[<prev] [next>] [<thread-prev] [day] [month] [year] [list]
Message-ID: <CADxym3bWqziz1-rHEZXC10JBgfO0Jc5S9nWEW2s3G09VvWioTA@mail.gmail.com>
Date: Wed, 4 Nov 2020 22:14:35 +0800
From: Menglong Dong <menglong8.dong@...il.com>
To: Nikolay Aleksandrov <nikolay@...dia.com>
Cc: Roopa Prabhu <roopa@...dia.com>,
"bridge@...ts.linux-foundation.org"
<bridge@...ts.linux-foundation.org>,
"dong.menglong@....com.cn" <dong.menglong@....com.cn>,
"davem@...emloft.net" <davem@...emloft.net>,
"kuba@...nel.org" <kuba@...nel.org>,
"linux-kernel@...r.kernel.org" <linux-kernel@...r.kernel.org>,
"netdev@...r.kernel.org" <netdev@...r.kernel.org>
Subject: Re: [PATCH] net: bridge: disable multicast while delete bridge
Dear Nik,
On Wed, Nov 4, 2020 at 12:26 AM Nikolay Aleksandrov <nikolay@...dia.com> wrote:
>
> On Mon, 2020-11-02 at 22:38 +0800, Menglong Dong wrote:
> > From: Menglong Dong <dong.menglong@....com.cn>
> >
> > This commit seems make no sense, as bridge is destroyed when
> > br_multicast_dev_del is called.
> >
> > In commit b1b9d366028f
> > ("bridge: move bridge multicast cleanup to ndo_uninit"), Xin Long
> > fixed the use-after-free panic in br_multicast_group_expired by
> > moving br_multicast_dev_del to ndo_uninit. However, that patch is
> > not applied to 4.4.X, and the bug exists.
> >
> > Fix that bug by disabling multicast in br_multicast_dev_del for
> > 4.4.X, and there is no harm for other branches.
> >
> > Signed-off-by: Menglong Dong <dong.menglong@....com.cn>
> > ---
> > net/bridge/br_multicast.c | 1 +
> > 1 file changed, 1 insertion(+)
> >
> > diff --git a/net/bridge/br_multicast.c b/net/bridge/br_multicast.c
> > index eae898c3cff7..9992fdff2951 100644
> > --- a/net/bridge/br_multicast.c
> > +++ b/net/bridge/br_multicast.c
> > @@ -3369,6 +3369,7 @@ void br_multicast_dev_del(struct net_bridge *br)
> > hlist_for_each_entry_safe(mp, tmp, &br->mdb_list, mdb_node)
> > br_multicast_del_mdb_entry(mp);
> > hlist_move_list(&br->mcast_gc_list, &deleted_head);
> > + br_opt_toggle(br, BROPT_MULTICAST_ENABLED, false);
> > spin_unlock_bh(&br->multicast_lock);
> >
> > br_multicast_gc(&deleted_head);
>
> This doesn't make any sense. It doesn't fix anything.
> If 4.4 has a problem then the relevant patches should get backported to it.
> We don't add random changes to fix older releases.
>
> Cheers,
> Nik
>
> Nacked-by: Nikolay Aleksandrov <nikolay@...dia.com>
Thanks for your patient explanation, and I see it now~
Cheers,
Menglong Dong
Powered by blists - more mailing lists