is IP specific and should not be messed with in the common code. The mask bits vary across SMU versions. Move all handling of smu->workload_mask in to the backends and simplify the code. Store the user's preference in smu->power_profile_mode which will be reflected in sysfs. For internal driver profile switches for KFD or VCN, just update the workload mask so that the user's preference is retained. Remove all of the extra now unused workload related elements in the smu structure. v2: use refcounts for workload profiles v3: rework based on feedback from Lijo v4: fix the refcount on failure, drop backend mask v5: rework custom handling v6: handle failure cleanup with custom profile v7: Update documentation Reviewed-by: Lijo Lazar Signed-off-by: Alex Deucher Cc: Kenneth Feng Cc: Lijo Lazar Cc: stable@vger.kernel.org # 6.11.x --- drivers/gpu/drm/amd/pm/amdgpu_pm.c | 6 +- drivers/gpu/drm/amd/pm/swsmu/amdgpu_smu.c | 150 ++++++++++------ drivers/gpu/drm/amd/pm/swsmu/inc/amdgpu_smu.h | 15 +- .../gpu/drm/amd/pm/swsmu/smu11/arcturus_ppt.c | 166 +++++++++-------- .../gpu/drm/amd/pm/swsmu/smu11/navi10_ppt.c | 167 ++++++++++------- .../amd/pm/swsmu/smu11/sienna_cichlid_ppt.c | 168 +++++++++++------- .../gpu/drm/amd/pm/swsmu/smu11/vangogh_ppt.c | 41 ++--- .../gpu/drm/amd/pm/swsmu/smu12/renoir_ppt.c | 43 ++--- .../drm/amd/pm/swsmu/smu13/smu_v13_0_0_ppt.c | 167 +++++++++-------- .../drm/amd/pm/swsmu/smu13/smu_v13_0_7_ppt.c | 138 ++++++++------ .../drm/amd/pm/swsmu/smu14/smu_v14_0_2_ppt.c | 168 +++++++++++------- drivers/gpu/drm/amd/pm/swsmu/smu_cmn.c | 25 +++ drivers/gpu/drm/amd/pm/swsmu/smu_cmn.h | 4 + 13 files changed, 741 insertions(+), 517 deletions(-) diff --git a/drivers/gpu/drm/amd/pm/amdgpu_pm.c b/drivers/gpu/drm/amd/pm/amdgpu_pm.c index 136e8193867c..e8ae7681bf0a 100644 --- a/drivers/gpu/drm/amd/pm/amdgpu_pm.c +++ b/drivers/gpu/drm/amd/pm/amdgpu_pm.c @@ -1361,7 +1361,11 @@ static ssize_t amdgpu_set_pp_mclk_od(struct device *dev, * create a custom set of heuristics, write a string of numbers to the file * starting with the number of the custom profile along with a setting * for each heuristic parameter. Due to differences across asic families - * the heuristic parameters vary from family to family. + * the heuristic parameters vary from family to family. Additionally, + * you can apply the custom heuristics to different clock domains. Each + * clock domain is considered a distinct operation so if you modify the + * gfxclk heuristics and then the memclk heuristics, the all of the + * custom heuristics will be retained until you switch to another profile. * */ diff --git a/drivers/gpu/drm/amd/pm/swsmu/amdgpu_smu.c b/drivers/gpu/drm/amd/pm/swsmu/amdgpu_smu.c index 563824f42b1d..5eae14fe79f1 100644 --- a/drivers/gpu/drm/amd/pm/swsmu/amdgpu_smu.c +++ b/drivers/gpu/drm/amd/pm/swsmu/amdgpu_smu.c @@ -72,6 +72,10 @@ static int smu_set_power_limit(void *handle, uint32_t limit); static int smu_set_fan_speed_rpm(void *handle, uint32_t speed); static int smu_set_gfx_cgpg(struct smu_context *smu, bool enabled); static int smu_set_mp1_state(void *handle, enum pp_mp1_state mp1_state); +static void smu_power_profile_mode_get(struct smu_context *smu, + enum PP_SMC_POWER_PROFILE profile_mode); +static void smu_power_profile_mode_put(struct smu_context *smu, + enum PP_SMC_POWER_PROFILE profile_mode); static int smu_sys_get_pp_feature_mask(void *handle, char *buf) @@ -1259,35 +1263,19 @@ static int smu_sw_init(struct amdgpu_ip_block *ip_block) INIT_WORK(&smu->interrupt_work, smu_interrupt_work_fn); atomic64_set(&smu->throttle_int_counter, 0); smu->watermarks_bitmap = 0; - smu->power_profile_mode = PP_SMC_POWER_PROFILE_BOOTUP_DEFAULT; - smu->default_power_profile_mode = PP_SMC_POWER_PROFILE_BOOTUP_DEFAULT; atomic_set(&smu->smu_power.power_gate.vcn_gated, 1); atomic_set(&smu->smu_power.power_gate.jpeg_gated, 1); atomic_set(&smu->smu_power.power_gate.vpe_gated, 1); atomic_set(&smu->smu_power.power_gate.umsch_mm_gated, 1); - smu->workload_prority[PP_SMC_POWER_PROFILE_BOOTUP_DEFAULT] = 0; - smu->workload_prority[PP_SMC_POWER_PROFILE_FULLSCREEN3D] = 1; - smu->workload_prority[PP_SMC_POWER_PROFILE_POWERSAVING] = 2; - smu->workload_prority[PP_SMC_POWER_PROFILE_VIDEO] = 3; - smu->workload_prority[PP_SMC_POWER_PROFILE_VR] = 4; - smu->workload_prority[PP_SMC_POWER_PROFILE_COMPUTE] = 5; - smu->workload_prority[PP_SMC_POWER_PROFILE_CUSTOM] = 6; - if (smu->is_apu || !smu_is_workload_profile_available(smu, PP_SMC_POWER_PROFILE_FULLSCREEN3D)) - smu->workload_mask = 1 << smu->workload_prority[PP_SMC_POWER_PROFILE_BOOTUP_DEFAULT]; + smu->power_profile_mode = PP_SMC_POWER_PROFILE_BOOTUP_DEFAULT; else - smu->workload_mask = 1 << smu->workload_prority[PP_SMC_POWER_PROFILE_FULLSCREEN3D]; - - smu->workload_setting[0] = PP_SMC_POWER_PROFILE_BOOTUP_DEFAULT; - smu->workload_setting[1] = PP_SMC_POWER_PROFILE_FULLSCREEN3D; - smu->workload_setting[2] = PP_SMC_POWER_PROFILE_POWERSAVING; - smu->workload_setting[3] = PP_SMC_POWER_PROFILE_VIDEO; - smu->workload_setting[4] = PP_SMC_POWER_PROFILE_VR; - smu->workload_setting[5] = PP_SMC_POWER_PROFILE_COMPUTE; - smu->workload_setting[6] = PP_SMC_POWER_PROFILE_CUSTOM; + smu->power_profile_mode = PP_SMC_POWER_PROFILE_FULLSCREEN3D; + smu_power_profile_mode_get(smu, smu->power_profile_mode); + smu->display_config = &adev->pm.pm_display_cfg; smu->smu_dpm.dpm_level = AMD_DPM_FORCED_LEVEL_AUTO; @@ -1340,6 +1328,11 @@ static int smu_sw_fini(struct amdgpu_ip_block *ip_block) return ret; } + if (smu->custom_profile_params) { + kfree(smu->custom_profile_params); + smu->custom_profile_params = NULL; + } + smu_fini_microcode(smu); return 0; @@ -2124,6 +2117,9 @@ static int smu_suspend(struct amdgpu_ip_block *ip_block) if (!ret) adev->gfx.gfx_off_entrycount = count; + /* clear this on suspend so it will get reprogrammed on resume */ + smu->workload_mask = 0; + return 0; } @@ -2236,25 +2232,49 @@ static int smu_enable_umd_pstate(void *handle, } static int smu_bump_power_profile_mode(struct smu_context *smu, - long *param, - uint32_t param_size) + long *custom_params, + u32 custom_params_max_idx) { - int ret = 0; + u32 workload_mask = 0; + int i, ret = 0; + + for (i = 0; i < PP_SMC_POWER_PROFILE_COUNT; i++) { + if (smu->workload_refcount[i]) + workload_mask |= 1 << i; + } + + if (smu->workload_mask == workload_mask) + return 0; if (smu->ppt_funcs->set_power_profile_mode) - ret = smu->ppt_funcs->set_power_profile_mode(smu, param, param_size); + ret = smu->ppt_funcs->set_power_profile_mode(smu, workload_mask, + custom_params, + custom_params_max_idx); + + if (!ret) + smu->workload_mask = workload_mask; return ret; } +static void smu_power_profile_mode_get(struct smu_context *smu, + enum PP_SMC_POWER_PROFILE profile_mode) +{ + smu->workload_refcount[profile_mode]++; +} + +static void smu_power_profile_mode_put(struct smu_context *smu, + enum PP_SMC_POWER_PROFILE profile_mode) +{ + if (smu->workload_refcount[profile_mode]) + smu->workload_refcount[profile_mode]--; +} + static int smu_adjust_power_state_dynamic(struct smu_context *smu, enum amd_dpm_forced_level level, - bool skip_display_settings, - bool init) + bool skip_display_settings) { int ret = 0; - int index = 0; - long workload[1]; struct smu_dpm_context *smu_dpm_ctx = &(smu->smu_dpm); if (!skip_display_settings) { @@ -2291,14 +2311,8 @@ static int smu_adjust_power_state_dynamic(struct smu_context *smu, } if (smu_dpm_ctx->dpm_level != AMD_DPM_FORCED_LEVEL_MANUAL && - smu_dpm_ctx->dpm_level != AMD_DPM_FORCED_LEVEL_PERF_DETERMINISM) { - index = fls(smu->workload_mask); - index = index > 0 && index <= WORKLOAD_POLICY_MAX ? index - 1 : 0; - workload[0] = smu->workload_setting[index]; - - if (init || smu->power_profile_mode != workload[0]) - smu_bump_power_profile_mode(smu, workload, 0); - } + smu_dpm_ctx->dpm_level != AMD_DPM_FORCED_LEVEL_PERF_DETERMINISM) + smu_bump_power_profile_mode(smu, NULL, 0); return ret; } @@ -2317,13 +2331,13 @@ static int smu_handle_task(struct smu_context *smu, ret = smu_pre_display_config_changed(smu); if (ret) return ret; - ret = smu_adjust_power_state_dynamic(smu, level, false, false); + ret = smu_adjust_power_state_dynamic(smu, level, false); break; case AMD_PP_TASK_COMPLETE_INIT: - ret = smu_adjust_power_state_dynamic(smu, level, true, true); + ret = smu_adjust_power_state_dynamic(smu, level, true); break; case AMD_PP_TASK_READJUST_POWER_STATE: - ret = smu_adjust_power_state_dynamic(smu, level, true, false); + ret = smu_adjust_power_state_dynamic(smu, level, true); break; default: break; @@ -2345,12 +2359,11 @@ static int smu_handle_dpm_task(void *handle, static int smu_switch_power_profile(void *handle, enum PP_SMC_POWER_PROFILE type, - bool en) + bool enable) { struct smu_context *smu = handle; struct smu_dpm_context *smu_dpm_ctx = &(smu->smu_dpm); - long workload[1]; - uint32_t index; + int ret; if (!smu->pm_enabled || !smu->adev->pm.dpm_enabled) return -EOPNOTSUPP; @@ -2358,21 +2371,21 @@ static int smu_switch_power_profile(void *handle, if (!