From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: X-Spam-Checker-Version: SpamAssassin 3.4.0 (2014-02-07) on aws-us-west-2-korg-lkml-1.web.codeaurora.org X-Spam-Level: X-Spam-Status: No, score=-2.2 required=3.0 tests=HEADER_FROM_DIFFERENT_DOMAINS, MAILING_LIST_MULTI,SPF_HELO_NONE,SPF_PASS,USER_AGENT_SANE_1 autolearn=no autolearn_force=no version=3.4.0 Received: from mail.kernel.org (mail.kernel.org [198.145.29.99]) by smtp.lore.kernel.org (Postfix) with ESMTP id 120FBC83001 for ; Thu, 30 Apr 2020 06:34:09 +0000 (UTC) Received: from vger.kernel.org (vger.kernel.org [23.128.96.18]) by mail.kernel.org (Postfix) with ESMTP id ED17220784 for ; Thu, 30 Apr 2020 06:34:08 +0000 (UTC) Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1726404AbgD3GeI (ORCPT ); Thu, 30 Apr 2020 02:34:08 -0400 Received: from mail104.syd.optusnet.com.au ([211.29.132.246]:38284 "EHLO mail104.syd.optusnet.com.au" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1726378AbgD3GeI (ORCPT ); Thu, 30 Apr 2020 02:34:08 -0400 Received: from dimstar.local.net (n175-34-64-112.sun1.vic.optusnet.com.au [175.34.64.112]) by mail104.syd.optusnet.com.au (Postfix) with SMTP id B4F5B820C58 for ; Thu, 30 Apr 2020 16:34:05 +1000 (AEST) Received: (qmail 11726 invoked by uid 501); 30 Apr 2020 06:34:04 -0000 Date: Thu, 30 Apr 2020 16:34:04 +1000 From: Duncan Roe To: Netfilter Development Cc: Pablo Neira Ayuso Subject: Re: [PATCH libnetfilter_queue 0/3] pktbuff API updates Message-ID: <20200430063404.GF3833@dimstar.local.net> Mail-Followup-To: Netfilter Development , Pablo Neira Ayuso References: <20200427170656.GA22296@salvia> <20200428043302.GB15436@dimstar.local.net> <20200428103407.GA1160@salvia> <20200428211452.GF15436@dimstar.local.net> <20200428225520.GA30421@salvia> <20200429132840.GA3833@dimstar.local.net> <20200429191047.GB3833@dimstar.local.net> <20200429191643.GA16749@salvia> <20200429203029.GD3833@dimstar.local.net> <20200429210512.GA14508@salvia> MIME-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: <20200429210512.GA14508@salvia> User-Agent: Mutt/1.10.1 (2018-07-13) X-Optus-CM-Score: 0 X-Optus-CM-Analysis: v=2.3 cv=W5xGqiek c=1 sm=1 tr=0 a=keeXcwCgVCrAuxOn72dlvA==:117 a=keeXcwCgVCrAuxOn72dlvA==:17 a=kj9zAlcOel0A:10 a=cl8xLZFz6L8A:10 a=RSmzAf-M6YYA:10 a=OLL_FvSJAAAA:8 a=voM4FWlXAAAA:8 a=3HDBlxybAAAA:8 a=_vdq4B85Ww93OoYmRX4A:9 a=32EQcqo6bRQ160ow:21 a=a9a8tTNdu4p-3ijc:21 a=CjuIK1q_8ugA:10 a=SAszQ4RR6nkA:10 a=Z7RzNMET8NMA:10 a=oIrB72frpwYPwTMnlWqB:22 a=IC2XNlieTeVoXbcui8wp:22 a=laEoCiVfU_Unz3mSdgXN:22 Sender: netfilter-devel-owner@vger.kernel.org Precedence: bulk List-ID: X-Mailing-List: netfilter-devel@vger.kernel.org On Wed, Apr 29, 2020 at 11:05:12PM +0200, Pablo Neira Ayuso wrote: > On Thu, Apr 30, 2020 at 06:30:29AM +1000, Duncan Roe wrote: > > Hi Pablo, > > > > I sent this email (explanation of how the system works) before I saw your email > > asking for that explanation. > > > > Then I replied to that email of yours before I saw this one. > > > > On Wed, Apr 29, 2020 at 09:16:43PM +0200, Pablo Neira Ayuso wrote: > > > On Thu, Apr 30, 2020 at 05:10:47AM +1000, Duncan Roe wrote: > > > [...] > > > > Sorry, I should have explained a bit more how the system would work: > > > > > > > > struct pkt_buff has 3 new members: > > > > > > > > bool copy_done; > > > > uint32_t extra; > > > > uint8_t *copy_buf; > > > > > > > > When extra > 0, pktb_alloc2 verifies that buflen is >= len + extra. It then > > > > stores extra and copy_buf in pktb, ready for use by pktb_mangle() (all the other > > > > manglers call this eventually). > > > > > > > > So that's how pktb_mangle() doesn't need to allocate a buffer. > > > > > > Thanks for the explaining. Given this is in userspace, it is easier if > > > > Tiny nit - this could be userspace on an embedded device where memory is really > > tight. So perhaps document with "If memory is at a premium, you really only need > > len + extra" otherwise a big buf is fine. > > This buffer is still relatively small, the reallocation also forces > the user to refetch pointers. Skipping all that complexity for a bit > of memory in userspace is fine. Oh well in that case, how about: > struct pkt_buff *pktb_alloc2(int family, void *buf, size_t buf_size, void *data, size_t len, size_t extra); I.e. exactly as you suggested in https://www.spinics.net/lists/netfilter-devel/msg65830.html except s/head/buf/ And we tell users to dimension buf to NFQ_BUFFER_SIZE. We don't even need to expose pktb_head_size(). > > > > the user allocates the maximum packet length that is possible: > > > > > > 0xffff + (MNL_SOCKET_BUFFER_SIZE/2); > > > > > > We can probably expose this to the header so they can pre-allocate a > > > buffer that is large enough and, hence, _mangle() is guaranteed to > > > have always enough room to add extra bytes. > > > > Yes I saw that expression in examples/nf-queue.c. How about > > > > #define COPY_BUF_SIZE (0xffff + (MNL_SOCKET_BUFFER_SIZE/2)) > > > > or what other name would you like? > > I'd suggest to add the NFQ_ prefix, probably NFQ_BUFFER_SIZE is fine. > > > --- Off-topic > > > > I'm intrigued that you ccan use MNL_SOCKET_BUFFER_SIZE when dimensioning static > > variables. The expansion is > > > > #define MNL_SOCKET_BUFFER_SIZE (sysconf(_SC_PAGESIZE) < 8192L ? > > sysconf(_SC_PAGESIZE) : 8192L) > > > > and sysconf is an actual function: > > > > unistd.h:622:extern long int sysconf (int __name) __THROW; > > > > If I try to dimension a static variable using pktb_head_size(), the compiler > > throws an error. Why is sysconf() ok but not pktb_head_size()? > > I'm fine to expose if you would like to add pktb_head_size() if you > prefer to store the pkt_buff in the stack. You will have to use > pktb_build_data() [1] to initialize the pkt_buff. Would that work for you? > > [1] https://patchwork.ozlabs.org/project/netfilter-devel/patch/20200426132356.8346-2-pablo@netfilter.org/