From 668d6808e9f23305e0eb108dbf75fc13ab344ae8 Mon Sep 17 00:00:00 2001 From: Robin Chen Date: Thu, 12 Sep 2024 20:59:36 +0800 Subject: [PATCH 01/16] drm/amd/display: Read Sink emission rate capability [WHY] To get sink emission rate information for future supported refresh rate calculation. Reviewed-by: ChunTao Tso Signed-off-by: Robin Chen Signed-off-by: Roman Li Tested-by: Daniel Wheeler Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/display/dc/dc_dp_types.h | 3 ++- .../dc/link/protocols/link_dp_capability.c | 8 ++++++++ .../gpu/drm/amd/display/include/dpcd_defs.h | 19 ++++++++++--------- 3 files changed, 20 insertions(+), 10 deletions(-) diff --git a/drivers/gpu/drm/amd/display/dc/dc_dp_types.h b/drivers/gpu/drm/amd/display/dc/dc_dp_types.h index 50fa8547d718..8dd6eb044829 100644 --- a/drivers/gpu/drm/amd/display/dc/dc_dp_types.h +++ b/drivers/gpu/drm/amd/display/dc/dc_dp_types.h @@ -1192,6 +1192,7 @@ struct dpcd_caps { struct edp_psr_info psr_info; struct replay_info pr_info; + uint16_t edp_oled_emission_rate; }; union dpcd_sink_ext_caps { @@ -1205,7 +1206,7 @@ union dpcd_sink_ext_caps { uint8_t oled : 1; uint8_t reserved_2 : 1; uint8_t miniled : 1; - uint8_t reserved : 1; + uint8_t emission_output : 1; } bits; uint8_t raw; }; diff --git a/drivers/gpu/drm/amd/display/dc/link/protocols/link_dp_capability.c b/drivers/gpu/drm/amd/display/dc/link/protocols/link_dp_capability.c index 2ec968f239f7..952e787493c4 100644 --- a/drivers/gpu/drm/amd/display/dc/link/protocols/link_dp_capability.c +++ b/drivers/gpu/drm/amd/display/dc/link/protocols/link_dp_capability.c @@ -2067,6 +2067,14 @@ void detect_edp_sink_caps(struct dc_link *link) core_link_read_dpcd(link, DP_SINK_PR_MAX_NUMBER_OF_DEVIATION_LINE, &link->dpcd_caps.pr_info.max_deviation_line, sizeof(link->dpcd_caps.pr_info.max_deviation_line)); + + /* + * OLED Emission Rate info + */ + if (link->dpcd_sink_ext_caps.bits.emission_output) + core_link_read_dpcd(link, DP_SINK_EMISSION_RATE, + (uint8_t *)&link->dpcd_caps.edp_oled_emission_rate, + sizeof(link->dpcd_caps.edp_oled_emission_rate)); } bool dp_get_max_link_enc_cap(const struct dc_link *link, struct dc_link_settings *max_link_enc_cap) diff --git a/drivers/gpu/drm/amd/display/include/dpcd_defs.h b/drivers/gpu/drm/amd/display/include/dpcd_defs.h index aee5170f5fb2..de8f3cfed6c8 100644 --- a/drivers/gpu/drm/amd/display/include/dpcd_defs.h +++ b/drivers/gpu/drm/amd/display/include/dpcd_defs.h @@ -164,18 +164,19 @@ enum dpcd_psr_sink_states { PSR_SINK_STATE_SINK_INTERNAL_ERROR = 7, }; -#define DP_SOURCE_SEQUENCE 0x30c -#define DP_SOURCE_TABLE_REVISION 0x310 -#define DP_SOURCE_PAYLOAD_SIZE 0x311 -#define DP_SOURCE_SINK_CAP 0x317 -#define DP_SOURCE_BACKLIGHT_LEVEL 0x320 -#define DP_SOURCE_BACKLIGHT_CURRENT_PEAK 0x326 -#define DP_SOURCE_BACKLIGHT_CONTROL 0x32E -#define DP_SOURCE_BACKLIGHT_ENABLE 0x32F -#define DP_SOURCE_MINIMUM_HBLANK_SUPPORTED 0x340 +#define DP_SOURCE_SEQUENCE 0x30C +#define DP_SOURCE_TABLE_REVISION 0x310 +#define DP_SOURCE_PAYLOAD_SIZE 0x311 +#define DP_SOURCE_SINK_CAP 0x317 +#define DP_SOURCE_BACKLIGHT_LEVEL 0x320 +#define DP_SOURCE_BACKLIGHT_CURRENT_PEAK 0x326 +#define DP_SOURCE_BACKLIGHT_CONTROL 0x32E +#define DP_SOURCE_BACKLIGHT_ENABLE 0x32F +#define DP_SOURCE_MINIMUM_HBLANK_SUPPORTED 0x340 #define DP_SINK_PR_REPLAY_STATUS 0x378 #define DP_SINK_PR_PIXEL_DEVIATION_PER_LINE 0x379 #define DP_SINK_PR_MAX_NUMBER_OF_DEVIATION_LINE 0x37A +#define DP_SINK_EMISSION_RATE 0x37E /* Remove once drm_dp_helper.h is updated upstream */ #ifndef DP_TOTAL_LTTPR_CNT -- 2.51.0 From c5a332f98da56faf54d6a647d903d7272de83320 Mon Sep 17 00:00:00 2001 From: Dillon Varone Date: Wed, 25 Sep 2024 20:18:07 -0400 Subject: [PATCH 02/16] drm/amd/display: resolve correct MALL size for dcn401 [WHY] Code for dcn401 to calculate available MALL size for display was shared with dcn32 and did not provide the correct result for all ASICs. [HOW] Add dcn401 specific function to properly calculate the available MALL for display. Reviewed-by: Chris Park Signed-off-by: Dillon Varone Signed-off-by: Roman Li Tested-by: Daniel Wheeler Signed-off-by: Alex Deucher --- .../dc/resource/dcn32/dcn32_resource.c | 4 +++ .../dc/resource/dcn401/dcn401_resource.c | 27 ++++++++++++++++--- 2 files changed, 28 insertions(+), 3 deletions(-) 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 a1890df29f5c..aaaa888d112d 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 @@ -1990,6 +1990,10 @@ unsigned int dcn32_calculate_mall_ways_from_bytes(const struct dc *dc, unsigned return 0; } + if (dc->caps.max_cab_allocation_bytes == 0) { + return 0xffffffff; + } + /* add 2 lines for worst case alignment */ cache_lines_used = total_size_in_mall_bytes / dc->caps.cache_line_size + 2; 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 59184abab1a7..7da81966b4df 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 @@ -1704,6 +1704,29 @@ static int dcn401_get_power_profile(const struct dc_state *context) return dpm_level; } +static unsigned int dcn401_calc_num_avail_chans_for_mall(struct dc *dc, unsigned int num_chans) +{ + unsigned int num_available_chans = 1; + + /* channels for MALL must be a power of 2 */ + while (num_chans > 1) { + num_available_chans = (num_available_chans << 1); + num_chans = (num_chans >> 1); + } + + /* cannot be odd */ + num_available_chans &= ~1; + + /* clamp to max available channels for MALL per ASIC */ + if (ASICREV_IS_GC_12_0_0_A0(dc->ctx->asic_id.hw_internal_rev)) { + num_available_chans = num_available_chans > 16 ? 16 : num_available_chans; + } else if (ASICREV_IS_GC_12_0_1_A0(dc->ctx->asic_id.hw_internal_rev)) { + num_available_chans = num_available_chans > 8 ? 8 : num_available_chans; + } + + return num_available_chans; +} + static struct resource_funcs dcn401_res_pool_funcs = { .destroy = dcn401_destroy_resource_pool, .link_enc_create = dcn401_link_encoder_create, @@ -1812,14 +1835,12 @@ static bool dcn401_resource_construct( dc->caps.min_horizontal_blanking_period = 80; dc->caps.dmdata_alloc_size = 2048; dc->caps.mall_size_per_mem_channel = 4; - /* total size = mall per channel * num channels * 1024 * 1024 */ - dc->caps.mall_size_total = dc->caps.mall_size_per_mem_channel * dc->ctx->dc_bios->vram_info.num_chans * 1048576; dc->caps.cursor_cache_size = dc->caps.max_cursor_size * dc->caps.max_cursor_size * 8; dc->caps.cache_line_size = 64; dc->caps.cache_num_ways = 16; /* Calculate the available MALL space */ - dc->caps.max_cab_allocation_bytes = dcn32_calc_num_avail_chans_for_mall( + dc->caps.max_cab_allocation_bytes = dcn401_calc_num_avail_chans_for_mall( dc, dc->ctx->dc_bios->vram_info.num_chans) * dc->caps.mall_size_per_mem_channel * 1024 * 1024; dc->caps.mall_size_total = dc->caps.max_cab_allocation_bytes; -- 2.51.0 From c6335dad37c1332124f8d6ed5dcb046614fc9902 Mon Sep 17 00:00:00 2001 From: Samson Tam Date: Wed, 2 Oct 2024 21:51:18 -0400 Subject: [PATCH 03/16] drm/amd/display: add sharpening policy to plane state [Why] Pass in sharpening policy through plane state from control side [How] Add sharpener support through dc_caps. Add sharpen policy to plane state and move to spl_input. Pass sharpen policy from plane state to SPL. Reviewed-by: Aric Cyr Signed-off-by: Samson Tam Signed-off-by: Roman Li Tested-by: Daniel Wheeler Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/display/dc/dc.h | 6 ++++++ drivers/gpu/drm/amd/display/dc/dc_spl_translate.c | 2 +- .../drm/amd/display/dc/resource/dcn401/dcn401_resource.c | 1 + drivers/gpu/drm/amd/display/dc/spl/dc_spl.c | 6 +++--- drivers/gpu/drm/amd/display/dc/spl/dc_spl_types.h | 2 +- 5 files changed, 12 insertions(+), 5 deletions(-) diff --git a/drivers/gpu/drm/amd/display/dc/dc.h b/drivers/gpu/drm/amd/display/dc/dc.h index 621444a1f190..ad8e2a91c306 100644 --- a/drivers/gpu/drm/amd/display/dc/dc.h +++ b/drivers/gpu/drm/amd/display/dc/dc.h @@ -227,6 +227,10 @@ struct dc_dmub_caps { uint8_t fams_ver; }; +struct dc_scl_caps { + bool sharpener_support; +}; + struct dc_caps { uint32_t max_streams; uint32_t max_links; @@ -292,6 +296,7 @@ struct dc_caps { bool sequential_ono; /* Conservative limit for DCC cases which require ODM4:1 to support*/ uint32_t dcc_plane_width_limit; + struct dc_scl_caps scl_caps; }; struct dc_bug_wa { @@ -1355,6 +1360,7 @@ struct dc_plane_state { enum mpcc_movable_cm_location mcm_location; struct