From mboxrd@z Thu Jan 1 00:00:00 1970 From: Richard Braun Date: Thu, 23 Mar 2006 22:35:10 +0000 Subject: Checksum problem on sparc64 Message-Id: <20060323223510.GA7683@sceen.net> MIME-Version: 1 Content-Type: multipart/mixed; boundary="2fHTh5uZTiUOsy+g" List-Id: To: sparclinux@vger.kernel.org --2fHTh5uZTiUOsy+g Content-Type: text/plain; charset=us-ascii Content-Disposition: inline Content-Transfer-Encoding: quoted-printable Hello, I'm using kernel 2.6.15.6 on my Ultra1 (vanilla kernel without modification= ). My network setup is a bit weird since I'm using a PPPoE link with MTU 1400 = (I have packet losses if using a higher MTU because of the link quality). My N= AT router is a 2.6.15.6 kernel too running on i386. The problems I experience were already present when I used 2.6.12.6 kernel. Now here is the problem : I have a SSH server on my sparc64 box, and port 22 on the public interface is redirected to that host. When downloading with sftp for example, everyth= ing runs fine until some point. Using tcpdump to discover what was going on, I captured these packets (on public interface) : 13:14:30.359867 > 0800 1416: IP (tos 0x8, ttl 63, id 26858, offset 0, flags [DF], length: 1400) 213.41.243.68.22 > 84.96.34.158.59002: . [tcp sum ok] 3703221:3704569(1348) ack 9968 win 6788 13:14:30.397874 > 0800 1416: IP (tos 0x8, ttl 63, id 26859, offset 0, flags [DF], length: 1400) 10.0.0.2.22 > 84.96.34.158.59002: . [bad tcp cksum 696d (->696c)!] 3016044:3017392(1348) ack 7729 win 6788 Packets were similar on the LAN interface with the exception of NAT for the first packet : 13:14:30.359816 08:00:20:86:c9:98 > 00:90:27:71:ed:ae, ethertype IPv4 (0x08= 00), length 1414: IP (tos 0x8, ttl 64, id 26858, offset 0, flags [DF], length: 1400) 10.0.0.2.22 > 84.96.34.158.59002: . [tcp sum ok] 3703221:3704569(1348) ack 9872 win 6788 13:14:30.397818 08:00:20:86:c9:98 > 00:90:27:71:ed:ae, ethertype IPv4 (0x08= 00), length 1414: IP (tos 0x8, ttl 64, id 26859, offset 0, flags [DF], length: 1400) 10.0.0.2.22 > 84.96.34.158.59002: . [bad tcp cksum 696d (->696c)!] 3643729:3645077(1348) ack 9872 win 6788 I'm not sure why the tcp checksum is bad. It could be a hardware problem or a kernel bug. Therefore I decided to report this. I think I have similar problems with IPv6, but I have not investigated yet. The problem looks quite random for now, but I can easily reproduce it (I begin a download over sftp, wait a few seconds/minutes, then the rate drops). Hope it helps (and sorry for previous private queries btw). --=20 Richard Braun --2fHTh5uZTiUOsy+g Content-Type: application/pgp-signature; name="signature.asc" Content-Description: Digital signature Content-Disposition: inline -----BEGIN PGP SIGNATURE----- Version: GnuPG v1.4.1 (GNU/Linux) iD8DBQFEIyKeBlWsEPLYRi8RAizpAJ9Hfu9+s1lPQ14vd2Bpuoj7srivHwCff4Gr KLmevCIn2ioSp2N/P+OV72E= =pdyE -----END PGP SIGNATURE----- --2fHTh5uZTiUOsy+g--