From mboxrd@z Thu Jan  1 00:00:00 1970
From: Eric Dumazet <eric.dumazet@gmail.com>
Subject: Re: a problem tcp_v4_err()
Date: Fri, 12 Nov 2010 19:33:23 +0100
Message-ID: <1289586803.3185.275.camel@edumazet-laptop>
References: <20101111210341.31350.86916.stgit@paris.rdu.redhat.com>
	 <00c201cb81eb$84e18160$8ea48420$@com>
	 <1289578108.3083.95.camel@localhost.localdomain>
	 <1289578532.3185.265.camel@edumazet-laptop>
	 <20101112163543.GB122902@jupiter.n2.diac24.net>
	 <4CDD7145.8070606@trash.net> <20101112175715.GB16544@ms2.inr.ac.ru>
	 <1289585578.3185.268.camel@edumazet-laptop>
	 <20101112182959.GA20459@ms2.inr.ac.ru>
Mime-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: QUOTED-PRINTABLE
Cc: Patrick McHardy <kaber@trash.net>,
	David Lamparter <equinox@diac24.net>,
	Eric Paris <eparis@redhat.com>, Hua Zhong <hzhong@gmail.com>,
	netdev@vger.kernel.org, linux-kernel@vger.kernel.org,
	davem@davemloft.net, pekkas@netcore.fi, jmorris@namei.org,
	yoshfuji@linux-ipv6.org, paul.moore@hp.com,
	Damian Lukowski <damian@tvk.rwth-aachen.de>
To: Alexey Kuznetsov <kuznet@ms2.inr.ac.ru>
Return-path: <linux-kernel-owner@vger.kernel.org>
In-Reply-To: <20101112182959.GA20459@ms2.inr.ac.ru>
Sender: linux-kernel-owner@vger.kernel.org
List-Id: netdev.vger.kernel.org

Le vendredi 12 novembre 2010 =C3=A0 21:29 +0300, Alexey Kuznetsov a =C3=
=A9crit :
> Hello!
>=20
> On Fri, Nov 12, 2010 at 07:12:58PM +0100, Eric Dumazet wrote:
> > I see socket is locked around line 368,
> >=20
> >         bh_lock_sock(sk);
> >         /* If too many ICMPs get dropped on busy
> >          * servers this needs to be solved differently.
> >          */
> >         if (sock_owned_by_user(sk))
> >                 NET_INC_STATS_BH(net, LINUX_MIB_LOCKDROPPEDICMPS);
> >=20
> >=20
> > Hmm, maybe some goto is missing ;)
>=20
> It is not missing, sock_owned_by_user() is checked later when some op=
eration which
> cannot be done without lock is required. It was done to save error in=
 sk_err_soft even
> when socket is locked.
>=20
> This code also _understands_ this: look at
>=20
>                 } else if (sock_owned_by_user(sk)) {
>                         /* RTO revert clocked out retransmission,
>                          * but socket is locked. Will defer. */
>                         inet_csk_reset_xmit_timer(sk, ICSK_TIME_RETRA=
NS,
>                                                   HZ/20, TCP_RTO_MAX)=
;
>=20
> but somehow it considers the manipulations with rto/backoff/write_que=
ue as safe.
> Seems, they are not.

Indeed, right you are, I came to same conclusion.

I CC Damian Lukowski in my previous answer (and this one too)

Thanks !