drm/amdgpu: add a BO metadata flag to disable write compression for Vulkan

Vulkan can't support DCC and Z/S compression on GFX12 without
WRITE_COMPRESS_DISABLE in this commit or a completely different DCC
interface.

AMDGPU_TILING_GFX12_SCANOUT is added because it's already used by userspace.

Cc: stable@vger.kernel.org # 6.12.x
Signed-off-by: Marek Olšák <marek.olsak@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

Author: Marek Olšák

drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c

@@ -119,9 +119,10 @@
  * - 3.57.0 - Compute tunneling on GFX10+
  * - 3.58.0 - Add GFX12 DCC support
  * - 3.59.0 - Cleared VRAM
+ * - 3.60.0 - Add AMDGPU_TILING_GFX12_DCC_WRITE_COMPRESS_DISABLE (Vulkan requirement)
  */
 #define KMS_DRIVER_MAJOR	3
-#define KMS_DRIVER_MINOR	59
+#define KMS_DRIVER_MINOR	60
 #define KMS_DRIVER_PATCHLEVEL	0
 
 /*

drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c

@@ -309,7 +309,7 @@ int amdgpu_ttm_copy_mem_to_mem(struct amdgpu_device *adev,
 	mutex_lock(&adev->mman.gtt_window_lock);
 	while (src_mm.remaining) {
 		uint64_t from, to, cur_size, tiling_flags;
-		uint32_t num_type, data_format, max_com;
+		uint32_t num_type, data_format, max_com, write_compress_disable;
 		struct dma_fence *next;
 
 		/* Never copy more than 256MiB at once to avoid a timeout */
@@ -340,9 +340,13 @@ int amdgpu_ttm_copy_mem_to_mem(struct amdgpu_device *adev,
 			max_com = AMDGPU_TILING_GET(tiling_flags, GFX12_DCC_MAX_COMPRESSED_BLOCK);
 			num_type = AMDGPU_TILING_GET(tiling_flags, GFX12_DCC_NUMBER_TYPE);
 			data_format = AMDGPU_TILING_GET(tiling_flags, GFX12_DCC_DATA_FORMAT);
+			write_compress_disable =
+				AMDGPU_TILING_GET(tiling_flags, GFX12_DCC_WRITE_COMPRESS_DISABLE);
 			copy_flags |= (AMDGPU_COPY_FLAGS_SET(MAX_COMPRESSED, max_com) |
 				       AMDGPU_COPY_FLAGS_SET(NUMBER_TYPE, num_type) |
-				       AMDGPU_COPY_FLAGS_SET(DATA_FORMAT, data_format));
+				       AMDGPU_COPY_FLAGS_SET(DATA_FORMAT, data_format) |
+				       AMDGPU_COPY_FLAGS_SET(WRITE_COMPRESS_DISABLE,
+							     write_compress_disable));
 		}
 
 		r = amdgpu_copy_buffer(ring, from, to, cur_size, resv,

drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.h

@@ -119,6 +119,8 @@ struct amdgpu_copy_mem {
 #define AMDGPU_COPY_FLAGS_NUMBER_TYPE_MASK		0x07
 #define AMDGPU_COPY_FLAGS_DATA_FORMAT_SHIFT		8
 #define AMDGPU_COPY_FLAGS_DATA_FORMAT_MASK		0x3f
+#define AMDGPU_COPY_FLAGS_WRITE_COMPRESS_DISABLE_SHIFT	14
+#define AMDGPU_COPY_FLAGS_WRITE_COMPRESS_DISABLE_MASK	0x1
 
