From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: from sc8-sf-mx2-b.sourceforge.net ([10.3.1.12] helo=sc8-sf-mx2.sourceforge.net) by sc8-sf-list1.sourceforge.net with esmtp (Exim 4.30) id 1CHAV2-0003lS-QO for user-mode-linux-devel@lists.sourceforge.net; Mon, 11 Oct 2004 17:26:44 -0700 Received: from smtp005.mail.ukl.yahoo.com ([217.12.11.36]) by sc8-sf-mx2.sourceforge.net with smtp (Exim 4.41) id 1CHAV1-0000PL-Jd for user-mode-linux-devel@lists.sourceforge.net; Mon, 11 Oct 2004 17:26:44 -0700 From: BlaisorBlade Subject: Re: [uml-devel] SOLVED: uml "ip header error" for large (fragmented) udp packets References: In-Reply-To: MIME-Version: 1.0 Content-Disposition: inline Content-Type: text/plain; charset="iso-8859-1" Content-Transfer-Encoding: 7bit Message-Id: <200410120227.00849.blaisorblade_spam@yahoo.it> Sender: user-mode-linux-devel-admin@lists.sourceforge.net Errors-To: user-mode-linux-devel-admin@lists.sourceforge.net List-Unsubscribe: , List-Id: The user-mode Linux development list List-Post: List-Help: List-Subscribe: , List-Archive: Date: Tue, 12 Oct 2004 02:27:00 +0200 To: user-mode-linux-devel@lists.sourceforge.net Cc: Lars Ellenberg On Tuesday 12 October 2004 02:03, Lars Ellenberg wrote: > / 2004-10-11 19:55:12 +0200 > > \ Lars Ellenberg: > > since UML => Host does not work, but > > Host => UML does, this suggests that > > the bug is somewhere on the sending side. > > > > I was not able to track it down. > > just so you know, finally: > ======================= > --- linux-2.6.6/arch/um/include/sysdep-i386/checksum.h.orig 2004-10-12 > 01:50:49.000000000 +0200 +++ > linux-2.6.6/arch/um/include/sysdep-i386/checksum.h 2004-10-12 > 01:50:58.000000000 +0200 @@ -102,8 +102,7 @@ > are modified, we must also specify them as outputs, or gcc > will assume they contain their original values. */ > > : "=r" (sum), "=r" (iph), "=r" (ihl) > > - : "1" (iph), "2" (ihl) > - : "memory"); > + : "1" (iph), "2" (ihl)); > return(sum); > } The patch you posted REMOVES a memory barrier - you reversed it. I actually checked the barrier is missing in the source code; but the strange thing is that you modified checksum.h.orig and not checksum.h! Are you sure that you compiled the corrected header? However, the patch is correct, and I assume you posted it the right way. I'm going to send it to Andrew Morton for 2.6.9 (hoping he wants to accept all these patches - they are rushing for 2.6.9). > that's all, folks. only a missing memory barrier. Thanks a lot, folks! Yes, the patch is right. > WTF :-/ > > same patch applies to 2.6.8.1 - uml and probably all other umls. > since that is a one-to-one copy anyways, maybe UML should better use the > original (in include/asm-i386/checksum.h) right away?? I could do this later - too much checks for the other header code for a one-minute patch. > there may be similar bugs hiding in various areas of uml... I think there are even worse ones. > btw, > anyone wants to give me a hint how to tune it best? > maybe how to up the mtu of the UML "nics"? Impossible with the current code - it emulates an ethernet card, so I don't think you can do anything for this. However I hope it does not cause too much problems. Search for the network howtos: someone explained how to set the interface packet scheduler (maybe even Jeff Dike on the main UML site). Bye -- Paolo Giarrusso, aka Blaisorblade Linux registered user n. 292729 ------------------------------------------------------- This SF.net email is sponsored by: IT Product Guide on ITManagersJournal Use IT products in your business? Tell us what you think of them. Give us Your Opinions, Get Free ThinkGeek Gift Certificates! Click to find out more http://productguide.itmanagersjournal.com/guidepromo.tmpl _______________________________________________ User-mode-linux-devel mailing list User-mode-linux-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/user-mode-linux-devel