> I tried every possible network setup, routed, bridged-new, bridged-old, > all to no avail. I ended up with the old style bridged setup because it > was the last one I tried.> Everything else works, I evenhave a DHCP server in one of the guest > domains, which serves machines on a different subnet from the rest of > the machines, and everything works fine (as it was before), but I > cannot, for instance, ssh from Domain_0 to a guest domain.> I can ssh to anyother machine and ssh back to the guest domain from > there, and it works fine, only the direct connection fails, it hangs > never completes. I even used ethereal to try to figure what was going > on, but couldn''t find anything, the connection just hangs there waiting > for a packet that never arrives.Sounds like the problem I have been experiencing. If you examine your packet capture carefully on the failed connections, you will probably see some transport layer checksum errors (tcp/udp.) If you want a routed setup, a workaround is to build gre tunnels over the top of the Dom0 vifX.0 <-> guest eth0 links. The new style bridged setup works, with the apparent limitation that Dom0 can now only be attached to one bridge group. If I remember correctly, vif0.0 should be bridged with vifX.0. veth0 should be assigned an appropriate address for your network. Dom0 veth0 is the equivalent of eth0 in a guest, vif0.0 being the matching end of the virtual link. Outside network access should work if you bridge eth0 in the same bridge group. Have fun if you want to run a firewall... Tim:> _______________________________________________ Xen-users mailing list Xen-users@lists.xensource.com http://lists.xensource.com/xen-users
On Wed, 2005-09-07 at 10:07 -0400, Tim Durack wrote:> > I tried every possible network setup, routed, bridged-new, bridged-old, > > all to no avail. I ended up with the old style bridged setup because it > > was the last one I tried. > > > Everything else works, I evenhave a DHCP server in one of the guest > > > domains, which serves machines on a different subnet from the rest of > > the machines, and everything works fine (as it was before), but I > > cannot, for instance, ssh from Domain_0 to a guest domain. > > > > I can ssh to anyother machine and ssh back to the guest domain from > > there, and it works fine, only the direct connection fails, it hangs > > never completes. I even used ethereal to try to figure what was going > > > on, but couldn''t find anything, the connection just hangs there waiting > > for a packet that never arrives. > > Sounds like the problem I have been experiencing. > > If you examine your packet capture carefully on the failed connections, you will probably see some transport layer checksum errors (tcp/udp.) > > > If you want a routed setup, a workaround is to build gre tunnels over the top of the Dom0 vifX.0 <-> guest eth0 links. > > The new style bridged setup works, with the apparent limitation that Dom0 can now only be attached to one bridge group. > > > If I remember correctly, vif0.0 should be bridged with vifX.0. veth0 should be assigned an appropriate address for your network. > Dom0 veth0 is the equivalent of eth0 in a guest, vif0.0 being the matching end of the virtual link. > > > Outside network access should work if you bridge eth0 in the same bridge group. Have fun if you want to run a firewall... > > Tim:><snip> Hmmm . . . I''ll take a stab in the dark at this. I''ve seen those types of errors with checksum offloading. Someone on this list in reply to a problem I was having mentioned that unstable implements checksum offloading. What happens if you disable it? - John -- John A. Sullivan III Open Source Development Corporation +1 207-985-7880 jsullivan@opensourcedevel.com If you would like to participate in the development of an open source enterprise class network security management system, please visit http://iscs.sourceforge.net _______________________________________________ Xen-users mailing list Xen-users@lists.xensource.com http://lists.xensource.com/xen-users
Tried various combinations of disabling offloading using ethtool. Hasn''t worked for me so far. Tim:> On 9/7/05, John A. Sullivan III <jsullivan@opensourcedevel.com> wrote:> > On Wed, 2005-09-07 at 10:07 -0400, Tim Durack wrote: > > > I tried every possible network setup, routed, bridged-new, > bridged-old, > > > all to no avail. I ended up with the old style bridged setup because > it > > > was the last one I tried. > > > > > Everything else works, I evenhave a DHCP server in one of the guest > > > > > domains, which serves machines on a different subnet from the rest of > > > the machines, and everything works fine (as it was before), but I > > > cannot, for instance, ssh from Domain_0 to a guest domain. > > > > > > > I can ssh to anyother machine and ssh back to the guest domain from > > > there, and it works fine, only the direct connection fails, it hangs > > > never completes. I even used ethereal to try to figure what was going > > > > > on, but couldn''t find anything, the connection just hangs there > waiting > > > for a packet that never arrives. > > > > Sounds like the problem I have been experiencing. > > > > If you examine your packet capture carefully on the failed connections, > you will probably see some transport layer checksum errors (tcp/udp.) > > > > > > If you want a routed setup, a workaround is to build gre tunnels over > the top of the Dom0 vifX.0 <-> guest eth0 links. > > > > The new style bridged setup works, with the apparent limitation that > Dom0 can now only be attached to one bridge group. > > > > > > If I remember correctly, vif0.0 should be bridged with vifX.0. veth0 > should be assigned an appropriate address for your network. > > Dom0 veth0 is the equivalent of eth0 in a guest, vif0.0 