dc_csc_transform cursor_csc_color_matrix; bool adaptive_sharpness_en; + int adaptive_sharpness_policy; int sharpness_level; enum linear_light_scaling linear_light_scaling; unsigned int sdr_white_level_nits; 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 20fd0afca8cb..eeffe44be9e3 100644 --- a/drivers/gpu/drm/amd/display/dc/dc_spl_translate.c +++ b/drivers/gpu/drm/amd/display/dc/dc_spl_translate.c @@ -188,7 +188,7 @@ void translate_SPL_in_params_from_pipe_ctx(struct pipe_ctx *pipe_ctx, struct spl spl_in->h_active = pipe_ctx->plane_res.scl_data.h_active; spl_in->v_active = pipe_ctx->plane_res.scl_data.v_active; - spl_in->debug.sharpen_policy = (enum sharpen_policy)pipe_ctx->stream->ctx->dc->debug.sharpen_policy; + spl_in->sharpen_policy = (enum sharpen_policy)plane_state->adaptive_sharpness_policy; spl_in->debug.scale_to_sharpness_policy = (enum scale_to_sharpness_policy)pipe_ctx->stream->ctx->dc->debug.scale_to_sharpness_policy; 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 7da81966b4df..46feff9348d9 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 @@ -2171,6 +2171,7 @@ static bool dcn401_resource_construct( /* SPL */ spl_init_easf_filter_coeffs(); spl_init_blur_scale_coeffs(); + dc->caps.scl_caps.sharpener_support = true; return true; diff --git a/drivers/gpu/drm/amd/display/dc/spl/dc_spl.c b/drivers/gpu/drm/amd/display/dc/spl/dc_spl.c index c38a5c8646e8..f043c7e32e16 100644 --- a/drivers/gpu/drm/amd/display/dc/spl/dc_spl.c +++ b/drivers/gpu/drm/amd/display/dc/spl/dc_spl.c @@ -848,13 +848,13 @@ static bool spl_get_isharp_en(struct spl_in *spl_in, * surfaces based on policy setting */ if (!spl_is_yuv420(spl_in->basic_in.format) && - (spl_in->debug.sharpen_policy == SHARPEN_YUV)) + (spl_in->sharpen_policy == SHARPEN_YUV)) return enable_isharp; else if ((spl_is_yuv420(spl_in->basic_in.format) && !fullscreen) && - (spl_in->debug.sharpen_policy == SHARPEN_RGB_FULLSCREEN_YUV)) + (spl_in->sharpen_policy == SHARPEN_RGB_FULLSCREEN_YUV)) return enable_isharp; else if (!spl_in->is_fullscreen && - spl_in->debug.sharpen_policy == SHARPEN_FULLSCREEN_ALL) + spl_in->sharpen_policy == SHARPEN_FULLSCREEN_ALL) return enable_isharp; /* diff --git a/drivers/gpu/drm/amd/display/dc/spl/dc_spl_types.h b/drivers/gpu/drm/amd/display/dc/spl/dc_spl_types.h index 48e4217555f8..fcb5d389592b 100644 --- a/drivers/gpu/drm/amd/display/dc/spl/dc_spl_types.h +++ b/drivers/gpu/drm/amd/display/dc/spl/dc_spl_types.h @@ -508,7 +508,6 @@ struct spl_funcs { struct spl_debug { int visual_confirm_base_offset; int visual_confirm_dpp_offset; - enum sharpen_policy sharpen_policy; enum scale_to_sharpness_policy scale_to_sharpness_policy; }; @@ -530,6 +529,7 @@ struct spl_in { int h_active; int v_active; int sdr_white_level_nits; + enum sharpen_policy sharpen_policy; }; // end of SPL inputs -- 2.51.0 From c0933f1de2a5eab2fefd07372b415583675cfe9f Mon Sep 17 00:00:00 2001 From: Joshua Aberback Date: Thu, 3 Oct 2024 17:28:11 -0400 Subject: [PATCH 04/16] drm/amd/display: Clear pipe pointers on pipe reset [Why] We want to clean up unnecessary asserts, one of which is an assert in resource_is_pipe_type that fires if a pipe has no stream and still has pointers to other pipes ("dangling state"). This gets hit because pipes are not properly cleaned up in reset_back_end_for_pipe. When resetting a pipe, the existing MPCC / ODM combine pointers are no longer valid, especially when we put ODM in bypass. [How] - reset pipe pointers in reset_back_end_for_pipe - remove useless code to avoid confusion (a long time ago it had a reason to be there, not anymore) Reviewed-by: Alvin Lee Signed-off-by: Joshua Aberback Signed-off-by: Roman Li Tested-by: Daniel Wheeler Signed-off-by: Alex Deucher --- .../gpu/drm/amd/display/dc/hwss/dcn20/dcn20_hwseq.c | 12 ++++-------- .../drm/amd/display/dc/hwss/dcn401/dcn401_hwseq.c | 12 ++++-------- 2 files changed, 8 insertions(+), 16 deletions(-) diff --git a/drivers/gpu/drm/amd/display/dc/hwss/dcn20/dcn20_hwseq.c b/drivers/gpu/drm/amd/display/dc/hwss/dcn20/dcn20_hwseq.c index 9da5b50bea8a..05424a9af58b 100644 --- a/drivers/gpu/drm/amd/display/dc/hwss/dcn20/dcn20_hwseq.c +++ b/drivers/gpu/drm/amd/display/dc/hwss/dcn20/dcn20_hwseq.c @@ -2766,7 +2766,6 @@ void dcn20_reset_back_end_for_pipe( struct pipe_ctx *pipe_ctx, struct dc_state *context) { - int i; struct dc_link *link = pipe_ctx->stream->link; const struct link_hwss *link_hwss = get_link_hwss(link, &pipe_ctx->link_res); @@ -2833,19 +2832,16 @@ void dcn20_reset_back_end_for_pipe( } } - for (i = 0; i < dc->res_pool->pipe_count; i++) - if (&dc->current_state->res_ctx.pipe_ctx[i] == pipe_ctx) - break; - - if (i == dc->res_pool->pipe_count) - return; - /* * In case of a dangling plane, setting this to NULL unconditionally * causes failures during reset hw ctx where, if stream is NULL, * it is expected that the pipe_ctx pointers to pipes and plane are NULL. */ pipe_ctx->stream = NULL; + pipe_ctx->top_pipe = NULL; + pipe_ctx->bottom_pipe = NULL; + pipe_ctx->next_odm_pipe = NULL; + pipe_ctx->prev_odm_pipe = NULL; DC_LOG_DEBUG("Reset back end for pipe %d, tg:%d\n", pipe_ctx->pipe_idx, pipe_ctx->stream_res.tg->inst); } 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 413dcbf4decf..3c70f40bf047 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 @@ -1883,7 +1883,6 @@ void dcn401_reset_back_end_for_pipe( struct pipe_ctx *pipe_ctx, struct dc_state *context) { - int i; struct dc_link *link = pipe_ctx->stream->link; const struct link_hwss *link_hwss = get_link_hwss(link, &pipe_ctx->link_res); @@ -1954,19 +1953,16 @@ void dcn401_reset_back_end_for_pipe( dc->res_pool->dccg->funcs->set_dtbclk_p_src(dc->res_pool->dccg, REFCLK, pipe_ctx->stream_res.tg->inst); } - for (i = 0; i < dc->res_pool->pipe_count; i++) - if (&dc->current_state->res_ctx.pipe_ctx[i] == pipe_ctx) - break; - - if (i == dc->res_pool->pipe_count) - return; - /* * In case of a dangling plane, setting this to NULL unconditionally * causes failures during reset hw ctx where, if stream is NULL, * it is expected that the pipe_ctx pointers to pipes and plane are NULL. */ pipe_ctx->stream = NULL; + pipe_ctx->top_pipe = NULL; + pipe_ctx->bottom_pipe = NULL; + pipe_ctx->next_odm_pipe = NULL; + pipe_ctx->prev_odm_pipe = NULL; DC_LOG_DEBUG("Reset back end for pipe %d, tg:%d\n", pipe_ctx->pipe_idx, pipe_ctx->stream_res.tg->inst); } -- 2.51.0 From 3ff3b48ac8eb303ffc8fd7952199adb302f179f8 Mon Sep 17 00:00:00 2001 From: Alex Hung Date: Thu, 3 Oct 2024 18:26:55 -0600 Subject: [PATCH 05/16] drm/amd/display: Check returned status from core_link_read_dpcd [WHAT] The function core_link_read_dpcd returns status which is not used at all, making them useless assignments. [HOW] Print error messages if core_link_read_dpcd does not return DC_OK. This fixes 2 UNUSED_VALUE issues reported by Coverity. Reviewed-by: Rodrigo Siqueira Signed-off-by: Alex Hung Signed-off-by: Roman Li Tested-by: Daniel Wheeler Signed-off-by: Alex Deucher --- .../drm/amd/display/dc/link/protocols/link_dp_capability.c | 6 ++++++ 1 file changed, 6 insertions(+) diff --git a/drivers/gpu/drm/amd/display/dc/link/protocols/link_dp_capability.c b/drivers/gpu/drm/amd/display/dc/link/protocols/link_dp_capability.c index 952e787493c4..518e5d1f3d90 100644 --- a/drivers/gpu/drm/amd/display/dc/link/protocols/link_dp_capability.c +++ b/drivers/gpu/drm/amd/display/dc/link/protocols/link_dp_capability.c @@ -1854,6 +1854,9 @@ static bool retrieve_link_cap(struct dc_link *link) DP_FEC_CAPABILITY, &link->dpcd_caps.fec_cap.raw, sizeof(link->dpcd_caps.fec_cap.raw)); + if (status != DC_OK) + DC_LOG_ERROR("%s:%d: core_link_read_dpcd (DP_FEC_CAPABILITY) failed\n", __func__, __LINE__); + status = core_link_read_dpcd( link, DP_DSC_SUPPORT, @@ -1876,6 +1879,9 @@ static bool retrieve_link_cap(struct dc_link *link) DP_DSC_BRANCH_OVERALL_THROUGHPUT_0, link->dpcd_caps.dsc_caps.dsc_branch_decoder_caps.raw, sizeof(link->dpcd_caps.dsc_caps.dsc_branch_decoder_caps.raw)); + if (status != DC_OK) + DC_LOG_ERROR("%s:%d: core_link_read_dpcd (DP_DSC_BRANCH_OVERALL_THROUGHPUT_0) failed\n", __func__, __LINE__); + DC_LOG_DSC("DSC branch decoder capability is read at link %d", link->link_index); DC_LOG_DSC("\tBRANCH_OVERALL_THROUGHPUT_0 = 0x%02x", link->dpcd_caps.dsc_caps.dsc_branch_decoder_caps.fields.BRANCH_OVERALL_THROUGHPUT_0); -- 2.51.0 From