I think you are missing some understanding of convolution. I recommend looking it up on wikipedia. When you convolve a matrix with a kernel, the result is another matrix with the same size as the original matrix. Above, dst is allocating only 6 elements when your input matrix is 42 elements. dst should be 42 elements as well.
I recommend for convolution that you look at vImageConvolve_Planar8 or vImageConvolve_PlanarF. It is much better built and takes care of a lot of corner edge scenarios.
Hope this helps