Commit 6333003b authored by Marek Olšák's avatar Marek Olšák Committed by Dave Airlie
Browse files

drm/radeon/kms: compute GPU addresses correctly on r600 



Signed-off-by: default avatarMarek Olšák <maraeo@gmail.com>
Reviewed-by: default avatarAlex Deucher <alexander.deucher@amd.com>
Signed-off-by: default avatarDave Airlie <airlied@redhat.com>
parent 747e42a1

drivers/gpu/drm/radeon/r600_cs.c

+68 −21
Original line number Diff line number Diff line
@@ -1538,6 +1538,8 @@ static int r600_packet3_check(struct radeon_cs_parser *p, 
	{

		int pred_op;

		int tmp;

		uint64_t offset;



		if (pkt->count != 1) {

			DRM_ERROR("bad SET PREDICATION\n");

			return -EINVAL;
@@ -1561,8 +1563,12 @@ static int r600_packet3_check(struct radeon_cs_parser *p, 
			return -EINVAL;

		}



		ib[idx + 0] = idx_value + (u32)(reloc->lobj.gpu_offset & 0xffffffff);

		ib[idx + 1] = tmp + (upper_32_bits(reloc->lobj.gpu_offset) & 0xff);

		offset = reloc->lobj.gpu_offset +

		         (idx_value & 0xfffffff0) +

		         ((u64)(tmp & 0xff) << 32);



		ib[idx + 0] = offset;

		ib[idx + 1] = (tmp & 0xffffff00) | (upper_32_bits(offset) & 0xff);

	}

	break;
@@ -1586,6 +1592,8 @@ static int r600_packet3_check(struct radeon_cs_parser *p, 
		}

		break;

	case PACKET3_DRAW_INDEX:

	{

		uint64_t offset;

		if (pkt->count != 3) {

			DRM_ERROR("bad DRAW_INDEX\n");

			return -EINVAL;
@@ -1595,14 +1603,21 @@ static int r600_packet3_check(struct radeon_cs_parser *p, 
			DRM_ERROR("bad DRAW_INDEX\n");

			return -EINVAL;

		}

		ib[idx+0] = idx_value + (u32)(reloc->lobj.gpu_offset & 0xffffffff);

		ib[idx+1] += upper_32_bits(reloc->lobj.gpu_offset) & 0xff;



		offset = reloc->lobj.gpu_offset +

		         idx_value +

		         ((u64)(radeon_get_ib_value(p, idx+1) & 0xff) << 32);



		ib[idx+0] = offset;

		ib[idx+1] = upper_32_bits(offset) & 0xff;



		r = r600_cs_track_check(p);

		if (r) {

			dev_warn(p->dev, "%s:%d invalid cmd stream\n", __func__, __LINE__);

			return r;

		}

		break;

	}

	case PACKET3_DRAW_INDEX_AUTO:

		if (pkt->count != 1) {

			DRM_ERROR("bad DRAW_INDEX_AUTO\n");
@@ -1633,13 +1648,20 @@ static int r600_packet3_check(struct radeon_cs_parser *p, 
		}

		/* bit 4 is reg (0) or mem (1) */

		if (idx_value & 0x10) {

			uint64_t offset;



			r = r600_cs_packet_next_reloc(p, &reloc);

			if (r) {

				DRM_ERROR("bad WAIT_REG_MEM\n");

				return -EINVAL;

			}

			ib[idx+1] += (u32)(reloc->lobj.gpu_offset & 0xffffffff);

			ib[idx+2] += upper_32_bits(reloc->lobj.gpu_offset) & 0xff;



			offset = reloc->lobj.gpu_offset +

			         (radeon_get_ib_value(p, idx+1) & 0xfffffff0) +

			         ((u64)(radeon_get_ib_value(p, idx+2) & 0xff) << 32);



			ib[idx+1] = (ib[idx+1] & 0x3) | (offset & 0xfffffff0);

			ib[idx+2] = upper_32_bits(offset) & 0xff;

		}

		break;

	case PACKET3_SURFACE_SYNC:
@@ -1664,16 +1686,25 @@ static int r600_packet3_check(struct radeon_cs_parser *p, 
			return -EINVAL;

		}

		if (pkt->count) {

			uint64_t offset;



			r = r600_cs_packet_next_reloc(p, &reloc);

			if (r) {

				DRM_ERROR("bad EVENT_WRITE\n");

				return -EINVAL;

			}

			ib[idx+1] += (u32)(reloc->lobj.gpu_offset & 0xffffffff);

			ib[idx+2] += upper_32_bits(reloc->lobj.gpu_offset) & 0xff;

			offset = reloc->lobj.gpu_offset +

			         (radeon_get_ib_value(p, idx+1) & 0xfffffff8) +

			         ((u64)(radeon_get_ib_value(p, idx+2) & 0xff) << 32);



			ib[idx+1] = offset & 0xfffffff8;

			ib[idx+2] = upper_32_bits(offset) & 0xff;

		}

		break;

	case PACKET3_EVENT_WRITE_EOP:

	{

		uint64_t offset;



		if (pkt->count != 4) {

			DRM_ERROR("bad EVENT_WRITE_EOP\n");

			return -EINVAL;
@@ -1683,9 +1714,15 @@ static int r600_packet3_check(struct radeon_cs_parser *p, 
			DRM_ERROR("bad EVENT_WRITE\n");

			return -EINVAL;