ef7e217f50159ca10d120a1c6ea09c6401019b3b Mon Sep 17 00:00:00 2001 From: Samson Tam Date: Fri, 4 Oct 2024 12:02:36 -0400 Subject: [PATCH 06/16] drm/amd/display: disable dynamic ODM when sharpness is enabled [Why & How] Disable dynamic ODM when sharpness is enabled Reviewed-by: Alvin Lee Signed-off-by: Samson Tam Signed-off-by: Roman Li Tested-by: Daniel Wheeler Signed-off-by: Alex Deucher --- .../gpu/drm/amd/display/dc/dml2/dml21/dml21_translation_helper.c | 1 + 1 file changed, 1 insertion(+) 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 7a01a956e4bb..d901b9f7bc6a 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 @@ -514,6 +514,7 @@ static void populate_dml21_stream_overrides_from_stream_state( break; } if (!stream->ctx->dc->debug.enable_single_display_2to1_odm_policy || + ((stream->ctx->dc->debug.force_sharpness > 1) && stream->ctx->dc->config.use_spl) || stream->debug.force_odm_combine_segments > 0) stream_desc->overrides.disable_dynamic_odm = true; stream_desc->overrides.disable_subvp = stream->ctx->dc->debug.force_disable_subvp || stream->hw_cursor_req; -- 2.51.0 From f9f3e8236e9e8fb1926297e23389905b63400a87 Mon Sep 17 00:00:00 2001 From: Aric Cyr Date: Mon, 7 Oct 2024 09:19:15 -0400 Subject: [PATCH 07/16] drm/amd/display: 3.2.305 - Add sharpening policy to plane state - Clear pipe pointers on pipe reset - Resolve correct MALL size for dcn401 - Read Sink emission rate capability - IPX fixes - Coverity fixes Reviewed-by: Roman Li Signed-off-by: Aric Cyr Signed-off-by: Roman Li Tested-by: Daniel Wheeler Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/display/dc/dc.h | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/drivers/gpu/drm/amd/display/dc/dc.h b/drivers/gpu/drm/amd/display/dc/dc.h index ad8e2a91c306..6ba167d4c785 100644 --- a/drivers/gpu/drm/amd/display/dc/dc.h +++ b/drivers/gpu/drm/amd/display/dc/dc.h @@ -55,7 +55,7 @@ struct aux_payload; struct set_config_cmd_payload; struct dmub_notification; -#define DC_VER "3.2.304" +#define DC_VER "3.2.305" #define MAX_SURFACES 3 #define MAX_PLANES 6 -- 2.51.0 From 62f38b4ccaa6aa063ca781d80b10aacd39dc5c76 Mon Sep 17 00:00:00 2001 From: Alex Deucher Date: Thu, 3 Oct 2024 10:09:50 -0400 Subject: [PATCH 08/16] drm/amdgpu/smu13: always apply the powersave optimization It can avoid margin issues in some very demanding applications. Closes: https://gitlab.freedesktop.org/drm/amd/-/issues/3618 Link: https://gitlab.freedesktop.org/drm/amd/-/issues/3131 Fixes: c50fe289ed72 ("drm/amdgpu/swsmu: always force a state reprogram on init") Reviewed-by: Kenneth Feng Signed-off-by: Alex Deucher --- .../drm/amd/pm/swsmu/smu13/smu_v13_0_0_ppt.c | 22 +++++++++---------- 1 file changed, 10 insertions(+), 12 deletions(-) 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 866dc77d1005..4e658abe3a20 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 @@ -2542,18 +2542,16 @@ static int smu_v13_0_0_set_power_profile_mode(struct smu_context *smu, workload_mask = 1 << workload_type; /* Add optimizations for SMU13.0.0/10. Reuse the power saving profile */ - if (smu->power_profile_mode == PP_SMC_POWER_PROFILE_COMPUTE) { - if ((amdgpu_ip_version(smu->adev, MP1_HWIP, 0) == IP_VERSION(13, 0, 0) && - ((smu->adev->pm.fw_version == 0x004e6601) || - (smu->adev->pm.fw_version >= 0x004e7300))) || - (amdgpu_ip_version(smu->adev, MP1_HWIP, 0) == IP_VERSION(13, 0, 10) && - smu->adev->pm.fw_version >= 0x00504500)) { - workload_type = smu_cmn_to_asic_specific_index(smu, - CMN2ASIC_MAPPING_WORKLOAD, - PP_SMC_POWER_PROFILE_POWERSAVING); - if (workload_type >= 0) - workload_mask |= 1 << workload_type; - } + if ((amdgpu_ip_version(smu->adev, MP1_HWIP, 0) == IP_VERSION(13, 0, 0) && + ((smu->adev->pm.fw_version == 0x004e6601) || + (smu->adev->pm.fw_version >= 0x004e7300))) || + (amdgpu_ip_version(smu->adev, MP1_HWIP, 0) == IP_VERSION(13, 0, 10) && + smu->adev->pm.fw_version >= 0x00504500)) { + workload_type = smu_cmn_to_asic_specific_index(smu, + CMN2ASIC_MAPPING_WORKLOAD, + PP_SMC_POWER_PROFILE_POWERSAVING); + if (workload_type >= 0) + workload_mask |= 1 << workload_type; } ret = smu_cmn_send_smc_msg_with_param(smu, -- 2.51.0 From 3b6e7d40649c0d75572039aff9d0911864c689db Mon Sep 17 00:00:00 2001 From: =?utf8?q?Ville=20Syrj=C3=A4l=C3=A4?= Date: Mon, 14 Oct 2024 19:09:36 +0300 Subject: [PATCH 09/16] drm/radeon: Fix encoder->possible_clones MIME-Version: 1.0 Content-Type: text/plain; charset=utf8 Content-Transfer-Encoding: 8bit Include the encoder itself in its possible_clones bitmask. In the past nothing validated that drivers were populating possible_clones correctly, but that changed in commit 74d2aacbe840 ("drm: Validate encoder->possible_clones"). Looks like radeon never got the memo and is still not following the rules 100% correctly. This results in some warnings during driver initialization: Bogus possible_clones: [ENCODER:46:TV-46] possible_clones=0x4 (full encoder mask=0x7) WARNING: CPU: 0 PID: 170 at drivers/gpu/drm/drm_mode_config.c:615 drm_mode_config_validate+0x113/0x39c ... Cc: Alex Deucher Cc: amd-gfx@lists.freedesktop.org Fixes: 74d2aacbe840 ("drm: Validate encoder->possible_clones") Reported-by: Erhard Furtner Closes: https://lore.kernel.org/dri-devel/20241009000321.418e4294@yea/ Tested-by: Erhard Furtner Signed-off-by: Ville Syrjälä Signed-off-by: Alex Deucher --- drivers/gpu/drm/radeon/radeon_encoders.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/drivers/gpu/drm/radeon/radeon_encoders.c b/drivers/gpu/drm/radeon/radeon_encoders.c index 0f723292409e..fafed331e0a0 100644 --- a/drivers/gpu/drm/radeon/radeon_encoders.c +++ b/drivers/gpu/drm/radeon/radeon_encoders.c @@ -43,7 +43,7 @@ static uint32_t radeon_encoder_clones(struct drm_encoder *encoder) struct radeon_device *rdev = dev->dev_private; struct radeon_encoder *radeon_encoder = to_radeon_encoder(encoder); struct drm_encoder *clone_encoder; - uint32_t index_mask = 0; + uint32_t index_mask = drm_encoder_mask(encoder); int count; /* DIG routing gets problematic */ -- 2.51.0 From 580ad7cbd4b7be8d2cb5ab5c1fca6bb76045eb0e Mon Sep 17 00:00:00 2001 From: Alex Deucher Date: Wed, 2 Oct 2024 10:22:30 -0400 Subject: [PATCH 10/16] drm/amdgpu/swsmu: Only force workload setup on init Needed to set the workload type at init time so that we can apply the navi3x margin optimization. Closes: https://gitlab.freedesktop.org/drm/amd/-/issues/3618 Link: https://gitlab.freedesktop.org/drm/amd/-/issues/3131 Fixes: c50fe289ed72 ("drm/amdgpu/swsmu: always force a state reprogram on init") Reviewed-by: Kenneth Feng Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/pm/swsmu/amdgpu_smu.c | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-) diff --git a/drivers/gpu/drm/amd/pm/swsmu/amdgpu_smu.c b/drivers/gpu/drm/amd/pm/swsmu/amdgpu_smu.c index ec92e24b7bda..89012029e8d7 100644 --- a/drivers/gpu/drm/amd/pm/swsmu/amdgpu_smu.c +++ b/drivers/gpu/drm/amd/pm/swsmu/amdgpu_smu.c @@ -2232,7 +2232,7 @@ static int smu_bump_power_profile_mode(struct smu_context *smu, static int smu_adjust_power_state_dynamic(struct smu_context *smu, enum amd_dpm_forced_level level, bool skip_display_settings, - bool force_update) + bool init) { int ret = 0; int index = 0; @@ -2261,7 +2261,7 @@ static int smu_adjust_power_state_dynamic(struct smu_context *smu, } } - if (force_update || smu_dpm_ctx->dpm_level != level) { + if (smu_dpm_ctx->dpm_level != level) { ret = smu_asic_set_performance_level(smu, level); if (ret) { dev_err(smu->adev->dev, "Failed to set performance level!"); @@ -2278,7 +2278,7 @@ static int smu_adjust_power_state_dynamic(struct smu_context *smu, index = index > 0 && index <= WORKLOAD_POLICY_MAX ? index - 1 : 0; workload[0] = smu->workload_setting[index]; - if (force_update || smu->power_profile_mode != workload[0]) + if (init || smu->power_profile_mode != workload[0]) smu_bump_power_profile_mode(smu, workload, 0); } -- 2.51.0 From 336568de918e08c825b3b1cbe2ec809f2fc26d94 Mon Sep 17 00:00:00 2001 From: Alex Deucher Date: Thu, 3 Oct 2024 09:57:38 -0400 Subject: [PATCH 11/16] drm/amdgpu/swsmu: default to fullscreen 3D profile for dGPUs This uses more aggressive hueristics than the the bootup default profile. On windows the OS has a special fullscreen 3D mode where this is used. Since we don't have the equivalent on Linux default to this profile for dGPUs. Closes: https://gitlab.freedesktop.org/drm/amd/-/issues/3618 