/**
 * cik_copy_dma - copy pages using the DMA engine
 *
 * @rdev: radeon_device pointer
 * @src_offset: src GPU address
 * @dst_offset: dst GPU address
 * @num_gpu_pages: number of GPU pages to xfer
 * @resv: reservation object to sync to
 *
 * Copy GPU paging using the DMA engine (CIK).
 * Used by the radeon ttm implementation to move pages if
 * registered as the asic copy callback.
 */
struct radeon_fence *cik_copy_dma(struct radeon_device *rdev,
				  uint64_t src_offset, uint64_t dst_offset,
				  unsigned num_gpu_pages,
				  struct reservation_object *resv)
{
	struct radeon_fence *fence;
	struct radeon_sync sync;
	int ring_index = rdev->asic->copy.dma_ring_index;
	struct radeon_ring *ring = &rdev->ring[ring_index];
	u32 size_in_bytes, cur_size_in_bytes;
	int i, num_loops;
	int r = 0;

	radeon_sync_create(&sync);

	size_in_bytes = (num_gpu_pages << RADEON_GPU_PAGE_SHIFT);
	num_loops = DIV_ROUND_UP(size_in_bytes, 0x1fffff);
	r = radeon_ring_lock(rdev, ring, num_loops * 7 + 14);
	if (r) {
		DRM_ERROR("radeon: moving bo (%d).\n", r);
		radeon_sync_free(rdev, &sync, NULL);
		return ERR_PTR(r);
	}

	radeon_sync_resv(rdev, &sync, resv, false);
	radeon_sync_rings(rdev, &sync, ring->idx);

	for (i = 0; i < num_loops; i++) {
		cur_size_in_bytes = size_in_bytes;
		if (cur_size_in_bytes > 0x1fffff)
			cur_size_in_bytes = 0x1fffff;
		size_in_bytes -= cur_size_in_bytes;
		radeon_ring_write(ring, SDMA_PACKET(SDMA_OPCODE_COPY, SDMA_COPY_SUB_OPCODE_LINEAR, 0));
		radeon_ring_write(ring, cur_size_in_bytes);
		radeon_ring_write(ring, 0); /* src/dst endian swap */
		radeon_ring_write(ring, lower_32_bits(src_offset));
		radeon_ring_write(ring, upper_32_bits(src_offset));
		radeon_ring_write(ring, lower_32_bits(dst_offset));
		radeon_ring_write(ring, upper_32_bits(dst_offset));
		src_offset += cur_size_in_bytes;
		dst_offset += cur_size_in_bytes;
	}

	r = radeon_fence_emit(rdev, &fence, ring->idx);
	if (r) {
		radeon_ring_unlock_undo(rdev, ring);
		radeon_sync_free(rdev, &sync, NULL);
		return ERR_PTR(r);
	}

	radeon_ring_unlock_commit(rdev, ring, false);
	radeon_sync_free(rdev, &sync, fence);

	return fence;
}
Exemple #2
0
/**
 * rv770_copy_dma - copy pages using the DMA engine
 *
 * @rdev: radeon_device pointer
 * @src_offset: src GPU address
 * @dst_offset: dst GPU address
 * @num_gpu_pages: number of GPU pages to xfer
 * @resv: reservation object to sync to
 *
 * Copy GPU paging using the DMA engine (r7xx).
 * Used by the radeon ttm implementation to move pages if
 * registered as the asic copy callback.
 */
struct radeon_fence *rv770_copy_dma(struct radeon_device *rdev,
		  uint64_t src_offset, uint64_t dst_offset,
		  unsigned num_gpu_pages,
				    struct reservation_object *resv)
{
	struct radeon_fence *fence;
	struct radeon_sync sync;
	int ring_index = rdev->asic->copy.dma_ring_index;
	struct radeon_ring *ring = &rdev->ring[ring_index];
	u32 size_in_dw, cur_size_in_dw;
	int i, num_loops;
	int r = 0;

	radeon_sync_create(&sync);

	size_in_dw = (num_gpu_pages << RADEON_GPU_PAGE_SHIFT) / 4;
	num_loops = DIV_ROUND_UP(size_in_dw, 0xFFFF);
	r = radeon_ring_lock(rdev, ring, num_loops * 5 + 8);
	if (r) {
		DRM_ERROR("radeon: moving bo (%d).\n", r);
		radeon_sync_free(rdev, &sync, NULL);
		return ERR_PTR(r);
	}

	radeon_sync_resv(rdev, &sync, resv, false);
	radeon_sync_rings(rdev, &sync, ring->idx);

	for (i = 0; i < num_loops; i++) {
		cur_size_in_dw = size_in_dw;
		if (cur_size_in_dw > 0xFFFF)
			cur_size_in_dw = 0xFFFF;
		size_in_dw -= cur_size_in_dw;
		radeon_ring_write(ring, DMA_PACKET(DMA_PACKET_COPY, 0, 0, cur_size_in_dw));
		radeon_ring_write(ring, dst_offset & 0xfffffffc);
		radeon_ring_write(ring, src_offset & 0xfffffffc);
		radeon_ring_write(ring, upper_32_bits(dst_offset) & 0xff);
		radeon_ring_write(ring, upper_32_bits(src_offset) & 0xff);
		src_offset += cur_size_in_dw * 4;
		dst_offset += cur_size_in_dw * 4;
	}

	r = radeon_fence_emit(rdev, &fence, ring->idx);
	if (r) {
		radeon_ring_unlock_undo(rdev, ring);
		radeon_sync_free(rdev, &sync, NULL);
		return ERR_PTR(r);
	}

	radeon_ring_unlock_commit(rdev, ring, false);
	radeon_sync_free(rdev, &sync, fence);

	return fence;
}