drm/msm/dpu: fix blend setup for DMA4 and DMA5 layers

SM8550 uses new register to map SSPP_DMA4 and SSPP_DMA5 units to blend
stages. Add proper support for this register to allow using these two
planes for image processing.

Fixes: efcd0107 ("drm/msm/dpu: add support for SM8550")
Cc: Neil Armstrong <neil.armstrong@linaro.org>
Signed-off-by: Dmitry Baryshkov <dmitry.baryshkov@linaro.org>
Reviewed-by: Neil Armstrong <neil.armstrong@linaro.org>
Tested-by: Neil Armstrong <neil.armstrong@linaro.org> # on SM8550
Patchwork: https://patchwork.freedesktop.org/patch/518481/
Link: https://lore.kernel.org/r/20230116063316.728496-1-dmitry.baryshkov@linaro.orgSigned-off-by: Dmitry Baryshkov <dmitry.baryshkov@linaro.org>

drivers/gpu/drm/msm/disp/dpu1/dpu_hw_catalog.c

@@ -67,6 +67,9 @@
 #define CTL_SC7280_MASK \
 	(BIT(DPU_CTL_ACTIVE_CFG) | BIT(DPU_CTL_FETCH_ACTIVE) | BIT(DPU_CTL_VM_CFG))
 
+#define CTL_SM8550_MASK \
+	(CTL_SC7280_MASK | BIT(DPU_CTL_HAS_LAYER_EXT4))
+
 #define MERGE_3D_SM8150_MASK (0)
 
 #define DSPP_MSM8998_MASK BIT(DPU_DSPP_PCC) | BIT(DPU_DSPP_GC)
@@ -996,37 +999,37 @@ static const struct dpu_ctl_cfg sm8550_ctl[] = {
 	{
 	.name = "ctl_0", .id = CTL_0,
 	.base = 0x15000, .len = 0x290,
-	.features = CTL_SC7280_MASK | BIT(DPU_CTL_SPLIT_DISPLAY),
+	.features = CTL_SM8550_MASK | BIT(DPU_CTL_SPLIT_DISPLAY),
 	.intr_start = DPU_IRQ_IDX(MDP_SSPP_TOP0_INTR2, 9),
 	},
 	{
 	.name = "ctl_1", .id = CTL_1,
 	.base = 0x16000, .len = 0x290,
-	.features = CTL_SC7280_MASK | BIT(DPU_CTL_SPLIT_DISPLAY),
+	.features = CTL_SM8550_MASK | BIT(DPU_CTL_SPLIT_DISPLAY),
 	.intr_start = DPU_IRQ_IDX(MDP_SSPP_TOP0_INTR2, 10),
 	},
 	{
 	.name = "ctl_2", .id = CTL_2,
 	.base = 0x17000, .len = 0x290,
-	.features = CTL_SC7280_MASK,
+	.features = CTL_SM8550_MASK,
 	.intr_start = DPU_IRQ_IDX(MDP_SSPP_TOP0_INTR2, 11),
 	},
 	{
 	.name = "ctl_3", .id = CTL_3,
 	.base = 0x18000, .len = 0x290,
-	.features = CTL_SC7280_MASK,
+	.features = CTL_SM8550_MASK,
 	.intr_start = DPU_IRQ_IDX(MDP_SSPP_TOP0_INTR2, 12),
 	},
 	{
 	.name = "ctl_4", .id = CTL_4,
 	.base = 0x19000, .len = 0x290,
-	.features = CTL_SC7280_MASK,
+	.features = CTL_SM8550_MASK,
 	.intr_start = DPU_IRQ_IDX(MDP_SSPP_TOP0_INTR2, 13),
 	},
 	{
 	.name = "ctl_5", .id = CTL_5,
 	.base = 0x1a000, .len = 0x290,
-	.features = CTL_SC7280_MASK,
+	.features = CTL_SM8550_MASK,
 	.intr_start = DPU_IRQ_IDX(MDP_SSPP_TOP0_INTR2, 23),
 	},
 };

drivers/gpu/drm/msm/disp/dpu1/dpu_hw_catalog.h

@@ -199,6 +199,7 @@ enum {
  * @DPU_CTL_SPLIT_DISPLAY:	CTL supports video mode split display
  * @DPU_CTL_FETCH_ACTIVE:	Active CTL for fetch HW (SSPPs)
  * @DPU_CTL_VM_CFG:		CTL config to support multiple VMs
+ * @DPU_CTL_HAS_LAYER_EXT4:	CTL has the CTL_LAYER_EXT4 register
  * @DPU_CTL_MAX
  */
 enum {
@@ -206,6 +207,7 @@ enum {
 	DPU_CTL_ACTIVE_CFG,
 	DPU_CTL_FETCH_ACTIVE,
 	DPU_CTL_VM_CFG,
+	DPU_CTL_HAS_LAYER_EXT4,
 	DPU_CTL_MAX
 };
 

drivers/gpu/drm/msm/disp/dpu1/dpu_hw_ctl.c

@@ -17,6 +17,8 @@
 	(0x70 + (((lm) - LM_0) * 0x004))
 #define   CTL_LAYER_EXT3(lm)             \
 	(0xA0 + (((lm) - LM_0) * 0x004))
+#define CTL_LAYER_EXT4(lm)             \
+	(0xB8 + (((lm) - LM_0) * 0x004))
 #define   CTL_TOP                       0x014
 #define   CTL_FLUSH                     0x018
 #define   CTL_START                     0x01C
@@ -383,6 +385,7 @@ static void dpu_hw_ctl_setup_blendstage(struct dpu_hw_ctl *ctx,
 	struct dpu_hw_blk_reg_map *c = &ctx->hw;
 	u32 mixercfg = 0, mixercfg_ext = 0, mix, ext;
 	u32 mixercfg_ext2 = 0, mixercfg_ext3 = 0;
+	u32 mixercfg_ext4 = 0;
 	int i, j;
 	int stages;
 	int pipes_per_stage;
@@ -492,6 +495,20 @@ static void dpu_hw_ctl_setup_blendstage(struct dpu_hw_ctl *ctx,
 					mixercfg_ext2 |= mix << 4;
 				}
 				break;
+			case SSPP_DMA4:
+				if (rect_index == DPU_SSPP_RECT_1) {
+					mixercfg_ext4 |= ((i + 1) & 0xF) << 8;
+				} else {
+					mixercfg_ext4 |= ((i + 1) & 0xF) << 0;
+				}
+				break;
+			case SSPP_DMA5:
+				if (rect_index == DPU_SSPP_RECT_1) {
+					mixercfg_ext4 |= ((i + 1) & 0xF) << 12;
+				} else {
+					mixercfg_ext4 |= ((i + 1) & 0xF) << 4;
+				}
+				break;
 			case SSPP_CURSOR0:
 				mixercfg_ext |= ((i + 1) & 0xF) << 20;
 				break;
@@ -509,6 +526,8 @@ static void dpu_hw_ctl_setup_blendstage(struct dpu_hw_ctl *ctx,
 	DPU_REG_WRITE(c, CTL_LAYER_EXT(lm), mixercfg_ext);
 	DPU_REG_WRITE(c, CTL_LAYER_EXT2(lm), mixercfg_ext2);
 	DPU_REG_WRITE(c, CTL_LAYER_EXT3(lm), mixercfg_ext3);
+	if ((test_bit(DPU_CTL_HAS_LAYER_EXT4, &ctx->caps->features)))
+		DPU_REG_WRITE(c, CTL_LAYER_EXT4(lm), mixercfg_ext4);
 }