From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1756445AbcANRZn (ORCPT ); Thu, 14 Jan 2016 12:25:43 -0500 Received: from Chamillionaire.breakpoint.cc ([80.244.247.6]:36021 "EHLO Chamillionaire.breakpoint.cc" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1756402AbcANRZj (ORCPT ); Thu, 14 Jan 2016 12:25:39 -0500 Date: Thu, 14 Jan 2016 18:25:27 +0100 From: Florian Westphal To: Sasha Levin Cc: pablo@netfilter.org, kaber@trash.net, kadlec@blackhole.kfki.hu, davem@davemloft.net, netfilter-devel@vger.kernel.org, coreteam@netfilter.org, netdev@vger.kernel.org, linux-kernel@vger.kernel.org Subject: Re: [PATCH v2] netfilter: nf_conntrack: use safer way to lock all buckets Message-ID: <20160114172527.GA19589@breakpoint.cc> References: <1452791036-2161-1-git-send-email-sasha.levin@oracle.com> MIME-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: <1452791036-2161-1-git-send-email-sasha.levin@oracle.com> User-Agent: Mutt/1.5.21 (2010-09-15) Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org Sasha Levin wrote: > When we need to lock all buckets in the connection hashtable we'd attempt to > lock 1024 spinlocks, which is way more preemption levels than supported by > the kernel. Furthermore, this behavior was hidden by checking if lockdep is > enabled, and if it was - use only 8 buckets(!). > > Fix this by using a global lock and synchronize all buckets on it when we > need to lock them all. This is pretty heavyweight, but is only done when we > need to resize the hashtable, and that doesn't happen often enough (or at all). > > diff --git a/net/netfilter/nf_conntrack_core.c b/net/netfilter/nf_conntrack_core.c > index 3cb3cb8..4ccf5ad 100644 > --- a/net/netfilter/nf_conntrack_core.c > +++ b/net/netfilter/nf_conntrack_core.c > @@ -66,6 +66,21 @@ EXPORT_SYMBOL_GPL(nf_conntrack_locks); > __cacheline_aligned_in_smp DEFINE_SPINLOCK(nf_conntrack_expect_lock); > EXPORT_SYMBOL_GPL(nf_conntrack_expect_lock); > > +spinlock_t nf_conntrack_locks_all_lock; > +bool nf_conntrack_locks_all; warning: symbol 'nf_conntrack_locks_all_lock' was not declared. Should it be static? warning: symbol 'nf_conntrack_locks_all' was not declared. Should it be static? > +void nf_conntrack_lock(spinlock_t *lock) __acquires(lock) It avoids 'context imbalance in 'nf_conntrack_lock' - wrong count at exit' sparse error. Other than that this looks good, thanks.