Message-ID: <20090520104838.GB25492@hmsreliant.think-freely.org>
Date:	Wed, 20 May 2009 06:48:38 -0400
From:	Neil Horman <nhorman@...driver.com>
To:	Eric Dumazet <dada1@...mosbay.com>
Cc:	David Miller <davem@...emloft.net>, jarkao2@...il.com, lav@....ru,
	shemminger@...ux-foundation.org, netdev@...r.kernel.org
Subject: Re: [PATCH] net: fix rtable leak in net/ipv4/route.c

On Wed, May 20, 2009 at 08:14:28AM +0200, Eric Dumazet wrote:
> Eric Dumazet a écrit :
> > David Miller a écrit :
> >> From: Neil Horman <nhorman@...driver.com>
> >> Date: Tue, 19 May 2009 15:24:50 -0400
> >>
> >>>> Moving whole group in front would defeat the purpose of move, actually,
> >>>> since rank in chain is used to decay the timeout in garbage collector.
> >>>> (search for tmo >>= 1; )
> >>>>
> >>> Argh, so the list is implicitly ordered by expiration time.  That
> >>> really defeats the entire purpose of doing grouping in the ilst at
> >>> all.  If thats the case, then I agree, its probably better to to
> >>> take the additional visitation hit in in check_expire above than to
> >>> try and preserve ordering.
> >> Yes, this seems best.
> >>
> >> I was worried that somehow the ordering also influences lookups,
> >> because the TOS bits don't go into the hash so I worried that it would
> >> be important that explicit TOS values appear before wildcard ones.
> >> But it doesn't appear that this is an issue, we don't have wildcard
> >> TOSs in the rtable entries, they are always explicit.
> >>
> >> So I would like to see an explicit final patch from Eric so we can get
> >> this fixed now.
> >>
> > 
> > I would like to split patches because we have two bugs indeed, and
> > I prefer to get attention for both problems, I dont remember Neil acknowledged
> > the length computation problem.
> > 
> > First and small patch, candidate for net-2.6 and stable (for 2.6.29) :
> > 
> 
> Then here is the patch on top on previous one.
> 
> Thanks to all
> 
> [PATCH] net: fix rtable leak in net/ipv4/route.c
> 
> Alexander V. Lukyanov found a regression in 2.6.29 and made a complete
> analysis found in http://bugzilla.kernel.org/show_bug.cgi?id=13339
> Quoted here because its a perfect one :
> 
> begin_of_quotation 
>  2.6.29 patch has introduced flexible route cache rebuilding. Unfortunately the
>  patch has at least one critical flaw, and another problem.
> 
>  rt_intern_hash calculates rthi pointer, which is later used for new entry
>  insertion. The same loop calculates cand pointer which is used to clean the
>  list. If the pointers are the same, rtable leak occurs, as first the cand is
>  removed then the new entry is appended to it.
> 
>  This leak leads to unregister_netdevice problem (usage count > 0).
> 
>  Another problem of the patch is that it tries to insert the entries in certain
>  order, to facilitate counting of entries distinct by all but QoS parameters.
>  Unfortunately, referencing an existing rtable entry moves it to list beginning,
>  to speed up further lookups, so the carefully built order is destroyed.
> 
>  For the first problem the simplest patch it to set rthi=0 when rthi==cand, but
>  it will also destroy the ordering.
> end_of_quotation
> 
> 
> Problematic commit is 1080d709fb9d8cd4392f93476ee46a9d6ea05a5b
> (net: implement emergency route cache rebulds when gc_elasticity is exceeded)
> 
> Trying to keep dst_entries ordered is too complex and breaks the fact that
> order should depend on the frequency of use for garbage collection.
> 
> A possible fix is to make rt_intern_hash() simpler, and only makes
> rt_check_expire() a litle bit smarter, being able to cope with an arbitrary
> entries order. The added loop is running on cache hot data, while cpu
> is prefetching next object, so should be unnoticied.
> 
> Reported-and-analyzed-by: Alexander V. Lukyanov <lav@....ru>
> Signed-off-by: Eric Dumazet <dada1@...mosbay.com>
Acked-by: Neil Horman <nhorman@...driver.com>

--
To unsubscribe from this list: send the line "unsubscribe netdev" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Powered by blists - more mailing lists