(type < PP_SMC_POWER_PROFILE_CUSTOM)) return -EINVAL; - if (!en) { - smu->workload_mask &= ~(1 << smu->workload_prority[type]); - index = fls(smu->workload_mask); - index = index > 0 && index <= WORKLOAD_POLICY_MAX ? index - 1 : 0; - workload[0] = smu->workload_setting[index]; - } else { - smu->workload_mask |= (1 << smu->workload_prority[type]); - index = fls(smu->workload_mask); - index = index <= WORKLOAD_POLICY_MAX ? index - 1 : 0; - workload[0] = smu->workload_setting[index]; - } - if (smu_dpm_ctx->dpm_level != AMD_DPM_FORCED_LEVEL_MANUAL && - smu_dpm_ctx->dpm_level != AMD_DPM_FORCED_LEVEL_PERF_DETERMINISM) - smu_bump_power_profile_mode(smu, workload, 0); + smu_dpm_ctx->dpm_level != AMD_DPM_FORCED_LEVEL_PERF_DETERMINISM) { + if (enable) + smu_power_profile_mode_get(smu, type); + else + smu_power_profile_mode_put(smu, type); + ret = smu_bump_power_profile_mode(smu, NULL, 0); + if (ret) { + if (enable) + smu_power_profile_mode_put(smu, type); + else + smu_power_profile_mode_get(smu, type); + return ret; + } + } return 0; } @@ -3064,12 +3077,35 @@ static int smu_set_power_profile_mode(void *handle, uint32_t param_size) { struct smu_context *smu = handle; + bool custom = false; + int ret = 0; if (!smu->pm_enabled || !smu->adev->pm.dpm_enabled || !smu->ppt_funcs->set_power_profile_mode) return -EOPNOTSUPP; - return smu_bump_power_profile_mode(smu, param, param_size); + if (param[param_size] == PP_SMC_POWER_PROFILE_CUSTOM) { + custom = true; + /* clear frontend mask so custom changes propogate */ + smu->workload_mask = 0; + } + + if ((param[param_size] != smu->power_profile_mode) || custom) { + /* clear the old user preference */ + smu_power_profile_mode_put(smu, smu->power_profile_mode); + /* set the new user preference */ + smu_power_profile_mode_get(smu, param[param_size]); + ret = smu_bump_power_profile_mode(smu, + custom ? param : NULL, + custom ? param_size : 0); + if (ret) + smu_power_profile_mode_put(smu, param[param_size]); + else + /* store the user's preference */ + smu->power_profile_mode = param[param_size]; + } + + return ret; } static int smu_get_fan_control_mode(void *handle, u32 *fan_mode) diff --git a/drivers/gpu/drm/amd/pm/swsmu/inc/amdgpu_smu.h b/drivers/gpu/drm/amd/pm/swsmu/inc/amdgpu_smu.h index 4ebcc1e53ea2..3925815358ce 100644 --- a/drivers/gpu/drm/amd/pm/swsmu/inc/amdgpu_smu.h +++ b/drivers/gpu/drm/amd/pm/swsmu/inc/amdgpu_smu.h @@ -556,11 +556,13 @@ struct smu_context { uint32_t hard_min_uclk_req_from_dal; bool disable_uclk_switch; + /* asic agnostic workload mask */ uint32_t workload_mask; - uint32_t workload_prority[WORKLOAD_POLICY_MAX]; - uint32_t workload_setting[WORKLOAD_POLICY_MAX]; + /* default/user workload preference */ uint32_t power_profile_mode; - uint32_t default_power_profile_mode; + uint32_t workload_refcount[PP_SMC_POWER_PROFILE_COUNT]; + /* backend specific custom workload settings */ + long *custom_profile_params; bool pm_enabled; bool is_apu; @@ -731,9 +733,12 @@ struct pptable_funcs { * @set_power_profile_mode: Set a power profile mode. Also used to * create/set custom power profile modes. * &input: Power profile mode parameters. - * &size: Size of &input. + * &workload_mask: mask of workloads to enable + * &custom_params: custom profile parameters + * &custom_params_max_idx: max valid idx into custom_params */ - int (*set_power_profile_mode)(struct smu_context *smu, long *input, uint32_t size); + int (*set_power_profile_mode)(struct smu_context *smu, u32 workload_mask, + long *custom_params, u32 custom_params_max_idx); /** * @dpm_set_vcn_enable: Enable/disable VCN engine dynamic power diff --git a/drivers/gpu/drm/amd/pm/swsmu/smu11/arcturus_ppt.c b/drivers/gpu/drm/amd/pm/swsmu/smu11/arcturus_ppt.c index a15754b1989f..8aa61a9f7778 100644 --- a/drivers/gpu/drm/amd/pm/swsmu/smu11/arcturus_ppt.c +++ b/drivers/gpu/drm/amd/pm/swsmu/smu11/arcturus_ppt.c @@ -1445,98 +1445,120 @@ static int arcturus_get_power_profile_mode(struct smu_context *smu, return size; } -static int arcturus_set_power_profile_mode(struct smu_context *smu, - long *input, - uint32_t size) +#define ARCTURUS_CUSTOM_PARAMS_COUNT 10 +#define ARCTURUS_CUSTOM_PARAMS_CLOCK_COUNT 2 +#define ARCTURUS_CUSTOM_PARAMS_SIZE (ARCTURUS_CUSTOM_PARAMS_CLOCK_COUNT * ARCTURUS_CUSTOM_PARAMS_COUNT * sizeof(long)) + +static int arcturus_set_power_profile_mode_coeff(struct smu_context *smu, + long *input) { DpmActivityMonitorCoeffInt_t activity_monitor; - int workload_type = 0; - uint32_t profile_mode = input[size]; - int ret = 0; + int ret, idx; - if (profile_mode > PP_SMC_POWER_PROFILE_CUSTOM) { - dev_err(smu->adev->dev, "Invalid power profile mode %d\n", profile_mode); - return -EINVAL; + ret = smu_cmn_update_table(smu, + SMU_TABLE_ACTIVITY_MONITOR_COEFF, + WORKLOAD_PPLIB_CUSTOM_BIT, + (void *)(&activity_monitor), + false); + if (ret) { + dev_err(smu->adev->dev, "[%s] Failed to get activity monitor!", __func__); + return ret; } + idx = 0 * ARCTURUS_CUSTOM_PARAMS_COUNT; + if (input[idx]) { + /* Gfxclk */ + activity_monitor.Gfx_FPS = input[idx + 1]; + activity_monitor.Gfx_UseRlcBusy = input[idx + 2]; + activity_monitor.Gfx_MinActiveFreqType = input[idx + 3]; + activity_monitor.Gfx_MinActiveFreq = input[idx + 4]; + activity_monitor.Gfx_BoosterFreqType = input[idx + 5]; + activity_monitor.Gfx_BoosterFreq = input[idx + 6]; + activity_monitor.Gfx_PD_Data_limit_c = input[idx + 7]; + activity_monitor.Gfx_PD_Data_error_coeff = input[idx + 8]; + activity_monitor.Gfx_PD_Data_error_rate_coeff = input[idx + 9]; + } + idx = 1 * ARCTURUS_CUSTOM_PARAMS_COUNT; + if (input[idx]) { + /* Uclk */ + activity_monitor.Mem_FPS = input[idx + 1]; + activity_monitor.Mem_UseRlcBusy = input[idx + 2]; + activity_monitor.Mem_MinActiveFreqType = input[idx + 3]; + activity_monitor.Mem_MinActiveFreq = input[idx + 4]; + activity_monitor.Mem_BoosterFreqType = input[idx + 5]; + activity_monitor.Mem_BoosterFreq = input[idx + 6]; + activity_monitor.Mem_PD_Data_limit_c = input[idx + 7]; + activity_monitor.Mem_PD_Data_error_coeff = input[idx + 8]; + activity_monitor.Mem_PD_Data_error_rate_coeff = input[idx + 9]; + } - if ((profile_mode == PP_SMC_POWER_PROFILE_CUSTOM) && - (smu->smc_fw_version >= 0x360d00)) { - if (size != 10) - return -EINVAL; + ret = smu_cmn_update_table(smu, + SMU_TABLE_ACTIVITY_MONITOR_COEFF, + WORKLOAD_PPLIB_CUSTOM_BIT, + (void *)(&activity_monitor), + true); + if (ret) { + dev_err(smu->adev->dev, "[%s] Failed to set activity monitor!", __func__); + return ret; + } - ret = smu_cmn_update_table(smu, - SMU_TABLE_ACTIVITY_MONITOR_COEFF, - WORKLOAD_PPLIB_CUSTOM_BIT, - (void *)(&activity_monitor), - false); - if (ret) { - dev_err(smu->adev->dev, "[%s] Failed to get activity monitor!", __func__); - return ret; - } + return ret; +} - switch (input[0]) { - case 0: /* Gfxclk */ - activity_monitor.Gfx_FPS = input[1]; - activity_monitor.Gfx_UseRlcBusy = input[2]; - activity_monitor.Gfx_MinActiveFreqType = input[3]; - activity_monitor.Gfx_MinActiveFreq = input[4]; - activity_monitor.Gfx_BoosterFreqType = input[5]; - activity_monitor.Gfx_BoosterFreq = input[6]; - activity_monitor.Gfx_PD_Data_limit_c = input[7]; - activity_monitor.Gfx_PD_Data_error_coeff = input[8]; - activity_monitor.Gfx_PD_Data_error_rate_coeff = input[9]; - break; - case 1: /* Uclk */ - activity_monitor.Mem_FPS = input[1]; - activity_monitor.Mem_UseRlcBusy = input[2]; - activity_monitor.Mem_MinActiveFreqType = input[3]; - activity_monitor.Mem_MinActiveFreq = input[4]; - activity_monitor.Mem_BoosterFreqType = input[5]; - activity_monitor.Mem_BoosterFreq = input[6]; - activity_monitor.Mem_PD_Data_limit_c = input[7]; - activity_monitor.Mem_PD_Data_error_coeff = input[8]; - activity_monitor.Mem_PD_Data_error_rate_coeff = input[9]; - break; - default: +static int arcturus_set_power_profile_mode(struct smu_context *smu, + u32 workload_mask, + long *custom_params, + u32 custom_params_max_idx) +{ + u32 backend_workload_mask = 0; + int ret, idx = -1, i; + + smu_cmn_get_backend_workload_mask(smu, workload_mask, + &backend_workload_mask); + + if (workload_mask & (1 << PP_SMC_POWER_PROFILE_CUSTOM)) { + if (smu->smc_fw_version < 0x360d00) return -EINVAL; + if (!smu->custom_profile_params) { + smu->custom_profile_params = + kzalloc(ARCTURUS_CUSTOM_PARAMS_SIZE, GFP_KERNEL); + if (!smu->custom_profile_params) + return -ENOMEM; } - - ret = smu_cmn_update_table(smu, - SMU_TABLE_ACTIVITY_MONITOR_COEFF, - WORKLOAD_PPLIB_CUSTOM_BIT, - (void *)(&activity_monitor), - true); + if (custom_params && custom_params_max_idx) { + if (custom_params_max_idx != ARCTURUS_CUSTOM_PARAMS_COUNT) + return -EINVAL; + if (custom_params[0] >= ARCTURUS_CUSTOM_PARAMS_CLOCK_COUNT) + return -EINVAL; + idx = custom_params[0] * ARCTURUS_CUSTOM_PARAMS_COUNT; + smu->custom_profile_params[idx] = 1; + for (i = 1; i < custom_params_max_idx; i++) + smu->custom_profile_params[idx + i] = custom_params[i]; + } + ret = arcturus_set_power_profile_mode_coeff(smu, + smu->custom_profile_params); if (ret) { - dev_err(smu->adev->dev, "[%s] Failed to set activity monitor!", __func__); + if (idx != -1) + smu->custom_profile_params[idx] = 0; return ret; } - } - - /* - * Conv PP_SMC_POWER_PROFILE* to WORKLOAD_PPLIB_*_BIT - * Not all profile modes are supported on arcturus. - */ - workload_type = smu_cmn_to_asic_specific_index(smu, - CMN2ASIC_MAPPING_WORKLOAD, - profile_mode); - if (workload_type < 0) { - dev_dbg(smu->adev->dev, "Unsupported power profile mode %d on arcturus\n", profile_mode); - return -EINVAL; + } else if (smu->custom_profile_params) { + memset(smu->custom_profile_params, 0, ARCTURUS_CUSTOM_PARAMS_SIZE); } ret = smu_cmn_send_smc_msg_with_param(smu, - SMU_MSG_SetWorkloadMask, - 1 << workload_type, - NULL); + SMU_MSG_SetWorkloadMask, + backend_workload_mask, + NULL); if (ret) { - dev_err(smu->adev->dev, "Fail to set workload type %d\n", workload_type); + dev_err(smu->adev->dev, "Failed to set workload mask 0x%08x\n", + workload_mask); + if (idx != -1) + smu->custom_profile_params[idx] = 0; return ret; } - smu->power_profile_mode = profile_mode; - - return 0; + return ret; } static int arcturus_set_performance_level(struct smu_context *smu, 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 faa8e7d9c3c6..7fad5dfb39c4 100644 --- a/drivers/gpu/drm/amd/pm/swsmu/smu11/navi10_ppt.c +++ b/drivers/gpu/drm/amd/pm/swsmu/smu11/navi10_ppt.c @@ -2006,87 +2006,122 @@ static int navi10_get_power_profile_mode(struct smu_context *smu, char *buf) return size; } -static int navi10_set_power_profile_mode(struct smu_context *smu, long *input, uint32_t size) +#define NAVI10_CUSTOM_PARAMS_COUNT 10 +#define NAVI10_CUSTOM_PARAMS_CLOCKS_COUNT 3 +#define NAVI10_CUSTOM_PARAMS_SIZE (NAVI10_CUSTOM_PARAMS_CLOCKS_COUNT * NAVI10_CUSTOM_PARAMS_COUNT * sizeof(long)) + +static int navi10_set_power_profile_mode_coeff(struct smu_context *smu, + long *input) { DpmActivityMonitorCoeffInt_t activity_monitor; - int workload_type, ret = 0; + int ret, idx; - smu->power_profile_mode = input[size]; + ret = smu_cmn_update_table(smu, + SMU_TABLE_ACTIVITY_MONITOR_COEFF, WORKLOAD_PPLIB_CUSTOM_BIT, + (void *)(&activity_monitor), false); + if (ret) { + dev_err(smu->adev->dev, "[%s] Failed to get activity monitor!", __func__); + return ret; + } - if (smu->power_profile_mode > PP_SMC_POWER_PROFILE_CUSTOM) { - dev_err(smu->adev->dev, "Invalid power profile mode %d\n", smu->power_profile_mode); - return -EINVAL; + idx = 0 * NAVI10_CUSTOM_PARAMS_COUNT; + if (input[idx]) { + /* Gfxclk */ + activity_monitor.Gfx_FPS = input[idx + 1]; + activity_monitor.Gfx_MinFreqStep = input[idx + 2]; + activity_monitor.Gfx_MinActiveFreqType = input[idx + 3]; + activity_monitor.Gfx_MinActiveFreq = input[idx + 4]; + activity_monitor.Gfx_BoosterFreqType = input[idx + 5]; + activity_monitor.Gfx_BoosterFreq = input[idx + 6]; + activity_monitor.Gfx_PD_Data_limit_c = input[idx + 7]; + activity_monitor.Gfx_PD_Data_error_coeff = input[idx + 8]; + activity_monitor.Gfx_PD_Data_error_rate_coeff = input[idx + 9]; + } + idx = 1 * NAVI10_CUSTOM_PARAMS_COUNT; + if (input[idx]) { + /* Socclk */ + activity_monitor.Soc_FPS = input[idx + 1]; + activity_monitor.Soc_MinFreqStep = input[idx + 2]; + activity_monitor.Soc_MinActiveFreqType = input[idx + 3]; + activity_monitor.Soc_MinActiveFreq = input[idx + 4]; + activity_monitor.Soc_BoosterFreqType = input[idx + 5]; + activity_monitor.Soc_BoosterFreq = input[idx + 6]; + activity_monitor.Soc_PD_Data_limit_c = input[idx + 7]; + activity_monitor.Soc_PD_Data_error_coeff = input[idx + 8]; + activity_monitor.Soc_PD_Data_error_rate_coeff = input[idx + 9]; + } + idx = 2 * NAVI10_CUSTOM_PARAMS_COUNT; + if (input[idx]) { + /* Memclk */ + activity_monitor.Mem_FPS = input[idx + 1]; + activity_monitor.Mem_MinFreqStep = input[idx + 2]; + activity_monitor.Mem_MinActiveFreqType = input[idx + 3]; + activity_monitor.Mem_MinActiveFreq = input[idx + 4]; + activity_monitor.Mem_BoosterFreqType = input[idx + 5]; + activity_monitor.Mem_BoosterFreq = input[idx + 6]; + activity_monitor.Mem_PD_Data_limit_c = input[idx + 7]; + activity_monitor.Mem_PD_Data_error_coeff = input[idx + 8]; + activity_monitor.Mem_PD_Data_error_rate_coeff = input[idx + 9]; + } + + ret = smu_cmn_update_table(smu, + SMU_TABLE_ACTIVITY_MONITOR_COEFF, WORKLOAD_PPLIB_CUSTOM_BIT, + (void *)(&activity_monitor), true); + if (ret) { + dev_err(smu->adev->dev, "[%s] Failed to set activity monitor!", __func__); + return ret; } - if (smu->power_profile_mode == PP_SMC_POWER_PROFILE_CUSTOM) { - if (size != 10) - return -EINVAL; + return ret; +} - ret = smu_cmn_update_table(smu, - SMU_TABLE_ACTIVITY_MONITOR_COEFF, WORKLOAD_PPLIB_CUSTOM_BIT, - (void *)(&activity_monitor), false); - if (ret) { - dev_err(smu->adev->dev, "[%s] Failed to get activity monitor!", __func__); - return ret; - } +static int navi10_set_power_profile_mode(struct smu_context *smu, + u32 workload_mask, + long *custom_params, + u32 custom_params_max_idx) +{ + u32 backend_workload_mask = 0; + int ret, idx = -1, i; - switch (input[0]) { - case 0: /* Gfxclk */ - activity_monitor.Gfx_FPS = input[1]; - activity_monitor.Gfx_MinFreqStep = input[2]; - activity_monitor.Gfx_MinActiveFreqType = input[3]; - activity_monitor.Gfx_MinActiveFreq = input[4]; - activity_monitor.Gfx_BoosterFreqType = input[5]; - activity_monitor.Gfx_BoosterFreq = input[6]; - activity_monitor.Gfx_PD_Data_limit_c = input[7]; - activity_monitor.Gfx_PD_Data_error_coeff = input[8]; - activity_monitor.Gfx_PD_Data_error_rate_coeff = input[9]; - break; - case 1: /* Socclk */ - activity_monitor.Soc_FPS = input[1]; - activity_monitor.Soc_MinFreqStep = input[2]; - activity_monitor.Soc_MinActiveFreqType = input[3]; - activity_monitor.Soc_MinActiveFreq = input[4]; - activity_monitor.Soc_BoosterFreqType = input[5]; - activity_monitor.Soc_BoosterFreq = input[6]; - activity_monitor.Soc_PD_Data_limit_c = input[7]; - activity_monitor.Soc_PD_Data_error_coeff = input[8]; - activity_monitor.Soc_PD_Data_error_rate_coeff = input[9]; - break; - case 2: /* Memclk */ - activity_monitor.Mem_FPS = input[1]; - activity_monitor.Mem_MinFreqStep = input[2]; - activity_monitor.Mem_MinActiveFreqType = input[3]; - activity_monitor.Mem_MinActiveFreq = input[4]; - activity_monitor.Mem_BoosterFreqType = input[5]; - activity_monitor.Mem_BoosterFreq = input[6]; - activity_monitor.Mem_PD_Data_limit_c = input[7]; - activity_monitor.Mem_PD_Data_error_coeff = input[8]; - activity_monitor.Mem_PD_Data_error_rate_coeff = input[9]; - break; - default: - return -EINVAL; - } + smu_cmn_get_backend_workload_mask(smu, workload_mask, + &backend_workload_mask); - ret = smu_cmn_update_table(smu, - SMU_TABLE_ACTIVITY_MONITOR_COEFF, WORKLOAD_PPLIB_CUSTOM_BIT, - (void *)(&activity_monitor), true); + if (workload_mask & (1 << PP_SMC_POWER_PROFILE_CUSTOM)) { + if (!smu->custom_profile_params) { + smu->custom_profile_params = kzalloc(NAVI10_CUSTOM_PARAMS_SIZE, GFP_KERNEL); + if (!smu->custom_profile_params) + return -ENOMEM; + } + if (custom_params && custom_params_max_idx) { + if (custom_params_max_idx != NAVI10_CUSTOM_PARAMS_COUNT) + return -EINVAL; + if (custom_params[0] >= NAVI10_CUSTOM_PARAMS_CLOCKS_COUNT) + return -EINVAL; + idx = custom_params[0] * NAVI10_CUSTOM_PARAMS_COUNT; + smu->custom_profile_params[idx] = 1; + for (i = 1; i < custom_params_max_idx; i++) + smu->custom_profile_params[idx + i] = custom_params[i]; + } + ret = navi10_set_power_profile_mode_coeff(smu, + smu->custom_profile_params); if (ret) { - dev_err(smu->adev->dev, "[%s] Failed to set activity monitor!", __func__); + if (idx != -1) + smu->custom_profile_params[idx] = 0; return ret; } + } else if (smu->custom_profile_params) { + memset(smu->custom_profile_params, 0, NAVI10_CUSTOM_PARAMS_SIZE); } - /* 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; ret = smu_cmn_send_smc_msg_with_param(smu, SMU_MSG_SetWorkloadMask, - 1 << workload_type, NULL); - if (ret) - dev_err(smu->adev->dev, "[%s] Failed to set work load mask!", __func__); + backend_workload_mask, NULL); + if (ret) { + dev_err(smu->adev->dev, "Failed to set workload mask 0x%08x\n", + workload_mask); + if (idx != -1) + smu->custom_profile_params[idx] = 0; + 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 a9cb28ce2133..286777ada1df 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 @@ -1708,90 +1708,126 @@ static int sienna_cichlid_get_power_profile_mode(struct smu_context *smu, char * return size; } -static int sienna_cichlid_set_power_profile_mode(struct smu_context *smu, long *input, uint32_t size) +#define SIENNA_CICHLID_CUSTOM_PARAMS_COUNT 10 +#define SIENNA_CICHLID_CUSTOM_PARAMS_CLOCK_COUNT 3 +#define SIENNA_CICHLID_CUSTOM_PARAMS_SIZE (SIENNA_CICHLID_CUSTOM_PARAMS_CLOCK_COUNT * SIENNA_CICHLID_CUSTOM_PARAMS_COUNT * sizeof(long)) + +static int sienna_cichlid_set_power_profile_mode_coeff(struct smu_context *smu, + long *input) { DpmActivityMonitorCoeffIntExternal_t activity_monitor_external; DpmActivityMonitorCoeffInt_t *activity_monitor = &(activity_monitor_external.DpmActivityMonitorCoeffInt); - int workload_type, ret = 0; + int ret, idx; - smu->power_profile_mode = input[size]; + ret = smu_cmn_update_table(smu, + SMU_TABLE_ACTIVITY_MONITOR_COEFF, WORKLOAD_PPLIB_CUSTOM_BIT, + (void *)(&activity_monitor_external), false); + if (ret) { + dev_err(smu->adev->dev, "[%s] Failed to get activity monitor!", __func__); + return ret; + } - if (smu->power_profile_mode > PP_SMC_POWER_PROFILE_CUSTOM) { - dev_err(smu->adev->dev, "Invalid power profile mode %d\n", smu->power_profile_mode); - return -EINVAL; + idx = 0 * SIENNA_CICHLID_CUSTOM_PARAMS_COUNT; + if (input[idx]) { + /* Gfxclk */ + activity_monitor->Gfx_FPS = input[idx + 1]; + activity_monitor->Gfx_MinFreqStep = input[idx + 2]; + activity_monitor->Gfx_MinActiveFreqType = input[idx + 3]; + activity_monitor->Gfx_MinActiveFreq = input[idx + 4]; + activity_monitor->Gfx_BoosterFreqType = input[idx + 5]; + activity_monitor->Gfx_BoosterFreq = input[idx + 6]; + activity_monitor->Gfx_PD_Data_limit_c = input[idx + 7]; + activity_monitor->Gfx_PD_Data_error_coeff = input[idx + 8]; + activity_monitor->Gfx_PD_Data_error_rate_coeff = input[idx + 9]; + } + idx = 1 * SIENNA_CICHLID_CUSTOM_PARAMS_COUNT; + if (input[idx]) { + /* Socclk */ + activity_monitor->Fclk_FPS = input[idx + 1]; + activity_monitor->Fclk_MinFreqStep = input[idx + 2]; + activity_monitor->Fclk_MinActiveFreqType = input[idx + 3]; + activity_monitor->Fclk_MinActiveFreq = input[idx + 4]; + activity_monitor->Fclk_BoosterFreqType = input[idx + 5]; + activity_monitor->Fclk_BoosterFreq = input[idx + 6]; + activity_monitor->Fclk_PD_Data_limit_c = input[idx + 7]; + activity_monitor->Fclk_PD_Data_error_coeff = input[idx + 8]; + activity_monitor->Fclk_PD_Data_error_rate_coeff = input[idx + 9]; + } + idx = 2 * SIENNA_CICHLID_CUSTOM_PARAMS_COUNT; + if (input[idx]) { + /* Memclk */ + activity_monitor->Mem_FPS = input[idx + 1]; + activity_monitor->Mem_MinFreqStep = input[idx + 2]; + activity_monitor->Mem_MinActiveFreqType = input[idx + 3]; + activity_monitor->Mem_MinActiveFreq = input[idx + 4]; + activity_monitor->Mem_BoosterFreqType = input[idx + 5]; + activity_monitor->Mem_BoosterFreq = input[idx + 6]; + activity_monitor->Mem_PD_Data_limit_c = input[idx + 7]; + activity_monitor->Mem_PD_Data_error_coeff = input[idx + 8]; + activity_monitor->Mem_PD_Data_error_rate_coeff = input[idx + 9]; } - if (smu->power_profile_mode == PP_SMC_POWER_PROFILE_CUSTOM) { - if (size != 10) - return -EINVAL; + 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; + } - ret = smu_cmn_update_table(smu, - SMU_TABLE_ACTIVITY_MONITOR_COEFF, WORKLOAD_PPLIB_CUSTOM_BIT, - (void *)(&activity_monitor_external), false); - if (ret) { - dev_err(smu->adev->dev, "[%s] Failed to get activity monitor!", __func__); - return ret; - } + return ret; +} - switch (input[0]) { - case 0: /* Gfxclk */ - activity_monitor->Gfx_FPS = input[1]; - activity_monitor->Gfx_MinFreqStep = input[2]; - activity_monitor->Gfx_MinActiveFreqType = input[3]; - activity_monitor->Gfx_MinActiveFreq = input[4]; - activity_monitor->Gfx_BoosterFreqType = input[5]; - activity_monitor->Gfx_BoosterFreq = input[6]; - activity_monitor->Gfx_PD_Data_limit_c = input[7]; - activity_monitor->Gfx_PD_Data_error_coeff = input[8]; - activity_monitor->Gfx_PD_Data_error_rate_coeff = input[9]; - break; - case 1: /* Socclk */ - activity_monitor->Fclk_FPS = input[1]; - activity_monitor->Fclk_MinFreqStep = input[2]; - activity_monitor->Fclk_MinActiveFreqType = input[3]; - activity_monitor->Fclk_MinActiveFreq = input[4]; - activity_monitor->Fclk_BoosterFreqType = input[5]; - activity_monitor->Fclk_BoosterFreq = input[6]; - activity_monitor->Fclk_PD_Data_limit_c = input[7]; - activity_monitor->Fclk_PD_Data_error_coeff = input[8]; - activity_monitor->Fclk_PD_Data_error_rate_coeff = input[9]; - break; - case 2: /* Memclk */ - activity_monitor->Mem_FPS = input[1]; - activity_monitor->Mem_MinFreqStep = input[2]; - activity_monitor->Mem_MinActiveFreqType = input[3]; - activity_monitor->Mem_MinActiveFreq = input[4]; - activity_monitor->Mem_BoosterFreqType = input[5]; - activity_monitor->Mem_BoosterFreq = input[6]; - activity_monitor->Mem_PD_Data_limit_c = input[7]; - activity_monitor->Mem_PD_Data_error_coeff = input[8]; - activity_monitor->Mem_PD_Data_error_rate_coeff = input[9]; - break; - default: - return -EINVAL; - } +static int sienna_cichlid_set_power_profile_mode(struct