From mboxrd@z Thu Jan 1 00:00:00 1970 From: Patrick McHardy Subject: Re: [PATCH] nf_conntrack_core.c: fix for dead connection after flushing conntrack cache Date: Fri, 14 May 2010 13:05:55 +0200 Message-ID: <4BED2E93.1030004@trash.net> References: <4BE3D31F.6000607@secunet.com> <4BE73DE3.6080304@netfilter.org> <4BE824EA.6000303@trash.net> <4BE83D93.8080909@secunet.com> Mime-Version: 1.0 Content-Transfer-Encoding: 7bit Return-path: In-Reply-To: <4BE83D93.8080909@secunet.com> Sender: netfilter-devel-owner@vger.kernel.org List-ID: Content-Type: text/plain; charset="us-ascii" To: Joerg Marx Cc: Pablo Neira Ayuso , Netfilter Development Mailinglist , Mail List - Netfilter Joerg Marx wrote: > On 05/10/2010 05:23 PM, Patrick McHardy wrote: > > ... >> I think this should be fine since the race you describe only affects >> unconfirmed conntracks, but it took me a while to realize that all >> the other spots where the DYING bit is set are fine without holding >> the conntrack lock. >> >> Could you please add a comment to the check in __nf_conntrack_confirm() >> stating that the dying check is supposed to prevent races against >> nf_ct_get_next_corpse()? The semantic of the DYING bit is unfortunately >> a bit overloaded. >> >> Also, since the condition unconfirmed + dying in nf_conntrack_confirm() >> is highly unlikely, I'd suggest to remove the dying check there and only >> perform it in __nf_conntrack_confirm(). >> > > I hope the comment is clearly pointing to the (solved) problem now. I > also removed the obsolete check in nf_conntrack_confirm. Thanks, this looks fine. But I need a formal submission, including a changelog and Signed-off-by: line. Thanks!