		}

		ib[idx+1] += (u32)(reloc->lobj.gpu_offset & 0xffffffff);

		ib[idx+2] += upper_32_bits(reloc->lobj.gpu_offset) & 0xff;



		offset = reloc->lobj.gpu_offset +

		         (radeon_get_ib_value(p, idx+1) & 0xfffffffc) +

		         ((u64)(radeon_get_ib_value(p, idx+2) & 0xff) << 32);



		ib[idx+1] = offset & 0xfffffffc;

		ib[idx+2] = (ib[idx+2] & 0xffffff00) | (upper_32_bits(offset) & 0xff);

		break;

	}

	case PACKET3_SET_CONFIG_REG:

		start_reg = (idx_value << 2) + PACKET3_SET_CONFIG_REG_OFFSET;

		end_reg = 4 * pkt->count + start_reg - 4;
@@ -1770,6 +1807,8 @@ static int r600_packet3_check(struct radeon_cs_parser *p, 
				ib[idx+1+(i*7)+3] += mip_offset;

				break;

			case SQ_TEX_VTX_VALID_BUFFER:

			{

				uint64_t offset64;

				/* vtx base */

				r = r600_cs_packet_next_reloc(p, &reloc);

				if (r) {
@@ -1782,11 +1821,15 @@ static int r600_packet3_check(struct radeon_cs_parser *p, 
					/* force size to size of the buffer */

					dev_warn(p->dev, "vbo resource seems too big (%d) for the bo (%ld)\n",

						 size + offset, radeon_bo_size(reloc->robj));

					ib[idx+1+(i*7)+1] = radeon_bo_size(reloc->robj);

					ib[idx+1+(i*7)+1] = radeon_bo_size(reloc->robj) - offset;

				}

				ib[idx+1+(i*7)+0] += (u32)((reloc->lobj.gpu_offset) & 0xffffffff);

				ib[idx+1+(i*7)+2] += upper_32_bits(reloc->lobj.gpu_offset) & 0xff;



				offset64 = reloc->lobj.gpu_offset + offset;

				ib[idx+1+(i*8)+0] = offset64;

				ib[idx+1+(i*8)+2] = (ib[idx+1+(i*8)+2] & 0xffffff00) |

						    (upper_32_bits(offset64) & 0xff);

				break;

			}

			case SQ_TEX_VTX_INVALID_TEXTURE:

			case SQ_TEX_VTX_INVALID_BUFFER:

			default:
@@ -1881,8 +1924,9 @@ static int r600_packet3_check(struct radeon_cs_parser *p, 
					  offset + 4, radeon_bo_size(reloc->robj));

				return -EINVAL;

			}

			ib[idx+1] += (u32)(reloc->lobj.gpu_offset & 0xffffffff);

			ib[idx+2] += upper_32_bits(reloc->lobj.gpu_offset) & 0xff;

			offset += reloc->lobj.gpu_offset;

			ib[idx+1] = offset;

			ib[idx+2] = upper_32_bits(offset) & 0xff;

		}

		/* Reading data from SRC_ADDRESS. */

		if (((idx_value >> 1) & 0x3) == 2) {
@@ -1899,8 +1943,9 @@ static int r600_packet3_check(struct radeon_cs_parser *p, 
					  offset + 4, radeon_bo_size(reloc->robj));

				return -EINVAL;

			}

			ib[idx+3] += (u32)(reloc->lobj.gpu_offset & 0xffffffff);

			ib[idx+4] += upper_32_bits(reloc->lobj.gpu_offset) & 0xff;

			offset += reloc->lobj.gpu_offset;

			ib[idx+3] = offset;

			ib[idx+4] = upper_32_bits(offset) & 0xff;

		}

		break;

	case PACKET3_COPY_DW:
@@ -1923,8 +1968,9 @@ static int r600_packet3_check(struct radeon_cs_parser *p, 
					  offset + 4, radeon_bo_size(reloc->robj));

				return -EINVAL;

			}

			ib[idx+1] += (u32)(reloc->lobj.gpu_offset & 0xffffffff);

			ib[idx+2] += upper_32_bits(reloc->lobj.gpu_offset) & 0xff;

			offset += reloc->lobj.gpu_offset;

			ib[idx+1] = offset;

			ib[idx+2] = upper_32_bits(offset) & 0xff;

		} else {

			/* SRC is a reg. */

			reg = radeon_get_ib_value(p, idx+1) << 2;
@@ -1946,8 +1992,9 @@ static int r600_packet3_check(struct radeon_cs_parser *p, 
					  offset + 4, radeon_bo_size(reloc->robj));

				return -EINVAL;

			}

			ib[idx+3] += (u32)(reloc->lobj.gpu_offset & 0xffffffff);

			ib[idx+4] += upper_32_bits(reloc->lobj.gpu_offset) & 0xff;

			offset += reloc->lobj.gpu_offset;

			ib[idx+3] = offset;

			ib[idx+4] = upper_32_bits(offset) & 0xff;

		} else {

			/* DST is a reg. */

			reg = radeon_get_ib_value(p, idx+3) << 2;