From dafc87dcdc3bc50ac72c59156d64ed5267ad28e2 Mon Sep 17 00:00:00 2001 From: Philip Yang Date: Thu, 3 Oct 2024 11:53:51 -0400 Subject: [PATCH 01/16] drm/amdkfd: Copy wave state only for compute queue get_wave_state is not defined for sdma queue, copy_context_work_handler calls it for sdma queue will crash. Signed-off-by: Philip Yang Reviewed-by: Jonathan Kim Tested-by: Jonathan Kim Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.c | 5 ++++- 1 file changed, 4 insertions(+), 1 deletion(-) diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.c b/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.c index 648f40091aa3..b2b16a812e73 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.c +++ b/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.c @@ -3173,7 +3173,7 @@ struct copy_context_work_handler_workarea { struct kfd_process *p; }; -static void copy_context_work_handler (struct work_struct *work) +static void copy_context_work_handler(struct work_struct *work) { struct copy_context_work_handler_workarea *workarea; struct mqd_manager *mqd_mgr; @@ -3200,6 +3200,9 @@ static void copy_context_work_handler (struct work_struct *work) struct qcm_process_device *qpd = &pdd->qpd; list_for_each_entry(q, &qpd->queues_list, list) { + if (q->properties.type != KFD_QUEUE_TYPE_COMPUTE) + continue; + mqd_mgr = dqm->mqd_mgrs[KFD_MQD_TYPE_CP]; /* We ignore the return value from get_wave_state -- 2.51.0 From e7cee5459517e49a540240b03d207d0f81625109 Mon Sep 17 00:00:00 2001 From: Srinivasan Shanmugam Date: Wed, 2 Oct 2024 09:58:15 +0530 Subject: [PATCH 02/16] drm/amdgpu/gfx12: Implement cleaner shader support for GFX12 hardware MIME-Version: 1.0 Content-Type: text/plain; charset=utf8 Content-Transfer-Encoding: 8bit This patch adds support for the PACKET3_RUN_CLEANER_SHADER packet in the gfx_v12_0 module. This packet is used to emit the cleaner shader, which is used to clear GPU memory before it's reused, helping to prevent data leakage between different processes. Finally, the patch updates the ring function structures to include the new gfx_v12_0_ring_emit_cleaner_shader function. This allows the cleaner shader to be emitted as part of the ring's operations. Cc: Christian König Cc: Alex Deucher Signed-off-by: Srinivasan Shanmugam Reviewed-by: Alex Deucher Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/gfx_v12_0.c | 27 ++++++++++++++++++++++++-- 1 file changed, 25 insertions(+), 2 deletions(-) diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v12_0.c b/drivers/gpu/drm/amd/amdgpu/gfx_v12_0.c index dc25fcb89b13..7c23f95ae186 100644 --- a/drivers/gpu/drm/amd/amdgpu/gfx_v12_0.c +++ b/drivers/gpu/drm/amd/amdgpu/gfx_v12_0.c @@ -1346,6 +1346,12 @@ static int gfx_v12_0_sw_init(struct amdgpu_ip_block *ip_block) break; } + switch (amdgpu_ip_version(adev, GC_HWIP, 0)) { + default: + adev->gfx.enable_cleaner_shader = false; + break; + } + /* recalculate compute rings to use based on hardware configuration */ num_compute_rings = (adev->gfx.mec.num_pipe_per_mec * adev->gfx.mec.num_queue_per_pipe) / 2; @@ -1460,6 +1466,10 @@ static int gfx_v12_0_sw_init(struct amdgpu_ip_block *ip_block) gfx_v12_0_alloc_ip_dump(adev); + r = amdgpu_gfx_sysfs_isolation_shader_init(adev); + if (r) + return r; + return 0; } @@ -1519,6 +1529,8 @@ static int gfx_v12_0_sw_fini(struct amdgpu_ip_block *ip_block) gfx_v12_0_free_microcode(adev); + amdgpu_gfx_sysfs_isolation_shader_fini(adev); + kfree(adev->gfx.ip_dump_core); kfree(adev->gfx.ip_dump_compute_queues); kfree(adev->gfx.ip_dump_gfx_queues); @@ -5038,6 +5050,13 @@ static void gfx_v12_ring_insert_nop(struct amdgpu_ring *ring, uint32_t num_nop) amdgpu_ring_write(ring, ring->funcs->nop); } +static void gfx_v12_0_ring_emit_cleaner_shader(struct amdgpu_ring *ring) +{ + /* Emit the cleaner shader */ + amdgpu_ring_write(ring, PACKET3(PACKET3_RUN_CLEANER_SHADER, 0)); + amdgpu_ring_write(ring, 0); /* RESERVED field, programmed to zero */ +} + static void gfx_v12_ip_print(struct amdgpu_ip_block *ip_block, struct drm_printer *p) { struct amdgpu_device *adev = ip_block->adev; @@ -5297,7 +5316,8 @@ static const struct amdgpu_ring_funcs gfx_v12_0_ring_funcs_gfx = { 3 + /* CNTX_CTRL */ 5 + /* HDP_INVL */ 8 + 8 + /* FENCE x2 */ - 8, /* gfx_v12_0_emit_mem_sync */ + 8 + /* gfx_v12_0_emit_mem_sync */ + 2, /* gfx_v12_0_ring_emit_cleaner_shader */ .emit_ib_size = 4, /* gfx_v12_0_ring_emit_ib_gfx */ .emit_ib = gfx_v12_0_ring_emit_ib_gfx, .emit_fence = gfx_v12_0_ring_emit_fence, @@ -5318,6 +5338,7 @@ static const struct amdgpu_ring_funcs gfx_v12_0_ring_funcs_gfx = { .soft_recovery = gfx_v12_0_ring_soft_recovery, .emit_mem_sync = gfx_v12_0_emit_mem_sync, .reset = gfx_v12_0_reset_kgq, + .emit_cleaner_shader = gfx_v12_0_ring_emit_cleaner_shader, }; static const struct amdgpu_ring_funcs gfx_v12_0_ring_funcs_compute = { @@ -5336,7 +5357,8 @@ static const struct amdgpu_ring_funcs gfx_v12_0_ring_funcs_compute = { SOC15_FLUSH_GPU_TLB_NUM_REG_WAIT * 7 + 2 + /* gfx_v12_0_ring_emit_vm_flush */ 8 + 8 + 8 + /* gfx_v12_0_ring_emit_fence x3 for user fence, vm fence */ - 8, /* gfx_v12_0_emit_mem_sync */ + 8 + /* gfx_v12_0_emit_mem_sync */ + 2, /* gfx_v12_0_ring_emit_cleaner_shader */ .emit_ib_size = 7, /* gfx_v12_0_ring_emit_ib_compute */ .emit_ib = gfx_v12_0_ring_emit_ib_compute, .emit_fence = gfx_v12_0_ring_emit_fence, @@ -5353,6 +5375,7 @@ static const struct amdgpu_ring_funcs gfx_v12_0_ring_funcs_compute = { .soft_recovery = gfx_v12_0_ring_soft_recovery, .emit_mem_sync = gfx_v12_0_emit_mem_sync, .reset = gfx_v12_0_reset_kcq, + .emit_cleaner_shader = gfx_v12_0_ring_emit_cleaner_shader, }; static const struct amdgpu_ring_funcs gfx_v12_0_ring_funcs_kiq = { -- 2.51.0 From fbca19695330ac78e9c14af05fc3e3f4c4278a52 Mon Sep 17 00:00:00 2001 From: Srinivasan Shanmugam Date: Wed, 2 Oct 2024 09:52:39 +0530 Subject: [PATCH 03/16] drm/amdgpu/gfx11: Apply Isolation Enforcement to GFX & Compute rings MIME-Version: 