Closes: https://gitlab.freedesktop.org/drm/amd/-/issues/1500 Link: https://gitlab.freedesktop.org/drm/amd/-/issues/3131 Fixes: c50fe289ed72 ("drm/amdgpu/swsmu: always force a state reprogram on init") Reviewed-by: Kenneth Feng Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/pm/swsmu/amdgpu_smu.c | 6 +++++- 1 file changed, 5 insertions(+), 1 deletion(-) diff --git a/drivers/gpu/drm/amd/pm/swsmu/amdgpu_smu.c b/drivers/gpu/drm/amd/pm/swsmu/amdgpu_smu.c index 89012029e8d7..0bd88212ff25 100644 --- a/drivers/gpu/drm/amd/pm/swsmu/amdgpu_smu.c +++ b/drivers/gpu/drm/amd/pm/swsmu/amdgpu_smu.c @@ -1265,7 +1265,11 @@ static int smu_sw_init(struct amdgpu_ip_block *ip_block) 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; - smu->workload_mask = 1 << smu->workload_prority[PP_SMC_POWER_PROFILE_BOOTUP_DEFAULT]; + + if (smu->is_apu) + 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_setting[0] = PP_SMC_POWER_PROFILE_BOOTUP_DEFAULT; smu->workload_setting[1] = PP_SMC_POWER_PROFILE_FULLSCREEN3D; -- 2.51.0 From 9f7e94af35996effead76ff2837b5ba322e38963 Mon Sep 17 00:00:00 2001 From: Dan Carpenter Date: Thu, 10 Oct 2024 21:35:36 +0300 Subject: [PATCH 12/16] drm/amdgpu: Fix off by one in current_memory_partition_show() The >= ARRAY_SIZE() should be > ARRAY_SIZE() to prevent an out of bounds read. Fixes: 012be6f22c01 ("drm/amdgpu: Add sysfs interfaces for NPS mode") Reviewed-by: Lijo Lazar Signed-off-by: Dan Carpenter Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.c index 4e0a2d86a726..1c19a65e6553 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.c @@ -1199,7 +1199,7 @@ static ssize_t current_memory_partition_show( enum amdgpu_memory_partition mode; mode = adev->gmc.gmc_funcs->query_mem_partition_mode(adev); - if ((mode > ARRAY_SIZE(nps_desc)) || + if ((mode >= ARRAY_SIZE(nps_desc)) || (BIT(mode) & AMDGPU_ALL_NPS_MASK) != BIT(mode)) return sysfs_emit(buf, "UNKNOWN\n"); -- 2.51.0 From 3d73327b748baaddf82cfad791fc25f8f1822ffe Mon Sep 17 00:00:00 2001 From: Alex Deucher Date: Tue, 1 Oct 2024 10:31:26 -0400 Subject: [PATCH 13/16] drm/amdgpu/swsmu: add automatic parameter to set_soft_freq_range On chips that support it, you can specificy 0 and 0xffff for min and max and the PMFW will use that to determine the optimal min and max. This enables optimal performance when the user manually switches between performance levels using sysfs. Previously we'd set soft min/max which could limit performance. Reviewed-by: Kenneth Feng Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/pm/swsmu/amdgpu_smu.c | 3 +- drivers/gpu/drm/amd/pm/swsmu/inc/amdgpu_smu.h | 3 +- drivers/gpu/drm/amd/pm/swsmu/inc/smu_v11_0.h | 2 +- drivers/gpu/drm/amd/pm/swsmu/inc/smu_v12_0.h | 2 +- drivers/gpu/drm/amd/pm/swsmu/inc/smu_v13_0.h | 2 +- drivers/gpu/drm/amd/pm/swsmu/inc/smu_v14_0.h | 2 +- .../gpu/drm/amd/pm/swsmu/smu11/navi10_ppt.c | 2 +- .../amd/pm/swsmu/smu11/sienna_cichlid_ppt.c | 2 +- .../gpu/drm/amd/pm/swsmu/smu11/smu_v11_0.c | 25 +++++++++---- .../gpu/drm/amd/pm/swsmu/smu11/vangogh_ppt.c | 19 +++++----- .../gpu/drm/amd/pm/swsmu/smu12/renoir_ppt.c | 14 ++++---- .../gpu/drm/amd/pm/swsmu/smu12/smu_v12_0.c | 2 +- .../drm/amd/pm/swsmu/smu13/aldebaran_ppt.c | 15 ++++---- .../gpu/drm/amd/pm/swsmu/smu13/smu_v13_0.c | 34 +++++++++++++----- .../drm/amd/pm/swsmu/smu13/smu_v13_0_0_ppt.c | 3 +- .../drm/amd/pm/swsmu/smu13/smu_v13_0_5_ppt.c | 22 +++++++----- .../drm/amd/pm/swsmu/smu13/smu_v13_0_6_ppt.c | 15 ++++---- .../drm/amd/pm/swsmu/smu13/smu_v13_0_7_ppt.c | 3 +- .../drm/amd/pm/swsmu/smu13/yellow_carp_ppt.c | 36 +++++++++++-------- .../gpu/drm/amd/pm/swsmu/smu14/smu_v14_0.c | 33 ++++++++++++----- .../drm/amd/pm/swsmu/smu14/smu_v14_0_2_ppt.c | 3 +- 21 files changed, 152 insertions(+), 90 deletions(-) diff --git a/drivers/gpu/drm/amd/pm/swsmu/amdgpu_smu.c b/drivers/gpu/drm/amd/pm/swsmu/amdgpu_smu.c index 0bd88212ff25..82b88326117f 100644 --- a/drivers/gpu/drm/amd/pm/swsmu/amdgpu_smu.c +++ b/drivers/gpu/drm/amd/pm/swsmu/amdgpu_smu.c @@ -140,7 +140,8 @@ int smu_set_soft_freq_range(struct smu_context *smu, ret = smu->ppt_funcs->set_soft_freq_limited_range(smu, clk_type, min, - max); + max, + false); return ret; } 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 727da0c37e06..8bb32b3f0d9c 100644 --- a/drivers/gpu/drm/amd/pm/swsmu/inc/amdgpu_smu.h +++ b/drivers/gpu/drm/amd/pm/swsmu/inc/amdgpu_smu.h @@ -1255,7 +1255,8 @@ struct pptable_funcs { * @set_soft_freq_limited_range: Set the soft frequency range of a clock * domain in MHz. */ - int (*set_soft_freq_limited_range)(struct smu_context *smu, enum smu_clk_type clk_type, uint32_t min, uint32_t max); + int (*set_soft_freq_limited_range)(struct smu_context *smu, enum smu_clk_type clk_type, uint32_t min, uint32_t max, + bool automatic); /** * @set_power_source: Notify the SMU of the current power source. diff --git a/drivers/gpu/drm/amd/pm/swsmu/inc/smu_v11_0.h b/drivers/gpu/drm/amd/pm/swsmu/inc/smu_v11_0.h index c2ab336bb530..ed8304d82831 100644 --- a/drivers/gpu/drm/amd/pm/swsmu/inc/smu_v11_0.h +++ b/drivers/gpu/drm/amd/pm/swsmu/inc/smu_v11_0.h @@ -255,7 +255,7 @@ int smu_v11_0_get_dpm_ultimate_freq(struct smu_context *smu, enum smu_clk_type c uint32_t *min, uint32_t *max); int smu_v11_0_set_soft_freq_limited_range(struct smu_context *smu, enum smu_clk_type clk_type, - uint32_t min, uint32_t max); + uint32_t min, uint32_t max, bool automatic); int smu_v11_0_set_hard_freq_limited_range(struct smu_context *smu, enum smu_clk_type clk_type, diff --git a/drivers/gpu/drm/amd/pm/swsmu/inc/smu_v12_0.h b/drivers/gpu/drm/amd/pm/swsmu/inc/smu_v12_0.h index 1ad2dff71090..0886d8cffbd0 100644 --- a/drivers/gpu/drm/amd/pm/swsmu/inc/smu_v12_0.h +++ b/drivers/gpu/drm/amd/pm/swsmu/inc/smu_v12_0.h @@ -56,7 +56,7 @@ int smu_v12_0_set_default_dpm_tables(struct smu_context *smu); int smu_v12_0_mode2_reset(struct smu_context *smu); int smu_v12_0_set_soft_freq_limited_range(struct smu_context *smu, enum smu_clk_type clk_type, - uint32_t min, uint32_t max); + uint32_t min, uint32_t max, bool automatic); int smu_v12_0_set_driver_table_location(struct smu_context *smu); diff --git a/drivers/gpu/drm/amd/pm/swsmu/inc/smu_v13_0.h b/drivers/gpu/drm/amd/pm/swsmu/inc/smu_v13_0.h index e58220a7ee2f..044d6893b43e 100644 --- a/drivers/gpu/drm/amd/pm/swsmu/inc/smu_v13_0.h +++ b/drivers/gpu/drm/amd/pm/swsmu/inc/smu_v13_0.h @@ -219,7 +219,7 @@ int smu_v13_0_get_dpm_ultimate_freq(struct smu_context *smu, enum smu_clk_type c uint32_t *min, uint32_t *max); int smu_v13_0_set_soft_freq_limited_range(struct smu_context *smu, enum smu_clk_type clk_type, - uint32_t min, uint32_t max); + uint32_t min, uint32_t max, bool automatic); int smu_v13_0_set_hard_freq_limited_range(struct smu_context *smu, enum smu_clk_type clk_type, diff --git a/drivers/gpu/drm/amd/pm/swsmu/inc/smu_v14_0.h b/drivers/gpu/drm/amd/pm/swsmu/inc/smu_v14_0.h index 46b456590a08..6cada19a8482 100644 --- a/drivers/gpu/drm/amd/pm/swsmu/inc/smu_v14_0.h +++ b/drivers/gpu/drm/amd/pm/swsmu/inc/smu_v14_0.h @@ -186,7 +186,7 @@ int smu_v14_0_get_dpm_ultimate_freq(struct smu_context *smu, enum smu_clk_type c uint32_t *min, uint32_t *max); int smu_v14_0_set_soft_freq_limited_range(struct smu_context *smu, enum smu_clk_type clk_type, - uint32_t min, uint32_t max); + uint32_t min, uint32_t max, bool automatic); int smu_v14_0_set_hard_freq_limited_range(struct smu_context *smu, enum smu_clk_type clk_type, 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 16af1a329621..9fa305ba6422 100644 --- a/drivers/gpu/drm/amd/pm/swsmu/smu11/navi10_ppt.c +++ b/drivers/gpu/drm/amd/pm/swsmu/smu11/navi10_ppt.c @@ -1689,7 +1689,7 @@ static int navi10_force_clk_levels(struct smu_context *smu, if (ret) return 0; - ret = smu_v11_0_set_soft_freq_limited_range(smu, clk_type, min_freq, max_freq); + ret = smu_v11_0_set_soft_freq_limited_range(smu, clk_type, min_freq, max_freq, false); if (ret) return 0; break; diff --git a/drivers/gpu/drm/amd/pm/swsmu/smu11/sienna_cichlid_ppt.c b/drivers/gpu/drm/amd/pm/swsmu/smu11/sienna_cichlid_ppt.c index 3b29cb74c47b..77e58eb46328 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 @@ -1469,7 +1469,7 @@ static int sienna_cichlid_force_clk_levels(struct smu_context *smu, if (ret) goto forec_level_out; - ret = smu_v11_0_set_soft_freq_limited_range(smu, clk_type, min_freq, max_freq); + ret = smu_v11_0_set_soft_freq_limited_range(smu, clk_type, min_freq, max_freq, false); if (ret) goto forec_level_out; break; diff --git a/drivers/gpu/drm/amd/pm/swsmu/smu11/smu_v11_0.c b/drivers/gpu/drm/amd/pm/swsmu/smu11/smu_v11_0.c index 8981302b19c8..480cf3cb204d 100644 --- a/drivers/gpu/drm/amd/pm/swsmu/smu11/smu_v11_0.c +++ b/drivers/gpu/drm/amd/pm/swsmu/smu11/smu_v11_0.c @@ -1764,7 +1764,8 @@ failed: int smu_v11_0_set_soft_freq_limited_range(struct smu_context *smu, enum smu_clk_type clk_type, uint32_t min, - uint32_t max) + uint32_t max, + bool automatic) { int ret = 0, clk_id = 0; uint32_t param; @@ -1779,7 +1780,10 @@ int smu_v11_0_set_soft_freq_limited_range(struct smu_context *smu, return clk_id; if (max > 0) { - param = (uint32_t)((clk_id << 16) | (max & 0xffff)); + if (automatic) + param = (uint32_t)((clk_id << 16) | 0xffff); + else + param = (uint32_t)((clk_id << 16) | (max & 0xffff)); ret = smu_cmn_send_smc_msg_with_param(smu, SMU_MSG_SetSoftMaxByFreq, param, NULL); if (ret) @@ -1787,7 +1791,10 @@ int smu_v11_0_set_soft_freq_limited_range(struct smu_context *smu, } if (min > 0) { - param = (uint32_t)((clk_id << 16) | (min & 0xffff)); + if (automatic) + param = (uint32_t)((clk_id << 16) | 0); + else + param = (uint32_t)((clk_id << 16) | (min & 0xffff)); ret = smu_cmn_send_smc_msg_with_param(smu, SMU_MSG_SetSoftMinByFreq, param, NULL); if (ret) @@ -1855,6 +1862,7 @@ int smu_v11_0_set_performance_level(struct smu_context *smu, uint32_t mclk_min = 0, mclk_max = 0; uint32_t socclk_min = 0, socclk_max = 0; int ret = 0; + bool auto_level = false; switch (level) { case AMD_DPM_FORCED_LEVEL_HIGH: @@ -1874,6 +1882,7 @@ int smu_v11_0_set_performance_level(struct smu_context *smu, mclk_max = mem_table->max; socclk_min = soc_table->min; socclk_max = soc_table->max; + auto_level = true; break; case AMD_DPM_FORCED_LEVEL_PROFILE_STANDARD: sclk_min = sclk_max = pstate_table->gfxclk_pstate.standard; @@ -1906,13 +1915,15 @@ int smu_v11_0_set_performance_level(struct smu_context *smu, if (amdgpu_ip_version(adev, MP1_HWIP, 0) == IP_VERSION(11, 0, 2)) { mclk_min = mclk_max = 0; socclk_min = socclk_max = 0; + auto_level = false; } if (sclk_min && sclk_max) { ret = smu_v11_0_set_soft_freq_limited_range(smu, SMU_GFXCLK, sclk_min, - sclk_max); + sclk_max, + auto_level); if (ret) return ret; } @@ -1921,7 +1932,8 @@ int smu_v11_0_set_performance_level(struct smu_context *smu, ret = smu_v11_0_set_soft_freq_limited_range(smu, SMU_MCLK, mclk_min, - mclk_max); + mclk_max, + auto_level); if (ret) return ret; } @@ -1930,7 +1942,8 @@ int smu_v11_0_set_performance_level(struct smu_context *smu, ret = smu_v11_0_set_soft_freq_limited_range(smu, SMU_SOCCLK, socclk_min, - socclk_max); + socclk_max, + auto_level); if (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 22737b11b1bf..a333ab827f48 100644 --- a/drivers/gpu/drm/amd/pm/swsmu/smu11/vangogh_ppt.c +++ b/drivers/gpu/drm/amd/pm/swsmu/smu11/vangogh_ppt.c @@ -1091,9 +1091,10 @@ static int vangogh_set_power_profile_mode(struct smu_context *smu, long *input, } static int vangogh_set_soft_freq_limited_range(struct smu_context *smu, - enum smu_clk_type clk_type, - uint32_t min, - uint32_t max) + enum smu_clk_type clk_type, + uint32_t min, + uint32_t max, + bool automatic) { int ret = 0; @@ -1299,7 +1300,7 @@ static int vangogh_force_dpm_limit_value(struct smu_context *smu, bool highest) return ret; force_freq = highest ? max_freq : min_freq; - ret = vangogh_set_soft_freq_limited_range(smu, clk_type, force_freq, force_freq); + ret = vangogh_set_soft_freq_limited_range(smu, clk_type, force_freq, force_freq, false); if (ret) return ret; } @@ -1335,7 +1336,7 @@ static int vangogh_unforce_dpm_levels(struct smu_context *smu) if (ret) return ret; - ret = vangogh_set_soft_freq_limited_range(smu, clk_type, min_freq, max_freq); + ret = vangogh_set_soft_freq_limited_range(smu, clk_type, min_freq, max_freq, false); if (ret) return ret; @@ -1354,7 +1355,7 @@ static int vangogh_set_peak_clock_by_device(struct smu_context *smu) if (ret) return ret; - ret = vangogh_set_soft_freq_limited_range(smu, SMU_FCLK, fclk_freq, fclk_freq); + ret = vangogh_set_soft_freq_limited_range(smu, SMU_FCLK, fclk_freq, fclk_freq, false); if (ret) return ret; @@ -1362,7 +1363,7 @@ static int vangogh_set_peak_clock_by_device(struct smu_context *smu) if (ret) return ret; - ret = vangogh_set_soft_freq_limited_range(smu, SMU_SOCCLK, socclk_freq, socclk_freq); + ret = vangogh_set_soft_freq_limited_range(smu, SMU_SOCCLK, socclk_freq, socclk_freq, false); if (ret) return ret; @@ -1370,7 +1371,7 @@ static int vangogh_set_peak_clock_by_device(struct smu_context *smu) if (ret) return ret; - ret = vangogh_set_soft_freq_limited_range(smu, SMU_VCLK, vclk_freq, vclk_freq); + ret = vangogh_set_soft_freq_limited_range(smu, SMU_VCLK, vclk_freq, vclk_freq, false); if (ret) return ret; @@ -1378,7 +1379,7 @@ static int vangogh_set_peak_clock_by_device(struct smu_context *smu) if (ret) return ret; - ret = vangogh_set_soft_freq_limited_range(smu, SMU_DCLK, dclk_freq, dclk_freq); + ret = vangogh_set_soft_freq_limited_range(smu, SMU_DCLK, dclk_freq, dclk_freq, false); if (ret) return ret; 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 cc0504b063fa..0b210b1f2628 100644 --- a/drivers/gpu/drm/amd/pm/swsmu/smu12/renoir_ppt.c +++ b/drivers/gpu/drm/amd/pm/swsmu/smu12/renoir_ppt.c @@ -707,7 +707,7 @@ static int renoir_force_dpm_limit_value(struct smu_context *smu, bool highest) return ret; force_freq = highest ? max_freq : min_freq; - ret = smu_v12_0_set_soft_freq_limited_range(smu, clk_type, force_freq, force_freq); + ret = smu_v12_0_set_soft_freq_limited_range(smu, clk_type, force_freq, force_freq, false); if (ret) return ret; } @@ -740,7 +740,7 @@ static int renoir_unforce_dpm_levels(struct smu_context *smu) { if (ret) return ret; - ret = smu_v12_0_set_soft_freq_limited_range(smu, clk_type, min_freq, max_freq); + ret = smu_v12_0_set_soft_freq_limited_range(smu, clk_type, min_freq, max_freq, false); if (ret) return ret; } @@ -911,7 +911,7 @@ static int renoir_set_peak_clock_by_device(struct smu_context *smu) if (ret) return ret; - ret = smu_v12_0_set_soft_freq_limited_range(smu, SMU_SCLK, sclk_freq, sclk_freq); + ret = smu_v12_0_set_soft_freq_limited_range(smu, SMU_SCLK, sclk_freq, sclk_freq, false); if (ret) return ret; @@ -919,7 +919,7 @@ static int renoir_set_peak_clock_by_device(struct smu_context *smu) if (ret) return ret; - ret = smu_v12_0_set_soft_freq_limited_range(smu, SMU_UCLK, uclk_freq, uclk_freq); + ret = smu_v12_0_set_soft_freq_limited_range(smu, SMU_UCLK, uclk_freq, uclk_freq, false); if (ret) return ret; @@ -961,13 +961,13 @@ static int renior_set_dpm_profile_freq(struct smu_context *smu, } if (sclk) - ret = smu_v12_0_set_soft_freq_limited_range(smu, SMU_SCLK, sclk, sclk); + ret = smu_v12_0_set_soft_freq_limited_range(smu, SMU_SCLK, sclk, sclk, false); if (socclk) - ret = smu_v12_0_set_soft_freq_limited_range(smu, SMU_SOCCLK, socclk, socclk); + ret = smu_v12_0_set_soft_freq_limited_range(smu, SMU_SOCCLK, socclk, socclk, false); if (fclk) - ret = smu_v12_0_set_soft_freq_limited_range(smu, SMU_FCLK, fclk, fclk); + ret = smu_v12_0_set_soft_freq_limited_range(smu, SMU_FCLK, fclk, fclk, false); return ret; } diff --git a/drivers/gpu/drm/amd/pm/swsmu/smu12/smu_v12_0.c b/drivers/gpu/drm/amd/pm/swsmu/smu12/smu_v12_0.c index ed15f5a0fd11..3d3cd546f0ad 100644 --- a/drivers/gpu/drm/amd/pm/swsmu/smu12/smu_v12_0.c +++ b/drivers/gpu/drm/amd/pm/swsmu/smu12/smu_v12_0.c @@ -211,7 +211,7 @@ int smu_v12_0_mode2_reset(struct smu_context *smu) } int smu_v12_0_set_soft_freq_limited_range(struct smu_context *smu, enum smu_clk_type clk_type, - uint32_t min, uint32_t max) + uint32_t min, uint32_t max, bool automatic) { int ret = 0; diff --git a/drivers/gpu/drm/amd/pm/swsmu/smu13/aldebaran_ppt.c b/drivers/gpu/drm/amd/pm/swsmu/smu13/aldebaran_ppt.c index 2c35eb31475a..f6b029354327 100644 --- a/drivers/gpu/drm/amd/pm/swsmu/smu13/aldebaran_ppt.c +++ b/drivers/gpu/drm/amd/pm/swsmu/smu13/aldebaran_ppt.c @@ -1297,9 +1297,10 @@ static int aldebaran_set_performance_level(struct smu_context *smu, } static int aldebaran_set_soft_freq_limited_range(struct smu_context *smu, - enum smu_clk_type clk_type, - uint32_t min, - uint32_t max) + enum smu_clk_type clk_type, + uint32_t min, + uint32_t max, + bool automatic) { struct