[openib-general] Re: [PATCH] check for valid MGID in user space

[Roland Dreier]

    Jack> The following patch checks validity of MGID when
    Jack> attaching/detaching a QP to/from a multicast group (for
    Jack> user-space only). IB spec demands that multicast gids start
    Jack> with 0xFF in 0-th byte. (IB Spec v1.2,section 4.1.1 (page
    Jack> 144)).

This seems like a strange place to check the condition.  Why do we
want to allow kernel callers to use invalid MGIDs?  In other words it
would seem more appropriate to me to put the check in verbs.c.

Also no Signed-off-by: line for your patch...

 - R.