1.0 Content-Type: text/plain; charset=utf8 Content-Transfer-Encoding: 8bit This commit applies isolation enforcement to the GFX and Compute rings in the gfx_v11_0 module. The commit sets `amdgpu_gfx_enforce_isolation_ring_begin_use` and `amdgpu_gfx_enforce_isolation_ring_end_use` as the functions to be called when a ring begins and ends its use, respectively. `amdgpu_gfx_enforce_isolation_ring_begin_use` is called when a ring begins its use. This function cancels any scheduled `enforce_isolation_work` and, if necessary, signals the Kernel Fusion Driver (KFD) to stop the runqueue. `amdgpu_gfx_enforce_isolation_ring_end_use` is called when a ring ends its use. This function schedules `enforce_isolation_work` to be run after a delay. These functions are part of the Enforce Isolation Handler, which enforces shader isolation on AMD GPUs to prevent data leakage between different processes. Cc: Christian König Cc: Alex Deucher Signed-off-by: Srinivasan Shanmugam Reviewed-by: Alex Deucher Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/amdgpu/gfx_v11_0.c | 4 ++++ 1 file changed, 4 insertions(+) diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v11_0.c b/drivers/gpu/drm/amd/amdgpu/gfx_v11_0.c index c78a63da8d2e..c3f98c46b39b 100644 --- a/drivers/gpu/drm/amd/amdgpu/gfx_v11_0.c +++ b/drivers/gpu/drm/amd/amdgpu/gfx_v11_0.c @@ -6875,6 +6875,8 @@ static const struct amdgpu_ring_funcs gfx_v11_0_ring_funcs_gfx = { .emit_mem_sync = gfx_v11_0_emit_mem_sync, .reset = gfx_v11_0_reset_kgq, .emit_cleaner_shader = gfx_v11_0_ring_emit_cleaner_shader, + .begin_use = amdgpu_gfx_enforce_isolation_ring_begin_use, + .end_use = amdgpu_gfx_enforce_isolation_ring_end_use, }; static const struct amdgpu_ring_funcs gfx_v11_0_ring_funcs_compute = { @@ -6915,6 +6917,8 @@ static const struct amdgpu_ring_funcs gfx_v11_0_ring_funcs_compute = { .emit_mem_sync = gfx_v11_0_emit_mem_sync, .reset = gfx_v11_0_reset_kcq, .emit_cleaner_shader = gfx_v11_0_ring_emit_cleaner_shader, + .begin_use = amdgpu_gfx_enforce_isolation_ring_begin_use, + .end_use = amdgpu_gfx_enforce_isolation_ring_end_use, }; static const struct amdgpu_ring_funcs gfx_v11_0_ring_funcs_kiq = { -- 2.51.0 From b4c1ad70e279bacbc772a468033bdecce2f5e0dc Mon Sep 17 00:00:00 2001 From: Wu Hoi Pok Date: Fri, 4 Oct 2024 09:05:59 +0800 Subject: [PATCH 04/16] drm/radeon: add late_register for connector The patch is to solve null dereference in 'aux.dev', which is introduced in recent radeon rework. By having 'late_register', the connector should be registered after 'drm_dev_register' automatically, where in before it is the opposite. Fixes: 90985660ba48 ("drm/radeon: remove load callback from kms_driver") Closes: https://gitlab.freedesktop.org/drm/amd/-/issues/3665 Tested-by: Hans de Goede Suggested-by: Christophe Leroy Signed-off-by: Wu Hoi Pok Signed-off-by: Alex Deucher --- drivers/gpu/drm/radeon/atombios_dp.c | 9 ++------- drivers/gpu/drm/radeon/radeon_connectors.c | 17 +++++++++++++++++ 2 files changed, 19 insertions(+), 7 deletions(-) diff --git a/drivers/gpu/drm/radeon/atombios_dp.c b/drivers/gpu/drm/radeon/atombios_dp.c index fca8b08535a5..6328627b7c34 100644 --- a/drivers/gpu/drm/radeon/atombios_dp.c +++ b/drivers/gpu/drm/radeon/atombios_dp.c @@ -228,10 +228,8 @@ void radeon_dp_aux_init(struct radeon_connector *radeon_connector) { struct drm_device *dev = radeon_connector->base.dev; struct radeon_device *rdev = dev->dev_private; - int ret; radeon_connector->ddc_bus->rec.hpd = radeon_connector->hpd.hpd; - radeon_connector->ddc_bus->aux.dev = radeon_connector->base.kdev; radeon_connector->ddc_bus->aux.drm_dev = radeon_connector->base.dev; if (ASIC_IS_DCE5(rdev)) { if (radeon_auxch) @@ -242,11 +240,8 @@ void radeon_dp_aux_init(struct radeon_connector *radeon_connector) radeon_connector->ddc_bus->aux.transfer = radeon_dp_aux_transfer_atom; } - ret = drm_dp_aux_register(&radeon_connector->ddc_bus->aux); - if (!ret) - radeon_connector->ddc_bus->has_aux = true; - - WARN(ret, "drm_dp_aux_register() failed with error %d\n", ret); + drm_dp_aux_init(&radeon_connector->ddc_bus->aux); + radeon_connector->ddc_bus->has_aux = true; } /***** general DP utility functions *****/ diff --git a/drivers/gpu/drm/radeon/radeon_connectors.c b/drivers/gpu/drm/radeon/radeon_connectors.c index 528a8f3677c2..f9c73c55f04f 100644 --- a/drivers/gpu/drm/radeon/radeon_connectors.c +++ b/drivers/gpu/drm/radeon/radeon_connectors.c @@ -1786,6 +1786,20 @@ static enum drm_mode_status radeon_dp_mode_valid(struct drm_connector *connector return MODE_OK; } +static int +radeon_connector_late_register(struct drm_connector *connector) +{ + struct radeon_connector *radeon_connector = to_radeon_connector(connector); + int r = 0; + + if (radeon_connector->ddc_bus->has_aux) { + radeon_connector->ddc_bus->aux.dev = radeon_connector->base.kdev; + r = drm_dp_aux_register(&radeon_connector->ddc_bus->aux); + } + + return r; +} + static const struct drm_connector_helper_funcs radeon_dp_connector_helper_funcs = { .get_modes = radeon_dp_get_modes, .mode_valid = radeon_dp_mode_valid, @@ -1800,6 +1814,7 @@ static const struct drm_connector_funcs radeon_dp_connector_funcs = { .early_unregister = radeon_connector_unregister, .destroy = radeon_connector_destroy, .force = radeon_dvi_force, + .late_register = radeon_connector_late_register, }; static const struct drm_connector_funcs radeon_edp_connector_funcs = { @@ -1810,6 +1825,7 @@ static const struct drm_connector_funcs radeon_edp_connector_funcs = { .early_unregister = radeon_connector_unregister, .destroy = radeon_connector_destroy, .force = radeon_dvi_force, + .late_register = radeon_connector_late_register, }; static const struct drm_connector_funcs radeon_lvds_bridge_connector_funcs = { @@ -1820,6 +1836,7 @@ static const struct drm_connector_funcs radeon_lvds_bridge_connector_funcs = { .early_unregister = radeon_connector_unregister, .destroy = radeon_connector_destroy, .force = radeon_dvi_force, + .late_register = radeon_connector_late_register, }; void -- 2.51.0 From 8cb028a1bbd960a1ff71b7300e557be9269f98fd Mon Sep 17 00:00:00 2001 From: Austin Zheng Date: Mon, 16 Sep 2024 13:09:19 -0400 Subject: [PATCH 05/16] drm/amd/display: Unify blank_phantom and blank_pixel_data [Why] dcn32_blank_phantom() does not consider the subVP+ODM case when blanking. Only one of the pipes will get blanked. Remaining pipes are not blanked. Will cause underflow in the phantom pipe when enabling the CRTC. [How] Use blank_pixel_data() instead of blank_phantom(). remove dcn32_blank_phantom() since logic is identical. Different DPG dimensions get programmed when blanking phantom pipes. Previously had phantom pipes use DPG dimensions of the main stream. Now use DPG dimensions of the phantom streams Reviewed-by: Alvin Lee Signed-off-by: Austin Zheng Signed-off-by: Rodrigo Siqueira Tested-by: Daniel Wheeler Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/display/dc/core/dc.c | 12 +---- .../amd/display/dc/hwss/dcn20/dcn20_hwseq.c | 15 ++---- .../amd/display/dc/hwss/dcn32/dcn32_hwseq.c | 46 ------------------- .../amd/display/dc/hwss/dcn32/dcn32_hwseq.h | 5 -- .../amd/display/dc/hwss/dcn32/dcn32_init.c | 1 - .../amd/display/dc/hwss/dcn401/dcn401_init.c | 1 - 6 files changed, 6 insertions(+), 74 deletions(-) diff --git a/drivers/gpu/drm/amd/display/dc/core/dc.c b/drivers/gpu/drm/amd/display/dc/core/dc.c index 7df748ec0433..f4a5accf1db9 100644 --- a/drivers/gpu/drm/amd/display/dc/core/dc.c +++ b/drivers/gpu/drm/amd/display/dc/core/dc.c @@ -1235,16 +1235,8 @@ static void disable_dangling_plane(struct dc *dc, struct dc_state *context) */ if (is_phantom) { if (tg->funcs->enable_crtc) { - int main_pipe_width = 0, main_pipe_height = 0; - struct dc_stream_state *old_paired_stream = dc_state_get_paired_subvp_stream(dc->current_state, old_stream); - - if (old_paired_stream) { - main_pipe_width = old_paired_stream->dst.width; - main_pipe_height = old_paired_stream->dst.height; - } - - if (dc->hwss.blank_phantom) - dc->hwss.blank_phantom(dc, tg, main_pipe_width, main_pipe_height); + if (dc->hwseq->funcs.blank_pixel_data) + dc->hwseq->funcs.blank_pixel_data(dc, pipe, true); tg->funcs->enable_crtc(tg); } } 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 e89499536c46..1a32e53c1b22 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 @@ -2054,22 +2054,15 @@ void dcn20_program_front_end_for_ctx( */ for (i = 0; i < dc->res_pool->pipe_count; i++) { struct dc_stream_state *stream = dc->current_state->res_ctx.pipe_ctx[i].stream; + pipe = &dc->current_state->res_ctx.pipe_ctx[i]; if (context->res_ctx.pipe_ctx[i].update_flags.bits.disable && stream && - dc_state_get_pipe_subvp_type(dc->current_state, &dc->current_state->res_ctx.pipe_ctx[i]) == SUBVP_PHANTOM) { + dc_state_get_pipe_subvp_type(dc->current_state, pipe) == SUBVP_PHANTOM) { struct timing_generator *tg = dc->current_state->res_ctx.pipe_ctx[i].stream_res.tg; if (tg->funcs->enable_crtc) { - if (dc->hwss.blank_phantom) { - int main_pipe_width = 0, main_pipe_height = 0; - struct dc_stream_state *phantom_stream = dc_state_get_paired_subvp_stream(dc->current_state, dc->current_state->res_ctx.pipe_ctx[i].stream); - - if (phantom_stream) { - main_pipe_width = phantom_stream->dst.width; - main_pipe_height = phantom_stream->dst.height; - } - - dc->hwss.blank_phantom(dc, tg, main_pipe_width, main_pipe_height); + if (dc->hwseq->funcs.blank_pixel_data) { + dc->hwseq->funcs.blank_pixel_data(dc, pipe, true); } tg->funcs->enable_crtc(tg); } 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 a985d191f500..119c6d08200f 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 @@ -1698,52 +1698,6 @@ void dcn32_init_blank( hws->funcs.wait_for_blank_complete(opp); } -void dcn32_blank_phantom(struct dc *dc, - struct timing_generator *tg, - int width, - int height) -{ - struct dce_hwseq *hws = dc->hwseq; - enum dc_color_space color_space; - struct tg_color black_color = {0}; - struct output_pixel_processor *opp = NULL; - uint32_t num_opps, opp_id_src0, opp_id_src1; - uint32_t otg_active_width, otg_active_height; - uint32_t i; - - /* program opp dpg blank color */ - color_space = COLOR_SPACE_SRGB; - color_space_to_black_color(dc, color_space, &black_color); - - otg_active_width = width; - otg_active_height = height; - - /* get the OPTC source */ - tg->funcs->get_optc_source(tg, &num_opps, &opp_id_src0, &opp_id_src1); - ASSERT(opp_id_src0 < dc->res_pool->res_cap->num_opp); - - for (i = 0; i < dc->res_pool->res_cap->num_opp; i++) { - if (dc->res_pool->opps[i] != NULL && dc->res_pool->opps[i]->inst == opp_id_src0) { - opp = dc->res_pool->opps[i]; - break; - } - } - - if (opp && opp->funcs->opp_set_disp_pattern_generator) - opp->funcs->opp_set_disp_pattern_generator( - opp, - CONTROLLER_DP_TEST_PATTERN_SOLID_COLOR, - CONTROLLER_DP_COLOR_SPACE_UDEFINED, - COLOR_DEPTH_UNDEFINED, - &black_color, - otg_active_width, - otg_active_height, - 0); - - if (tg->funcs->is_tg_enabled(tg)) - hws->funcs.wait_for_blank_complete(opp); -} - /* phantom stream id's can change often, but can be identical between contexts. * This function checks for the condition the streams are identical to avoid * redundant pipe transitions. diff --git a/drivers/gpu/drm/amd/display/dc/hwss/dcn32/dcn32_hwseq.h b/drivers/gpu/drm/amd/display/dc/hwss/dcn32/dcn32_hwseq.h index cac4a08b92a4..0303a5953673 100644 --- a/drivers/gpu/drm/amd/display/dc/hwss/dcn32/dcn32_hwseq.h +++ b/drivers/gpu/drm/amd/display/dc/hwss/dcn32/dcn32_hwseq.h @@ -119,11 +119,6 @@ void dcn32_init_blank( struct dc *dc, struct timing_generator *tg); -void dcn32_blank_phantom(struct dc *dc, - struct timing_generator *tg, - int width, - int height); - bool dcn32_is_pipe_topology_transition_seamless(struct dc *dc, const struct dc_state *cur_ctx, const struct dc_state *new_ctx); diff --git a/drivers/gpu/drm/amd/display/dc/hwss/dcn32/dcn32_init.c