From mboxrd@z Thu Jan 1 00:00:00 1970 From: "David S. Miller" Subject: Re: [IPX]: Fix checksum computation. Date: Fri, 31 Oct 2003 13:53:28 -0800 Sender: netdev-bounce@oss.sgi.com Message-ID: <20031031135328.2a997f6a.davem@redhat.com> References: <200310312006.h9VK62Hh005910@hera.kernel.org> <1067635446.11564.92.camel@localhost.localdomain> <20031031132331.35a9aaca.davem@redhat.com> <1067637004.11564.98.camel@localhost.localdomain> Mime-Version: 1.0 Content-Type: text/plain; charset=US-ASCII Content-Transfer-Encoding: 7bit Cc: netdev@oss.sgi.com Return-path: To: Joe Perches In-Reply-To: <1067637004.11564.98.camel@localhost.localdomain> Errors-to: netdev-bounce@oss.sgi.com List-Id: netdev.vger.kernel.org On Fri, 31 Oct 2003 13:50:04 -0800 Joe Perches wrote: > Was an old NG Sniffer being used to verify this? > Sniffer had a long term problem with IPX checksums. No, Arnaldo would verify the checksum by running the old code and the new code, they produced different checksums on every sendmsg() call. He then tested it further by making sure he could use netatalk successfully between a 2.4.x Linux appletalk box and a 2.6.x system with the checksum patch applied. Without the patch the 2.4.x system would reject all packets sent by the 2.6.x box. Don't assume that we're a bunch of fucknuts and didn't verify things to the best of our abilities ok? Thanks. > Has the gcc team been contacted? Why would we contact them before we even know if it's a gcc bug or not? It could be a sign extension issue or something else that our brains are not grokking at the moment. Contacting the gcc team would be utterly premature. Here is something you could do for us instead of your current blabbering. Why don't you take a look at the assembler diff I posted and try to figure out how the before code produces a different checksum than the after code?