Re: [PATCH] ipv4: remove all rt cache entries on UNREGISTER event

[Octavian Purdila <opurdila@ixiacom.com>]

From: Eric Dumazet <eric.dumazet@gmail.com>
Date: Tuesday 28 September 2010, 19:33:49

> Le mardi 28 septembre 2010 à 17:24 +0200, Nicolas Dichtel a écrit :
> > Hi,
> > 
> > I face a problem when I try to remove an interface,
> > netdev_wait_allrefs() complains about refcount.
> > 
> > Here is a trivial scenario to reproduce the problem:
> > # ip tunnel add mode ipip remote 10.16.0.164 local 10.16.0.72 dev eth0
> > # ./a.out tunl1
> > # ip tunnel del tunl1
> > 
> > Note: a.out binary create an IPv4 raw socket, attach it to tunl1
> > (SO_BINDTODEVICE), set it as multicast (IP_MULTICAST_LOOP), set the
> > multicast interface to tunl1 (IP_MULTICAST_IF), build the IP header
> > (IP_HDRINCL) and then send a single packet (192.168.6.1 -> 224.0.0.18).
> > 
> > Note2: when a.out is executed, tunl1 has no ip address and is down.
> 
> CC Octavian Purdila, the patch author.
> 
> I am just wondering why this route is created in the first place.
> 
> Maybe a fix would be to forbid this ?
> 
> Some machines have a giant route cache, so its very important to avoid
> expensive scans.
> 
> > Then, I got a serie of "kernel:[1206699.728010] unregister_netdevice:
> > waiting for tunl1 to become free. Usage count = 3" and after some time,
> > interface is removed.
> > 
> > The problem is that route cache entries are only invalidate on
> > UNREGISTER event, and not removed (introduced by commit
> > e2ce146848c81af2f6d42e67990191c284bf0c33). We must wait that
> > rt_check_expire() remove the remaining route cache entries.
> > 
> > To fix the problem, I propose to remove a part of the previous commit.
> > 


Hi Nicolas,

The purpose of my original patch was to speed up interfaces deregistration 
even more after Eric's batch work. Reverting it might slow things down again, 
but since this is breaking things we should probably revert it and think a 
proper optimization afterward. I know that Eric B has done some more work in 
this area, for batch namespace cleanup, maybe the issue is not even there 
anymore. 

So, Ack from me.

We might even fully revert the patch, since the bit that is left doesn't have 
any value anymore.

Thanks,
tavi