drm/amd/display: Disable MALL when TMZ surface

[Description]
- Don't use MALL buffering of any kind when the
  surface is TMZ
- Workaround for a HW bug

Reviewed-by: Jun Lei <Jun.Lei@amd.com>
Acked-by: Jasdeep Dhillon <jdhillon@amd.com>
Signed-off-by: Alvin Lee <Alvin.Lee2@amd.com>
Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

diff --git a/drivers/gpu/drm/amd/display/dc/core/dc.c b/drivers/gpu/drm/amd/display/dc/core/dc.c
index eeaa347..cfe207e 100644 (file)
--- a/drivers/gpu/drm/amd/display/dc/core/dc.c
+++ b/drivers/gpu/drm/amd/display/dc/core/dc.c
@@ -2322,9 +2322,13 @@ static enum surface_update_type det_surface_update(const struct dc *dc,
        type = get_scaling_info_update_type(u);
        elevate_update_type(&overall_type, type);
 
-       if (u->flip_addr)
+       if (u->flip_addr) {
                update_flags->bits.addr_update = 1;
-
+               if (u->flip_addr->address.tmz_surface != u->surface->address.tmz_surface) {
+                       update_flags->bits.tmz_changed = 1;
+                       elevate_update_type(&overall_type, UPDATE_TYPE_FULL);
+               }
+       }
        if (u->in_transfer_func)
                update_flags->bits.in_transfer_func_change = 1;
 

diff --git a/drivers/gpu/drm/amd/display/dc/dc.h b/drivers/gpu/drm/amd/display/dc/dc.h
index 8e392eb..e01c8f1 100644 (file)
--- a/drivers/gpu/drm/amd/display/dc/dc.h
+++ b/drivers/gpu/drm/amd/display/dc/dc.h
@@ -1120,6 +1120,7 @@ union surface_update_flags {
                uint32_t clock_change:1;
                uint32_t stereo_format_change:1;
                uint32_t lut_3d:1;
+               uint32_t tmz_changed:1;
                uint32_t full_update:1;
        } bits;
 

diff --git a/drivers/gpu/drm/amd/display/dc/dcn32/dcn32_hwseq.c b/drivers/gpu/drm/amd/display/dc/dcn32/dcn32_hwseq.c
index ba45b22..2242c87 100644 (file)
--- a/drivers/gpu/drm/amd/display/dc/dcn32/dcn32_hwseq.c
+++ b/drivers/gpu/drm/amd/display/dc/dcn32/dcn32_hwseq.c
@@ -369,7 +369,7 @@ bool dcn32_apply_idle_power_optimizations(struct dc *dc, bool enable)
        union dmub_rb_cmd cmd;
        uint8_t ways, i;
        int j;
-       bool stereo_in_use = false;
+       bool mall_ss_unsupported = false;
        struct dc_plane_state *plane = NULL;
 
        if (!dc->ctx->dmub_srv)
@@ -400,22 +400,23 @@ bool dcn32_apply_idle_power_optimizations(struct dc *dc, bool enable)
                         */
                        ways = dcn32_calculate_cab_allocation(dc, dc->current_state);
 
-                       /* MALL not supported with Stereo3D. If any plane is using stereo,
-                        * don't try to enter MALL.
+                       /* MALL not supported with Stereo3D or TMZ surface. If any plane is using stereo,
+                        * or TMZ surface, don't try to enter MALL.
                         */
                        for (i = 0; i < dc->current_state->stream_count; i++) {
                                for (j = 0; j < dc->current_state->stream_status[i].plane_count; j++) {
                                        plane = dc->current_state->stream_status[i].plane_states[j];
 
-                                       if (plane->address.type == PLN_ADDR_TYPE_GRPH_STEREO) {
-                                               stereo_in_use = true;
+                                       if (plane->address.type == PLN_ADDR_TYPE_GRPH_STEREO ||
+                                                       plane->address.tmz_surface) {
+                                               mall_ss_unsupported = true;
                                                break;
                                        }
                                }
-                               if (stereo_in_use)
+                               if (mall_ss_unsupported)
                                        break;
                        }
-                       if (ways <= dc->caps.cache_num_ways && !stereo_in_use) {
+                       if (ways <= dc->caps.cache_num_ways && !mall_ss_unsupported) {
                                memset(&cmd, 0, sizeof(cmd));
                                cmd.cab.header.type = DMUB_CMD__CAB_FOR_SS;
                                cmd.cab.header.sub_type = DMUB_CMD__CAB_DCN_SS_FIT_IN_CAB;
@@ -773,7 +774,8 @@ void dcn32_update_mall_sel(struct dc *dc, struct dc_state *context)
                                hubp->funcs->hubp_update_mall_sel(hubp,
                                        num_ways <= dc->caps.cache_num_ways &&
                                        pipe->stream->link->psr_settings.psr_version == DC_PSR_VERSION_UNSUPPORTED &&
-                                       pipe->plane_state->address.type !=  PLN_ADDR_TYPE_GRPH_STEREO ? 2 : 0,
+                                       pipe->plane_state->address.type !=  PLN_ADDR_TYPE_GRPH_STEREO &&
+                                       !pipe->plane_state->address.tmz_surface ? 2 : 0,
                                                        cache_cursor);
                        }
                }

diff --git a/drivers/gpu/drm/amd/display/dc/dml/dcn32/dcn32_fpu.c b/drivers/gpu/drm/amd/display/dc/dml/dcn32/dcn32_fpu.c
index 6687cfe..821474e 100644 (file)
--- a/drivers/gpu/drm/amd/display/dc/dml/dcn32/dcn32_fpu.c
+++ b/drivers/gpu/drm/amd/display/dc/dml/dcn32/dcn32_fpu.c
@@ -674,9 +674,10 @@ static bool dcn32_assign_subvp_pipe(struct dc *dc,
                 * - Not able to switch in vactive naturally (switching in active means the
                 *   DET provides enough buffer to hide the P-State switch latency -- trying
                 *   to combine this with SubVP can cause issues with the scheduling).
+                * - Not TMZ surface
                 */
                if (pipe->plane_state && !pipe->top_pipe &&
-                               pipe->stream->mall_stream_config.type == SUBVP_NONE && refresh_rate < 120 &&
+                               pipe->stream->mall_stream_config.type == SUBVP_NONE && refresh_rate < 120 && !pipe->plane_state->address.tmz_surface &&
                                vba->ActiveDRAMClockChangeLatencyMarginPerState[vba->VoltageLevel][vba->maxMpcComb][vba->pipe_plane[pipe_idx]] <= 0) {
                        while (pipe) {
                                num_pipes++;