[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-Id: <20250505221419.2672473-468-sashal@kernel.org>
Date: Mon, 5 May 2025 18:11:24 -0400
From: Sasha Levin <sashal@...nel.org>
To: linux-kernel@...r.kernel.org,
stable@...r.kernel.org
Cc: Dillon Varone <dillon.varone@....com>,
Austin Zheng <Austin.Zheng@....com>,
Alex Hung <alex.hung@....com>,
Daniel Wheeler <daniel.wheeler@....com>,
Alex Deucher <alexander.deucher@....com>,
Sasha Levin <sashal@...nel.org>,
austin.zheng@....com,
jun.lei@....com,
harry.wentland@....com,
sunpeng.li@....com,
christian.koenig@....com,
airlied@...il.com,
simona@...ll.ch,
hamzamahfooz@...ux.microsoft.com,
ryanseto@....com,
rostrows@....com,
aurabindo.pillai@....com,
alvin.lee2@....com,
chenhuacai@...nel.org,
roheetchavan@...il.com,
chris.park@....com,
george.shen@....com,
jerry.zuo@....com,
jiapeng.chong@...ux.alibaba.com,
rodrigo.siqueira@....com,
chaitanya.dhere@....com,
okuzhyln@....com,
ruanjinjie@...wei.com,
colin.i.king@...il.com,
wenjing.liu@....com,
amd-gfx@...ts.freedesktop.org,
dri-devel@...ts.freedesktop.org
Subject: [PATCH AUTOSEL 6.14 468/642] drm/amd/display: Fixes for mcache programming in DML21
From: Dillon Varone <dillon.varone@....com>
[ Upstream commit c909a49128a31bced8cfbd2dfb0a4fe56e01a6d0 ]
[WHY & HOW]
- Fix indexing phantom planes for mcache programming in the wrapper
- Fix phantom mcache allocations to align with HW guidance
- Fix mcache assignment for chroma plane for multi-planar formats
Reviewed-by: Austin Zheng <Austin.Zheng@....com>
Signed-off-by: Dillon Varone <dillon.varone@....com>
Signed-off-by: Alex Hung <alex.hung@....com>
Tested-by: Daniel Wheeler <daniel.wheeler@....com>
Signed-off-by: Alex Deucher <alexander.deucher@....com>
Signed-off-by: Sasha Levin <sashal@...nel.org>
---
.../amd/display/dc/dml2/dml21/dml21_utils.c | 1 -
.../amd/display/dc/dml2/dml21/dml21_wrapper.c | 14 +++++++++
.../display/dc/dml2/dml21/inc/dml_top_types.h | 1 +
.../dml2/dml21/src/dml2_core/dml2_core_dcn4.c | 30 ++++++++++++++++++-
.../src/dml2_core/dml2_core_dcn4_calcs.c | 3 ++
.../dml21/src/dml2_pmo/dml2_pmo_dcn4_fams2.c | 21 +++++++++++++
.../dml2/dml21/src/dml2_top/dml2_top_soc15.c | 8 -----
7 files changed, 68 insertions(+), 10 deletions(-)
diff --git a/drivers/gpu/drm/amd/display/dc/dml2/dml21/dml21_utils.c b/drivers/gpu/drm/amd/display/dc/dml2/dml21/dml21_utils.c
index 1e56d995cd0e7..930e86cdb88a2 100644
--- a/drivers/gpu/drm/amd/display/dc/dml2/dml21/dml21_utils.c
+++ b/drivers/gpu/drm/amd/display/dc/dml2/dml21/dml21_utils.c
@@ -232,7 +232,6 @@ void dml21_program_dc_pipe(struct dml2_context *dml_ctx, struct dc_state *contex
context->bw_ctx.bw.dcn.clk.dppclk_khz = pipe_ctx->plane_res.bw.dppclk_khz;
dml21_populate_mall_allocation_size(context, dml_ctx, pln_prog, pipe_ctx);
- memcpy(&context->bw_ctx.bw.dcn.mcache_allocations[pipe_ctx->pipe_idx], &pln_prog->mcache_allocation, sizeof(struct dml2_mcache_surface_allocation));
bool sub_vp_enabled = is_sub_vp_enabled(pipe_ctx->stream->ctx->dc, context);
diff --git a/drivers/gpu/drm/amd/display/dc/dml2/dml21/dml21_wrapper.c b/drivers/gpu/drm/amd/display/dc/dml2/dml21/dml21_wrapper.c
index d6fd13f43c08f..5d16f36ec95c8 100644
--- a/drivers/gpu/drm/amd/display/dc/dml2/dml21/dml21_wrapper.c
+++ b/drivers/gpu/drm/amd/display/dc/dml2/dml21/dml21_wrapper.c
@@ -129,6 +129,7 @@ static void dml21_calculate_rq_and_dlg_params(const struct dc *dc, struct dc_sta
struct pipe_ctx *dc_main_pipes[__DML2_WRAPPER_MAX_STREAMS_PLANES__];
struct pipe_ctx *dc_phantom_pipes[__DML2_WRAPPER_MAX_STREAMS_PLANES__] = {0};
int num_pipes;
+ unsigned int dml_phantom_prog_idx;
