Date: Thu, 23 May 2013 11:01:04 -0700 From: Colin Percival <cperciva@freebsd.org> To: Jay West <jwest@ezwind.net> Cc: freebsd-xen@freebsd.org, Andre Oppermann <andre@freebsd.org> Subject: Re: tcp offloading Message-ID: <519E5960.40707@freebsd.org> In-Reply-To: <001901ce57be$192d3940$4b87abc0$@ezwind.net> References: <001901ce57be$192d3940$4b87abc0$@ezwind.net>
next in thread | previous in thread | raw e-mail | index | archive | help
On 05/23/13 07:02, Jay West wrote: > Environment is Citrix Xenserver Advanced, with (mostly) freebsd 9.1 64bit > HVM with PV drivers. > > I've heard much mention around the net that in pretty much all cases, TCP > offloading should be disabled inside the VM. I have also seen other posts > that TCP offloading should be disabled on the Dom0. > > So my question is; for best performance should TCP offloading be disabled in > the VM, or the Xenserver (dom0), or should it be disabled in BOTH? There's an incompatibility between FreeBSD's TCP stack, TSO, and the Linux Xen netback driver: With TSO enabled on netfront, we can generate longer mbuf chains than the Linux netback is able to handle. I have an ugly patch for this on EC2, but andre will be fixing this better in the near future. For now I recommend # ifconfig xn0 -tso unless you need multi-Gbps networking performance. -- Colin Percival Security Officer Emeritus, FreeBSD | The power to serve Founder, Tarsnap | www.tarsnap.com | Online backups for the truly paranoid
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?519E5960.40707>