[openib-general] Re: User MAD support for cancel MAD

[Sean Hefty]

shaharf wrote:
> The interface issue is not so trivial. An IOCTL may do, but the problem
> is the parameter of the IOCTL. The most straight forward way is to
> specify a TID to cancel. This requires the usermode to avoid sending the
> same mad until it is timeouts. While this is reasonable, I would not
> force such a limitation,

Why would a client send the same MAD before it times out?  I think that 
enforcing such a limitation is perfectly acceptable.  We could argue 
that cancel MAD should work based on TID, management class, and 
SGID/SLID, rather than simply the TID.

> For example, if you have a retries mechanism that work over IB_MGT
> that just retires the mad after X msecs, I think you want to let it work
> also above openib. This means that either you allow no timeout/no kernel
> matching semantics

Clients are required to use the access layer timeout mechanism to 
properly match responses with requests.  If a timeout is not specified, 
the access layer will drop the response, since it cannot uniquely 
identify which client to hand the response to.  This is done to avoid 
making multiple data copies of a receive MAD in order to hand it to all 
clients.

- Sean