From owner-freebsd-net@FreeBSD.ORG Wed Aug 21 20:59:25 2013 Return-Path: Delivered-To: freebsd-net@freebsd.org Received: from mx1.freebsd.org (mx1.freebsd.org [IPv6:2001:1900:2254:206a::19:1]) (using TLSv1 with cipher ADH-AES256-SHA (256/256 bits)) (No client certificate requested) by hub.freebsd.org (Postfix) with ESMTP id 0F46CA15 for ; Wed, 21 Aug 2013 20:59:25 +0000 (UTC) (envelope-from andre@freebsd.org) Received: from c00l3r.networx.ch (c00l3r.networx.ch [62.48.2.2]) (using TLSv1 with cipher DHE-RSA-AES256-SHA (256/256 bits)) (No client certificate requested) by mx1.freebsd.org (Postfix) with ESMTPS id 3063E2A97 for ; Wed, 21 Aug 2013 20:59:23 +0000 (UTC) Received: (qmail 74377 invoked from network); 21 Aug 2013 21:42:25 -0000 Received: from c00l3r.networx.ch (HELO [127.0.0.1]) ([62.48.2.2]) (envelope-sender ) by c00l3r.networx.ch (qmail-ldap-1.03) with SMTP for ; 21 Aug 2013 21:42:25 -0000 Message-ID: <52152A24.8030109@freebsd.org> Date: Wed, 21 Aug 2013 22:59:16 +0200 From: Andre Oppermann User-Agent: Mozilla/5.0 (Windows NT 6.1; WOW64; rv:17.0) Gecko/20130801 Thunderbird/17.0.8 MIME-Version: 1.0 To: Navdeep Parhar Subject: Re: M_NOFREE removal (was Re: svn commit: r254520 - in head/sys: kern sys) References: <201308191116.r7JBGsc6065793@svn.freebsd.org> <521256CE.6070706@FreeBSD.org> <5212587A.2080202@freebsd.org> <52128937.1010407@freebsd.org> <52129E55.30803@freebsd.org> <5214D7E8.1080106@freebsd.org> <5214ED19.40907@FreeBSD.org> <5215047C.8080107@freebsd.org> <521505A3.4020103@FreeBSD.org> <52151368.40803@freebsd.org> <52151799.3020809@FreeBSD.org> <52152697.9070500@freebsd.org> <5215288E.6070708@FreeBSD.org> In-Reply-To: <5215288E.6070708@FreeBSD.org> Content-Type: text/plain; charset=UTF-8; format=flowed Content-Transfer-Encoding: 7bit Cc: svn-src-head@freebsd.org, freebsd-net@freebsd.org, Peter Grehan X-BeenThere: freebsd-net@freebsd.org X-Mailman-Version: 2.1.14 Precedence: list List-Id: Networking and TCP/IP with FreeBSD List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , X-List-Received-Date: Wed, 21 Aug 2013 20:59:25 -0000 On 21.08.2013 22:52, Navdeep Parhar wrote: > On 08/21/13 13:44, Andre Oppermann wrote: >> On 21.08.2013 21:40, Navdeep Parhar wrote: >>> On 08/21/13 12:22, Andre Oppermann wrote: >>>> On 21.08.2013 20:23, Navdeep Parhar wrote: >>>>> I believe we need an extra patch to get M_NOFREE correct. I've had it >>>>> forever in some of my internal repos but never committed it upstream >>>>> (just plain forgot). Since this stuff is fresh in your mind, can you >>>>> review this: >>>>> >>>>> diff -r cd78031b7885 sys/sys/mbuf.h >>>>> --- a/sys/sys/mbuf.h Fri Aug 16 13:35:26 2013 -0700 >>>>> +++ b/sys/sys/mbuf.h Wed Aug 21 10:55:57 2013 -0700 >>>>> @@ -535,6 +535,8 @@ m_free(struct mbuf *m) >>>>> { >>>>> struct mbuf *n = m->m_next; >>>>> >>>>> + if ((m->m_flags & (M_PKTHDR|M_NOFREE)) == (M_PKTHDR|M_NOFREE)) >>>>> + m_tag_delete_chain(m, NULL); >>>>> if (m->m_flags & M_EXT) >>>>> mb_free_ext(m); >>>>> else if ((m->m_flags & M_NOFREE) == 0) >>>>> >>>>> It prevents leaks of tags from M_NOFREE+pkthdr mbufs. >>>> >>>> Good point. Looks correct. >>>> >>>> But then I wonder if it is really a smart thing to allow single >>>> mbufs (w/o M_EXT) to be M_NOFREE at the same time. They easily >>>> get lost. If it doesn't have an external buffer attached there >>>> is no way to know when and if it was freed. >>>> >>>> If M_NOFREE is only allowed together with M_EXT then the tag chain >>>> delete should happen in mb_ext_free() next to 'skipmbuf' instead. >>>> >>>> Index: kern/uipc_mbuf.c >>>> =================================================================== >>>> --- kern/uipc_mbuf.c (revision 254605) >>>> +++ kern/uipc_mbuf.c (working copy) >>>> @@ -283,11 +283,6 @@ >>>> KASSERT((m->m_flags & M_EXT) == M_EXT, ("%s: M_EXT not set", >>>> __func__)); >>>> KASSERT(m->m_ext.ref_cnt != NULL, ("%s: ref_cnt not set", >>>> __func__)); >>>> >>>> - /* >>>> - * check if the header is embedded in the cluster >>>> - */ >>>> - skipmbuf = (m->m_flags & M_NOFREE); >>>> - >>>> /* Free attached storage if this mbuf is the only reference to >>>> it. */ >>>> if (*(m->m_ext.ref_cnt) == 1 || >>>> atomic_fetchadd_int(m->m_ext.ref_cnt, -1) == 1) { >>>> @@ -328,8 +323,14 @@ >>>> ("%s: unknown ext_type", __func__)); >>>> } >>>> } >>>> - if (skipmbuf) >>>> + >>>> + /* >>>> + * Do not free if the mbuf is embedded in the cluster. >>>> + */ >>>> + if (m->m_flags & M_NOFREE) { >>>> + m_tag_delete_chain(m, NULL); >>>> return; >>>> + } >>> >>> The problem with this is that the mbuf may already be gone if it was >>> embedded in the cluster that was just freed by the ext free. That's why >>> skipmbuf is used to cache the M_NOFREE bit. >> >> Next try: ;) > > It is most flexible to let M_NOFREE work without any assumptions > attached (must be M_EXT, etc.) So I still prefer my patch to this. If > you don't have any strong preferences may I commit that one instead? I don't mind having your patch. Though I don't see how it possibly can't leak mbufs if they are not M_EXT. -- Andre