smu_dpm_context *smu_dpm = &(smu->smu_dpm); struct smu_13_0_dpm_context *dpm_context = smu_dpm->dpm_context; @@ -1328,7 +1329,7 @@ static int aldebaran_set_soft_freq_limited_range(struct smu_context *smu, return 0; ret = smu_v13_0_set_soft_freq_limited_range(smu, SMU_GFXCLK, - min, max); + min, max, false); if (!ret) { pstate_table->gfxclk_pstate.curr.min = min; pstate_table->gfxclk_pstate.curr.max = max; @@ -1348,7 +1349,7 @@ static int aldebaran_set_soft_freq_limited_range(struct smu_context *smu, /* Restore default min/max clocks and enable determinism */ min_clk = dpm_context->dpm_tables.gfx_table.min; max_clk = dpm_context->dpm_tables.gfx_table.max; - ret = smu_v13_0_set_soft_freq_limited_range(smu, SMU_GFXCLK, min_clk, max_clk); + ret = smu_v13_0_set_soft_freq_limited_range(smu, SMU_GFXCLK, min_clk, max_clk, false); if (!ret) { usleep_range(500, 1000); ret = smu_cmn_send_smc_msg_with_param(smu, @@ -1422,7 +1423,7 @@ static int aldebaran_usr_edit_dpm_table(struct smu_context *smu, enum PP_OD_DPM_ min_clk = dpm_context->dpm_tables.gfx_table.min; max_clk = dpm_context->dpm_tables.gfx_table.max; - return aldebaran_set_soft_freq_limited_range(smu, SMU_GFXCLK, min_clk, max_clk); + return aldebaran_set_soft_freq_limited_range(smu, SMU_GFXCLK, min_clk, max_clk, false); } break; case PP_OD_COMMIT_DPM_TABLE: @@ -1441,7 +1442,7 @@ static int aldebaran_usr_edit_dpm_table(struct smu_context *smu, enum PP_OD_DPM_ min_clk = pstate_table->gfxclk_pstate.custom.min; max_clk = pstate_table->gfxclk_pstate.custom.max; - return aldebaran_set_soft_freq_limited_range(smu, SMU_GFXCLK, min_clk, max_clk); + return aldebaran_set_soft_freq_limited_range(smu, SMU_GFXCLK, min_clk, max_clk, false); } break; default: diff --git a/drivers/gpu/drm/amd/pm/swsmu/smu13/smu_v13_0.c b/drivers/gpu/drm/amd/pm/swsmu/smu13/smu_v13_0.c index e17466cc1952..6cfd66363915 100644 --- a/drivers/gpu/drm/amd/pm/swsmu/smu13/smu_v13_0.c +++ b/drivers/gpu/drm/amd/pm/swsmu/smu13/smu_v13_0.c @@ -1608,7 +1608,8 @@ failed: int smu_v13_0_set_soft_freq_limited_range(struct smu_context *smu, enum smu_clk_type clk_type, uint32_t min, - uint32_t max) + uint32_t max, + bool automatic) { int ret = 0, clk_id = 0; uint32_t param; @@ -1623,7 +1624,10 @@ int smu_v13_0_set_soft_freq_limited_range(struct smu_context *smu, return clk_id; if (max > 0) { - param = (uint32_t)((clk_id << 16) | (max & 0xffff)); + if (automatic) + param = (uint32_t)((clk_id << 16) | 0xffff); + else + param = (uint32_t)((clk_id << 16) | (max & 0xffff)); ret = smu_cmn_send_smc_msg_with_param(smu, SMU_MSG_SetSoftMaxByFreq, param, NULL); if (ret) @@ -1631,7 +1635,10 @@ int smu_v13_0_set_soft_freq_limited_range(struct smu_context *smu, } if (min > 0) { - param = (uint32_t)((clk_id << 16) | (min & 0xffff)); + if (automatic) + param = (uint32_t)((clk_id << 16) | 0); + else + param = (uint32_t)((clk_id << 16) | (min & 0xffff)); ret = smu_cmn_send_smc_msg_with_param(smu, SMU_MSG_SetSoftMinByFreq, param, NULL); if (ret) @@ -1708,6 +1715,7 @@ int smu_v13_0_set_performance_level(struct smu_context *smu, uint32_t dclk_min = 0, dclk_max = 0; uint32_t fclk_min = 0, fclk_max = 0; int ret = 0, i; + bool auto_level = false; switch (level) { case AMD_DPM_FORCED_LEVEL_HIGH: @@ -1739,6 +1747,7 @@ int smu_v13_0_set_performance_level(struct smu_context *smu, dclk_max = dclk_table->max; fclk_min = fclk_table->min; fclk_max = fclk_table->max; + auto_level = true; break; case AMD_DPM_FORCED_LEVEL_PROFILE_STANDARD: sclk_min = sclk_max = pstate_table->gfxclk_pstate.standard; @@ -1780,13 +1789,15 @@ int smu_v13_0_set_performance_level(struct smu_context *smu, vclk_min = vclk_max = 0; dclk_min = dclk_max = 0; fclk_min = fclk_max = 0; + auto_level = false; } if (sclk_min && sclk_max) { ret = smu_v13_0_set_soft_freq_limited_range(smu, SMU_GFXCLK, sclk_min, - sclk_max); + sclk_max, + auto_level); if (ret) return ret; @@ -1798,7 +1809,8 @@ int smu_v13_0_set_performance_level(struct smu_context *smu, ret = smu_v13_0_set_soft_freq_limited_range(smu, SMU_MCLK, mclk_min, - mclk_max); + mclk_max, + auto_level); if (ret) return ret; @@ -1810,7 +1822,8 @@ int smu_v13_0_set_performance_level(struct smu_context *smu, ret = smu_v13_0_set_soft_freq_limited_range(smu, SMU_SOCCLK, socclk_min, - socclk_max); + socclk_max, + auto_level); if (ret) return ret; @@ -1825,7 +1838,8 @@ int smu_v13_0_set_performance_level(struct smu_context *smu, ret = smu_v13_0_set_soft_freq_limited_range(smu, i ? SMU_VCLK1 : SMU_VCLK, vclk_min, - vclk_max); + vclk_max, + auto_level); if (ret) return ret; } @@ -1840,7 +1854,8 @@ int smu_v13_0_set_performance_level(struct smu_context *smu, ret = smu_v13_0_set_soft_freq_limited_range(smu, i ? SMU_DCLK1 : SMU_DCLK, dclk_min, - dclk_max); + dclk_max, + auto_level); if (ret) return ret; } @@ -1852,7 +1867,8 @@ int smu_v13_0_set_performance_level(struct smu_context *smu, ret = smu_v13_0_set_soft_freq_limited_range(smu, SMU_FCLK, fclk_min, - fclk_max); + fclk_max, + auto_level); if (ret) return ret; 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 4e658abe3a20..3e2277abc754 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 @@ -1962,7 +1962,8 @@ static int smu_v13_0_0_force_clk_levels(struct smu_context *smu, ret = smu_v13_0_set_soft_freq_limited_range(smu, clk_type, min_freq, - max_freq); + max_freq, + false); break; case SMU_DCEFCLK: case SMU_PCIE: diff --git a/drivers/gpu/drm/amd/pm/swsmu/smu13/smu_v13_0_5_ppt.c b/drivers/gpu/drm/amd/pm/swsmu/smu13/smu_v13_0_5_ppt.c index 9c2c43bfed0b..a71b7c0803f1 100644 --- a/drivers/gpu/drm/amd/pm/swsmu/smu13/smu_v13_0_5_ppt.c +++ b/drivers/gpu/drm/amd/pm/swsmu/smu13/smu_v13_0_5_ppt.c @@ -811,9 +811,10 @@ failed: } static int smu_v13_0_5_set_soft_freq_limited_range(struct smu_context *smu, - enum smu_clk_type clk_type, - uint32_t min, - uint32_t max) + enum smu_clk_type clk_type, + uint32_t min, + uint32_t max, + bool automatic) { enum smu_message_type msg_set_min, msg_set_max; uint32_t min_clk = min; @@ -950,7 +951,7 @@ static int smu_v13_0_5_force_clk_levels(struct smu_context *smu, if (ret) goto force_level_out; - ret = smu_v13_0_5_set_soft_freq_limited_range(smu, clk_type, min_freq, max_freq); + ret = smu_v13_0_5_set_soft_freq_limited_range(smu, clk_type, min_freq, max_freq, false); if (ret) goto force_level_out; break; @@ -1046,9 +1047,10 @@ static int smu_v13_0_5_set_performance_level(struct smu_context *smu, if (sclk_min && sclk_max) { ret = smu_v13_0_5_set_soft_freq_limited_range(smu, - SMU_SCLK, - sclk_min, - sclk_max); + SMU_SCLK, + sclk_min, + sclk_max, + false); if (ret) return ret; @@ -1060,7 +1062,8 @@ static int smu_v13_0_5_set_performance_level(struct smu_context *smu, ret = smu_v13_0_5_set_soft_freq_limited_range(smu, SMU_VCLK, vclk_min, - vclk_max); + vclk_max, + false); if (ret) return ret; } @@ -1069,7 +1072,8 @@ static int smu_v13_0_5_set_performance_level(struct smu_context *smu, ret = smu_v13_0_5_set_soft_freq_limited_range(smu, SMU_DCLK, dclk_min, - dclk_max); + dclk_max, + false); if (ret) return ret; } diff --git a/drivers/gpu/drm/amd/pm/swsmu/smu13/smu_v13_0_6_ppt.c b/drivers/gpu/drm/amd/pm/swsmu/smu13/smu_v13_0_6_ppt.c index 91297590f006..7ebb675c5786 100644 --- a/drivers/gpu/drm/amd/pm/swsmu/smu13/smu_v13_0_6_ppt.c +++ b/drivers/gpu/drm/amd/pm/swsmu/smu13/smu_v13_0_6_ppt.c @@ -1758,7 +1758,7 @@ static int smu_v13_0_6_set_performance_level(struct smu_context *smu, if (uclk_table->max != pstate_table->uclk_pstate.curr.max) { /* Min UCLK is not expected to be changed */ ret = smu_v13_0_set_soft_freq_limited_range( - smu, SMU_UCLK, 0, uclk_table->max); + smu, SMU_UCLK, 0, uclk_table->max, false); if (ret) return ret; pstate_table->uclk_pstate.curr.max = uclk_table->max; @@ -1777,7 +1777,8 @@ static int smu_v13_0_6_set_performance_level(struct smu_context *smu, static int smu_v13_0_6_set_soft_freq_limited_range(struct smu_context *smu, enum smu_clk_type clk_type, - uint32_t min, uint32_t max) + uint32_t min, uint32_t max, + bool automatic) { struct smu_dpm_context *smu_dpm = &(smu->smu_dpm); struct smu_13_0_dpm_context *dpm_context = smu_dpm->dpm_context; @@ -1825,7 +1826,7 @@ static int smu_v13_0_6_set_soft_freq_limited_range(struct smu_context *smu, return -EOPNOTSUPP; /* Only max clock limiting is allowed for UCLK */ ret = smu_v13_0_set_soft_freq_limited_range( - smu, SMU_UCLK, 