From mboxrd@z Thu Jan 1 00:00:00 1970 From: Patrick McHardy Subject: Re: [PATCH] netfilter: use per-cpu spinlock rather than RCU (v3) Date: Thu, 16 Apr 2009 14:31:48 +0200 Message-ID: <49E72534.1060302@trash.net> References: <20090411174801.GG6822@linux.vnet.ibm.com> <18913.53699.544083.320542@cargo.ozlabs.ibm.com> <20090412173108.GO6822@linux.vnet.ibm.com> <20090412.181330.23529546.davem@davemloft.net> <20090413040413.GQ6822@linux.vnet.ibm.com> <20090413095309.631cf395@nehalam> <49E48136.5060700@trash.net> <49E49C65.8060808@cosmosbay.com> <20090414074554.7fa73e2f@nehalam> <49E4B0A5.70404@cosmosbay.com> <20090414111716.28585806@nehalam> <49E4E3E8.5090201@cosmosbay.com> <20090414141351.0f63ac98@nehalam> <49E502B5.7070700@cosmosbay.com> <49E5BDF7.8090502@trash.net> <20090415135526.2afc4d18@nehalam> <49E64C91.5020708@cosmosbay.com> <49E720BE.2000200@trash.net> Mime-Version: 1.0 Content-Type: text/plain; charset=ISO-8859-15; format=flowed Content-Transfer-Encoding: 7bit Cc: Eric Dumazet , Stephen Hemminger , Jeff Chua , paulmck@linux.vnet.ibm.com, David Miller , paulus@samba.org, mingo@elte.hu, torvalds@linux-foundation.org, laijs@cn.fujitsu.com, r000n@r000n.net, linux-kernel@vger.kernel.org, netfilter-devel@vger.kernel.org, netdev@vger.kernel.org, benh@kernel.crashing.org To: Jan Engelhardt Return-path: Received: from stinky.trash.net ([213.144.137.162]:55048 "EHLO stinky.trash.net" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1754771AbZDPMbv (ORCPT ); Thu, 16 Apr 2009 08:31:51 -0400 In-Reply-To: Sender: netfilter-devel-owner@vger.kernel.org List-ID: Jan Engelhardt wrote: > On Thursday 2009-04-16 14:12, Patrick McHardy wrote: >> Jan Engelhardt wrote: >>>> So change per-cpu spinlocks to per-cpu rwlocks >>>> and use read_lock() in ipt_do_table() to allow recursion... >>>> >>> iptables cannot quite recurse into itself due to the comefrom stuff. >> Actually it can by using the REJECT target: > > Yes, but it has to return an absolute verdict (which REJECT does), > so it's not really a recursion, it's more like a goto without return. Its recursion in the sense that we reenter the same code path, while holding a lock. The verdict is issued *after* recursing. A (quite ugly) workaround would be to have ipt_REJECT queue the packets to a temporary queue and have ipt_do_table call dst_output() after dropping the lock.