svn commit: r316309 - head/sys/dev/qlxgbe

[Somayajulu, David]

>> > I know this is not a new topic but a little more descriptive commit-log would have been nicer. Also, you should update the manpage reflecting this change. i.e. now it also supports software LRO when h/w LRO is disabled.
>> Will do. Sorry about that.
Minor correction. Please note that the driver provides the ability to choose between SoftwareLRO and HW LRO, when LRO is enabled - it is HW LRO by default. If LRO is turned off via ifconfig, neither Software nor HW LRO is enabled.

Cheers
David S.

-----Original Message-----
From: hiren panchasara [mailto:hiren at strugglingcoder.info] 
Sent: Thursday, March 30, 2017 4:12 PM
To: Somayajulu, David <David.Somayajulu at cavium.com>
Cc: David C Somayajulu <davidcs at FreeBSD.org>; src-committers at freebsd.org; svn-src-all at freebsd.org; svn-src-head at freebsd.org
Subject: Re: svn commit: r316309 - head/sys/dev/qlxgbe

On 03/30/17 at 11:07P, Somayajulu, David wrote:
> Hi Hiren,
> > I know this is not a new topic but a little more descriptive commit-log would have been nicer. Also, you should update the manpage reflecting this change. i.e. now it also supports software LRO when h/w LRO is disabled.
> Will do. Sorry about that.

Thanks!
> 
> > Do you know of a case where one would want to disable h/w lro and enable s/w lro? I guess where you want to free up nic and make cpu do more work?
> I was under the impression as well, that s/w lro is moot, if h/w lro was available, till one costumer asked for it. Not sure what the use case is.

I thought this idea (like gro in linux) was popular when lro in h/w was considered buggy and couldn't correctly do batching (i.e. hide useful flags). Not sure if that's still the case.

cheers,
Hiren