 #define AMDGPU_COPY_FLAGS_SET(field, value) \
 	(((__u32)(value) & AMDGPU_COPY_FLAGS_##field##_MASK) << AMDGPU_COPY_FLAGS_##field##_SHIFT)

drivers/gpu/drm/amd/amdgpu/sdma_v7_0.c

@@ -1741,11 +1741,12 @@ static void sdma_v7_0_emit_copy_buffer(struct amdgpu_ib *ib,
 				       uint32_t byte_count,
 				       uint32_t copy_flags)
 {
-	uint32_t num_type, data_format, max_com;
+	uint32_t num_type, data_format, max_com, write_cm;
 
 	max_com = AMDGPU_COPY_FLAGS_GET(copy_flags, MAX_COMPRESSED);
 	data_format = AMDGPU_COPY_FLAGS_GET(copy_flags, DATA_FORMAT);
 	num_type = AMDGPU_COPY_FLAGS_GET(copy_flags, NUMBER_TYPE);
+	write_cm = AMDGPU_COPY_FLAGS_GET(copy_flags, WRITE_COMPRESS_DISABLE) ? 2 : 1;
 
 	ib->ptr[ib->length_dw++] = SDMA_PKT_COPY_LINEAR_HEADER_OP(SDMA_OP_COPY) |
 		SDMA_PKT_COPY_LINEAR_HEADER_SUB_OP(SDMA_SUBOP_COPY_LINEAR) |
@@ -1762,7 +1763,7 @@ static void sdma_v7_0_emit_copy_buffer(struct amdgpu_ib *ib,
 	if ((copy_flags & (AMDGPU_COPY_FLAGS_READ_DECOMPRESSED | AMDGPU_COPY_FLAGS_WRITE_COMPRESSED)))
 		ib->ptr[ib->length_dw++] = SDMA_DCC_DATA_FORMAT(data_format) | SDMA_DCC_NUM_TYPE(num_type) |
 			((copy_flags & AMDGPU_COPY_FLAGS_READ_DECOMPRESSED) ? SDMA_DCC_READ_CM(2) : 0) |
-			((copy_flags & AMDGPU_COPY_FLAGS_WRITE_COMPRESSED) ? SDMA_DCC_WRITE_CM(1) : 0) |
+			((copy_flags & AMDGPU_COPY_FLAGS_WRITE_COMPRESSED) ? SDMA_DCC_WRITE_CM(write_cm) : 0) |
 			SDMA_DCC_MAX_COM(max_com) | SDMA_DCC_MAX_UCOM(1);
 	else
 		ib->ptr[ib->length_dw++] = 0;

include/uapi/drm/amdgpu_drm.h

@@ -411,13 +411,20 @@ struct drm_amdgpu_gem_userptr {
 /* GFX12 and later: */
 #define AMDGPU_TILING_GFX12_SWIZZLE_MODE_SHIFT			0
 #define AMDGPU_TILING_GFX12_SWIZZLE_MODE_MASK			0x7
-/* These are DCC recompression setting for memory management: */
+/* These are DCC recompression settings for memory management: */
 #define AMDGPU_TILING_GFX12_DCC_MAX_COMPRESSED_BLOCK_SHIFT	3
 #define AMDGPU_TILING_GFX12_DCC_MAX_COMPRESSED_BLOCK_MASK	0x3 /* 0:64B, 1:128B, 2:256B */
 #define AMDGPU_TILING_GFX12_DCC_NUMBER_TYPE_SHIFT		5
 #define AMDGPU_TILING_GFX12_DCC_NUMBER_TYPE_MASK		0x7 /* CB_COLOR0_INFO.NUMBER_TYPE */
 #define AMDGPU_TILING_GFX12_DCC_DATA_FORMAT_SHIFT		8
 #define AMDGPU_TILING_GFX12_DCC_DATA_FORMAT_MASK		0x3f /* [0:4]:CB_COLOR0_INFO.FORMAT, [5]:MM */
+/* When clearing the buffer or moving it from VRAM to GTT, don't compress and set DCC metadata
+ * to uncompressed. Set when parts of an allocation bypass DCC and read raw data. */
+#define AMDGPU_TILING_GFX12_DCC_WRITE_COMPRESS_DISABLE_SHIFT	14
+#define AMDGPU_TILING_GFX12_DCC_WRITE_COMPRESS_DISABLE_MASK	0x1
+/* bit gap */
+#define AMDGPU_TILING_GFX12_SCANOUT_SHIFT			63
+#define AMDGPU_TILING_GFX12_SCANOUT_MASK			0x1
 
 /* Set/Get helpers for tiling flags. */
 #define AMDGPU_TILING_SET(field, value) \