[ovs-dev] [PATCH] ofproto-dpif-xlate: Work around Linux netdev_max_backlog limit.

[Ben Pfaff]

On Wed, Sep 10, 2014 at 11:46:55AM -0300, Flavio Leitner wrote:
> On Wed, Sep 10, 2014 at 07:39:17AM -0700, Ben Pfaff wrote:
> > On Wed, Sep 10, 2014 at 08:42:50AM -0300, Flavio Leitner wrote:
> > > On Tue, Sep 09, 2014 at 03:07:28PM -0700, Ben Pfaff wrote:
> > > > Linux has an internal queue that temporarily holds packets transmitted to
> > > > certain network devices.  If too many packets are transmitted to such
> > > > network devices within a single list of actions, then packets tend to get
> > > > dropped.  Broadcast or flooded or multicast packets on bridges with
> > > > thousands of ports are examples of how this can occur.
> > > > 
> > > > This commit avoids the problem by implementing a flow in userspace when it
> > > > outputs its packet more times than the maximum length of the queue.
> > > > 
> > > > CC: Flavio Leitner <fbl at redhat.com>
> > > > Signed-off-by: Ben Pfaff <blp at nicira.com>
> > > > ---
> > > 
> > > Hi Ben,
> > > 
> > > The code looks good to me.
> > > 
> > > I used netdev_max_backlog  = 110 here and I was able to go all
> > > the 4k ports in the test.  Without the patch, the failure starts
> > > with 112 ports.
> > > 
> > > Thanks!
> > > 
> > > Acked-by: Flavio Leitner <fbl at redhat.com>
> > 
> > Thank you for testing!  I forgot to mention that I tested that the code
> > correctly read the netdev_max_backlog value but that I did not test that
> > it actually avoided the problem, so I am pleased to hear that it did.
> > 
> > I added your ack and also a Tested-by: since you tested it, and applied
> > this to master.
> 
> Thanks, could you please push it to branch-2.3 as well?

Done.

> Or how do you actually prefer? Requests like this one or
> backported patches posted to the list?

A request is usually good enough for me.