context->bw_ctx.bw.dcn.clk.dppclk_khz = 0;
@@ -142,6 +143,9 @@ static void dml21_calculate_rq_and_dlg_params(const struct dc *dc, struct dc_sta
context->bw_ctx.bw.dcn.mall_ss_psr_active_size_bytes = 0;
context->bw_ctx.bw.dcn.mall_subvp_size_bytes = 0;
+ /* phantom's start after main planes */
+ dml_phantom_prog_idx = in_ctx->v21.mode_programming.programming->display_config.num_planes;
+
for (dml_prog_idx = 0; dml_prog_idx < DML2_MAX_PLANES; dml_prog_idx++) {
pln_prog = &in_ctx->v21.mode_programming.programming->plane_programming[dml_prog_idx];
@@ -167,6 +171,16 @@ static void dml21_calculate_rq_and_dlg_params(const struct dc *dc, struct dc_sta
dml21_program_dc_pipe(in_ctx, context, dc_phantom_pipes[dc_pipe_index], pln_prog, stream_prog);
}
}
+
+ /* copy per plane mcache allocation */
+ memcpy(&context->bw_ctx.bw.dcn.mcache_allocations[dml_prog_idx], &pln_prog->mcache_allocation, sizeof(struct dml2_mcache_surface_allocation));
+ if (pln_prog->phantom_plane.valid) {
+ memcpy(&context->bw_ctx.bw.dcn.mcache_allocations[dml_phantom_prog_idx],
+ &pln_prog->phantom_plane.mcache_allocation,
+ sizeof(struct dml2_mcache_surface_allocation));
+
+ dml_phantom_prog_idx++;
+ }
}
/* assign global clocks */
diff --git a/drivers/gpu/drm/amd/display/dc/dml2/dml21/inc/dml_top_types.h b/drivers/gpu/drm/amd/display/dc/dml2/dml21/inc/dml_top_types.h
index d2d053f2354d0..0ab19cf4d2421 100644
--- a/drivers/gpu/drm/amd/display/dc/dml2/dml21/inc/dml_top_types.h
+++ b/drivers/gpu/drm/amd/display/dc/dml2/dml21/inc/dml_top_types.h
@@ -245,6 +245,7 @@ struct dml2_per_plane_programming {
struct {
bool valid;
struct dml2_plane_parameters descriptor;
+ struct dml2_mcache_surface_allocation mcache_allocation;
struct dml2_dchub_per_pipe_register_set *pipe_regs[DML2_MAX_PLANES];
} phantom_plane;
};
diff --git a/drivers/gpu/drm/amd/display/dc/dml2/dml21/src/dml2_core/dml2_core_dcn4.c b/drivers/gpu/drm/amd/display/dc/dml2/dml21/src/dml2_core/dml2_core_dcn4.c
index 7216d25c783e6..44d2969a904ea 100644
--- a/drivers/gpu/drm/amd/display/dc/dml2/dml21/src/dml2_core/dml2_core_dcn4.c
+++ b/drivers/gpu/drm/amd/display/dc/dml2/dml21/src/dml2_core/dml2_core_dcn4.c
@@ -253,7 +253,8 @@ static void expand_implict_subvp(const struct display_configuation_with_meta *di
static void pack_mode_programming_params_with_implicit_subvp(struct dml2_core_instance *core, const struct display_configuation_with_meta *display_cfg,
const struct dml2_display_cfg *svp_expanded_display_cfg, struct dml2_display_cfg_programming *programming, struct dml2_core_scratch *scratch)
{
- unsigned int stream_index, plane_index, pipe_offset, stream_already_populated_mask, main_plane_index;
+ unsigned int stream_index, plane_index, pipe_offset, stream_already_populated_mask, main_plane_index, mcache_index;
+ unsigned int total_main_mcaches_required = 0;
int total_pipe_regs_copied = 0;
int dml_internal_pipe_index = 0;
const struct dml2_plane_parameters *main_plane;
@@ -324,6 +325,13 @@ static void pack_mode_programming_params_with_implicit_subvp(struct dml2_core_in
dml2_core_calcs_get_mall_allocation(&core->clean_me_up.mode_lib, &programming->plane_programming[plane_index].surface_size_mall_bytes, dml_internal_pipe_index);
+ memcpy(&programming->plane_programming[plane_index].mcache_allocation,
+ &display_cfg->stage2.mcache_allocations[plane_index],
+ sizeof(struct dml2_mcache_surface_allocation));
+ total_main_mcaches_required += programming->plane_programming[plane_index].mcache_allocation.num_mcaches_plane0 +
+ programming->plane_programming[plane_index].mcache_allocation.num_mcaches_plane1 -
