[Openib-windows] [PATCH] mad with GRH valid

[Fabian Tillier]

Hi Yossi,

On 4/3/06, Yossi Leybovich <sleybo at mellanox.co.il> wrote:
> Fab
>
> While debuging problems with new mthca,
> we found that ib_get_mad create p_mad struct with p_mad_buf and p_grh
> which points to the same address.
> (the assert we add was fired)
>
> ual_mgr.c line 879
> ".....
> if( p_mad )
>  {
>   /* Save off the pointers since the proxy overwrites the element. */
>   p_mad_buf = p_mad->p_mad_buf;
>   p_grh = p_mad->p_grh;
>   CL_ASSERT((VOID *)p_mad_buf != (VOID *)p_grh);
>
>
>   ioctl_buf.in.p_mad_buf = p_mad_buf;
>   ioctl_buf.in.p_grh = p_grh;
>  }
> ....."
>
> This cause the kernel to copy the GRH data on the mad data and the mad
> droped by the OpenSM
>
> this patch fix the problem:

Good catch, committed in revision 274.

- Fab