b/drivers/gpu/drm/amd/display/dc/hwss/dcn32/dcn32_init.c index 8e0946fd5b7f..dbcd2dfb19c1 100644 --- a/drivers/gpu/drm/amd/display/dc/hwss/dcn32/dcn32_init.c +++ b/drivers/gpu/drm/amd/display/dc/hwss/dcn32/dcn32_init.c @@ -117,7 +117,6 @@ static const struct hw_sequencer_funcs dcn32_funcs = { .update_phantom_vp_position = dcn32_update_phantom_vp_position, .update_dsc_pg = dcn32_update_dsc_pg, .apply_update_flags_for_phantom = dcn32_apply_update_flags_for_phantom, - .blank_phantom = dcn32_blank_phantom, .is_pipe_topology_transition_seamless = dcn32_is_pipe_topology_transition_seamless, .calculate_pix_rate_divider = dcn32_calculate_pix_rate_divider, .program_outstanding_updates = dcn32_program_outstanding_updates, diff --git a/drivers/gpu/drm/amd/display/dc/hwss/dcn401/dcn401_init.c b/drivers/gpu/drm/amd/display/dc/hwss/dcn401/dcn401_init.c index af0d40a5cb77..a1392e776709 100644 --- a/drivers/gpu/drm/amd/display/dc/hwss/dcn401/dcn401_init.c +++ b/drivers/gpu/drm/amd/display/dc/hwss/dcn401/dcn401_init.c @@ -93,7 +93,6 @@ static const struct hw_sequencer_funcs dcn401_funcs = { .update_phantom_vp_position = dcn32_update_phantom_vp_position, .update_dsc_pg = dcn32_update_dsc_pg, .apply_update_flags_for_phantom = dcn32_apply_update_flags_for_phantom, - .blank_phantom = dcn32_blank_phantom, .wait_for_dcc_meta_propagation = dcn401_wait_for_dcc_meta_propagation, .is_pipe_topology_transition_seamless = dcn32_is_pipe_topology_transition_seamless, .fams2_global_control_lock = dcn401_fams2_global_control_lock, -- 2.51.0 From 0e37e4b9afbd08df1f00a70bbb4d1ec273d18c9e Mon Sep 17 00:00:00 2001 From: Fudongwang Date: Sat, 14 Sep 2024 09:33:44 +0800 Subject: [PATCH 06/16] drm/amd/display: skip disable CRTC in seemless bootup case Resync FIFO is a workaround to write the same value to DENTIST_DISPCLK_CNTL register after programming OTG_PIXEL_RATE_DIV register, in case seemless boot, there is no OTG_PIXEL_RATE_DIV register update, so skip CRTC disable when resync FIFO to avoid random FIFO error and garbage. Reviewed-by: Nicholas Kazlauskas Signed-off-by: Fudongwang Signed-off-by: Rodrigo Siqueira Tested-by: Daniel Wheeler Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/display/dc/hwss/dcn314/dcn314_hwseq.c | 4 +++- 1 file changed, 3 insertions(+), 1 deletion(-) diff --git a/drivers/gpu/drm/amd/display/dc/hwss/dcn314/dcn314_hwseq.c b/drivers/gpu/drm/amd/display/dc/hwss/dcn314/dcn314_hwseq.c index 5b6cf2a8e38d..e0054e654db6 100644 --- a/drivers/gpu/drm/amd/display/dc/hwss/dcn314/dcn314_hwseq.c +++ b/drivers/gpu/drm/amd/display/dc/hwss/dcn314/dcn314_hwseq.c @@ -371,7 +371,9 @@ void dcn314_resync_fifo_dccg_dio(struct dce_hwseq *hws, struct dc *dc, struct dc if (pipe->top_pipe || pipe->prev_odm_pipe) continue; - if (pipe->stream && (pipe->stream->dpms_off || dc_is_virtual_signal(pipe->stream->signal))) { + if (pipe->stream && (pipe->stream->dpms_off || dc_is_virtual_signal(pipe->stream->signal)) && + !pipe->stream->apply_seamless_boot_optimization && + !pipe->stream->apply_edp_fast_boot_optimization) { pipe->stream_res.tg->funcs->disable_crtc(pipe->stream_res.tg); reset_sync_context_for_pipe(dc, context, i); otg_disabled[i] = true; -- 2.51.0 From bd5230b1a6f24447aa604e631e70ffe62834ff74 Mon Sep 17 00:00:00 2001 From: Austin Zheng Date: Fri, 20 Sep 2024 14:22:38 -0400 Subject: [PATCH 07/16] drm/amd/display: Flip All Planes Under OTG Master When Flip Immediate [Why] The MPO plane will receive a flip but desktop plane may not receive a flip when GSL is enabled. As a result, system will be stuck waiting for a flip that was never sent. [How] Set update address update flag of all flip_immediate planes if there are multiple planes. Reviewed-by: Dillon Varone Signed-off-by: Austin Zheng Signed-off-by: Rodrigo Siqueira Tested-by: Daniel Wheeler Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/display/dc/core/dc.c | 33 ++++++++++++++++++++++++ 1 file changed, 33 insertions(+) diff --git a/drivers/gpu/drm/amd/display/dc/core/dc.c b/drivers/gpu/drm/amd/display/dc/core/dc.c index f4a5accf1db9..48057ac22cbd 100644 --- a/drivers/gpu/drm/amd/display/dc/core/dc.c +++ b/drivers/gpu/drm/amd/display/dc/core/dc.c @@ -2661,6 +2661,29 @@ static enum surface_update_type det_surface_update(const struct dc *dc, return overall_type; } +/* May need to flip the desktop plane in cases where MPO plane receives a flip but desktop plane doesn't + * while both planes are flip_immediate + */ +static void force_immediate_gsl_plane_flip(struct dc *dc, struct dc_surface_update *updates, int surface_count) +{ + bool has_flip_immediate_plane = false; + int i; + + for (i = 0; i < surface_count; i++) { + if (updates[i].surface->flip_immediate) { + has_flip_immediate_plane = true; + break; + } + } + + if (has_flip_immediate_plane && surface_count > 1) { + for (i = 0; i < surface_count; i++) { + if (updates[i].surface->flip_immediate) + updates[i].surface->update_flags.bits.addr_update = 1; + } + } +} + static enum surface_update_type check_update_surfaces_for_stream( struct dc *dc, struct dc_surface_update *updates, @@ -3177,6 +3200,11 @@ static bool update_planes_and_stream_state(struct dc *dc, context = dc->current_state; update_type = dc_check_update_surfaces_for_stream( dc, srf_updates, surface_count, stream_update, stream_status); + /* It is possible to receive a flip for one plane while there are multiple flip_immediate planes in the same stream. + * E.g. Desktop and MPO plane are flip_immediate but only the MPO plane received a flip + * Force the other flip_immediate planes to flip so GSL doesn't wait for a flip that won't come. + */ + force_immediate_gsl_plane_flip(dc, srf_updates, surface_count); if (update_type == UPDATE_TYPE_FULL) backup_planes_and_stream_state(&dc->scratch.current_state, stream); @@ -4807,6 +4835,11 @@ static bool update_planes_and_stream_v1(struct dc *dc, update_type = dc_check_update_surfaces_for_stream( dc, srf_updates, surface_count, stream_update, stream_status); + /* It is possible to receive a flip for one plane while there are multiple flip_immediate planes in the same stream. + * E.g. Desktop and MPO plane are flip_immediate but only the MPO plane received a flip + * Force the other flip_immediate planes to flip so GSL doesn't wait for a flip that won't come. + */ + force_immediate_gsl_plane_flip(dc, srf_updates, surface_count); if (update_type >= UPDATE_TYPE_FULL) { -- 2.51.0 From 0e93b76cf92f229409e8da85c2a143868835fec3 Mon Sep 17 00:00:00 2001 From: Austin Zheng Date: Mon, 23 Sep 2024 10:07:32 -0400 Subject: [PATCH 08/16] drm/amd/display: Revert commit Update Interface to Check UCLK DPM This reverts commit b8d046985c2dc41a0e264a391da4606099f8d44f. Reverting as regression discovered on certain systems and golden values need to updated. Reviewed-by: Alvin Lee Signed-off-by: Austin Zheng Signed-off-by: Rodrigo Siqueira Tested-by: Daniel Wheeler Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/display/dc/core/dc.c | 9 +-------- .../dml21/src/dml2_pmo/dml2_pmo_dcn4_fams2.c | 1 - drivers/gpu/drm/amd/display/dc/inc/core_types.h | 4 ---- .../dc/resource/dcn315/dcn315_resource.c | 6 ------ .../dc/resource/dcn401/dcn401_resource.c | 17 ----------------- 5 files changed, 1 insertion(+), 36 deletions(-) diff --git a/drivers/gpu/drm/amd/display/dc/core/dc.c b/drivers/gpu/drm/amd/display/dc/core/dc.c index 48057ac22cbd..57ad6ce88f3f 100644 --- a/drivers/gpu/drm/amd/display/dc/core/dc.c +++ b/drivers/gpu/drm/amd/display/dc/core/dc.c @@ -6038,15 +6038,8 @@ void dc_set_edp_power(const struct dc *dc, struct dc_link *edp_link, struct dc_power_profile dc_get_power_profile_for_dc_state(const struct dc_state *context) { struct dc_power_profile profile = { 0 }; - struct dc *dc = NULL; - if (!context || !context->clk_mgr || !context->clk_mgr->ctx || !context->clk_mgr->ctx->dc) - return profile; - - dc = context->clk_mgr->ctx->dc; - - if (dc->res_pool->funcs->get_power_profile) - profile.power_level = dc->res_pool->funcs->get_power_profile(context); + profile.power_level += !context->bw_ctx.bw.dcn.clk.p_state_change_support; return profile; } diff --git a/drivers/gpu/drm/amd/display/dc/dml2/dml21/src/dml2_pmo/dml2_pmo_dcn4_fams2.c b/drivers/gpu/drm/amd/display/dc/dml2/dml21/src/dml2_pmo/dml2_pmo_dcn4_fams2.c index 2e9c59e9e0c1..1cf9015e854a 100644 --- a/drivers/gpu/drm/amd/display/dc/dml2/dml21/src/dml2_pmo/dml2_pmo_dcn4_fams2.c +++ b/drivers/gpu/drm/amd/display/dc/dml2/dml21/src/dml2_pmo/dml2_pmo_dcn4_fams2.c @@ -1798,7 +1798,6 @@ bool pmo_dcn4_fams2_init_for_pstate_support(struct dml2_pmo_init_for_pstate_supp } if (s->pmo_dcn4.num_pstate_candidates > 0) { - s->pmo_dcn4.pstate_strategy_candidates[s->pmo_dcn4.num_pstate_candidates-1].allow_state_increase = true; s->pmo_dcn4.cur_pstate_candidate = -1; return true; } else { 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 8597e866bfe6..bfb8b8502d20 100644 --- a/drivers/gpu/drm/amd/display/dc/inc/core_types.h +++ b/drivers/gpu/drm/amd/display/dc/inc/core_types.h @@ -215,10 +215,6 @@ struct resource_funcs { void (*get_panel_config_defaults)(struct dc_panel_config *panel_config); void (*build_pipe_pix_clk_params)(struct pipe_ctx *pipe_ctx); - /* - * Get indicator of power from a context that went through full validation - */ - int (*get_power_profile)(const struct dc_state *context); }; struct audio_support{ diff --git a/drivers/gpu/drm/amd/display/dc/resource/dcn315/dcn315_resource.c b/drivers/gpu/drm/amd/display/dc/resource/dcn315/dcn315_resource.c index f6b840f046a5..3f4b9dba4112 100644 --- a/drivers/gpu/drm/amd/display/dc/resource/dcn315/dcn315_resource.c +++ b/drivers/gpu/drm/amd/display/dc/resource/dcn315/dcn315_resource.c @@ -1812,11 +1812,6 @@ static void dcn315_get_panel_config_defaults(struct dc_panel_config *panel_confi *panel_config = panel_config_defaults; } -static int dcn315_get_power_profile(const struct dc_state *context) -{ - return !context->bw_ctx.bw.dcn.clk.p_state_change_support; -} - static struct dc_cap_funcs cap_funcs = { .get_dcc_compression_cap = dcn20_get_dcc_compression_cap }; @@ -1845,7 +1840,6 @@ static struct resource_funcs dcn315_res_pool_funcs = { .update_bw_bounding_box = dcn315_update_bw_bounding_box, .patch_unknown_plane_state = dcn20_patch_unknown_plane_state, .get_panel_config_defaults = dcn315_get_panel_config_defaults, - .get_power_profile = dcn315_get_power_profile, }; static bool dcn315_resource_construct( 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..f2653a86d3e7 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 @@ -1688,22 +1688,6 @@ static void dcn401_build_pipe_pix_clk_params(struct pipe_ctx *pipe_ctx) } } -static int dcn401_get_power_profile(const struct dc_state *context) -{ - int uclk_mhz = context->bw_ctx.bw.dcn.clk.dramclk_khz / 1000; - int dpm_level = 0; - - for (int i = 0; i < context->clk_mgr->bw_params->clk_table.num_entries_per_clk.num_memclk_levels; i++) { - if (context->clk_mgr->bw_params->clk_table.entries[i].memclk_mhz == 0 || - uclk_mhz < context->clk_mgr->bw_params->clk_table.entries[i].memclk_mhz) - break; - if (uclk_mhz > context->clk_mgr->bw_params->clk_table.entries[i].memclk_mhz) - dpm_level++; - } - - return dpm_level; -} - static struct resource_funcs dcn401_res_pool_funcs = { .destroy = dcn401_destroy_resource_pool, .link_enc_create = dcn401_link_encoder_create, @@ -1730,7 +1714,6 @@ static struct resource_funcs dcn401_res_pool_funcs = { .prepare_mcache_programming = dcn401_prepare_mcache_programming, .build_pipe_pix_clk_params = dcn401_build_pipe_pix_clk_params, .calculate_mall_ways_from_bytes = dcn32_calculate_mall_ways_from_bytes, - .get_power_profile = dcn401_get_power_profile, }; static uint32_t read_pipe_fuses(struct dc_context *ctx) -- 2.51.0 From 915e2ad6dc9ab8c7d3fa997ff349aa027a212b8d Mon Sep 17 00:00:00 2001 From: Fudongwang Date: Sat, 14 Sep 