drm/amd/display: Update plane tiling attributes for stream updates
authorNicholas Kazlauskas <nicholas.kazlauskas@amd.com>
Mon, 28 Jan 2019 16:02:08 +0000 (11:02 -0500)
committerAlex Deucher <alexander.deucher@amd.com>
Tue, 19 Mar 2019 20:04:03 +0000 (15:04 -0500)
[Why]
Tiling and DCC attributes can change when swapping framebuffers but
these will only ever get updated on full commits where
state->allow_modeset is true. But for the page-flip IOCTL
state->allow_modeset = false so these aren't updated and DCC changes
aren't being programmed.

[How]
Split out updating the tiling and DCC attributes into its own function
that's called when creating the plane for the first time but also
called when switching framebuffers before commit stream updates.

These functions also don't modify the plane state directly since the
plane state being used as reference is often the current one
being used by the hardware.

DC needs to be able to compare plane_state to the stream update
structures to determine if the hardware programming should happen. If
it's modified in place then nothing will be programmed and we can run
into concurrency issues.

Signed-off-by: Nicholas Kazlauskas <nicholas.kazlauskas@amd.com>
Reviewed-by: Harry Wentland <Harry.Wentland@amd.com>
Acked-by: Bhawanpreet Lakha <Bhawanpreet.Lakha@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm.c

index 3bf1157..0fbc27e 100644 (file)
@@ -2316,9 +2316,11 @@ static inline uint64_t get_dcc_address(uint64_t address, uint64_t tiling_flags)
        return offset ? (address + offset * 256) : 0;
 }
 
-static bool fill_plane_dcc_attributes(struct amdgpu_device *adev,
+static int fill_plane_dcc_attributes(struct amdgpu_device *adev,
                                      const struct amdgpu_framebuffer *afb,
-                                     struct dc_plane_state *plane_state,
+                                     const struct dc_plane_state *plane_state,
+                                     struct dc_plane_dcc_param *dcc,
+                                     struct dc_plane_address *address,
                                      uint64_t info)
 {
        struct dc *dc = adev->dm.dc;
@@ -2332,10 +2334,13 @@ static bool fill_plane_dcc_attributes(struct amdgpu_device *adev,
        memset(&output, 0, sizeof(output));
 
        if (!offset)
-               return false;
+               return 0;
+
+       if (plane_state->address.type != PLN_ADDR_TYPE_GRAPHICS)
+               return 0;
 
        if (!dc->cap_funcs.get_dcc_compression_cap)
-               return false;
+               return -EINVAL;
 
        input.format = plane_state->format;
        input.surface_size.width =
@@ -2352,26 +2357,96 @@ static bool fill_plane_dcc_attributes(struct amdgpu_device *adev,
                input.scan = SCAN_DIRECTION_VERTICAL;
 
        if (!dc->cap_funcs.get_dcc_compression_cap(dc, &input, &output))
-               return false;
+               return -EINVAL;
 
        if (!output.capable)
-               return false;
+               return -EINVAL;
 
        if (i64b == 0 && output.grph.rgb.independent_64b_blks != 0)
-               return false;
+               return -EINVAL;
 
-       plane_state->dcc.enable = 1;
-       plane_state->dcc.grph.meta_pitch =
+       dcc->enable = 1;
+       dcc->grph.meta_pitch =
                AMDGPU_TILING_GET(info, DCC_PITCH_MAX) + 1;
-       plane_state->dcc.grph.independent_64b_blks = i64b;
+       dcc->grph.independent_64b_blks = i64b;
 
        dcc_address = get_dcc_address(afb->address, info);
-       plane_state->address.grph.meta_addr.low_part =
-               lower_32_bits(dcc_address);
-       plane_state->address.grph.meta_addr.high_part =
-               upper_32_bits(dcc_address);
+       address->grph.meta_addr.low_part = lower_32_bits(dcc_address);
+       address->grph.meta_addr.high_part = upper_32_bits(dcc_address);
 
