On Tue, Apr 16, 2013 at 9:02 AM, Roland Dreier <[email protected]> wrote:
> On Mon, Apr 15, 2013 at 2:10 PM, Wendy Cheng <[email protected]> wrote:
>> Before closing this thread ..... I see there is a clear warning in
>> IPOIB code that enabling connected mode will cause multicast packet to
>> drop. I assume this is due to message ordering requirement and the
>> multicast packet could fall behind the large packet fragment that ends
>> up causing multicast logic to time out. If my assumption is not
>> correct, please do kindly correct me.
>
> No, it's because multicasts can't be sent over the RC connections, so
> they need to be sent via UD.  And the datagram size is limited by the
> underlying IB MTU, which is going to be either 4K or 2K most likely.
> However if you enable CM and then set your IPoIB MTU to something like
> 64K, the IP stack will happily create multicast messages bigger than
> the datagram MTU, and the IPoIB driver really can't do anything with
> those except throw them away.
>

Great thanks (for the clarification) !

Now, assuming CM is only turned on with devices (e.g. MLX card) that
are capable of doing segmentation off-load, i.e. skb_is_gso() is true,
will this limitation (losing multicast packets) still exist ? I see
multicast and normal data packets are directed to different QPs,
aren't they ?

I'm trying to understand the ramifications of running IPOIB-CM in a
production environment, as it does perform (much) better.

Thanks,
Wendy
--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
the body of a message to [email protected]
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Reply via email to