[openib-general] user_mad questions

[Roland Dreier]

    Sean> First, the only failed sends that are reported are response
    Sean> timeouts.  Is there a reason that other failures are not
    Sean> reported?  (I can understand treating the MADs as
    Sean> unreliable, but just wanted to check that this was the
    Sean> intent.)

No reason really except that all other errors "should never happen"
and therefore there's not much that a consumer can do with the error.

    Sean> Second, when a failed send is reported, there's a mismatch
    Sean> between the size of the data copied to the user and the size
    Sean> returned from ib_umad_read().  The copied data is
    Sean> sizeof(ib_user_mad) + sizeof(ib_mad).  The reported size
    Sean> turns out to be sizeof(ib_user_mad) + sizeof(ib_mad_hdr).
    Sean> It appears that the data copy is off.

Yes, that seems to be a long-standing bug.  Not sure when that was introduced.

 - R.