smu_context *smu, + u32 workload_mask, + long *custom_params, + u32 custom_params_max_idx) +{ + u32 backend_workload_mask = 0; + int ret, idx = -1, i; - ret = smu_cmn_update_table(smu, - SMU_TABLE_ACTIVITY_MONITOR_COEFF, WORKLOAD_PPLIB_CUSTOM_BIT, - (void *)(&activity_monitor_external), true); + smu_cmn_get_backend_workload_mask(smu, workload_mask, + &backend_workload_mask); + + if (workload_mask & (1 << PP_SMC_POWER_PROFILE_CUSTOM)) { + if (!smu->custom_profile_params) { + smu->custom_profile_params = + kzalloc(SIENNA_CICHLID_CUSTOM_PARAMS_SIZE, GFP_KERNEL); + if (!smu->custom_profile_params) + return -ENOMEM; + } + if (custom_params && custom_params_max_idx) { + if (custom_params_max_idx != SIENNA_CICHLID_CUSTOM_PARAMS_COUNT) + return -EINVAL; + if (custom_params[0] >= SIENNA_CICHLID_CUSTOM_PARAMS_CLOCK_COUNT) + return -EINVAL; + idx = custom_params[0] * SIENNA_CICHLID_CUSTOM_PARAMS_COUNT; + smu->custom_profile_params[idx] = 1; + for (i = 1; i < custom_params_max_idx; i++) + smu->custom_profile_params[idx + i] = custom_params[i]; + } + ret = sienna_cichlid_set_power_profile_mode_coeff(smu, + smu->custom_profile_params); if (ret) { - dev_err(smu->adev->dev, "[%s] Failed to set activity monitor!", __func__); + if (idx != -1) + smu->custom_profile_params[idx] = 0; return ret; } + } else if (smu->custom_profile_params) { + memset(smu->custom_profile_params, 0, SIENNA_CICHLID_CUSTOM_PARAMS_SIZE); } - /* 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; ret = smu_cmn_send_smc_msg_with_param(smu, SMU_MSG_SetWorkloadMask, - 1 << workload_type, NULL); - if (ret) - dev_err(smu->adev->dev, "[%s] Failed to set work load mask!", __func__); + backend_workload_mask, NULL); + if (ret) { + dev_err(smu->adev->dev, "Failed to set workload mask 0x%08x\n", + workload_mask); + if (idx != -1) + smu->custom_profile_params[idx] = 0; + return ret; + } return ret; } 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 cd3e9ba3eff4..a55ea76d7399 100644 --- a/drivers/gpu/drm/amd/pm/swsmu/smu11/vangogh_ppt.c +++ b/drivers/gpu/drm/amd/pm/swsmu/smu11/vangogh_ppt.c @@ -1056,42 +1056,27 @@ static int vangogh_get_power_profile_mode(struct smu_context *smu, return size; } -static int vangogh_set_power_profile_mode(struct smu_context *smu, long *input, uint32_t size) +static int vangogh_set_power_profile_mode(struct smu_context *smu, + u32 workload_mask, + long *custom_params, + u32 custom_params_max_idx) { - int workload_type, ret; - uint32_t profile_mode = input[size]; + u32 backend_workload_mask = 0; + int ret; - if (profile_mode >= PP_SMC_POWER_PROFILE_COUNT) { - dev_err(smu->adev->dev, "Invalid power profile mode %d\n", profile_mode); - return -EINVAL; - } - - if (profile_mode == PP_SMC_POWER_PROFILE_BOOTUP_DEFAULT || - profile_mode == PP_SMC_POWER_PROFILE_POWERSAVING) - return 0; - - /* conv PP_SMC_POWER_PROFILE* to WORKLOAD_PPLIB_*_BIT */ - workload_type = smu_cmn_to_asic_specific_index(smu, - CMN2ASIC_MAPPING_WORKLOAD, - profile_mode); - if (workload_type < 0) { - dev_dbg(smu->adev->dev, "Unsupported power profile mode %d on VANGOGH\n", - profile_mode); - return -EINVAL; - } + smu_cmn_get_backend_workload_mask(smu, workload_mask, + &backend_workload_mask); ret = smu_cmn_send_smc_msg_with_param(smu, SMU_MSG_ActiveProcessNotify, - 1 << workload_type, - NULL); + backend_workload_mask, + NULL); if (ret) { - dev_err_once(smu->adev->dev, "Fail to set workload type %d\n", - workload_type); + dev_err_once(smu->adev->dev, "Fail to set workload mask 0x%08x\n", + workload_mask); return ret; } - smu->power_profile_mode = profile_mode; - - return 0; + return ret; } static int vangogh_set_soft_freq_limited_range(struct smu_context *smu, 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 a34797f3576b..37d82a71a2d7 100644 --- a/drivers/gpu/drm/amd/pm/swsmu/smu12/renoir_ppt.c +++ b/drivers/gpu/drm/amd/pm/swsmu/smu12/renoir_ppt.c @@ -864,44 +864,27 @@ static int renoir_force_clk_levels(struct smu_context *smu, return ret; } -static int renoir_set_power_profile_mode(struct smu_context *smu, long *input, uint32_t size) +static int renoir_set_power_profile_mode(struct smu_context *smu, + u32 workload_mask, + long *custom_params, + u32 custom_params_max_idx) { - int workload_type, ret; - uint32_t profile_mode = input[size]; + int ret; + u32 backend_workload_mask = 0; - if (profile_mode > PP_SMC_POWER_PROFILE_CUSTOM) { - dev_err(smu->adev->dev, "Invalid power profile mode %d\n", profile_mode); - return -EINVAL; - } - - if (profile_mode == PP_SMC_POWER_PROFILE_BOOTUP_DEFAULT || - profile_mode == PP_SMC_POWER_PROFILE_POWERSAVING) - return 0; - - /* conv PP_SMC_POWER_PROFILE* to WORKLOAD_PPLIB_*_BIT */ - workload_type = smu_cmn_to_asic_specific_index(smu, - CMN2ASIC_MAPPING_WORKLOAD, - profile_mode); - if (workload_type < 0) { - /* - * TODO: If some case need switch to powersave/default power mode - * then can consider enter WORKLOAD_COMPUTE/WORKLOAD_CUSTOM for power saving. - */ - dev_dbg(smu->adev->dev, "Unsupported power profile mode %d on RENOIR\n", profile_mode); - return -EINVAL; - } + smu_cmn_get_backend_workload_mask(smu, workload_mask, + &backend_workload_mask); ret = smu_cmn_send_smc_msg_with_param(smu, SMU_MSG_ActiveProcessNotify, - 1 << workload_type, - NULL); + backend_workload_mask, + NULL); if (ret) { - dev_err_once(smu->adev->dev, "Fail to set workload type %d\n", workload_type); + dev_err_once(smu->adev->dev, "Failed to set workload mask 0x08%x\n", + workload_mask); return ret; } - smu->power_profile_mode = profile_mode; - - return 0; + return ret; } static int renoir_set_peak_clock_by_device(struct smu_context *smu) 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 199bdd9720d3..3aa705aae4c0 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 @@ -2571,82 +2571,76 @@ static int smu_v13_0_0_get_power_profile_mode(struct smu_context *smu, return size; } -static int smu_v13_0_0_set_power_profile_mode(struct smu_context *smu, - long *input, - uint32_t size) +#define SMU_13_0_0_CUSTOM_PARAMS_COUNT 9 +#define SMU_13_0_0_CUSTOM_PARAMS_CLOCK_COUNT 2 +#define SMU_13_0_0_CUSTOM_PARAMS_SIZE (SMU_13_0_0_CUSTOM_PARAMS_CLOCK_COUNT * SMU_13_0_0_CUSTOM_PARAMS_COUNT * sizeof(long)) + +static int smu_v13_0_0_set_power_profile_mode_coeff(struct smu_context *smu, + long *input) { DpmActivityMonitorCoeffIntExternal_t activity_monitor_external; DpmActivityMonitorCoeffInt_t *activity_monitor = &(activity_monitor_external.DpmActivityMonitorCoeffInt); - int workload_type, ret = 0; - u32 workload_mask, selected_workload_mask; - - smu->power_profile_mode = input[size]; + int ret, idx; - if (smu->power_profile_mode >= PP_SMC_POWER_PROFILE_COUNT) { - dev_err(smu->adev->dev, "Invalid power profile mode %d\n", smu->power_profile_mode); - return -EINVAL; + ret = smu_cmn_update_table(smu, + SMU_TABLE_ACTIVITY_MONITOR_COEFF, + WORKLOAD_PPLIB_CUSTOM_BIT, + (void *)(&activity_monitor_external), + false); + if (ret) { + dev_err(smu->adev->dev, "[%s] Failed to get activity monitor!", __func__); + return ret; } - if (smu->power_profile_mode == PP_SMC_POWER_PROFILE_CUSTOM) { - if (size != 9) - return -EINVAL; - - ret = smu_cmn_update_table(smu, - SMU_TABLE_ACTIVITY_MONITOR_COEFF, - WORKLOAD_PPLIB_CUSTOM_BIT, - (void *)(&activity_monitor_external), - false); - if (ret) { - dev_err(smu->adev->dev, "[%s] Failed to get activity monitor!", __func__); - return ret; - } - - switch (input[0]) { - case 0: /* Gfxclk */ - activity_monitor->Gfx_FPS = input[1]; - activity_monitor->Gfx_MinActiveFreqType = input[2]; - activity_monitor->Gfx_MinActiveFreq = input[3]; - activity_monitor->Gfx_BoosterFreqType = input[4]; - activity_monitor->Gfx_BoosterFreq = input[5]; - activity_monitor->Gfx_PD_Data_limit_c = input[6]; - activity_monitor->Gfx_PD_Data_error_coeff = input[7]; - activity_monitor->Gfx_PD_Data_error_rate_coeff = input[8]; - break; - case 1: /* Fclk */ - activity_monitor->Fclk_FPS = input[1]; - activity_monitor->Fclk_MinActiveFreqType = input[2]; - activity_monitor->Fclk_MinActiveFreq = input[3]; - activity_monitor->Fclk_BoosterFreqType = input[4]; - activity_monitor->Fclk_BoosterFreq = input[5]; - activity_monitor->Fclk_PD_Data_limit_c = input[6]; - activity_monitor->Fclk_PD_Data_error_coeff = input[7]; - activity_monitor->Fclk_PD_Data_error_rate_coeff = input[8]; - break; - default: - return -EINVAL; - } + idx = 0 * SMU_13_0_0_CUSTOM_PARAMS_COUNT; + if (input[idx]) { + /* Gfxclk */ + activity_monitor->Gfx_FPS = input[idx + 1]; + activity_monitor->Gfx_MinActiveFreqType = input[idx + 2]; + activity_monitor->Gfx_MinActiveFreq = input[idx + 3]; + activity_monitor->Gfx_BoosterFreqType = input[idx + 4]; + activity_monitor->Gfx_BoosterFreq = input[idx + 5]; + activity_monitor->Gfx_PD_Data_limit_c = input[idx + 6]; + activity_monitor->Gfx_PD_Data_error_coeff = input[idx + 7]; + activity_monitor->Gfx_PD_Data_error_rate_coeff = input[idx + 8]; + } + idx = 1 * SMU_13_0_0_CUSTOM_PARAMS_COUNT; + if (input[idx]) { + /* Fclk */ + activity_monitor->Fclk_FPS = input[idx + 1]; + activity_monitor->Fclk_MinActiveFreqType = input[idx + 2]; + activity_monitor->Fclk_MinActiveFreq = input[idx + 3]; + activity_monitor->Fclk_BoosterFreqType = input[idx + 4]; + activity_monitor->Fclk_BoosterFreq = input[idx + 5]; + activity_monitor->Fclk_PD_Data_limit_c = input[idx + 6]; + activity_monitor->Fclk_PD_Data_error_coeff = input[idx + 7]; + activity_monitor->Fclk_PD_Data_error_rate_coeff = input[idx + 8]; + } - 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; - } + 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; } - /* 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); + return ret; +} - if (workload_type < 0) - return -EINVAL; +static int smu_v13_0_0_set_power_profile_mode(struct smu_context *smu, + u32 workload_mask, + long *custom_params, + u32 custom_params_max_idx) +{ + u32 backend_workload_mask = 0; + int workload_type, ret, idx = -1, i; - selected_workload_mask = workload_mask = 1 << workload_type; + smu_cmn_get_backend_workload_mask(smu, workload_mask, + &backend_workload_mask); /* Add optimizations for SMU13.0.0/10. Reuse the power saving profile */ if ((amdgpu_ip_version(smu->adev, MP1_HWIP, 0) == IP_VERSION(13, 0, 0) && @@ -2658,15 +2652,48 @@ static int smu_v13_0_0_set_power_profile_mode(struct smu_context *smu, CMN2ASIC_MAPPING_WORKLOAD, PP_SMC_POWER_PROFILE_POWERSAVING); if (workload_type >= 0) - workload_mask |= 1 << workload_type; + backend_workload_mask |= 1 << workload_type; + } + + if (workload_mask & (1 << PP_SMC_POWER_PROFILE_CUSTOM)) { + if (!smu->custom_profile_params) { + smu->custom_profile_params = + kzalloc(SMU_13_0_0_CUSTOM_PARAMS_SIZE, GFP_KERNEL); + if (!smu->custom_profile_params) + return -ENOMEM; + } + if (custom_params && custom_params_max_idx) { + if (custom_params_max_idx != SMU_13_0_0_CUSTOM_PARAMS_COUNT) + return -EINVAL; + if (custom_params[0] >= SMU_13_0_0_CUSTOM_PARAMS_CLOCK_COUNT) + return -EINVAL; + idx = custom_params[0] * SMU_13_0_0_CUSTOM_PARAMS_COUNT; + smu->custom_profile_params[idx] = 1; + for (i = 1; i < custom_params_max_idx; i++) + smu->custom_profile_params[idx + i] = custom_params[i]; + } + ret = smu_v13_0_0_set_power_profile_mode_coeff(smu, + smu->custom_profile_params); + if (ret) { + if (idx != -1) + smu->custom_profile_params[idx] = 0; + return ret; + } + } else if (smu->custom_profile_params) { + memset(smu->custom_profile_params, 0, SMU_13_0_0_CUSTOM_PARAMS_SIZE); } ret = smu_cmn_send_smc_msg_with_param(smu, - SMU_MSG_SetWorkloadMask, - workload_mask, - NULL); - if (!ret) - smu->workload_mask = selected_workload_mask; + SMU_MSG_SetWorkloadMask, + backend_workload_mask, + NULL); + if (ret) { + dev_err(smu->adev->dev, "Failed to set workload mask 0x%08x\n", + workload_mask); + if (idx != -1) + smu->custom_profile_params[idx] = 0; + return ret; + } return ret; } 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 34c1e0c7e1e4..f4ac403b8b36 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 @@ -2530,78 +2530,110 @@ out: return result; } -static int smu_v13_0_7_set_power_profile_mode(struct smu_context *smu, long *input, uint32_t size) +#define SMU_13_0_7_CUSTOM_PARAMS_COUNT 8 +#define SMU_13_0_7_CUSTOM_PARAMS_CLOCK_COUNT 2 +#define SMU_13_0_7_CUSTOM_PARAMS_SIZE (SMU_13_0_7_CUSTOM_PARAMS_CLOCK_COUNT * SMU_13_0_7_CUSTOM_PARAMS_COUNT * sizeof(long)) + +static int smu_v13_0_7_set_power_profile_mode_coeff(struct smu_context *smu, + long *input) { DpmActivityMonitorCoeffIntExternal_t activity_monitor_external; DpmActivityMonitorCoeffInt_t *activity_monitor = &(activity_monitor_external.DpmActivityMonitorCoeffInt); - int workload_type, ret = 0; + int ret, idx; - smu->power_profile_mode = input[size]; + ret = smu_cmn_update_table(smu, + SMU_TABLE_ACTIVITY_MONITOR_COEFF, WORKLOAD_PPLIB_CUSTOM_BIT, + (void *)(&activity_monitor_external), false); + if (ret) { + dev_err(smu->adev->dev, "[%s] Failed to get activity monitor!", __func__); + return ret; + } - if (smu->power_profile_mode > PP_SMC_POWER_PROFILE_WINDOW3D) { - dev_err(smu->adev->dev, "Invalid power profile mode %d\n", smu->power_profile_mode); - return -EINVAL; + idx = 0 * SMU_13_0_7_CUSTOM_PARAMS_COUNT; + if (input[idx]) { + /* Gfxclk */ + activity_monitor->Gfx_ActiveHystLimit = input[idx + 1]; + activity_monitor->Gfx_IdleHystLimit = input[idx + 2]; + activity_monitor->Gfx_FPS = input[idx + 3]; + activity_monitor->Gfx_MinActiveFreqType = input[idx + 4]; + activity_monitor->Gfx_BoosterFreqType = input[idx + 5]; + activity_monitor->Gfx_MinActiveFreq = input[idx + 6]; + activity_monitor->Gfx_BoosterFreq = input[idx + 7]; + } + idx = 1 * SMU_13_0_7_CUSTOM_PARAMS_COUNT; + if (input[idx]) { + /* Fclk */ + activity_monitor->Fclk_ActiveHystLimit = input[idx + 1]; + activity_monitor->Fclk_IdleHystLimit = input[idx + 2]; + activity_monitor->Fclk_FPS = input[idx + 3]; + activity_monitor->Fclk_MinActiveFreqType = input[idx + 4]; + activity_monitor->Fclk_BoosterFreqType = input[idx + 5]; + activity_monitor->Fclk_MinActiveFreq = input[idx + 6]; + activity_monitor->Fclk_BoosterFreq = input[idx + 7]; } - if (smu->power_profile_mode == PP_SMC_POWER_PROFILE_CUSTOM) { - if (size != 8) - return -EINVAL; + 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; + } - ret = smu_cmn_update_table(smu, - SMU_TABLE_ACTIVITY_MONITOR_COEFF, WORKLOAD_PPLIB_CUSTOM_BIT, - (void *)(&activity_monitor_external), false); - if (ret) { - dev_err(smu->adev->dev, "[%s] Failed to get activity monitor!", __func__); - return ret; - } + return ret; +} - switch (input[0]) { - case 0: /* Gfxclk */ - activity_monitor->Gfx_ActiveHystLimit = input[1]; - activity_monitor->Gfx_IdleHystLimit = input[2]; - activity_monitor->Gfx_FPS = input[3]; - activity_monitor->Gfx_MinActiveFreqType = input[4]; - activity_monitor->Gfx_BoosterFreqType = input[5]; - activity_monitor->Gfx_MinActiveFreq = input[6]; - activity_monitor->Gfx_BoosterFreq = input[7]; - break; - case 1: /* Fclk */ - activity_monitor->Fclk_ActiveHystLimit = input[1]; - activity_monitor->Fclk_IdleHystLimit = input[2]; - activity_monitor->Fclk_FPS = input[3]; - activity_monitor->Fclk_MinActiveFreqType = input[4]; - activity_monitor->Fclk_BoosterFreqType = input[5]; - activity_monitor->Fclk_MinActiveFreq = input[6]; - activity_monitor->Fclk_BoosterFreq = input[7]; - break; - default: - return -EINVAL; +static int smu_v13_0_7_set_power_profile_mode(struct smu_context *smu, + u32 workload_mask, + long *custom_params, + u32 custom_params_max_idx) +{ + u32 backend_workload_mask = 0; + int ret, idx = -1, i; + + smu_cmn_get_backend_workload_mask(smu, workload_mask, + &backend_workload_mask); + + if (workload_mask & (1 << PP_SMC_POWER_PROFILE_CUSTOM)) { + if (!smu->custom_profile_params) { + smu->custom_profile_params = + kzalloc(SMU_13_0_7_CUSTOM_PARAMS_SIZE, GFP_KERNEL); + if (!smu->custom_profile_params) + return -ENOMEM; } - - ret = smu_cmn_update_table(smu, - SMU_TABLE_ACTIVITY_MONITOR_COEFF, WORKLOAD_PPLIB_CUSTOM_BIT, - (void *)(&activity_monitor_external), true); + if (custom_params && custom_params_max_idx) { + if (custom_params_max_idx != SMU_13_0_7_CUSTOM_PARAMS_COUNT) + return -EINVAL; + if (custom_params[0] >= SMU_13_0_7_CUSTOM_PARAMS_CLOCK_COUNT) + return -EINVAL; + idx = custom_params[0] * SMU_13_0_7_CUSTOM_PARAMS_COUNT; + smu->custom_profile_params[idx] = 1; + for (i = 1; i < custom_params_max_idx; i++) + smu->custom_profile_params[idx + i] = custom_params[i]; + } + ret = smu_v13_0_7_set_power_profile_mode_coeff(smu, + smu->custom_profile_params); if (ret) { - dev_err(smu->adev->dev, "[%s] Failed to set activity monitor!", __func__); + if (idx != -1) + smu->custom_profile_params[idx] = 0; return ret; } + } else if (smu->custom_profile_params) { + memset(smu->custom_profile_params, 0, SMU_13_0_7_CUSTOM_PARAMS_SIZE); } - /* 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; ret = smu_cmn_send_smc_msg_with_param(smu, SMU_MSG_SetWorkloadMask, - 1 << workload_type, NULL); + backend_workload_mask, NULL); - if (ret) - dev_err(smu->adev->dev, "[%s] Failed to set work load mask!", __func__); - else - smu->workload_mask = (1 << workload_type); + if (ret) { + dev_err(smu->adev->dev, "Failed to set workload mask 0x%08x\n", + workload_mask); + if (idx != -1) + smu->custom_profile_params[idx] = 0; + return ret; + } return ret; } diff --git a/drivers/gpu/drm/amd/pm/swsmu/smu14/smu_v14_0_2_ppt.c b/drivers/gpu/drm/amd/pm/swsmu/smu14/smu_v14_0_2_ppt.c index 98e01a06add8..6a565ce74d5b 100644 --- a/drivers/gpu/drm/amd/pm/swsmu/smu14/smu_v14_0_2_ppt.c +++ b/drivers/gpu/drm/amd/pm/swsmu/smu14/smu_v14_0_2_ppt.c @@ -1739,90 +1739,120 @@ static int smu_v14_0_2_get_power_profile_mode(struct smu_context *smu, return size; } -static int smu_v14_0_2_set_power_profile_mode(struct smu_context *smu, - long *input, - uint32_t size) +#define SMU_14_0_2_CUSTOM_PARAMS_COUNT 9 +#define SMU_14_0_2_CUSTOM_PARAMS_CLOCK_COUNT 2 +#define SMU_14_0_2_CUSTOM_PARAMS_SIZE (SMU_14_0_2_CUSTOM_PARAMS_CLOCK_COUNT * SMU_14_0_2_CUSTOM_PARAMS_COUNT * sizeof(long)) + +static int smu_v14_0_2_set_power_profile_mode_coeff(struct smu_context *smu, + long *input) { DpmActivityMonitorCoeffIntExternal_t activity_monitor_external; DpmActivityMonitorCoeffInt_t *activity_monitor = &(activity_monitor_external.DpmActivityMonitorCoeffInt); - int workload_type, ret = 0; - uint32_t current_profile_mode = smu->power_profile_mode; - smu->power_profile_mode = input[size]; + int ret, idx; - if (smu->power_profile_mode >= PP_SMC_POWER_PROFILE_COUNT) { - dev_err(smu->adev->dev, "Invalid power profile mode %d\n", smu->power_profile_mode); - return -EINVAL; + ret = smu_cmn_update_table(smu, + SMU_TABLE_ACTIVITY_MONITOR_COEFF, + WORKLOAD_PPLIB_CUSTOM_BIT, + (void *)(&activity_monitor_external), + false); + if (ret) { + dev_err(smu->adev->dev, "[%s] Failed to get activity monitor!", __func__); + return ret; } - if (smu->power_profile_mode == PP_SMC_POWER_PROFILE_CUSTOM) { - if (size != 9) - return -EINVAL; + idx = 0 * SMU_14_0_2_CUSTOM_PARAMS_COUNT; + if (input[idx]) { + /* Gfxclk */ + activity_monitor->Gfx_FPS = input[idx + 1]; + activity_monitor->Gfx_MinActiveFreqType = input[idx + 2]; + activity_monitor->Gfx_MinActiveFreq = input[idx + 3]; + activity_monitor->Gfx_BoosterFreqType = input[idx + 4]; + activity_monitor->Gfx_BoosterFreq = input[idx + 5]; + activity_monitor->Gfx_PD_Data_limit_c = input[idx + 6]; + activity_monitor->Gfx_PD_Data_error_coeff = input[idx + 7]; + activity_monitor->Gfx_PD_Data_error_rate_coeff = input[idx + 8]; + } + idx = 1 * SMU_14_0_2_CUSTOM_PARAMS_COUNT; + if (input[idx]) { + /* Fclk */ + activity_monitor->Fclk_FPS = input[idx + 1]; + activity_monitor->Fclk_MinActiveFreqType = input[idx + 2]; + activity_monitor->Fclk_MinActiveFreq = input[idx + 3]; + activity_monitor->Fclk_BoosterFreqType = input[idx + 4]; + activity_monitor->Fclk_BoosterFreq = input[idx + 5]; + activity_monitor->Fclk_PD_Data_limit_c = input[idx + 6]; + activity_monitor->Fclk_PD_Data_error_coeff = input[idx + 7]; + activity_monitor->Fclk_PD_Data_error_rate_coeff = input[idx + 8]; + } - ret = smu_cmn_update_table(smu, - SMU_TABLE_ACTIVITY_MONITOR_COEFF, - WORKLOAD_PPLIB_CUSTOM_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; + } - switch (input[0]) { - case 0: /* Gfxclk */ - activity_monitor->Gfx_FPS = input[1]; - activity_monitor->Gfx_MinActiveFreqType = input[2]; - activity_monitor->Gfx_MinActiveFreq = input[3]; - activity_monitor->Gfx_BoosterFreqType = input[4]; - activity_monitor->Gfx_BoosterFreq = input[5]; - activity_monitor->Gfx_PD_Data_limit_c = input[6]; - activity_monitor->Gfx_PD_Data_error_coeff = input[7]; - activity_monitor->Gfx_PD_Data_error_rate_coeff = input[8]; - break; - case 1: /* Fclk */ - activity_monitor->Fclk_FPS = input[1]; - activity_monitor->Fclk_MinActiveFreqType = input[2]; - activity_monitor->Fclk_MinActiveFreq = input[3]; - activity_monitor->Fclk_BoosterFreqType = input[4]; - activity_monitor->Fclk_BoosterFreq = input[5]; - activity_monitor->Fclk_PD_Data_limit_c = input[6]; - activity_monitor->Fclk_PD_Data_error_coeff = input[7]; - activity_monitor->Fclk_PD_Data_error_rate_coeff = input[8]; - break; - default: - return -EINVAL; - } + 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; - } - } +static int smu_v14_0_2_set_power_profile_mode(struct smu_context *smu, + u32 workload_mask, + long *custom_params, + u32 custom_params_max_idx) +{ + u32 backend_workload_mask = 0; + int ret, idx = -1, i; + + smu_cmn_get_backend_workload_mask(smu, workload_mask, + &backend_workload_mask); - if (smu->power_profile_mode == PP_SMC_POWER_PROFILE_COMPUTE) + /* disable deep sleep if compute is enabled */ + if (workload_mask & (1 << PP_SMC_POWER_PROFILE_COMPUTE)) smu_v14_0_deep_sleep_control(smu, false); - else if (current_profile_mode == PP_SMC_POWER_PROFILE_COMPUTE) + else smu_v14_0_deep_sleep_control(smu, true); - /* 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; + if (workload_mask & (1 << PP_SMC_POWER_PROFILE_CUSTOM)) { + if (!smu->custom_profile_params) { + smu->custom_profile_params = + kzalloc(SMU_14_0_2_CUSTOM_PARAMS_SIZE, GFP_KERNEL); + if (!smu->custom_profile_params) + return -ENOMEM; + } + if (custom_params && custom_params_max_idx) { + if (custom_params_max_idx != SMU_14_0_2_CUSTOM_PARAMS_COUNT) + return -EINVAL; + if (custom_params[0] >= SMU_14_0_2_CUSTOM_PARAMS_CLOCK_COUNT) + return -EINVAL; + idx = custom_params[0] * SMU_14_0_2_CUSTOM_PARAMS_COUNT; + smu->custom_profile_params[idx] = 1; + for (i = 1; i < custom_params_max_idx; i++) + smu->custom_profile_params[idx + i] = custom_params[i]; + } + ret = smu_v14_0_2_set_power_profile_mode_coeff(smu, + smu->custom_profile_params); + if (ret) { + if (idx != -1) + smu->custom_profile_params[idx] = 0; + return ret; + } + } else if (smu->custom_profile_params) { + memset(smu->custom_profile_params, 0, SMU_14_0_2_CUSTOM_PARAMS_SIZE); + } - ret = smu_cmn_send_smc_msg_with_param(smu, - SMU_MSG_SetWorkloadMask, - 1 << workload_type, - NULL); - if (!ret) - smu->workload_mask = 1 << workload_type; + ret = smu_cmn_send_smc_msg_with_param(smu, SMU_MSG_SetWorkloadMask, + backend_workload_mask, NULL); + if (ret) { + dev_err(smu->adev->dev, "Failed to set workload mask 0x%08x\n", + workload_mask); + if (idx != -1) + smu->custom_profile_params[idx] = 0; + return ret; + } return ret; } diff --git a/drivers/gpu/drm/amd/pm/swsmu/smu_cmn.c b/drivers/gpu/drm/amd/pm/swsmu/smu_cmn.c index 0bc6fb7d7223..9f55207ea9bc 100644 --- a/drivers/gpu/drm/amd/pm/swsmu/smu_cmn.c +++ b/drivers/gpu/drm/amd/pm/swsmu/smu_cmn.c @@ -1221,3 +1221,28 @@ void smu_cmn_generic_plpd_policy_desc(struct smu_dpm_policy *policy) { policy->desc = &xgmi_plpd_policy_desc; } + +void smu_cmn_get_backend_workload_mask(struct smu_context *smu, + u32 workload_mask, + u32 *backend_workload_mask) +{ + int workload_type; + u32 profile_mode; + + *backend_workload_mask = 0; + + for (profile_mode = 0; profile_mode < PP_SMC_POWER_PROFILE_COUNT; profile_mode++) { + if (!(workload_mask & (1 << profile_mode))) + continue; + + /* conv PP_SMC_POWER_PROFILE* to WORKLOAD_PPLIB_*_BIT */ + workload_type = smu_cmn_to_asic_specific_index(smu, + CMN2ASIC_MAPPING_WORKLOAD, + profile_mode); + + if (workload_type < 0) + continue; + + *backend_workload_mask |= 1 << workload_type; + } +} diff --git a/drivers/gpu/drm/amd/pm/swsmu/smu_cmn.h b/drivers/gpu/drm/amd/pm/swsmu/smu_cmn.h index 1de685defe85..a020277dec3e 100644 --- a/drivers/gpu/drm/amd/pm/swsmu/smu_cmn.h +++ b/drivers/gpu/drm/amd/pm/swsmu/smu_cmn.h @@ -147,5 +147,9 @@ bool smu_cmn_is_audio_func_enabled(struct amdgpu_device *adev); void smu_cmn_generic_soc_policy_desc(struct smu_dpm_policy *policy); void smu_cmn_generic_plpd_policy_desc(struct smu_dpm_policy *policy); +void smu_cmn_get_backend_workload_mask(struct smu_context *smu, + u32 workload_mask, + u32 *backend_workload_mask); + #endif #endif -- 2.50.1 From 73dae652dcac776296890da215ee7dec357a1032 Mon Sep 17 00:00:00 2001 From: Alex Deucher Date: Mon, 25 Nov 2024 13:59:09 -0500 Subject: [PATCH 07/16] drm/amdgpu: rework resume handling for display (v2) MIME-Version: 1.0 Content-Type: text/plain; charset=utf8 Content-Transfer-Encoding: 8bit Split resume into a 3rd step to handle displays when DCC is enabled on DCN 4.0.1. Move display after the buffer funcs have been re-enabled so that the GPU will do the move and properly set the DCC metadata for DCN. v2: fix fence irq resume ordering Reviewed-by: Christian König Signed-off-by: Alex Deucher Cc: stable@vger.kernel.org # 6.11.x --- drivers/gpu/drm/amd/amdgpu/amdgpu_device.c | 45 +++++++++++++++++++++- 1 file changed, 43 insertions(+), 2 deletions(-) diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c index 913eb4f08ee6..96316111300a 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c @@ -3762,7 +3762,7 @@ static int amdgpu_device_ip_resume_phase1(struct amdgpu_device *adev) * * @adev: amdgpu_device pointer * - * First resume function for hardware IPs. The list of all the hardware + * Second resume function for hardware IPs. The list of all the hardware * IPs that make up the asic is walked and the resume callbacks are run for * all blocks except COMMON, GMC, and IH. resume puts the hardware into a * functional state after a suspend and updates the software state as @@ -3780,6 +3780,7 @@ static int amdgpu_device_ip_resume_phase2(struct amdgpu_device *adev) if (adev->ip_blocks[i].version->type == AMD_IP_BLOCK_TYPE_COMMON || adev->ip_blocks[i].version->type == AMD_IP_BLOCK_TYPE_GMC || adev->ip_blocks[i].version->type == AMD_IP_BLOCK_TYPE_IH || + adev->ip_blocks[i].version->type == AMD_IP_BLOCK_TYPE_DCE || adev->ip_blocks[i].version->type == AMD_IP_BLOCK_TYPE_PSP) continue; r = amdgpu_ip_block_resume(&adev->ip_blocks[i]); @@ -3790,6 +3791,36 @@ static int amdgpu_device_ip_resume_phase2(struct amdgpu_device *adev) return 0; } +/** + * amdgpu_device_ip_resume_phase3 - run resume for hardware IPs + * + * @adev: amdgpu_device pointer + * + * Third resume function for hardware IPs. The list of all the hardware + * IPs that make up the asic is walked and the resume callbacks are run for + * all DCE. resume puts the hardware into a functional state after a suspend + * and updates the software state as necessary. This function is also used + * for restoring the GPU after a GPU reset. + * + * Returns 0 on success, negative error code on failure. + */ +static int amdgpu_device_ip_resume_phase3(struct amdgpu_device *adev) +{ + int i, r; + + for (i = 0; i < adev->num_ip_blocks; i++) { + if (!adev->ip_blocks[i].status.valid || adev->ip_blocks[i].status.hw) + continue; + if (adev->ip_blocks[i].version->type == AMD_IP_BLOCK_TYPE_DCE) { + r = amdgpu_ip_block_resume(&adev->ip_blocks[i]); + if (r) + return r; + } + } + + return 0; +} + /** * amdgpu_device_ip_resume - run resume for hardware IPs * @@ -3819,6 +3850,13 @@ static int amdgpu_device_ip_resume(struct amdgpu_device *adev) if (adev->mman.buffer_funcs_ring->sched.ready) amdgpu_ttm_set_buffer_funcs_status(adev, true); + if (r) + return r; + + amdgpu_fence_driver_hw_init(adev); + + r = amdgpu_device_ip_resume_phase3(adev); + return r; } @@ -4899,7 +4937,6 @@ int amdgpu_device_resume(struct drm_device *dev, bool notify_clients) dev_err(adev->dev, "amdgpu_device_ip_resume failed (%d).\n", r); goto exit; } - amdgpu_fence_driver_hw_init(adev); if (!adev->in_s0ix) { r = amdgpu_amdkfd_resume(adev, adev->in_runpm); @@ -5484,6 +5521,10 @@ int amdgpu_device_reinit_after_reset(struct amdgpu_reset_context *reset_context) if (tmp_adev->mman.buffer_funcs_ring->sched.ready) amdgpu_ttm_set_buffer_funcs_status(tmp_adev, true); + r = amdgpu_device_ip_resume_phase3(tmp_adev); + if (r) + goto out; + if (vram_lost) amdgpu_device_fill_reset_magic(tmp_adev); -- 2.50.1 From 9f4ddfdc2c03956d278bdafca6adc21cf90cc834 Mon Sep 17 00:00:00 2001 From: Pratap Nirujogi Date: Mon, 2 Dec 2024 23:29:31 -0500 Subject: [PATCH 08/16] Revert "drm/amdgpu: Fix ISP hw init issue" This reverts commit 274e3f4596446955bf17680fd4eb5489f5ecac00. Additional review comments to address. Will resubmit. Reviewed-by: Mario Limonciello Signed-off-by: Pratap Nirujogi Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_device.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c index 96316111300a..23ad65e336d5 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c @@ -145,7 +145,7 @@ const char *amdgpu_asic_name[] = { "LAST", }; -#define AMDGPU_IP_BLK_MASK_ALL GENMASK(AMDGPU_MAX_IP_NUM, 0) +#define AMDGPU_IP_BLK_MASK_ALL GENMASK(AMDGPU_MAX_IP_NUM - 1, 0) /* * Default init level where all blocks are expected to be initialized. This is * the level of initialization expected by default and also after a full reset -- 2.50.1 From 9a4ab400f1fad0e6e8686b8f5fc5376383860ce8 Mon Sep 17 00:00:00 2001 From: Victor Zhao Date: Sat, 23 Nov 2024 18:37:43 +0800 Subject: [PATCH 09/16] drm/amdgpu: use sjt mec fw on gfx943 for sriov Use second jump table in sriov for live migration or mulitple VF support so different VF can load different version of MEC as long as they support sjt Signed-off-by: Victor Zhao Reviewed-by: Yang Wang Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/gfx_v9_4_3.c | 10 ++++++++-- 1 file changed, 8 insertions(+), 2 deletions(-) diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v9_4_3.c b/drivers/gpu/drm/amd/amdgpu/gfx_v9_4_3.c index e2b3dda57030..54459254bd37 100644 --- a/drivers/gpu/drm/amd/amdgpu/gfx_v9_4_3.c +++ b/drivers/gpu/drm/amd/amdgpu/gfx_v9_4_3.c @@ -45,6 +45,8 @@ MODULE_FIRMWARE("amdgpu/gc_9_4_3_mec.bin"); MODULE_FIRMWARE("amdgpu/gc_9_4_4_mec.bin"); MODULE_FIRMWARE("amdgpu/gc_9_4_3_rlc.bin"); MODULE_FIRMWARE("amdgpu/gc_9_4_4_rlc.bin"); +MODULE_FIRMWARE("amdgpu/gc_9_4_3_sjt_mec.bin"); +MODULE_FIRMWARE("amdgpu/gc_9_4_4_sjt_mec.bin"); #define GFX9_MEC_HPD_SIZE 4096 #define RLCG_UCODE_LOADING_START_ADDRESS 0x00002000L @@ -574,8 +576,12 @@ static int gfx_v9_4_3_init_cp_compute_microcode(struct amdgpu_device *adev, { int err; - err = amdgpu_ucode_request(adev, &adev->gfx.mec_fw, - "amdgpu/%s_mec.bin", chip_name); + if (amdgpu_sriov_vf(adev)) + err = amdgpu_ucode_request(adev, &adev->gfx.mec_fw, + "amdgpu/%s_sjt_mec.bin", chip_name); + else + err = amdgpu_ucode_request(adev, &adev->gfx.mec_fw, + "amdgpu/%s_mec.bin", chip_name); if (err) goto out; amdgpu_gfx_cp_init_microcode(adev, AMDGPU_UCODE_ID_CP_MEC1); -- 2.50.1 From 12f325bcd2411e571dbb500bf6862c812c479735 Mon Sep 17 00:00:00 2001 From: =?utf8?q?Christian=20K=C3=B6nig?