+ (programming->plane_programming[plane_index].mcache_allocation.last_slice_sharing.plane0_plane1 ? 1 : 0);
+
for (pipe_offset = 0; pipe_offset < programming->plane_programming[plane_index].num_dpps_required; pipe_offset++) {
// Assign storage for this pipe's register values
programming->plane_programming[plane_index].pipe_regs[pipe_offset] = &programming->pipe_regs[total_pipe_regs_copied];
@@ -362,6 +370,22 @@ static void pack_mode_programming_params_with_implicit_subvp(struct dml2_core_in
memcpy(&programming->plane_programming[main_plane_index].phantom_plane.descriptor, phantom_plane, sizeof(struct dml2_plane_parameters));
dml2_core_calcs_get_mall_allocation(&core->clean_me_up.mode_lib, &programming->plane_programming[main_plane_index].svp_size_mall_bytes, dml_internal_pipe_index);
+
+ /* generate mcache allocation, phantoms use identical mcache configuration, but in the MALL set and unique mcache ID's beginning after all main ID's */
+ memcpy(&programming->plane_programming[main_plane_index].phantom_plane.mcache_allocation,
+ &programming->plane_programming[main_plane_index].mcache_allocation,
+ sizeof(struct dml2_mcache_surface_allocation));
+ for (mcache_index = 0; mcache_index < programming->plane_programming[main_plane_index].phantom_plane.mcache_allocation.num_mcaches_plane0; mcache_index++) {
+ programming->plane_programming[main_plane_index].phantom_plane.mcache_allocation.global_mcache_ids_plane0[mcache_index] += total_main_mcaches_required;
+ programming->plane_programming[main_plane_index].phantom_plane.mcache_allocation.global_mcache_ids_mall_plane0[mcache_index] =
+ programming->plane_programming[main_plane_index].phantom_plane.mcache_allocation.global_mcache_ids_plane0[mcache_index];
+ }
+ for (mcache_index = 0; mcache_index < programming->plane_programming[main_plane_index].phantom_plane.mcache_allocation.num_mcaches_plane1; mcache_index++) {
+ programming->plane_programming[main_plane_index].phantom_plane.mcache_allocation.global_mcache_ids_plane1[mcache_index] += total_main_mcaches_required;
+ programming->plane_programming[main_plane_index].phantom_plane.mcache_allocation.global_mcache_ids_mall_plane1[mcache_index] =
+ programming->plane_programming[main_plane_index].phantom_plane.mcache_allocation.global_mcache_ids_plane1[mcache_index];
+ }
+
for (pipe_offset = 0; pipe_offset < programming->plane_programming[main_plane_index].num_dpps_required; pipe_offset++) {
// Assign storage for this pipe's register values
programming->plane_programming[main_plane_index].phantom_plane.pipe_regs[pipe_offset] = &programming->pipe_regs[total_pipe_regs_copied];
@@ -571,6 +595,10 @@ bool core_dcn4_mode_programming(struct dml2_core_mode_programming_in_out *in_out
dml2_core_calcs_get_mall_allocation(&core->clean_me_up.mode_lib, &in_out->programming->plane_programming[plane_index].surface_size_mall_bytes, dml_internal_pipe_index);
+ memcpy(&in_out->programming->plane_programming[plane_index].mcache_allocation,
+ &in_out->display_cfg->stage2.mcache_allocations[plane_index],
+ sizeof(struct dml2_mcache_surface_allocation));
+
for (pipe_offset = 0; pipe_offset < in_out->programming->plane_programming[plane_index].num_dpps_required; pipe_offset++) {
in_out->programming->plane_programming[plane_index].plane_descriptor = &in_out->programming->display_config.plane_descriptors[plane_index];
diff --git a/drivers/gpu/drm/amd/display/dc/dml2/dml21/src/dml2_core/dml2_core_dcn4_calcs.c b/drivers/gpu/drm/amd/display/dc/dml2/dml21/src/dml2_core/dml2_core_dcn4_calcs.c
index c1ff869512f27..8ad7704b76691 100644
--- a/drivers/gpu/drm/amd/display/dc/dml2/dml21/src/dml2_core/dml2_core_dcn4_calcs.c