2024 17:57:06 +0800 Subject: [PATCH 09/16] drm/amd/display: force TBT4 dock dsc on [why] TBT4 dock have bandwidth limitation, need dsc always on to support all modes. [how] force dsc always on when detect TBT4 dock. Reviewed-by: Aric Cyr Signed-off-by: Fudongwang Signed-off-by: Rodrigo Siqueira Tested-by: Daniel Wheeler Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/display/dc/dc_dp_types.h | 4 ++++ drivers/gpu/drm/amd/display/dc/link/link_detection.c | 12 ++++++++++++ .../display/dc/link/protocols/link_dp_capability.c | 7 +++++++ 3 files changed, 23 insertions(+) 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 41bd95e9177a..50fa8547d718 100644 --- a/drivers/gpu/drm/amd/display/dc/dc_dp_types.h +++ b/drivers/gpu/drm/amd/display/dc/dc_dp_types.h @@ -1166,6 +1166,7 @@ struct dpcd_caps { int8_t branch_dev_name[6]; int8_t branch_hw_revision; int8_t branch_fw_revision[2]; + int8_t branch_vendor_specific_data[4]; bool allow_invalid_MSA_timing_param; bool panel_mode_edp; @@ -1358,6 +1359,9 @@ struct dp_trace { #ifndef DP_TUNNELING_IRQ #define DP_TUNNELING_IRQ (1 << 5) #endif +#ifndef DP_BRANCH_VENDOR_SPECIFIC_START +#define DP_BRANCH_VENDOR_SPECIFIC_START 0x50C +#endif /** USB4 DPCD BW Allocation Registers Chapter 10.7 **/ #ifndef DP_TUNNELING_CAPABILITIES #define DP_TUNNELING_CAPABILITIES 0xE000D /* 1.4a */ diff --git a/drivers/gpu/drm/amd/display/dc/link/link_detection.c b/drivers/gpu/drm/amd/display/dc/link/link_detection.c index d21ee9d12d26..e026c728042a 100644 --- a/drivers/gpu/drm/amd/display/dc/link/link_detection.c +++ b/drivers/gpu/drm/amd/display/dc/link/link_detection.c @@ -48,6 +48,9 @@ #include "dm_helpers.h" #include "clk_mgr.h" + // Offset DPCD 050Eh == 0x5A +#define MST_HUB_ID_0x5A 0x5A + #define DC_LOGGER \ link->ctx->logger #define DC_LOGGER_INIT(logger) @@ -692,6 +695,15 @@ static void apply_dpia_mst_dsc_always_on_wa(struct dc_link *link) link->dpcd_caps.dsc_caps.dsc_basic_caps.fields.dsc_support.DSC_SUPPORT && !link->dc->debug.dpia_debug.bits.disable_mst_dsc_work_around) link->wa_flags.dpia_mst_dsc_always_on = true; + + if (link->ep_type == DISPLAY_ENDPOINT_USB4_DPIA && + link->type == dc_connection_mst_branch && + link->dpcd_caps.branch_dev_id == DP_BRANCH_DEVICE_ID_90CC24 && + link->dpcd_caps.branch_vendor_specific_data[2] == MST_HUB_ID_0x5A && + link->dpcd_caps.dsc_caps.dsc_basic_caps.fields.dsc_support.DSC_SUPPORT && + !link->dc->debug.dpia_debug.bits.disable_mst_dsc_work_around) { + link->wa_flags.dpia_mst_dsc_always_on = true; + } } static void revert_dpia_mst_dsc_always_on_wa(struct dc_link *link) 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 fe4aa2c158ea..286841456157 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 @@ -1207,6 +1207,13 @@ static void get_active_converter_info( dp_hw_fw_revision.ieee_fw_rev, sizeof(dp_hw_fw_revision.ieee_fw_rev)); } + + core_link_read_dpcd( + link, + DP_BRANCH_VENDOR_SPECIFIC_START, + (uint8_t *)link->dpcd_caps.branch_vendor_specific_data, + sizeof(link->dpcd_caps.branch_vendor_specific_data)); + if (link->dpcd_caps.dpcd_rev.raw >= DPCD_REV_14 && link->dpcd_caps.dongle_type != DISPLAY_DONGLE_NONE) { union dp_dfp_cap_ext dfp_cap_ext; -- 2.51.0 From 608c50afa1bebb1d21859bd45e602a5ed9e1802d Mon Sep 17 00:00:00 2001 From: Sridevi Arvindekar Date: Fri, 20 Sep 2024 14:43:09 -0400 Subject: [PATCH 10/16] drm/amd/display: Assign socclk in dml Assign socclk_khz value from dcn4x. Reviewed-by: Ariel Bernstein Signed-off-by: Sridevi Arvindekar Signed-off-by: Rodrigo Siqueira 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 8697eac1e1f7..7a01a956e4bb 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 @@ -1036,6 +1036,7 @@ void dml21_copy_clocks_to_dc_state(struct dml2_context *in_ctx, struct dc_state context->bw_ctx.bw.dcn.clk.p_state_change_support = in_ctx->v21.mode_programming.programming->uclk_pstate_supported; context->bw_ctx.bw.dcn.clk.dtbclk_en = in_ctx->v21.mode_programming.programming->min_clocks.dcn4x.dtbrefclk_khz > 0; context->bw_ctx.bw.dcn.clk.ref_dtbclk_khz = in_ctx->v21.mode_programming.programming->min_clocks.dcn4x.dtbrefclk_khz; + context->bw_ctx.bw.dcn.clk.socclk_khz = in_ctx->v21.mode_programming.programming->min_clocks.dcn4x.socclk_khz; } void dml21_extract_legacy_watermark_set(const struct dc *in_dc, struct dcn_watermarks *watermark, enum dml2_dchub_watermark_reg_set_index reg_set_idx, struct dml2_context *in_ctx) -- 2.51.0 From ffa1e31f70d2e97c121709b44a8960f5d7becb10 Mon Sep 17 00:00:00 2001 From: Zhongwei Date: Wed, 18 Sep 2024 14:43:49 +0800 Subject: [PATCH 11/16] drm/amd/display: Fix garbage or black screen when resetting otg [Why] For some EDP to MIPI panel, disabling OTG when link is alive like boot case, the converter might output garbage or show no display because our GPU is not sending required pixel data. Alos Dig fifo underflow was found which might cause garbage, when resetting otg for other types of EDP panels. [How] Skipping resetting OTG if the dig fifo is on. Make sure that the otg for the pipe is the one that the dig fifo is selecting via the FE mask. Reviewed-by: Nicholas Kazlauskas Signed-off-by: Zhongwei Signed-off-by: Rodrigo Siqueira Tested-by: Daniel Wheeler Signed-off-by: Alex Deucher --- .../dc/dio/dcn314/dcn314_dio_stream_encoder.c | 10 ++++++++++ .../amd/display/dc/hwss/dcn314/dcn314_hwseq.c | 16 ++++++++++++++++ 2 files changed, 26 insertions(+) diff --git a/drivers/gpu/drm/amd/display/dc/dio/dcn314/dcn314_dio_stream_encoder.c b/drivers/gpu/drm/amd/display/dc/dio/dcn314/dcn314_dio_stream_encoder.c index 5b343f745cf3..ae81451a3a72 100644 --- a/drivers/gpu/drm/amd/display/dc/dio/dcn314/dcn314_dio_stream_encoder.c +++ b/drivers/gpu/drm/amd/display/dc/dio/dcn314/dcn314_dio_stream_encoder.c @@ -83,6 +83,15 @@ void enc314_disable_fifo(struct stream_encoder *enc) REG_UPDATE(DIG_FIFO_CTRL0, DIG_FIFO_ENABLE, 0); } +static bool enc314_is_fifo_enabled(struct stream_encoder *enc) +{ + struct dcn10_stream_encoder *enc1 = DCN10STRENC_FROM_STRENC(enc); + uint32_t reset_val; + + REG_GET(DIG_FIFO_CTRL0, DIG_FIFO_ENABLE, &reset_val); + return (reset_val != 0); +} + void enc314_dp_set_odm_combine( struct stream_encoder *enc, bool odm_combine) @@ -468,6 +477,7 @@ static const struct stream_encoder_funcs dcn314_str_enc_funcs = { .enable_fifo = enc314_enable_fifo, .disable_fifo = enc314_disable_fifo, + .is_fifo_enabled = enc314_is_fifo_enabled, .set_input_mode = enc314_set_dig_input_mode, }; diff --git a/drivers/gpu/drm/amd/display/dc/hwss/dcn314/dcn314_hwseq.c b/drivers/gpu/drm/amd/display/dc/hwss/dcn314/dcn314_hwseq.c index e0054e654db6..9b88eb72086d 100644 --- a/drivers/gpu/drm/amd/display/dc/hwss/dcn314/dcn314_hwseq.c +++ b/drivers/gpu/drm/amd/display/dc/hwss/dcn314/dcn314_hwseq.c @@ -355,6 +355,20 @@ void dcn314_calculate_pix_rate_divider( } } +static bool dcn314_is_pipe_dig_fifo_on(struct pipe_ctx *pipe) +{ + return pipe && pipe->stream + // Check dig's otg instance. + && pipe->stream_res.stream_enc + && pipe->stream_res.stream_enc->funcs->dig_source_otg + && pipe->stream_res.tg->inst == pipe->stream_res.stream_enc->funcs->dig_source_otg(pipe->stream_res.stream_enc) + && pipe->stream->link && pipe->stream->link->link_enc + && pipe->stream->link->link_enc->funcs->is_dig_enabled + && pipe->stream->link->link_enc->funcs->is_dig_enabled(pipe->stream->link->link_enc) + && pipe->stream_res.stream_enc->funcs->is_fifo_enabled + && pipe->stream_res.stream_enc->funcs->is_fifo_enabled(pipe->stream_res.stream_enc); +} + void dcn314_resync_fifo_dccg_dio(struct dce_hwseq *hws, struct dc *dc, struct dc_state *context, unsigned int current_pipe_idx) { unsigned int i; @@ -374,6 +388,8 @@ void dcn314_resync_fifo_dccg_dio(struct dce_hwseq *hws, struct dc *dc, struct dc if (pipe->stream && (pipe->stream->dpms_off || dc_is_virtual_signal(pipe->stream->signal)) && !pipe->stream->apply_seamless_boot_optimization && !pipe->stream->apply_edp_fast_boot_optimization) { + if (dcn314_is_pipe_dig_fifo_on(pipe)) + continue; pipe->stream_res.tg->funcs->disable_crtc(pipe->stream_res.tg); reset_sync_context_for_pipe(dc, context, i); otg_disabled[i] = true; -- 2.51.0 From aacbed5b41d93bb741d8dab6e3e008a732f3e3df Mon Sep 17 00:00:00 2001 From: po-tchen Date: Fri, 20 Sep 2024 15:41:55 +0800 Subject: [PATCH 12/16] drm/amd/display: Display lost signal on playing video [Why] When Source extend the vblank to reach the minimum panel refresh rate, the vtotal length could have 1 line longer than the maximum supported vtotal. The reason is we optimized the vtotal/refresh-rate calculation to get more accurate vtotal number by rounding the calculation result. But when the target refresh rate is the minimum refresh rate, the vtotal result could be round up and over the maximum supported vtotal. Reviewed-by: Anthony Koo Signed-off-by: po-tchen Signed-off-by: Rodrigo Siqueira Tested-by: Daniel Wheeler Signed-off-by: Alex Deucher --- .../drm/amd/display/modules/freesync/freesync.c | 17 ++++++++++++++--- 1 file changed, 14 insertions(+), 3 deletions(-) diff --git a/drivers/gpu/drm/amd/display/modules/freesync/freesync.c b/drivers/gpu/drm/amd/display/modules/freesync/freesync.c index bbd259cea4f4..fc4268729017 100644 --- a/drivers/gpu/drm/amd/display/modules/freesync/freesync.c +++ b/drivers/gpu/drm/amd/display/modules/freesync/freesync.c @@ -48,6 +48,7 @@ #define VSYNCS_BETWEEN_FLIP_THRESHOLD 2 #define FREESYNC_CONSEC_FLIP_AFTER_VSYNC 5 #define FREESYNC_VSYNC_TO_FLIP_DELTA_IN_US 500 +#define MICRO_HZ_TO_HZ(x) (x / 1000000) struct core_freesync { struct mod_freesync public; @@ -132,9 +133,19 @@ unsigned int mod_freesync_calc_v_total_from_refresh( ((unsigned int)(div64_u64((1000000000ULL * 1000000), refresh_in_uhz))); - v_total = div64_u64(div64_u64(((unsigned long long)( - frame_duration_in_ns) * (stream->timing.pix_clk_100hz / 10)), - stream->timing.h_total) + 500000, 1000000); + if (MICRO_HZ_TO_HZ(refresh_in_uhz) <= stream->timing.min_refresh_in_uhz) { + /* When the target refresh rate is the minimum panel refresh rate, + * round down the vtotal value to avoid stretching vblank over + * panel's vtotal boundary. + */ + v_total = div64_u64(div64_u64(((unsigned long long)( + frame_duration_in_ns) * (stream->timing.pix_clk_100hz / 10)), + stream->timing.h_total), 1000000); + } else { + v_total = div64_u64(div64_u64(((unsigned long long)( + frame_duration_in_ns) * (stream->timing.pix_clk_100hz / 10)), + stream->timing.h_total) + 500000, 1000000); + } /* v_total cannot be less than nominal */ if (v_total < stream->timing.v_total) { -- 2.51.0 From 21b673da5192e1212059caf1154c3749a9be66fa Mon Sep 17 00:00:00 2001 From: "JinZe.Xu" Date: Thu, 26 Sep 2024 16:31:21 +0800 Subject: [PATCH 13/16] drm/amd/display: Noitfy DMCUB of D0/D3 state in hardware init [Why] Missing a dc_dmub_srv_notify_fw_dc_power_state in driver init. [How] Notify DMCUB of D0 state in hardware_init. Reviewed-by: Nicholas Kazlauskas Signed-off-by: JinZe.Xu Signed-off-by: Rodrigo Siqueira Tested-by: Daniel Wheeler Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/display/dc/core/dc.c | 1 + 1 file changed, 1 insertion(+) diff --git a/drivers/gpu/drm/amd/display/dc/core/dc.c b/drivers/gpu/drm/amd/display/dc/core/dc.c index 57ad6ce88f3f..981d9a327daf 100644 --- a/drivers/gpu/drm/amd/display/dc/core/dc.c +++ b/drivers/gpu/drm/amd/display/dc/core/dc.c @@ -1431,6 +1431,7 @@ void dc_hardware_init(struct dc *dc) detect_edp_presence(dc); if (dc->ctx->dce_environment != DCE_ENV_VIRTUAL_HW) dc->hwss.init_hw(dc); + dc_dmub_srv_notify_fw_dc_power_state(dc->ctx->dmub_srv, DC_ACPI_CM_POWER_STATE_D0); } void dc_init_callbacks(struct dc *dc, -- 2.51.0 From eda7f2e9bb3b6ba16cb19e521b433b22b220c0d0 Mon Sep 17 00:00:00 2001 From: Peterson Date: Mon, 23 Sep 2024 10:30:50 -0400 Subject: [PATCH 14/16] drm/amd/display: Fix low black values by increasing error [WHY] Regamma