= Date: Fri, 29 Nov 2024 14:19:21 +0100 Subject: [PATCH 10/16] drm/amdgpu: fix UVD contiguous CS mapping problem MIME-Version: 1.0 Content-Type: text/plain; charset=utf8 Content-Transfer-Encoding: 8bit When starting the mpv player, Radeon R9 users are observing the below error in dmesg. [drm:amdgpu_uvd_cs_pass2 [amdgpu]] *ERROR* msg/fb buffer ff00f7c000-ff00f7e000 out of 256MB segment! The patch tries to set the TTM_PL_FLAG_CONTIGUOUS for both user flag(AMDGPU_GEM_CREATE_VRAM_CONTIGUOUS) set and not set cases. v2: Make the TTM_PL_FLAG_CONTIGUOUS mandatory for user BO's. v3: revert back to v1, but fix the check instead (chk). Closes:https://gitlab.freedesktop.org/drm/amd/-/issues/3599 Closes:https://gitlab.freedesktop.org/drm/amd/-/issues/3501 Signed-off-by: Arunpravin Paneer Selvam Signed-off-by: Christian König Reviewed-by: Arunpravin Paneer Selvam Signed-off-by: Alex Deucher Cc: stable@vger.kernel.org # 6.10+ --- drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c | 17 +++++++++++------ drivers/gpu/drm/amd/amdgpu/amdgpu_uvd.c | 2 ++ 2 files changed, 13 insertions(+), 6 deletions(-) diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c index d891ab779ca7..5df21529b3b1 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c @@ -1801,13 +1801,18 @@ int amdgpu_cs_find_mapping(struct amdgpu_cs_parser *parser, if (dma_resv_locking_ctx((*bo)->tbo.base.resv) != &parser->exec.ticket) return -EINVAL; + /* Make sure VRAM is allocated contigiously */ (*bo)->flags |= AMDGPU_GEM_CREATE_VRAM_CONTIGUOUS; - amdgpu_bo_placement_from_domain(*bo, (*bo)->allowed_domains); - for (i = 0; i < (*bo)->placement.num_placement; i++) - (*bo)->placements[i].flags |= TTM_PL_FLAG_CONTIGUOUS; - r = ttm_bo_validate(&(*bo)->tbo, &(*bo)->placement, &ctx); - if (r) - return r; + if ((*bo)->tbo.resource->mem_type == TTM_PL_VRAM && + !((*bo)->tbo.resource->placement & TTM_PL_FLAG_CONTIGUOUS)) { + + amdgpu_bo_placement_from_domain(*bo, (*bo)->allowed_domains); + for (i = 0; i < (*bo)->placement.num_placement; i++) + (*bo)->placements[i].flags |= TTM_PL_FLAG_CONTIGUOUS; + r = ttm_bo_validate(&(*bo)->tbo, &(*bo)->placement, &ctx); + if (r) + return r; + } return amdgpu_ttm_alloc_gart(&(*bo)->tbo); } diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_uvd.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_uvd.c index 31fd30dcd593..65bb26215e86 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_uvd.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_uvd.c @@ -551,6 +551,8 @@ static void amdgpu_uvd_force_into_uvd_segment(struct amdgpu_bo *abo) for (i = 0; i < abo->placement.num_placement; ++i) { abo->placements[i].fpfn = 0 >> PAGE_SHIFT; abo->placements[i].lpfn = (256 * 1024 * 1024) >> PAGE_SHIFT; + if (abo->placements[i].mem_type == TTM_PL_VRAM) + abo->placements[i].flags |= TTM_PL_FLAG_CONTIGUOUS; } } -- 2.50.1 From 47f402a3e08113e0f5d8e1e6fcc197667a16022f Mon Sep 17 00:00:00 2001 From: "David (Ming Qiang) Wu" Date: Wed, 4 Dec 2024 11:30:01 -0500 Subject: [PATCH 11/16] amdgpu/uvd: get ring reference from rq scheduler MIME-Version: 1.0 Content-Type: text/plain; charset=utf8 Content-Transfer-Encoding: 8bit base.sched may not be set for each instance and should not be used for cases such as non-IB tests. Fixes: 2320c9e6a768 ("drm/sched: memset() 'job' in drm_sched_job_init()") Signed-off-by: David (Ming Qiang) Wu Reviewed-by: Christian König Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/uvd_v7_0.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/drivers/gpu/drm/amd/amdgpu/uvd_v7_0.c b/drivers/gpu/drm/amd/amdgpu/uvd_v7_0.c index 079131aeb2f7..3c8ab8698af8 100644 --- a/drivers/gpu/drm/amd/amdgpu/uvd_v7_0.c +++ b/drivers/gpu/drm/amd/amdgpu/uvd_v7_0.c @@ -1288,7 +1288,7 @@ static int uvd_v7_0_ring_patch_cs_in_place(struct amdgpu_cs_parser *p, struct amdgpu_job *job, struct amdgpu_ib *ib) { - struct amdgpu_ring *ring = to_amdgpu_ring(job->base.sched); + struct amdgpu_ring *ring = amdgpu_job_ring(job); unsigned i; /* No patching necessary for the first instance */ -- 2.50.1 From 8eb966f2403abb844e972fb4eb1348640111f121 Mon Sep 17 00:00:00 2001 From: Lijo Lazar Date: Wed, 4 Dec 2024 13:11:28 +0530 Subject: [PATCH 12/16] drm/amd/pm: Initialize power profile mode Refactor such that individual SMU IP versions can choose the startup power profile mode. If no preference, then use the generic default power profile selection logic. Signed-off-by: Lijo Lazar Reviewed-by: Yang Wang Signed-off-by: Alex Deucher Cc: stable@vger.kernel.org # 6.11.x --- .../gpu/drm/amd/include/kgd_pp_interface.h | 1 + drivers/gpu/drm/amd/pm/swsmu/amdgpu_smu.c | 24 +++++++++++++------ 2 files changed, 18 insertions(+), 7 deletions(-) diff --git a/drivers/gpu/drm/amd/include/kgd_pp_interface.h b/drivers/gpu/drm/amd/include/kgd_pp_interface.h index 67a5de573943..d7acdd42d80f 100644 --- a/drivers/gpu/drm/amd/include/kgd_pp_interface.h +++ b/drivers/gpu/drm/amd/include/kgd_pp_interface.h @@ -164,6 +164,7 @@ enum amd_pp_task { }; enum PP_SMC_POWER_PROFILE { + PP_SMC_POWER_PROFILE_UNKNOWN = -1, PP_SMC_POWER_PROFILE_BOOTUP_DEFAULT = 0x0, PP_SMC_POWER_PROFILE_FULLSCREEN3D = 0x1, PP_SMC_POWER_PROFILE_POWERSAVING = 0x2, diff --git a/drivers/gpu/drm/amd/pm/swsmu/amdgpu_smu.c b/drivers/gpu/drm/amd/pm/swsmu/amdgpu_smu.c index 5eae14fe79f1..21bd635bcdfc 100644 --- a/drivers/gpu/drm/amd/pm/swsmu/amdgpu_smu.c +++ b/drivers/gpu/drm/amd/pm/swsmu/amdgpu_smu.c @@ -764,6 +764,7 @@ static int smu_early_init(struct amdgpu_ip_block *ip_block) smu->smu_baco.platform_support = false; smu->smu_baco.maco_support = false; smu->user_dpm_profile.fan_mode = -1; + smu->power_profile_mode = PP_SMC_POWER_PROFILE_UNKNOWN; mutex_init(&smu->message_lock); @@ -1248,6 +1249,21 @@ static bool smu_is_workload_profile_available(struct smu_context *smu, return smu->workload_map && smu->workload_map[profile].valid_mapping; } +static void smu_init_power_profile(struct smu_context *smu) +{ + if (smu->power_profile_mode == PP_SMC_POWER_PROFILE_UNKNOWN) { + if (smu->is_apu || + !smu_is_workload_profile_available( + smu, PP_SMC_POWER_PROFILE_FULLSCREEN3D)) + smu->power_profile_mode = + PP_SMC_POWER_PROFILE_BOOTUP_DEFAULT; + else + smu->power_profile_mode = + PP_SMC_POWER_PROFILE_FULLSCREEN3D; + } + smu_power_profile_mode_get(smu, smu->power_profile_mode); +} + static int smu_sw_init(struct amdgpu_ip_block *ip_block) { struct amdgpu_device *adev = ip_block->adev; @@ -1269,13 +1285,7 @@ static int smu_sw_init(struct amdgpu_ip_block *ip_block) atomic_set(&smu->smu_power.power_gate.vpe_gated, 1); atomic_set(&smu->smu_power.power_gate.umsch_mm_gated, 1); - if (smu->is_apu || - !smu_is_workload_profile_available(smu, PP_SMC_POWER_PROFILE_FULLSCREEN3D)) - smu->power_profile_mode = PP_SMC_POWER_PROFILE_BOOTUP_DEFAULT; - else - smu->power_profile_mode = PP_SMC_POWER_PROFILE_FULLSCREEN3D; - smu_power_profile_mode_get(smu, smu->power_profile_mode); - + smu_init_power_profile(smu); smu->display_config = &adev->pm.pm_display_cfg; smu->smu_dpm.dpm_level = AMD_DPM_FORCED_LEVEL_AUTO; -- 2.50.1 From 3912a78cf72eb45f8153a395162b08fef9c5ec3d Mon Sep 17 00:00:00 2001 From: Kenneth Feng Date: Wed, 4 Dec 2024 13:22:10 +0530 Subject: [PATCH 13/16] drm/amd/pm: Set SMU v13.0.7 default workload type Set the default workload type to bootup type on smu v13.0.7. This is because of the constraint on smu v13.0.7. Gfx activity has an even higher set point on 3D fullscreen mode than the one on bootup mode. This causes the 3D fullscreen mode's performance is worse than the bootup mode's performance for the lightweighted/medium workload. For the high workload, the performance is the same between 3D fullscreen mode and bootup mode. v2: set the default workload in ASIC specific file Signed-off-by: Kenneth Feng Reviewed-by: Yang Wang Signed-off-by: Alex Deucher Cc: stable@vger.kernel.org # 6.11.x --- drivers/gpu/drm/amd/pm/swsmu/smu13/smu_v13_0_7_ppt.c | 1 + 1 file changed, 1 insertion(+) 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 f4ac403b8b36..aabb94796005 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 @@ -2810,4 +2810,5 @@ void smu_v13_0_7_set_ppt_funcs(struct smu_context *smu) 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); + smu->power_profile_mode = PP_SMC_POWER_PROFILE_BOOTUP_DEFAULT; } -- 2.50.1 From 5c3de6b02d38eb9386edf50490e050bb44398e40 Mon Sep 17 00:00:00 2001 From: Prike Liang Date: Tue, 5 Nov 2024 09:57:42 +0800 Subject: [PATCH 14/16] drm/amdkfd: Correct the migration DMA map direction The SVM DMA device map direction should be set the same as the DMA unmap setting, otherwise the DMA core will report the following warning. Before finialize this solution, there're some discussion on the DMA mapping type(stream-based or coherent) in this KFD migration case, followed by https://lore.kernel.org/all/04d4ab32 -45a1-4b88-86ee-fb0f35a0ca40@amd.com/T/. As there's no dma_sync_single_for_*() in the DMA buffer accessed that because this migration operation should be sync properly and automatically. Give that there's might not be a performance problem in various cache sync policy of DMA sync. Therefore, in order to simplify the DMA direction setting alignment, let's set the DMA map direction as BIDIRECTIONAL. [ 150.834218] WARNING: CPU: 8 PID: 1812 at kernel/dma/debug.c:1028 check_unmap+0x1cc/0x930 [ 150.834225] Modules linked in: amdgpu(OE) amdxcp drm_exec(OE) gpu_sched drm_buddy(OE) drm_ttm_helper(OE) ttm(OE) drm_suballoc_helper(OE) drm_display_helper(OE) drm_kms_helper(OE) i2c_algo_bit rpcsec_gss_krb5 auth_rpcgss nfsv4 nfs lockd grace netfs xt_conntrack xt_MASQUERADE nf_conntrack_netlink xfrm_user xfrm_algo iptable_nat xt_addrtype iptable_filter br_netfilter nvme_fabrics overlay nfnetlink_cttimeout nfnetlink openvswitch nsh nf_conncount nf_nat nf_conntrack nf_defrag_ipv6 nf_defrag_ipv4 libcrc32c bridge stp llc sch_fq_codel intel_rapl_msr amd_atl intel_rapl_common snd_hda_codec_realtek snd_hda_codec_generic snd_hda_scodec_component snd_hda_codec_hdmi snd_hda_intel snd_intel_dspcfg edac_mce_amd snd_pci_acp6x snd_hda_codec snd_acp_config snd_hda_core snd_hwdep snd_soc_acpi kvm_amd sunrpc snd_pcm kvm binfmt_misc snd_seq_midi crct10dif_pclmul snd_seq_midi_event ghash_clmulni_intel sha512_ssse3 snd_rawmidi nls_iso8859_1 sha256_ssse3 sha1_ssse3 snd_seq aesni_intel snd_seq_device crypto_simd snd_timer cryptd input_leds [ 150.834310] wmi_bmof serio_raw k10temp rapl snd sp5100_tco