[rds-devel] RDS: updates for net-next

[Or Gerlitz]

On Tue, Jun 9, 2009 at 8:42 PM, Andy Grover <andy.grover at oracle.com> wrote:
> Remember I'm not targeting Linus' merge window, I'm targeting DaveM's,
> which is earlier -- he likes to have all the new stuff sitting in his
> net-next for a while before Linus' merge window opens. That branch adds
> TCP support that I basically revived from git, and haven't tested beyond
> a ping test. It needs more testing and have checkpatch.pl errors fixed.
> I expect DaveM's .32 net-next tree will be opening by July so that's
> pretty soon, and we'll be in the pipeline for .32.

Andy, the patches that I was referring to sit in the
code-drop/20090513 branch, were the TCP support is in the
future/20090428 branch. From your comment I understand that the TCP
code isn't ready for merge anyway, anytime soon since its untested.

Basically, its correct that Dave would like to have patches sit enough
time in net-next before he pushes to Linus, but its not always the
case, specifically as you can see over netdev, nowadays he gets
patches for net-next everyday.

>> Looking forward, lets max the effort to enable distros which are to be
>> released in the next months/quarters (e.g FC11, RHEL6, OEL6, SLES11 SP1,
>> etc) to base their RDS code on mainline bits and not on other venues,

> None of these will be using 2.6.31 anyways :)

mmm, interesting, okay, you may know that even if they don't use .31
as its still possible to try and negotiate .31 bits / patchset of this
or that driver. For example, if one of them uses .29 and rds isn't
there, you can still try to push it there.

> the rds IB wire protocol changed between ofed 1.3 and 1.4. RDS in OFED 1.4.1 > has compatibility code to talk to 1.3 hosts but what's in mainline doesn't.
> Do you think I should also get that into mainline, or should mainline just pretend > RDS prior to 1.4 never existed?

best if you can get it in

Or.