[openib-general] [RFC] MAD API changes to fix DMA mapping issues

[Roland Dreier]

    Sean> I'd like to make supporting very large sends a separate
    Sean> change at this point though.  That change could possibly
    Sean> combine the ib_mad_send_buf and ib_mad_recv_wc structures
    Sean> together.

Fair enough.  One approach would be something like struct scsi_cmnd:
it has a void *buffer member, which is either a pointer to the data
buffer itself, or a pointer to an array of struct sg_list that point
to the pieces of the buffer.

 - R.