cvs commit: src/sys/sys mbuf.h src/sys/net if_ethersubr.c src/sys/dev/mxge mxge_lro.c

[Scott Long]

Andre Oppermann wrote:
> Andrew Gallatin wrote:
>> gallatin    2007-06-11 14:59:56 UTC
>>
>>   FreeBSD src repository
>>
>>   Modified files:
>>     sys/sys              mbuf.h     sys/net             
>> if_ethersubr.c     sys/dev/mxge         mxge_lro.c   Log:
>>   Allow drivers, such as cxgb and mxge, which support LRO to bypass
>>   the MTU check in ether_input() on LRO merged frames.
>>     Discussed with: kmacy
> 
> Not discussed with: andre
> 
> Your change isn't the right way to make this work.  LRO is an interface
> capability (that should have the option to disable it) and the test in
> ether_input() should go on that instead.  LRO is not an information
> that is needed beyond ether_input() and thus doesn't have to be a mbuf
> flag.
> 
> I've indicated that I'm working in this area as well and at least
> dropping an email or a ping IRC would have been nice.  I would have
> told you the above right away.  My common version of LRO isn't ready
> yet as I'm a bit short on time and I chose to concentrate on TCP it-
> self.  We only have to make sure that we don't exclude a common LRO
> implementation due to API/ABI issues for 7.1R.
> 

Drew's commit looks simple and non-obtrusive enough that it can likely
be replaced once your perfected LRO implementation is done and in the
tree.  Until that happens, I can't imagine a good reason to block his
and Kip's work.

Scott