resolution for the first few black levels can have problems for calibration. [HOW] HW LUT is divided into N power-of-2 regions each with K segments. For SDR mode we set min point at 2^-10 and increments of 2^-13. It's generally more than 8-bit SDR needs, but some calibration tools and API use 12-bit curves. The fix shifts starting point to 2^-12 and starting increments at 2^-16. Reviewed-by: Krunoslav Kovac Signed-off-by: Peterson Signed-off-by: Rodrigo Siqueira Tested-by: Daniel Wheeler Signed-off-by: Alex Deucher --- .../amd/display/dc/dcn10/dcn10_cm_common.c | 25 ++++++++----------- .../amd/display/dc/dcn30/dcn30_cm_common.c | 25 ++++++++----------- 2 files changed, 20 insertions(+), 30 deletions(-) diff --git a/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_cm_common.c b/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_cm_common.c index eaed5d1c398a..dcd2cdfe91eb 100644 --- a/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_cm_common.c +++ b/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_cm_common.c @@ -365,23 +365,18 @@ bool cm_helper_translate_curve_to_hw_format(struct dc_context *ctx, region_start = -MAX_LOW_POINT; region_end = NUMBER_REGIONS - MAX_LOW_POINT; } else { - /* 11 segments - * segment is from 2^-10 to 2^1 + /* 13 segments + * segment is from 2^-12 to 2^0 * There are less than 256 points, for optimization */ - seg_distr[0] = 3; - seg_distr[1] = 4; - seg_distr[2] = 4; - seg_distr[3] = 4; - seg_distr[4] = 4; - seg_distr[5] = 4; - seg_distr[6] = 4; - seg_distr[7] = 4; - seg_distr[8] = 4; - seg_distr[9] = 4; - seg_distr[10] = 1; - - region_start = -10; + const uint8_t SEG_COUNT = 12; + + for (i = 0; i < SEG_COUNT; i++) + seg_distr[i] = 4; + + seg_distr[SEG_COUNT] = 1; + + region_start = -SEG_COUNT; region_end = 1; } diff --git a/drivers/gpu/drm/amd/display/dc/dcn30/dcn30_cm_common.c b/drivers/gpu/drm/amd/display/dc/dcn30/dcn30_cm_common.c index f31f0e3abfc0..1e1038fb04e8 100644 --- a/drivers/gpu/drm/amd/display/dc/dcn30/dcn30_cm_common.c +++ b/drivers/gpu/drm/amd/display/dc/dcn30/dcn30_cm_common.c @@ -140,23 +140,18 @@ bool cm3_helper_translate_curve_to_hw_format( region_start = -MAX_LOW_POINT; region_end = NUMBER_REGIONS - MAX_LOW_POINT; } else { - /* 11 segments - * segment is from 2^-10 to 2^0 + /* 13 segments + * segment is from 2^-12 to 2^0 * There are less than 256 points, for optimization */ - seg_distr[0] = 3; - seg_distr[1] = 4; - seg_distr[2] = 4; - seg_distr[3] = 4; - seg_distr[4] = 4; - seg_distr[5] = 4; - seg_distr[6] = 4; - seg_distr[7] = 4; - seg_distr[8] = 4; - seg_distr[9] = 4; - seg_distr[10] = 1; - - region_start = -10; + const uint8_t SEG_COUNT = 12; + + for (i = 0; i < SEG_COUNT; i++) + seg_distr[i] = 4; + + seg_distr[SEG_COUNT] = 1; + + region_start = -SEG_COUNT; region_end = 1; } -- 2.51.0 From 10ce17b7fb7079e164afe8d208fcbfac6ebd0e28 Mon Sep 17 00:00:00 2001 From: Dillon Varone Date: Thu, 26 Sep 2024 12:45:41 -0400 Subject: [PATCH 15/16] drm/amd/display: Remove programming outstanding updates for dcn35 [WHY&HOW] Programming outstanding updates is causing hangs on dcn35, so remove for now. Reviewed-by: Martin Leung Signed-off-by: Dillon Varone Signed-off-by: Rodrigo Siqueira Tested-by: Daniel Wheeler Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/display/dc/hwss/dcn35/dcn35_init.c | 1 - drivers/gpu/drm/amd/display/dc/hwss/dcn351/dcn351_init.c | 1 - 2 files changed, 2 deletions(-) diff --git a/drivers/gpu/drm/amd/display/dc/hwss/dcn35/dcn35_init.c b/drivers/gpu/drm/amd/display/dc/hwss/dcn35/dcn35_init.c index 2bbf1fef94fd..55dc5799e725 100644 --- a/drivers/gpu/drm/amd/display/dc/hwss/dcn35/dcn35_init.c +++ b/drivers/gpu/drm/amd/display/dc/hwss/dcn35/dcn35_init.c @@ -123,7 +123,6 @@ static const struct hw_sequencer_funcs dcn35_funcs = { .root_clock_control = dcn35_root_clock_control, .set_long_vtotal = dcn35_set_long_vblank, .calculate_pix_rate_divider = dcn32_calculate_pix_rate_divider, - .program_outstanding_updates = dcn32_program_outstanding_updates, }; static const struct hwseq_private_funcs dcn35_private_funcs = { diff --git a/drivers/gpu/drm/amd/display/dc/hwss/dcn351/dcn351_init.c b/drivers/gpu/drm/amd/display/dc/hwss/dcn351/dcn351_init.c index d00822e8daa5..a93864b63d48 100644 --- a/drivers/gpu/drm/amd/display/dc/hwss/dcn351/dcn351_init.c +++ b/drivers/gpu/drm/amd/display/dc/hwss/dcn351/dcn351_init.c @@ -122,7 +122,6 @@ static const struct hw_sequencer_funcs dcn351_funcs = { .root_clock_control = dcn35_root_clock_control, .set_long_vtotal = dcn35_set_long_vblank, .calculate_pix_rate_divider = dcn32_calculate_pix_rate_divider, - .program_outstanding_updates = dcn32_program_outstanding_updates, .setup_hpo_hw_control = dcn35_setup_hpo_hw_control, }; -- 2.51.0 From de08e41930e5a521d19924aaa2859361eb209df0 Mon Sep 17 00:00:00 2001 From: Charlene Liu Date: Wed, 25 Sep 2024 20:57:00 -0400 Subject: [PATCH 16/16] drm/amd/display: update sr_exit latency for z8 This is based on real asic performance result. Reviewed-by: Alvin Lee Signed-off-by: Charlene Liu Signed-off-by: Rodrigo Siqueira Tested-by: Daniel Wheeler Signed-off-by: Alex Deucher --- drivers/gpu/drm/amd/display/dc/dml/dcn351/dcn351_fpu.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) diff --git a/drivers/gpu/drm/amd/display/dc/dml/dcn351/dcn351_fpu.c b/drivers/gpu/drm/amd/display/dc/dml/dcn351/dcn351_fpu.c index a201dbb743d7..d9e63c4fdd95 100644 --- a/drivers/gpu/drm/amd/display/dc/dml/dcn351/dcn351_fpu.c +++ b/drivers/gpu/drm/amd/display/dc/dml/dcn351/dcn351_fpu.c @@ -204,8 +204,8 @@ struct _vcs_dpi_soc_bounding_box_st dcn3_51_soc = { .num_states = 8, .sr_exit_time_us = 28.0, .sr_enter_plus_exit_time_us = 30.0, - .sr_exit_z8_time_us = 250.0, - .sr_enter_plus_exit_z8_time_us = 350.0, + .sr_exit_z8_time_us = 263.0, + .sr_enter_plus_exit_z8_time_us = 363.0, .fclk_change_latency_us = 24.0, .usr_retraining_latency_us = 2, .writeback_latency_us = 12.0, -- 2.51.0