ipmi_devintf soundcore ccp ipmi_msghandler cm32181 industrialio mac_hid msr parport_pc ppdev lp parport efi_pstore drm(OE) ip_tables x_tables pci_stub crc32_pclmul nvme ahci libahci i2c_piix4 r8169 nvme_core i2c_designware_pci realtek i2c_ccgx_ucsi video wmi hid_generic cdc_ether usbnet usbhid hid r8152 mii [ 150.834354] CPU: 8 PID: 1812 Comm: rocrtst64 Tainted: G OE 6.10.0-custom #492 [ 150.834358] Hardware name: AMD Majolica-RN/Majolica-RN, BIOS RMJ1009A 06/13/2021 [ 150.834360] RIP: 0010:check_unmap+0x1cc/0x930 [ 150.834363] Code: c0 4c 89 4d c8 e8 34 bf 86 00 4c 8b 4d c8 4c 8b 45 c0 48 8b 4d b8 48 89 c6 41 57 4c 89 ea 48 c7 c7 80 49 b4 84 e8 b4 81 f3 ff <0f> 0b 48 c7 c7 04 83 ac 84 e8 76 ba fc ff 41 8b 76 4c 49 8d 7e 50 [ 150.834365] RSP: 0018:ffffaac5023739e0 EFLAGS: 00010086 [ 150.834368] RAX: 0000000000000000 RBX: ffffffff8566a2e0 RCX: 0000000000000027 [ 150.834370] RDX: ffff8f6a8f621688 RSI: 0000000000000001 RDI: ffff8f6a8f621680 [ 150.834372] RBP: ffffaac502373a30 R08: 00000000000000c9 R09: ffffaac502373850 [ 150.834373] R10: ffffaac502373848 R11: ffffffff84f46328 R12: ffffaac502373a40 [ 150.834375] R13: ffff8f6741045330 R14: ffff8f6741a77700 R15: ffffffff84ac831b [ 150.834377] FS: 00007faf0fc94c00(0000) GS:ffff8f6a8f600000(0000) knlGS:0000000000000000 [ 150.834379] CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033 [ 150.834381] CR2: 00007faf0b600020 CR3: 000000010a52e000 CR4: 0000000000350ef0 [ 150.834383] Call Trace: [ 150.834385] [ 150.834387] ? show_regs+0x6d/0x80 [ 150.834393] ? __warn+0x8c/0x140 [ 150.834397] ? check_unmap+0x1cc/0x930 [ 150.834400] ? report_bug+0x193/0x1a0 [ 150.834406] ? handle_bug+0x46/0x80 [ 150.834410] ? exc_invalid_op+0x1d/0x80 [ 150.834413] ? asm_exc_invalid_op+0x1f/0x30 [ 150.834420] ? check_unmap+0x1cc/0x930 [ 150.834425] debug_dma_unmap_page+0x86/0x90 [ 150.834431] ? srso_return_thunk+0x5/0x5f [ 150.834435] ? rmap_walk+0x28/0x50 [ 150.834438] ? srso_return_thunk+0x5/0x5f [ 150.834441] ? remove_migration_ptes+0x79/0x80 [ 150.834445] ? srso_return_thunk+0x5/0x5f [ 150.834448] dma_unmap_page_attrs+0xfa/0x1d0 [ 150.834453] svm_range_dma_unmap_dev+0x8a/0xf0 [amdgpu] [ 150.834710] svm_migrate_ram_to_vram+0x361/0x740 [amdgpu] [ 150.834914] svm_migrate_to_vram+0xa8/0xe0 [amdgpu] [ 150.835111] svm_range_set_attr+0xff2/0x1450 [amdgpu] [ 150.835311] svm_ioctl+0x4a/0x50 [amdgpu] [ 150.835510] kfd_ioctl_svm+0x54/0x90 [amdgpu] [ 150.835701] kfd_ioctl+0x3c2/0x530 [amdgpu] [ 150.835888] ? __pfx_kfd_ioctl_svm+0x10/0x10 [amdgpu] [ 150.836075] ? srso_return_thunk+0x5/0x5f [ 150.836080] ? tomoyo_file_ioctl+0x20/0x30 [ 150.836086] __x64_sys_ioctl+0x9c/0xd0 [ 150.836091] x64_sys_call+0x1219/0x20d0 [ 150.836095] do_syscall_64+0x51/0x120 [ 150.836098] entry_SYSCALL_64_after_hwframe+0x76/0x7e [ 150.836102] RIP: 0033:0x7faf0f11a94f [ 150.836105] Code: 00 48 89 44 24 18 31 c0 48 8d 44 24 60 c7 04 24 10 00 00 00 48 89 44 24 08 48 8d 44 24 20 48 89 44 24 10 b8 10 00 00 00 0f 05 <41> 89 c0 3d 00 f0 ff ff 77 1f 48 8b 44 24 18 64 48 2b 04 25 28 00 [ 150.836107] RSP: 002b:00007ffeced26bc0 EFLAGS: 00000246 ORIG_RAX: 0000000000000010 [ 150.836110] RAX: ffffffffffffffda RBX: 000055c683528fb0 RCX: 00007faf0f11a94f [ 150.836112] RDX: 00007ffeced26c60 RSI: 00000000c0484b20 RDI: 0000000000000003 [ 150.836114] RBP: 00007ffeced26c50 R08: 0000000000000000 R09: 0000000000000001 [ 150.836115] R10: 0000000000000032 R11: 0000000000000246 R12: 000055c683528bd0 [ 150.836117] R13: 0000000000000000 R14: 0000000000000021 R15: 0000000000000000 [ 150.836122] [ 150.836124] ---[ end trace 0000000000000000 ]--- Signed-off-by: Prike Liang Reviewed-by: Felix Kuehling Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdkfd/kfd_migrate.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_migrate.c b/drivers/gpu/drm/amd/amdkfd/kfd_migrate.c index eacfeb32f35d..4b275937d05e 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_migrate.c +++ b/drivers/gpu/drm/amd/amdkfd/kfd_migrate.c @@ -306,7 +306,7 @@ svm_migrate_copy_to_vram(struct kfd_node *node, struct svm_range *prange, spage = migrate_pfn_to_page(migrate->src[i]); if (spage && !is_zone_device_page(spage)) { src[i] = dma_map_page(dev, spage, 0, PAGE_SIZE, - DMA_TO_DEVICE); + DMA_BIDIRECTIONAL); r = dma_mapping_error(dev, src[i]); if (r) { dev_err(dev, "%s: fail %d dma_map_page\n", @@ -629,7 +629,7 @@ svm_migrate_copy_to_ram(struct amdgpu_device *adev, struct svm_range *prange, goto out_oom; } - dst[i] = dma_map_page(dev, dpage, 0, PAGE_SIZE, DMA_FROM_DEVICE); + dst[i] = dma_map_page(dev, dpage, 0, PAGE_SIZE, DMA_BIDIRECTIONAL); r = dma_mapping_error(dev, dst[i]); if (r) { dev_err(adev->dev, "%s: fail %d dma_map_page\n", __func__, r); -- 2.50.1 From a592bb19abdc2072875c87da606461bfd7821b08 Mon Sep 17 00:00:00 2001 From: Andrew Martin Date: Tue, 26 Nov 2024 12:10:59 -0500 Subject: [PATCH 15/16] drm/amdkfd: Dereference null return value In the function pqm_uninit there is a call-assignment of "pdd = kfd_get_process_device_data" which could be null, and this value was later dereferenced without checking. Fixes: fb91065851cd ("drm/amdkfd: Refactor queue wptr_bo GART mapping") Signed-off-by: Andrew Martin Reviewed-by: Felix Kuehling Signed-off-by: Alex Deucher Cc: stable@vger.kernel.org --- .../gpu/drm/amd/amdkfd/kfd_process_queue_manager.c | 12 ++++++++---- 1 file changed, 8 insertions(+), 4 deletions(-) diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_process_queue_manager.c b/drivers/gpu/drm/amd/amdkfd/kfd_process_queue_manager.c index c76db22a1000..59b92d66e958 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_process_queue_manager.c +++ b/drivers/gpu/drm/amd/amdkfd/kfd_process_queue_manager.c @@ -212,13 +212,17 @@ static void pqm_clean_queue_resource(struct process_queue_manager *pqm, void pqm_uninit(struct process_queue_manager *pqm) { struct process_queue_node *pqn, *next; - struct kfd_process_device *pdd; list_for_each_entry_safe(pqn, next, &pqm->queues, process_queue_list) { if (pqn->q) { - pdd = kfd_get_process_device_data(pqn->q->device, pqm->process); - kfd_queue_unref_bo_vas(pdd, &pqn->q->properties); - kfd_queue_release_buffers(pdd, &pqn->q->properties); + struct kfd_process_device *pdd = kfd_get_process_device_data(pqn->q->device, + pqm->process); + if (pdd) { + kfd_queue_unref_bo_vas(pdd, &pqn->q->properties); + kfd_queue_release_buffers(pdd, &pqn->q->properties); + } else { + WARN_ON(!pdd); + } pqm_clean_queue_resource(pqm, pqn); } -- 2.50.1 From 321048c4a3e375416b51b4093978f9ce2aa4d391 Mon Sep 17 00:00:00 2001 From: Harish Kasiviswanathan Date: Wed, 27 Nov 2024 14:01:35 -0500 Subject: [PATCH 16/16] drm/amdkfd: hard-code cacheline size for gfx11 This information is not available in ip discovery table. Signed-off-by: Harish Kasiviswanathan Reviewed-by: David Belanger Signed-off-by: Alex Deucher Cc: stable@vger.kernel.org --- drivers/gpu/drm/amd/amdkfd/kfd_crat.c | 22 ++++++++++++++++++++-- 1 file changed, 20 insertions(+), 2 deletions(-) diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_crat.c b/drivers/gpu/drm/amd/amdkfd/kfd_crat.c index 7b826a136ceb..d73510c52286 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_crat.c +++ b/drivers/gpu/drm/amd/amdkfd/kfd_crat.c @@ -1423,6 +1423,7 @@ err: static int kfd_fill_gpu_cache_info_from_gfx_config(struct kfd_dev *kdev, + bool cache_line_size_missing, struct kfd_gpu_cache_info *pcache_info) { struct amdgpu_device *adev = kdev->adev; @@ -1437,6 +1438,8 @@ static int kfd_fill_gpu_cache_info_from_gfx_config(struct kfd_dev *kdev, CRAT_CACHE_FLAGS_SIMD_CACHE); pcache_info[i].num_cu_shared = adev->gfx.config.gc_num_tcp_per_wpg / 2; pcache_info[i].cache_line_size = adev->gfx.config.gc_tcp_cache_line_size; + if (cache_line_size_missing && !pcache_info[i].cache_line_size) + pcache_info[i].cache_line_size = 128; i++; } /* Scalar L1 Instruction Cache per SQC */ @@ -1449,6 +1452,8 @@ static int kfd_fill_gpu_cache_info_from_gfx_config(struct kfd_dev *kdev, CRAT_CACHE_FLAGS_SIMD_CACHE); pcache_info[i].num_cu_shared = adev->gfx.config.gc_num_sqc_per_wgp * 2; pcache_info[i].cache_line_size = adev->gfx.config.gc_instruction_cache_line_size; + if (cache_line_size_missing && !pcache_info[i].cache_line_size) + pcache_info[i].cache_line_size = 128; i++; } /* Scalar L1 Data Cache per SQC */ @@ -1460,6 +1465,8 @@ static int kfd_fill_gpu_cache_info_from_gfx_config(struct kfd_dev *kdev, CRAT_CACHE_FLAGS_SIMD_CACHE); pcache_info[i].num_cu_shared = adev->gfx.config.gc_num_sqc_per_wgp * 2; pcache_info[i].cache_line_size = adev->gfx.config.gc_scalar_data_cache_line_size; + if (cache_line_size_missing && !pcache_info[i].cache_line_size) + pcache_info[i].cache_line_size = 64; i++; } /* GL1 Data Cache per SA */ @@ -1472,7 +1479,8 @@ static int kfd_fill_gpu_cache_info_from_gfx_config(struct kfd_dev *kdev, CRAT_CACHE_FLAGS_DATA_CACHE | CRAT_CACHE_FLAGS_SIMD_CACHE); pcache_info[i].num_cu_shared = adev->gfx.config.max_cu_per_sh; - pcache_info[i].cache_line_size = 0; + if (cache_line_size_missing) + pcache_info[i].cache_line_size = 128; i++; } /* L2 Data Cache per GPU (Total Tex Cache) */ @@ -1484,6 +1492,8 @@ static int kfd_fill_gpu_cache_info_from_gfx_config(struct kfd_dev *kdev, CRAT_CACHE_FLAGS_SIMD_CACHE); pcache_info[i].num_cu_shared = adev->gfx.config.max_cu_per_sh; pcache_info[i].cache_line_size = adev->gfx.config.gc_tcc_cache_line_size; + if (cache_line_size_missing && !pcache_info[i].cache_line_size) + pcache_info[i].cache_line_size = 128; i++; } /* L3 Data Cache per GPU */ @@ -1569,6 +1579,7 @@ static int kfd_fill_gpu_cache_info_from_gfx_config_v2(struct kfd_dev *kdev, int kfd_get_gpu_cache_info(struct kfd_node *kdev, struct kfd_gpu_cache_info **pcache_info) { int num_of_cache_types = 0; + bool cache_line_size_missing = false; switch (kdev->adev->asic_type) { case CHIP_KAVERI: @@ -1692,10 +1703,17 @@ int kfd_get_gpu_cache_info(struct kfd_node *kdev, struct kfd_gpu_cache_info **pc case IP_VERSION(11, 5, 0): case IP_VERSION(11, 5, 1): case IP_VERSION(11, 5, 2): + /* Cacheline size not available in IP discovery for gc11. + * kfd_fill_gpu_cache_info_from_gfx_config to hard code it + */ + cache_line_size_missing = true; + fallthrough; case IP_VERSION(12, 0, 0): case IP_VERSION(12, 0, 1): num_of_cache_types = - kfd_fill_gpu_cache_info_from_gfx_config(kdev->kfd, *pcache_info); + kfd_fill_gpu_cache_info_from_gfx_config(kdev->kfd, + cache_line_size_missing, + *pcache_info); break; default: *pcache_info = dummy_cache_info; -- 2.50.1