svn commit: r294327 - in head/sys: dev/cxgb dev/cxgbe dev/e1000 dev/hyperv/netvsc dev/ixgbe dev/mxge netinet sys

[Ryan Stone]

On Tue, Jan 19, 2016 at 3:23 PM, Adrian Chadd <adrian.chadd at gmail.com>
wrote:

> Sure, but TCP fragments and non-fragments can hash to different
> values, so suddenly you get interleaved packets.
>

If the NIC is hashing fragments and non-fragments to different values,
aren't we already potentially getting out-of-order packets?  If the same
stream is spread across multiple queues all bets are already off.  I'm not
sure that this change has made things significantly worse in this regard.

> So hopefully the LRO code
> handles seeing that hole in the TCP stream and will eject the whole
> stream up.

The LRO code is aware of sequence numbers and will not merge frames if
there is a hole in the sequence numbers.