0, max); + smu, SMU_UCLK, 0, max, false); if (!ret) pstate_table->uclk_pstate.curr.max = max; } @@ -1965,7 +1966,7 @@ static int smu_v13_0_6_usr_edit_dpm_table(struct smu_context *smu, max_clk = dpm_context->dpm_tables.gfx_table.max; ret = smu_v13_0_6_set_soft_freq_limited_range( - smu, SMU_GFXCLK, min_clk, max_clk); + smu, SMU_GFXCLK, min_clk, max_clk, false); if (ret) return ret; @@ -1973,7 +1974,7 @@ static int smu_v13_0_6_usr_edit_dpm_table(struct smu_context *smu, min_clk = dpm_context->dpm_tables.uclk_table.min; max_clk = dpm_context->dpm_tables.uclk_table.max; ret = smu_v13_0_6_set_soft_freq_limited_range( - smu, SMU_UCLK, min_clk, max_clk); + smu, SMU_UCLK, min_clk, max_clk, false); if (ret) return ret; pstate_table->uclk_pstate.custom.max = 0; @@ -1997,7 +1998,7 @@ static int smu_v13_0_6_usr_edit_dpm_table(struct smu_context *smu, max_clk = pstate_table->gfxclk_pstate.custom.max; ret = smu_v13_0_6_set_soft_freq_limited_range( - smu, SMU_GFXCLK, min_clk, max_clk); + smu, SMU_GFXCLK, min_clk, max_clk, false); if (ret) return ret; @@ -2008,7 +2009,7 @@ static int smu_v13_0_6_usr_edit_dpm_table(struct smu_context *smu, min_clk = pstate_table->uclk_pstate.curr.min; max_clk = pstate_table->uclk_pstate.custom.max; return smu_v13_0_6_set_soft_freq_limited_range( - smu, SMU_UCLK, min_clk, max_clk); + smu, SMU_UCLK, min_clk, max_clk, false); } break; default: 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 f7d7cae16d9b..23f13388455f 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 @@ -1951,7 +1951,8 @@ static int smu_v13_0_7_force_clk_levels(struct smu_context *smu, ret = smu_v13_0_set_soft_freq_limited_range(smu, clk_type, min_freq, - max_freq); + max_freq, + false); break; case SMU_DCEFCLK: case SMU_PCIE: diff --git a/drivers/gpu/drm/amd/pm/swsmu/smu13/yellow_carp_ppt.c b/drivers/gpu/drm/amd/pm/swsmu/smu13/yellow_carp_ppt.c index 260c339f89c5..71d58c8c8cc0 100644 --- a/drivers/gpu/drm/amd/pm/swsmu/smu13/yellow_carp_ppt.c +++ b/drivers/gpu/drm/amd/pm/swsmu/smu13/yellow_carp_ppt.c @@ -945,9 +945,10 @@ failed: } static int yellow_carp_set_soft_freq_limited_range(struct smu_context *smu, - enum smu_clk_type clk_type, - uint32_t min, - uint32_t max) + enum smu_clk_type clk_type, + uint32_t min, + uint32_t max, + bool automatic) { enum smu_message_type msg_set_min, msg_set_max; uint32_t min_clk = min; @@ -1134,7 +1135,7 @@ static int yellow_carp_force_clk_levels(struct smu_context *smu, if (ret) goto force_level_out; - ret = yellow_carp_set_soft_freq_limited_range(smu, clk_type, min_freq, max_freq); + ret = yellow_carp_set_soft_freq_limited_range(smu, clk_type, min_freq, max_freq, false); if (ret) goto force_level_out; break; @@ -1254,9 +1255,10 @@ static int yellow_carp_set_performance_level(struct smu_context *smu, if (sclk_min && sclk_max) { ret = yellow_carp_set_soft_freq_limited_range(smu, - SMU_SCLK, - sclk_min, - sclk_max); + SMU_SCLK, + sclk_min, + sclk_max, + false); if (ret) return ret; @@ -1266,18 +1268,20 @@ static int yellow_carp_set_performance_level(struct smu_context *smu, if (fclk_min && fclk_max) { ret = yellow_carp_set_soft_freq_limited_range(smu, - SMU_FCLK, - fclk_min, - fclk_max); + SMU_FCLK, + fclk_min, + fclk_max, + false); if (ret) return ret; } if (socclk_min && socclk_max) { ret = yellow_carp_set_soft_freq_limited_range(smu, - SMU_SOCCLK, - socclk_min, - socclk_max); + SMU_SOCCLK, + socclk_min, + socclk_max, + false); if (ret) return ret; } @@ -1286,7 +1290,8 @@ static int yellow_carp_set_performance_level(struct smu_context *smu, ret = yellow_carp_set_soft_freq_limited_range(smu, SMU_VCLK, vclk_min, - vclk_max); + vclk_max, + false); if (ret) return ret; } @@ -1295,7 +1300,8 @@ static int yellow_carp_set_performance_level(struct smu_context *smu, ret = yellow_carp_set_soft_freq_limited_range(smu, SMU_DCLK, dclk_min, - dclk_max); + dclk_max, + false); if (ret) return ret; } diff --git a/drivers/gpu/drm/amd/pm/swsmu/smu14/smu_v14_0.c b/drivers/gpu/drm/amd/pm/swsmu/smu14/smu_v14_0.c index 865e916fc425..f7745eaf118e 100644 --- a/drivers/gpu/drm/amd/pm/swsmu/smu14/smu_v14_0.c +++ b/drivers/gpu/drm/amd/pm/swsmu/smu14/smu_v14_0.c @@ -1102,7 +1102,8 @@ failed: int smu_v14_0_set_soft_freq_limited_range(struct smu_context *smu, enum smu_clk_type clk_type, uint32_t min, - uint32_t max) + uint32_t max, + bool automatic) { int ret = 0, clk_id = 0; uint32_t param; @@ -1117,7 +1118,10 @@ int smu_v14_0_set_soft_freq_limited_range(struct smu_context *smu, return clk_id; if (max > 0) { - param = (uint32_t)((clk_id << 16) | (max & 0xffff)); + if (automatic) + param = (uint32_t)((clk_id << 16) | 0xffff); + else + param = (uint32_t)((clk_id << 16) | (max & 0xffff)); ret = smu_cmn_send_smc_msg_with_param(smu, SMU_MSG_SetSoftMaxByFreq, param, NULL); if (ret) @@ -1125,7 +1129,10 @@ int smu_v14_0_set_soft_freq_limited_range(struct smu_context *smu, } if (min > 0) { - param = (uint32_t)((clk_id << 16) | (min & 0xffff)); + if (automatic) + param = (uint32_t)((clk_id << 16) | 0); + else + param = (uint32_t)((clk_id << 16) | (min & 0xffff)); ret = smu_cmn_send_smc_msg_with_param(smu, SMU_MSG_SetSoftMinByFreq, param, NULL); if (ret) @@ -1202,6 +1209,7 @@ int smu_v14_0_set_performance_level(struct smu_context *smu, uint32_t dclk_min = 0, dclk_max = 0; uint32_t fclk_min = 0, fclk_max = 0; int ret = 0, i; + bool auto_level = false; switch (level) { case AMD_DPM_FORCED_LEVEL_HIGH: @@ -1233,6 +1241,7 @@ int smu_v14_0_set_performance_level(struct smu_context *smu, dclk_max = dclk_table->max; fclk_min = fclk_table->min; fclk_max = fclk_table->max; + auto_level = true; break; case AMD_DPM_FORCED_LEVEL_PROFILE_STANDARD: sclk_min = sclk_max = pstate_table->gfxclk_pstate.standard; @@ -1268,7 +1277,8 @@ int smu_v14_0_set_performance_level(struct smu_context *smu, ret = smu_v14_0_set_soft_freq_limited_range(smu, SMU_GFXCLK, sclk_min, - sclk_max); + sclk_max, + auto_level); if (ret) return ret; @@ -1280,7 +1290,8 @@ int smu_v14_0_set_performance_level(struct smu_context *smu, ret = smu_v14_0_set_soft_freq_limited_range(smu, SMU_MCLK, mclk_min, - mclk_max); + mclk_max, + auto_level); if (ret) return ret; @@ -1292,7 +1303,8 @@ int smu_v14_0_set_performance_level(struct smu_context *smu, ret = smu_v14_0_set_soft_freq_limited_range(smu, SMU_SOCCLK, socclk_min, - socclk_max); + socclk_max, + auto_level); if (ret) return ret; @@ -1307,7 +1319,8 @@ int smu_v14_0_set_performance_level(struct smu_context *smu, ret = smu_v14_0_set_soft_freq_limited_range(smu, i ? SMU_VCLK1 : SMU_VCLK, vclk_min, - vclk_max); + vclk_max, + auto_level); if (ret) return ret; } @@ -1322,7 +1335,8 @@ int smu_v14_0_set_performance_level(struct smu_context *smu, ret = smu_v14_0_set_soft_freq_limited_range(smu, i ? SMU_DCLK1 : SMU_DCLK, dclk_min, - dclk_max); + dclk_max, + auto_level); if (ret) return ret; } @@ -1334,7 +1348,8 @@ int smu_v14_0_set_performance_level(struct smu_context *smu, ret = smu_v14_0_set_soft_freq_limited_range(smu, SMU_FCLK, fclk_min, - fclk_max); + fclk_max, + auto_level); if (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 ade3ee398e3d..27f4e0ce447d 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 @@ -1503,7 +1503,8 @@ static int smu_v14_0_2_force_clk_levels(struct smu_context *smu, ret = smu_v14_0_set_soft_freq_limited_range(smu, clk_type, min_freq, - max_freq); + max_freq, + false); break; case SMU_DCEFCLK: case SMU_PCIE: -- 2.51.0 From c9b7c809b89f24e9372a4e7f02d64c950b07fdee Mon Sep 17 00:00:00 2001 From: Mario Limonciello Date: Fri, 11 Oct 2024 12:23:15 -0500 Subject: [PATCH 14/16] drm/amd: Guard against bad data for ATIF ACPI method If a BIOS provides bad data in response to an ATIF method call this causes a NULL pointer dereference in the caller. ``` ? show_regs (arch/x86/kernel/dumpstack.c:478 (discriminator 1)) ? __die (arch/x86/kernel/dumpstack.c:423 arch/x86/kernel/dumpstack.c:434) ? page_fault_oops (arch/x86/mm/fault.c:544 (discriminator 2) arch/x86/mm/fault.c:705 (discriminator 2)) ? do_user_addr_fault (arch/x86/mm/fault.c:440 (discriminator 1) arch/x86/mm/fault.c:1232 (discriminator 1)) ? acpi_ut_update_object_reference (drivers/acpi/acpica/utdelete.c:642) ? exc_page_fault (arch/x86/mm/fault.c:1542) ? asm_exc_page_fault (./arch/x86/include/asm/idtentry.h:623) ? amdgpu_atif_query_backlight_caps.constprop.0 (drivers/gpu/drm/amd/amdgpu/amdgpu_acpi.c:387 (discriminator 2)) amdgpu ? amdgpu_atif_query_backlight_caps.constprop.0 (drivers/gpu/drm/amd/amdgpu/amdgpu_acpi.c:386 (discriminator 1)) amdgpu ``` It has been encountered on at least one system, so guard for it. Fixes: d38ceaf99ed0 ("drm/amdgpu: add core driver (v4)") Acked-by: Alex Deucher Signed-off-by: Mario Limonciello Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_acpi.c | 15 ++++++++++++--- 1 file changed, 12 insertions(+), 3 deletions(-) diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_acpi.