linux-kernel - Re: [PATCH net] ipv4: route: Prevent rt_bind

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives
Hash Suite: Windows password security audit tool. GUI, reports in PDF.
[<prev] [next>] [<thread-prev] [day] [month] [year] [list]
Message-ID: <CACueBy6LKYmusLjQPnQGCoSZQLEVAo5_X47B-gaH-2dSx6xDuw@mail.gmail.com>
Date: Thu, 6 Nov 2025 08:31:00 +0800
From: chuang <nashuiliang@...il.com>
To: Ido Schimmel <idosch@...sch.org>
Cc: Eric Dumazet <edumazet@...gle.com>, stable@...r.kernel.org, 
	"David S. Miller" <davem@...emloft.net>, David Ahern <dsahern@...nel.org>, 
	Jakub Kicinski <kuba@...nel.org>, Paolo Abeni <pabeni@...hat.com>, Simon Horman <horms@...nel.org>, 
	Networking <netdev@...r.kernel.org>, open list <linux-kernel@...r.kernel.org>
Subject: Re: [PATCH net] ipv4: route: Prevent rt_bind_exception() from
 rebinding stale fnhe

Thanks, your analysis is excellent and makes perfect sense. I can
briefly describe the issue.

This problem took quite some time to analyze overall — we enabled
netdev refcnt, added dst tracepoints, and eventually captured a race
condition between fnhe deletion and rt_bind_exception.

In our environment, we use the sit driver(ip tunnel). During the xmit
path, it records the PMTU for each destination, creating or updating
fnhe entries (even when the MTU is already appropriate). Because there
are many data flows, the sit driver updates PMTU very frequently,
which leads to the race condition mentioned above.

Sorry for the brief summary — I’ll provide a more detailed explanation
later, along with the patch verification method.

On Wed, Nov 5, 2025 at 23:34 Ido Schimmel <idosch@...sch.org> wrote:
>
> On Wed, Nov 05, 2025 at 06:26:22AM -0800, Eric Dumazet wrote:
> > On Mon, Nov 3, 2025 at 7:09 PM chuang <nashuiliang@...il.com> wrote:
> > >
> > > From 35dbc9abd8da820007391b707bd2c1a9c99ee67d Mon Sep 17 00:00:00 2001
> > > From: Chuang Wang <nashuiliang@...il.com>
> > > Date: Tue, 4 Nov 2025 02:52:11 +0000
> > > Subject: [PATCH net] ipv4: route: Prevent rt_bind_exception() from rebinding
> > >  stale fnhe
> > >
> > > A race condition exists between fnhe_remove_oldest() and
> > > rt_bind_exception() where a fnhe that is scheduled for removal can be
> > > rebound to a new dst.
> > >
> > > The issue occurs when fnhe_remove_oldest() selects an fnhe (fnheX)
> > > for deletion, but before it can be flushed and freed via RCU,
> > > CPU 0 enters rt_bind_exception() and attempts to reuse the entry.
> > >
> > > CPU 0                             CPU 1
> > > __mkroute_output()
> > >   find_exception() [fnheX]
> > >                                   update_or_create_fnhe()
> > >                                     fnhe_remove_oldest() [fnheX]
> > >   rt_bind_exception() [bind dst]
> > >                                   RCU callback [fnheX freed, dst leak]
> > >
> > > If rt_bind_exception() successfully binds fnheX to a new dst, the
> > > newly bound dst will never be properly freed because fnheX will
> > > soon be released by the RCU callback, leading to a permanent
> > > reference count leak on the old dst and the device.
> > >
> > > This issue manifests as a device reference count leak and a
> > > warning in dmesg when unregistering the net device:
> > >
> > >   unregister_netdevice: waiting for ethX to become free. Usage count = N
> > >
> > > Fix this race by clearing 'oldest->fnhe_daddr' before calling
> > > fnhe_flush_routes(). Since rt_bind_exception() checks this field,
> > > setting it to zero prevents the stale fnhe from being reused and
> > > bound to a new dst just before it is freed.
> > >
> > > Cc: stable@...r.kernel.org
> > > Fixes: 67d6d681e15b ("ipv4: make exception cache less predictible")
> >
> > I do not see how this commit added the bug you are looking at ?
>
> Not the author, but my understanding is that the issue is that an
> exception entry which is queued for deletion allows a dst entry to be
> bound to it. As such, nobody will ever release the reference from the
> dst entry and the associated net device.
>
> Before 67d6d681e15b, exception entries were only queued for deletion by
> ip_del_fnhe() and it prevented dst entries from binding themselves to
> the deleted exception entry by clearing 'fnhe->fnhe_daddr' which is
> checked in rt_bind_exception(). See ee60ad219f5c7.
>
> 67d6d681e15b added another point in the code that queues exception
> entries for deletion, but without clearing 'fnhe->fnhe_daddr' first.
> Therefore, it added another instance of the bug that was fixed in
> ee60ad219f5c7.
>
> >
> > > Signed-off-by: Chuang Wang <nashuiliang@...il.com>
> > > ---
> > >  net/ipv4/route.c | 5 +++++
> > >  1 file changed, 5 insertions(+)
> > >
> > > diff --git a/net/ipv4/route.c b/net/ipv4/route.c
> > > index 6d27d3610c1c..b549d6a57307 100644
> > > --- a/net/ipv4/route.c
> > > +++ b/net/ipv4/route.c
> > > @@ -607,6 +607,11 @@ static void fnhe_remove_oldest(struct
> > > fnhe_hash_bucket *hash)
> > >                         oldest_p = fnhe_p;
> > >                 }
> > >         }
> > > +
> > > +       /* Clear oldest->fnhe_daddr to prevent this fnhe from being
> > > +        * rebound with new dsts in rt_bind_exception().
> > > +        */
> > > +       oldest->fnhe_daddr = 0;
> > >         fnhe_flush_routes(oldest);
> > >         *oldest_p = oldest->fnhe_next;
> > >         kfree_rcu(oldest, rcu);
> > > --
> >