Tony Hoyle wrote: > -----BEGIN PGP SIGNED MESSAGE----- > Hash: SHA1 > > Ronciak, John wrote: >> Is this being seen on the bare metal (without KVM at all)? Is that >> what you meant by "testing on the maching itself"? What's type of >> system is this? This is a low performance networking device (i.e. >> x1 PCIe) but if it's working OK on IPv4 and ant IPv6 then it could >> be lots of things other then the NIC/driver. > > Yes.. It happens on the bare metal. ipv6 is consistently about 1/10th > of the transfer speed of ipv4, between two machines on the same > network, due to retransmissions. I'm not definitively saying it's > the driver, > only that switching checksum offloading off cures the problem > completely.. > > I can narrow it down to tx-checksum (enabling rx checksums doesn't > appear to impact performance). > > The system is an intel dual core 6...@2.4ghz, ICH9 chipset, 4G Memory, > 1TB of raid1. Can't find out much more detail without ripping the > case apart.. > > Tony
I ran a quick test with 2.6.32 kernel (it's net-2.6 pull from git) and was not able to reproduce the issue. My test was on host OS (no VMs). If it's possible can you try and reproduce on 2.6.32 kernel? Also can you provide the output of: lspci -vvv ethtool -e ethX ethtool ethX Thanks, Emil ------------------------------------------------------------------------------ This SF.Net email is sponsored by the Verizon Developer Community Take advantage of Verizon's best-in-class app development support A streamlined, 14 day to market process makes app distribution fast and easy Join now and get one step closer to millions of Verizon customers http://p.sf.net/sfu/verizon-dev2dev _______________________________________________ E1000-devel mailing list E1000-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/e1000-devel
