Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S932151Ab2BAOrt (ORCPT ); Wed, 1 Feb 2012 09:47:49 -0500 Received: from seven.medozas.de ([188.40.89.202]:35109 "EHLO seven.medozas.de" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1752978Ab2BAOrs (ORCPT ); Wed, 1 Feb 2012 09:47:48 -0500 Date: Wed, 1 Feb 2012 15:47:44 +0100 (CET) From: Jan Engelhardt To: richard -rw- weinberger cc: Netfilter Developer Mailing List , Linux Networking Developer Mailing List , Linux Kernel Mailing List , eric.dumazet@gmail.com, Pablo Neira Ayuso Subject: Re: [PATCH 1/6] Netfilter: Merge ipt_LOG and ip6_LOG into xt_LOG In-Reply-To: Message-ID: References: <1327269381-20748-1-git-send-email-richard@nod.at> <1327269381-20748-2-git-send-email-richard@nod.at> User-Agent: Alpine 2.01 (LNX 1266 2009-07-14) MIME-Version: 1.0 Content-Type: TEXT/PLAIN; charset=US-ASCII Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org Content-Length: 2568 Lines: 72 On Wednesday 2012-02-01 10:14, richard -rw- weinberger wrote: >> >> Do you have the patchset as a retrievable git tree, btw? > >Only for kernel stuff: >git://git.kernel.org/pub/scm/linux/kernel/git/rw/misc.git xt_LOG > >If you want it for the user space part too, I can setup something... That would be welcome. >>>[PATCH 6/6] iptables: xt_LOG: Add ring buffer support >>> >>>in libxt_LOG.c: >>>+.size = XT_ALIGN(sizeof(struct xt_log_info_v1) + sizeof(void *)), >> >> Why the extra void*? > >The kernel space part of xt_log_info_v1 differs from the user space struct, >it contains an extra pointer to the ring buffer context. >Iptables must not see this extra pointer because the kernel writes to it. I don't see that extra pointer (outside xt_log_info_v1) that you talk about. +struct xt_log_info_v1 { + unsigned char level; + unsigned char logflags; + char prefix[30]; + + char ring_name[30]; + unsigned long ring_size; + struct xt_LOG_ring_ctx *rctx; +}; This is the struct as you defined it. First, you must not use unsigned long (due to its flexible size, which is an ABI breaker; cf. "Writing Netfilter Modules" PDF), and the pointer must be tagged for 8-alignment as well for the same reason. Second, the kernel will reject a ruleset if (kernel) .targetsize != (user) .size. This is likely to occur, since you essentially have .targetsize=72 in xt_LOG.c and .size=72+sizeof(void*) in libxt_LOG. Check commit 8a252e8941cd50e4390e6fedf280ba1c7ff9ac43 for trailing whitespace, git log -p shows them. + if (par->family == NFPROTO_IPV4) + m = ipt_log_packet(NFPROTO_IPV4, par->hooknum, skb, par->in, + par->out, &li, loginfo->prefix); +#if IS_ENABLED(CONFIG_IPV6) + else if (par->family == NFPROTO_IPV6) + m = ip6t_log_packet(NFPROTO_IPV6, par->hooknum, skb, par->in, + par->out, &li, loginfo->prefix); +#endif + else + BUG(); Please avoid BUG. Either do nothing in the 'else' case, or print a courtesy message with some unkn_log_packet() or WARN_ON_ONCE(). The same goes for xt_LOG_ring_exit, which I vote for being reduced to WARN_ON as well. If the condition triggers, one will have a little memory leak, but at least the machine still runs. -- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/