-       return true;
+       return 0;
+}
+
+static int
+fill_plane_tiling_attributes(struct amdgpu_device *adev,
+                            const struct amdgpu_framebuffer *afb,
+                            const struct dc_plane_state *plane_state,
+                            union dc_tiling_info *tiling_info,
+                            struct dc_plane_dcc_param *dcc,
+                            struct dc_plane_address *address,
+                            uint64_t tiling_flags)
+{
+       int ret;
+
+       memset(tiling_info, 0, sizeof(*tiling_info));
+       memset(dcc, 0, sizeof(*dcc));
+
+       /* Fill GFX8 params */
+       if (AMDGPU_TILING_GET(tiling_flags, ARRAY_MODE) == DC_ARRAY_2D_TILED_THIN1) {
+               unsigned int bankw, bankh, mtaspect, tile_split, num_banks;
+
+               bankw = AMDGPU_TILING_GET(tiling_flags, BANK_WIDTH);
+               bankh = AMDGPU_TILING_GET(tiling_flags, BANK_HEIGHT);
+               mtaspect = AMDGPU_TILING_GET(tiling_flags, MACRO_TILE_ASPECT);
+               tile_split = AMDGPU_TILING_GET(tiling_flags, TILE_SPLIT);
+               num_banks = AMDGPU_TILING_GET(tiling_flags, NUM_BANKS);
+
+               /* XXX fix me for VI */
+               tiling_info->gfx8.num_banks = num_banks;
+               tiling_info->gfx8.array_mode =
+                               DC_ARRAY_2D_TILED_THIN1;
+               tiling_info->gfx8.tile_split = tile_split;
+               tiling_info->gfx8.bank_width = bankw;
+               tiling_info->gfx8.bank_height = bankh;
+               tiling_info->gfx8.tile_aspect = mtaspect;
+               tiling_info->gfx8.tile_mode =
+                               DC_ADDR_SURF_MICRO_TILING_DISPLAY;
+       } else if (AMDGPU_TILING_GET(tiling_flags, ARRAY_MODE)
+                       == DC_ARRAY_1D_TILED_THIN1) {
+               tiling_info->gfx8.array_mode = DC_ARRAY_1D_TILED_THIN1;
+       }
+
+       tiling_info->gfx8.pipe_config =
+                       AMDGPU_TILING_GET(tiling_flags, PIPE_CONFIG);
+
+       if (adev->asic_type == CHIP_VEGA10 ||
+           adev->asic_type == CHIP_VEGA12 ||
+           adev->asic_type == CHIP_VEGA20 ||
+           adev->asic_type == CHIP_RAVEN) {
+               /* Fill GFX9 params */
+               tiling_info->gfx9.num_pipes =
+                       adev->gfx.config.gb_addr_config_fields.num_pipes;
+               tiling_info->gfx9.num_banks =
+                       adev->gfx.config.gb_addr_config_fields.num_banks;
+               tiling_info->gfx9.pipe_interleave =
+                       adev->gfx.config.gb_addr_config_fields.pipe_interleave_size;
+               tiling_info->gfx9.num_shader_engines =
+                       adev->gfx.config.gb_addr_config_fields.num_se;
+               tiling_info->gfx9.max_compressed_frags =
+                       adev->gfx.config.gb_addr_config_fields.max_compress_frags;
+               tiling_info->gfx9.num_rb_per_se =
+                       adev->gfx.config.gb_addr_config_fields.num_rb_per_se;
+               tiling_info->gfx9.swizzle =
+                       AMDGPU_TILING_GET(tiling_flags, SWIZZLE_MODE);
+               tiling_info->gfx9.shaderEnable = 1;
+
+               ret = fill_plane_dcc_attributes(adev, afb, plane_state, dcc,
+                                               address, tiling_flags);
+               if (ret)
+                       return ret;
+       }
+
+       return 0;
 }
 
 static int fill_plane_attributes_from_fb(struct amdgpu_device *adev,
@@ -2427,8 +2502,6 @@ static int fill_plane_attributes_from_fb(struct amdgpu_device *adev,
        }
 
        memset(&plane_state->address, 0, sizeof(plane_state->address));
-       memset(&plane_state->tiling_info, 0, sizeof(plane_state->tiling_info));
-       memset(&plane_state->dcc, 0, sizeof(plane_state->dcc));
 
        if (plane_state->format < SURFACE_PIXEL_FORMAT_VIDEO_BEGIN) {
                plane_state->address.type = PLN_ADDR_TYPE_GRAPHICS;
@@ -2461,58 +2534,11 @@ static int fill_plane_attributes_from_fb(struct amdgpu_device *adev,
                plane_state->color_space = COLOR_SPACE_YCBCR709;
        }
 
