From mboxrd@z Thu Jan 1 00:00:00 1970 From: Jon Maloy Subject: Re: skb_try_coalesce bug? Date: Tue, 22 Apr 2014 17:28:57 -0400 Message-ID: <5356DF19.8050709@ericsson.com> References: <20140422120125.GC7019@eerihug-hybrid.rnd.ki.sw.ericsson.se> <1398172293.29946.37.camel@edumazet-glaptop2.roam.corp.google.com> <5356C518.7050602@ericsson.com> <1398197116.29946.60.camel@edumazet-glaptop2.roam.corp.google.com> <5356D2AF.7030401@ericsson.com> Mime-Version: 1.0 Content-Type: text/plain; charset="UTF-8" Content-Transfer-Encoding: 7bit Cc: Erik Hugne , To: Eric Dumazet Return-path: Received: from usevmg21.ericsson.net ([198.24.6.65]:62389 "EHLO usevmg21.ericsson.net" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1757973AbaDVV3b (ORCPT ); Tue, 22 Apr 2014 17:29:31 -0400 In-Reply-To: <5356D2AF.7030401@ericsson.com> Sender: netdev-owner@vger.kernel.org List-ID: On 04/22/2014 04:35 PM, Jon Maloy wrote: > On 04/22/2014 04:05 PM, Eric Dumazet wrote: >> On Tue, 2014-04-22 at 15:38 -0400, Jon Maloy wrote: >> >>> >>> In the case I encountered, our head buffer is linear (skb->data_len == 0), >>> so it is the real tailroom value that is returned. An alas, that one is big >>> enough to contain the last (small) fragment of the message. >> >> >> Whole point of skb_try_coalesce() is to coalesce as much as possible, >> without guarantee of keeping some sort of 'segments' >> >> skb_try_coalesce - try to merge skb to prior one >> >> If you do not want this to happen, (you seem to want nothing else in >> your head buffer skb->head), you need to add some logic. > > Ok. I should have given a little background. > > 1: We send a message of 3041 bytes, inclusive TIPC header, via loopback interface. > > 2: This one gets chopped up in three fragments: 1420, 1420,and 201 bytes. > (The mtu was of course wrong, but this is how I discovered the problem). > > 3: First fragment is received, uncloned, and serves as head. > > 4; Second fragment (a clone) is received. skb_try_coalesce() fails at > the skb_head_is_locked() test, because the buffer is a clone. > Because of this, we add the buffer to skb_shinfo(head)->frag_list > instead. > > 5: Third fragment (also a clone) is received. Now, since we i.e., skb_try_coalesce(head, frag) check for > space in tailroom of header before we do anything else, it slips > in there, and bypasses the already chained-up second segment. More background: our reassembly code is based on the one found in ip_fragment.c::ip_frag_reasm(), which always first try to coalecse a buffer with head. That is a bad idea, I guess, but I wonder why they don't see this problem in ipv4. > > Regards > ///jon > > >> >> A helper temporarily setting head->tail = head->end would do it I guess. That would work. Or just check skb_has_frag_list(head) first, and make the call to skb_try_coalesce() conditional to the the result. It just feels a little unnecessary, since that test is done inside skb_try_coalesce() anyway. ///jon >> >> > > -- > To unsubscribe from this list: send the line "unsubscribe netdev" in > the body of a message to majordomo@vger.kernel.org > More majordomo info at http://vger.kernel.org/majordomo-info.html >