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_acpi.c index f85ace0384d2..1f5a296f5ed2 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_acpi.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_acpi.c @@ -147,6 +147,7 @@ static union acpi_object *amdgpu_atif_call(struct amdgpu_atif *atif, struct acpi_buffer *params) { acpi_status status; + union acpi_object *obj; union acpi_object atif_arg_elements[2]; struct acpi_object_list atif_arg; struct acpi_buffer buffer = { ACPI_ALLOCATE_BUFFER, NULL }; @@ -169,16 +170,24 @@ static union acpi_object *amdgpu_atif_call(struct amdgpu_atif *atif, status = acpi_evaluate_object(atif->handle, NULL, &atif_arg, &buffer); + obj = (union acpi_object *)buffer.pointer; - /* Fail only if calling the method fails and ATIF is supported */ + /* Fail if calling the method fails and ATIF is supported */ if (ACPI_FAILURE(status) && status != AE_NOT_FOUND) { DRM_DEBUG_DRIVER("failed to evaluate ATIF got %s\n", acpi_format_exception(status)); - kfree(buffer.pointer); + kfree(obj); return NULL; } - return buffer.pointer; + if (obj->type != ACPI_TYPE_BUFFER) { + DRM_DEBUG_DRIVER("bad object returned from ATIF: %d\n", + obj->type); + kfree(obj); + return NULL; + } + + return obj; } /** -- 2.51.0 From d2f57b6d8913df979a8cf719164c372b8e52c709 Mon Sep 17 00:00:00 2001 From: Alex Deucher Date: Wed, 25 Sep 2024 11:09:50 +0800 Subject: [PATCH 15/16] drm/amdgpu/discovery: add ISP discovery entries for old APUs Raven1/2 and Picasso have ISP 2.0.0, however their ISP blocks are not in the IP discovery table yet. This commit fixes this issue by adding new ISP entries for Raven and Picasso in the IP discovery table. Signed-off-by: Alex Deucher Signed-off-by: Zhu Lingshan Acked-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/amdgpu_discovery.c | 2 ++ 1 file changed, 2 insertions(+) diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_discovery.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_discovery.c index 9f9a1867da72..73f4d56c5de4 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_discovery.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_discovery.c @@ -2530,6 +2530,7 @@ int amdgpu_discovery_set_ip_blocks(struct amdgpu_device *adev) adev->ip_versions[GC_HWIP][0] = IP_VERSION(9, 2, 2); adev->ip_versions[UVD_HWIP][0] = IP_VERSION(1, 0, 1); adev->ip_versions[DCE_HWIP][0] = IP_VERSION(1, 0, 1); + adev->ip_versions[ISP_HWIP][0] = IP_VERSION(2, 0, 0); } else { adev->ip_versions[MMHUB_HWIP][0] = IP_VERSION(9, 1, 0); adev->ip_versions[ATHUB_HWIP][0] = IP_VERSION(9, 1, 0); @@ -2546,6 +2547,7 @@ int amdgpu_discovery_set_ip_blocks(struct amdgpu_device *adev) adev->ip_versions[GC_HWIP][0] = IP_VERSION(9, 1, 0); adev->ip_versions[UVD_HWIP][0] = IP_VERSION(1, 0, 0); adev->ip_versions[DCE_HWIP][0] = IP_VERSION(1, 0, 0); + adev->ip_versions[ISP_HWIP][0] = IP_VERSION(2, 0, 0); } break; case CHIP_VEGA20: -- 2.51.0 From 9ee8ab245c4f4d6a38598b4f14d33c6581c4cce6 Mon Sep 17 00:00:00 2001 From: Zhu Lingshan Date: Wed, 25 Sep 2024 11:09:51 +0800 Subject: [PATCH 16/16] drm/amdgpu: init saw registers for mmhub v1.0 This commits init registers in the Stand Along Walker for mmhub v1.0, to support ISP use cases. Signed-off-by: Zhu Lingshan Reported-and-tested-by: Du Bin Acked-by: Alex Deucher Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/mmhub_v1_0.c | 49 +++++++++++++++++++ .../include/asic_reg/mmhub/mmhub_1_0_offset.h | 23 +++++++++ 2 files changed, 72 insertions(+) diff --git a/drivers/gpu/drm/amd/amdgpu/mmhub_v1_0.c b/drivers/gpu/drm/amd/amdgpu/mmhub_v1_0.c index e3ddd22aa172..e9a6f33ca710 100644 --- a/drivers/gpu/drm/amd/amdgpu/mmhub_v1_0.c +++ b/drivers/gpu/drm/amd/amdgpu/mmhub_v1_0.c @@ -229,6 +229,52 @@ static void mmhub_v1_0_disable_identity_aperture(struct amdgpu_device *adev) 0); } +static void mmhub_v1_0_init_saw(struct amdgpu_device *adev) +{ + uint64_t pt_base = amdgpu_gmc_pd_addr(adev->gart.bo); + uint32_t tmp; + + /* VM_9_X_REGISTER_VM_L2_SAW_CONTEXT0_PAGE_TABLE_BASE_ADDR_LO32 */ + WREG32_SOC15(MMHUB, 0, mmVM_L2_SAW_CONTEXT0_PAGE_TABLE_BASE_ADDR_LO32, + lower_32_bits(pt_base >> 12)); + + /* VM_9_X_REGISTER_VM_L2_SAW_CONTEXT0_PAGE_TABLE_BASE_ADDR_HI32 */ + WREG32_SOC15(MMHUB, 0, mmVM_L2_SAW_CONTEXT0_PAGE_TABLE_BASE_ADDR_HI32, + upper_32_bits(pt_base >> 12)); + + /* VM_9_X_REGISTER_VM_L2_SAW_CONTEXT0_PAGE_TABLE_START_ADDR_LO32 */ + WREG32_SOC15(MMHUB, 0, mmVM_L2_SAW_CONTEXT0_PAGE_TABLE_START_ADDR_LO32, + (u32)(adev->gmc.gart_start >> 12)); + + /* VM_9_X_REGISTER_VM_L2_SAW_CONTEXT0_PAGE_TABLE_START_ADDR_HI32 */ + WREG32_SOC15(MMHUB, 0, mmVM_L2_SAW_CONTEXT0_PAGE_TABLE_START_ADDR_HI32, + (u32)(adev->gmc.gart_start >> 44)); + + /* VM_9_X_REGISTER_VM_L2_SAW_CONTEXT0_PAGE_TABLE_END_ADDR_LO32 */ + WREG32_SOC15(MMHUB, 0, mmVM_L2_SAW_CONTEXT0_PAGE_TABLE_END_ADDR_LO32, + (u32)(adev->gmc.gart_end >> 12)); + + /* VM_9_X_REGISTER_VM_L2_SAW_CONTEXT0_PAGE_TABLE_END_ADDR_HI32 */ + WREG32_SOC15(MMHUB, 0, mmVM_L2_SAW_CONTEXT0_PAGE_TABLE_END_ADDR_HI32, + (u32)(adev->gmc.gart_end >> 44)); + + /* Program SAW CONTEXT0 CNTL */ + tmp = RREG32_SOC15(MMHUB, 0, mmVM_L2_SAW_CONTEXT0_CNTL); + tmp |= 1 << CONTEXT0_CNTL_ENABLE_OFFSET; + tmp &= ~(3 << CONTEXT0_CNTL_PAGE_TABLE_DEPTH_OFFSET); + WREG32_SOC15(MMHUB, 0, mmVM_L2_SAW_CONTEXT0_CNTL, tmp); + + /* Disable all Contexts except Context0 */ + tmp = 0xfffe; + WREG32_SOC15(MMHUB, 0, mmVM_L2_SAW_CONTEXTS_DISABLE, tmp); + + /* Program SAW CNTL4 */ + tmp = RREG32_SOC15(MMHUB, 0, mmVM_L2_SAW_CNTL4); + tmp |= 1 << VMC_TAP_PDE_REQUEST_SNOOP_OFFSET; + tmp |= 1 << VMC_TAP_PTE_REQUEST_SNOOP_OFFSET; + WREG32_SOC15(MMHUB, 0, mmVM_L2_SAW_CNTL4, tmp); +} + static void mmhub_v1_0_setup_vmid_config(struct amdgpu_device *adev) { struct amdgpu_vmhub *hub = &adev->vmhub[AMDGPU_MMHUB0(0)]; @@ -283,6 +329,9 @@ static void mmhub_v1_0_setup_vmid_config(struct amdgpu_device *adev) i * hub->ctx_addr_distance, upper_32_bits(adev->vm_manager.max_pfn - 1)); } + + if (amdgpu_ip_version(adev, ISP_HWIP, 0)) + mmhub_v1_0_init_saw(adev); } static void mmhub_v1_0_program_invalidation(struct amdgpu_device *adev) diff --git a/drivers/gpu/drm/amd/include/asic_reg/mmhub/mmhub_1_0_offset.h b/drivers/gpu/drm/amd/include/asic_reg/mmhub/mmhub_1_0_offset.h index 2c3ce243861a..380e44230bda 100644 --- a/drivers/gpu/drm/amd/include/asic_reg/mmhub/mmhub_1_0_offset.h +++ b/drivers/gpu/drm/amd/include/asic_reg/mmhub/mmhub_1_0_offset.h @@ -1232,6 +1232,29 @@ #define mmMC_VM_MX_L1_PERFCOUNTER_HI 0x059d #define mmMC_VM_MX_L1_PERFCOUNTER_HI_BASE_IDX 0 +// Stand Alone Walker Registers +#define VMC_TAP_PDE_REQUEST_SNOOP_OFFSET 8 +#define VMC_TAP_PTE_REQUEST_SNOOP_OFFSET 11 +#define mmVM_L2_SAW_CONTEXT0_PAGE_TABLE_BASE_ADDR_LO32 0x0606 +#define mmVM_L2_SAW_CONTEXT0_PAGE_TABLE_BASE_ADDR_LO32_BASE_IDX 0 +#define mmVM_L2_SAW_CONTEXT0_PAGE_TABLE_BASE_ADDR_HI32 0x0607 +#define mmVM_L2_SAW_CONTEXT0_PAGE_TABLE_BASE_ADDR_HI32_BASE_IDX 0 +#define mmVM_L2_SAW_CONTEXT0_PAGE_TABLE_START_ADDR_LO32 0x0608 +#define mmVM_L2_SAW_CONTEXT0_PAGE_TABLE_START_ADDR_LO32_BASE_IDX 0 +#define mmVM_L2_SAW_CONTEXT0_PAGE_TABLE_START_ADDR_HI32 0x0609 +#define mmVM_L2_SAW_CONTEXT0_PAGE_TABLE_START_ADDR_HI32_BASE_IDX 0 +#define mmVM_L2_SAW_CONTEXT0_PAGE_TABLE_END_ADDR_LO32 0x060a +#define mmVM_L2_SAW_CONTEXT0_PAGE_TABLE_END_ADDR_LO32_BASE_IDX 0 +#define mmVM_L2_SAW_CONTEXT0_PAGE_TABLE_END_ADDR_HI32 0x060b +#define mmVM_L2_SAW_CONTEXT0_PAGE_TABLE_END_ADDR_HI32_BASE_IDX 0 +#define mmVM_L2_SAW_CONTEXT0_CNTL 0x0604 +#define mmVM_L2_SAW_CONTEXT0_CNTL_BASE_IDX 0 +#define CONTEXT0_CNTL_ENABLE_OFFSET 0 +#define CONTEXT0_CNTL_PAGE_TABLE_DEPTH_OFFSET 1 +#define mmVM_L2_SAW_CONTEXTS_DISABLE 0x060c +#define mmVM_L2_SAW_CONTEXTS_DISABLE_BASE_IDX 0 +#define mmVM_L2_SAW_CNTL4 0x0603 +#define mmVM_L2_SAW_CNTL4_BASE_IDX 0 // addressBlock: mmhub_utcl2_atcl2dec // base address: 0x69900 -- 2.51.0