computational MHD flows, boundary conditions

packer

I am trying to make my own solver for hypersonic mhd flow field,especially for the flow over blunt body,but i found it is really difficult to give the solid wall boundary conditions. How to give the initial magnetic field? According to some references, magnetic field is imposed on the solid wall at the beginning, while the other part of the flow field has no magnetic field. so i give the cells cling to the solid wall a initial magnetic field ,and plus the magnetic pressure to the inner energy at the same time, because we know in the MHD equations, e=p/(r-1)+Q+B*B/mu. As the computation proceeding, the magnetic field diffuse to the upstream slowly, and decrease the pressure greatly, and if the magnetic field is strong enough, negative pressure occurs(P=(r-1)*(e-Q-B*B/mu)),i am not sure giving initial magnetic field like this is right. Another problem: I am using cell centered scheme, if i do not give the initial magnetic field to the cells cling to the wall directly, and using the ghost cell method instead, to impose the boundary conditions.How to reallize it? If anybody can give me some advises,i will be very appreciate of you~ Thanks in advance~