+++ b/drivers/gpu/drm/amd/display/dc/dml2/dml21/src/dml2_core/dml2_core_dcn4_calcs.c
@@ -2638,6 +2638,9 @@ static void calculate_mcache_setting(
// Luma/Chroma combine in the last mcache
// In the case of Luma/Chroma combine-mCache (with lc_comb_mcache==1), all mCaches except the last segment are filled as much as possible, when stay aligned to mvmpg boundary
if (*p->lc_comb_mcache && l->is_dual_plane) {
+ /* if luma and chroma planes share an mcache, increase total chroma mcache count */
+ *p->num_mcaches_c = *p->num_mcaches_c + 1;
+
for (n = 0; n < *p->num_mcaches_l - 1; n++)
p->mcache_offsets_l[n] = (n + 1) * l->mvmpg_per_mcache_lb_l * l->mvmpg_access_width_l;
p->mcache_offsets_l[*p->num_mcaches_l - 1] = l->full_vp_access_width_l;
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 a3324f7b9ba68..15c906c42ec45 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
@@ -1082,12 +1082,21 @@ static bool all_timings_support_svp(const struct dml2_pmo_instance *pmo,
const struct dml2_fams2_meta *stream_fams2_meta;
unsigned int microschedule_vlines;
unsigned int i;
+ unsigned int mcaches_per_plane;
+ unsigned int total_mcaches_required = 0;
unsigned int num_planes_per_stream[DML2_MAX_PLANES] = { 0 };
/* confirm timing it is not a centered timing */
for (i = 0; i < display_config->display_config.num_planes; i++) {
plane_descriptor = &display_config->display_config.plane_descriptors[i];
+ mcaches_per_plane = 0;
+
+ if (plane_descriptor->surface.dcc.enable) {
+ mcaches_per_plane += display_config->stage2.mcache_allocations[i].num_mcaches_plane0 +
+ display_config->stage2.mcache_allocations[i].num_mcaches_plane1 -
+ (display_config->stage2.mcache_allocations[i].last_slice_sharing.plane0_plane1 ? 1 : 0);
+ }
if (is_bit_set_in_bitfield(mask, (unsigned char)plane_descriptor->stream_index)) {
num_planes_per_stream[plane_descriptor->stream_index]++;
@@ -1098,7 +1107,19 @@ static bool all_timings_support_svp(const struct dml2_pmo_instance *pmo,
plane_descriptor->composition.rotation_angle != dml2_rotation_0) {
return false;
}
+
+ /* phantom requires same number of mcaches as main */
+ if (plane_descriptor->surface.dcc.enable) {
+ mcaches_per_plane *= 2;
+ }
}
+
+ total_mcaches_required += mcaches_per_plane;
+ }
+
+ if (total_mcaches_required > pmo->soc_bb->num_dcc_mcaches) {
+ /* too many mcaches required */
+ return false;
}
for (i = 0; i < DML2_MAX_PLANES; i++) {
diff --git a/drivers/gpu/drm/amd/display/dc/dml2/dml21/src/dml2_top/dml2_top_soc15.c b/drivers/gpu/drm/amd/display/dc/dml2/dml21/src/dml2_top/dml2_top_soc15.c
index a8f58f8448e42..dc2ce5e77f579 100644
--- a/drivers/gpu/drm/amd/display/dc/dml2/dml21/src/dml2_top/dml2_top_soc15.c
+++ b/drivers/gpu/drm/amd/display/dc/dml2/dml21/src/dml2_top/dml2_top_soc15.c
@@ -831,7 +831,6 @@ static bool dml2_top_soc15_build_mode_programming(struct dml2_build_mode_program
bool uclk_pstate_success = false;
bool vmin_success = false;
bool stutter_success = false;
- unsigned int i;
memset(l, 0, sizeof(struct dml2_build_mode_programming_locals));
memset(in_out->programming, 0, sizeof(struct dml2_display_cfg_programming));
@@ -976,13 +975,6 @@ static bool dml2_top_soc15_build_mode_programming(struct dml2_build_mode_program
l->base_display_config_with_meta.stage5.success = true;
}
- /*
- * Populate mcache programming
- */
- for (i = 0; i < in_out->display_config->num_planes; i++) {
- in_out->programming->plane_programming[i].mcache_allocation = l->base_display_config_with_meta.stage2.mcache_allocations[i];
- }
-
/*
* Call DPMM to map all requirements to minimum clock state
*/
--
2.39.5
Powered by blists - more mailing lists