-       /* Fill GFX8 params */
-       if (AMDGPU_TILING_GET(tiling_flags, ARRAY_MODE) == DC_ARRAY_2D_TILED_THIN1) {
-               unsigned int bankw, bankh, mtaspect, tile_split, num_banks;
-
-               bankw = AMDGPU_TILING_GET(tiling_flags, BANK_WIDTH);
-               bankh = AMDGPU_TILING_GET(tiling_flags, BANK_HEIGHT);
-               mtaspect = AMDGPU_TILING_GET(tiling_flags, MACRO_TILE_ASPECT);
-               tile_split = AMDGPU_TILING_GET(tiling_flags, TILE_SPLIT);
-               num_banks = AMDGPU_TILING_GET(tiling_flags, NUM_BANKS);
-
-               /* XXX fix me for VI */
-               plane_state->tiling_info.gfx8.num_banks = num_banks;
-               plane_state->tiling_info.gfx8.array_mode =
-                               DC_ARRAY_2D_TILED_THIN1;
-               plane_state->tiling_info.gfx8.tile_split = tile_split;
-               plane_state->tiling_info.gfx8.bank_width = bankw;
-               plane_state->tiling_info.gfx8.bank_height = bankh;
-               plane_state->tiling_info.gfx8.tile_aspect = mtaspect;
-               plane_state->tiling_info.gfx8.tile_mode =
-                               DC_ADDR_SURF_MICRO_TILING_DISPLAY;
-       } else if (AMDGPU_TILING_GET(tiling_flags, ARRAY_MODE)
-                       == DC_ARRAY_1D_TILED_THIN1) {
-               plane_state->tiling_info.gfx8.array_mode = DC_ARRAY_1D_TILED_THIN1;
-       }
-
-       plane_state->tiling_info.gfx8.pipe_config =
-                       AMDGPU_TILING_GET(tiling_flags, PIPE_CONFIG);
-
-       if (adev->asic_type == CHIP_VEGA10 ||
-           adev->asic_type == CHIP_VEGA12 ||
-           adev->asic_type == CHIP_VEGA20 ||
-           adev->asic_type == CHIP_RAVEN) {
-               /* Fill GFX9 params */
-               plane_state->tiling_info.gfx9.num_pipes =
-                       adev->gfx.config.gb_addr_config_fields.num_pipes;
-               plane_state->tiling_info.gfx9.num_banks =
-                       adev->gfx.config.gb_addr_config_fields.num_banks;
-               plane_state->tiling_info.gfx9.pipe_interleave =
-                       adev->gfx.config.gb_addr_config_fields.pipe_interleave_size;
-               plane_state->tiling_info.gfx9.num_shader_engines =
-                       adev->gfx.config.gb_addr_config_fields.num_se;
-               plane_state->tiling_info.gfx9.max_compressed_frags =
-                       adev->gfx.config.gb_addr_config_fields.max_compress_frags;
-               plane_state->tiling_info.gfx9.num_rb_per_se =
-                       adev->gfx.config.gb_addr_config_fields.num_rb_per_se;
-               plane_state->tiling_info.gfx9.swizzle =
-                       AMDGPU_TILING_GET(tiling_flags, SWIZZLE_MODE);
-               plane_state->tiling_info.gfx9.shaderEnable = 1;
-
-               fill_plane_dcc_attributes(adev, amdgpu_fb, plane_state,
-                                         tiling_flags);
-       }
+       fill_plane_tiling_attributes(adev, amdgpu_fb, plane_state,
+                                    &plane_state->tiling_info,
+                                    &plane_state->dcc,
+                                    &plane_state->address,
+                                    tiling_flags);
 
        plane_state->visible = true;
        plane_state->scaling_quality.h_taps_c = 0;
@@ -4668,7 +4694,7 @@ static void amdgpu_dm_commit_planes(struct drm_atomic_state *state,
        int planes_count = 0, vpos, hpos;
        unsigned long flags;
        struct amdgpu_bo *abo;
-       uint64_t tiling_flags, dcc_address;
+       uint64_t tiling_flags;
        uint32_t target, target_vblank;
        uint64_t last_flip_vblank;
        bool vrr_active = acrtc_state->freesync_config.state == VRR_STATE_ACTIVE_VARIABLE;
@@ -4771,9 +4797,11 @@ static void amdgpu_dm_commit_planes(struct drm_atomic_state *state,
                bundle->flip_addrs[planes_count].address.grph.addr.low_part = lower_32_bits(afb->address);
                bundle->flip_addrs[planes_count].address.grph.addr.high_part = upper_32_bits(afb->address);
 
-               dcc_address = get_dcc_address(afb->address, tiling_flags);
-               bundle->flip_addrs[planes_count].address.grph.meta_addr.low_part = lower_32_bits(dcc_address);
-               bundle->flip_addrs[planes_count].address.grph.meta_addr.high_part = upper_32_bits(dcc_address);
+               fill_plane_tiling_attributes(dm->adev, afb, dc_plane,
+                       &bundle->plane_infos[planes_count].tiling_info,
+                       &bundle->plane_infos[planes_count].dcc,
+                       &bundle->flip_addrs[planes_count].address,
+                       tiling_flags);
 
                bundle->flip_addrs[planes_count].flip_immediate =
                                (crtc->state->pageflip_flags & DRM_MODE_PAGE_FLIP_ASYNC) != 0;