being the > matching end of the virtual link. > > > > > > Outside network access should work if you bridge eth0 in the same bridge > group. Have fun if you want to run a firewall... > > > > Tim:> > <snip> > Hmmm . . . I''ll take a stab in the dark at this. I''ve seen those types > of errors with checksum offloading. Someone on this list in reply to a > problem I was having mentioned that unstable implements checksum > offloading. What happens if you disable it? - John > -- > John A. Sullivan III > Open Source Development Corporation > +1 207-985-7880 > jsullivan@opensourcedevel.com > > If you would like to participate in the development of an open source > enterprise class network security management system, please visit > http://iscs.sourceforge.net > >_______________________________________________ Xen-users mailing list Xen-users@lists.xensource.com http://lists.xensource.com/xen-users
Okay, I''ll eat my words (just not my hat yet): guest> ethtool -K eth0 tx off fixed the checksum problem (at least on my system with an e100 driver.) I''ll try it again on my tg3 based system. I will eat my hat if it works there... Apologies if I mislead anyone. Tim:> On 9/7/05, Tim Durack <tdurack@gmail.com> wrote:> > Tried various combinations of disabling offloading using ethtool. Hasn''t > worked for me so far. > > Tim:> > > On 9/7/05, John A. Sullivan III <jsullivan@opensourcedevel.com> wrote: > > > > On Wed, 2005-09-07 at 10:07 -0400, Tim Durack wrote: > > > > I tried every possible network setup, routed, bridged-new, > > bridged-old, > > > > all to no avail. I ended up with the old style bridged setup because > > it > > > > was the last one I tried. > > > > > > > Everything else works, I evenhave a DHCP server in one of the guest > > > > > > > domains, which serves machines on a different subnet from the rest > > of > > > > the machines, and everything works fine (as it was before), but I > > > > cannot, for instance, ssh from Domain_0 to a guest domain. > > > > > > > > > > I can ssh to anyother machine and ssh back to the guest domain from > > > > there, and it works fine, only the direct connection fails, it hangs > > > > never completes. I even used ethereal to try to figure what was > > going > > > > > > > on, but couldn''t find anything, the connection just hangs there > > waiting > > > > for a packet that never arrives. > > > > > > Sounds like the problem I have been experiencing. > > > > > > If you examine your packet capture carefully on the failed > > connections, you will probably see some transport layer checksum errors > > (tcp/udp.) > > > > > > > > > If you want a routed setup, a workaround is to build gre tunnels over > > the top of the Dom0 vifX.0 <-> guest eth0 links. > > > > > > The new style bridged setup works, with the apparent limitation that > > Dom0 can now only be attached to one bridge group. > > > > > > > > > If I remember correctly, vif0.0 should be bridged with vifX.0. veth0 > > should be assigned an appropriate address for your network. > > > Dom0 veth0 is the equivalent of eth0 in a guest, vif0.0 being the > > matching end of the virtual link. > > > > > > > > > Outside network access should work if you bridge eth0 in the same > > bridge group. Have fun if you want to run a firewall... > > > > > > Tim:> > > <snip> > > Hmmm . . . I''ll take a stab in the dark at this. I''ve seen those types > > of errors with checksum offloading. Someone on this list in reply to a > > problem I was having mentioned that unstable implements checksum > > offloading. What happens if you disable it? - John > > -- > > John A. Sullivan III > > Open Source Development Corporation > > +1 207-985-7880 > > jsullivan@opensourcedevel.com > > > > If you would like to participate in the development of an open source > > enterprise class network security management system, please visit > > http://iscs.sourceforge.net > > > > >_______________________________________________ Xen-users mailing list Xen-users@lists.xensource.com http://lists.xensource.com/xen-users
Same thing doesn''t work on a tg3 based system. Perhaps a driver bug? Might explain why Ramiro is having problems too. Tim:> On 9/7/05, Tim Durack <tdurack@gmail.com> wrote:> > Okay, I''ll eat my words (just not my hat yet): > > guest> ethtool -K eth0 tx off > > fixed the checksum problem (at least on my system with an e100 driver.) > > I''ll try it again on my tg3 based system. I will eat my hat if it works > there... > > Apologies if I mislead anyone. > > Tim:> > > > On 9/7/05, Tim Durack <tdurack@gmail.com> wrote: > > > > Tried various combinations of disabling offloading using ethtool. Hasn''t > > worked for me so far. > > > > Tim:> > > > > On 9/7/05, John A. Sullivan III < jsullivan@opensourcedevel.com> wrote: > > > > > > On Wed, 2005-09-07 at 10:07 -0400, Tim Durack wrote: > > > > > I tried every possible network setup, routed, bridged-new, > > > bridged-old, > > > > > all to no avail. I ended up with the old style bridged setup > > > because it > > > > > was the last one I tried. > > > > > > > > > Everything else works, I evenhave a DHCP server in one of the > > > guest > > > > > > > > > domains, which serves machines on a different subnet from the rest > > > of > > > > > the machines, and everything works fine (as it was before), but I > > > > > cannot, for instance, ssh from Domain_0 to a guest domain. > > > > > > > > > > > > > I can ssh to anyother machine and ssh back to the guest domain > > > from > > > > > there, and it works fine, only the direct connection fails, it > > > hangs > > > > > never completes. I even used ethereal to try to figure what was > > > going > > > > > > > > > on, but couldn''t find anything, the connection just hangs there > > > waiting > > > > > for a packet that never arrives. > > > > > > > > Sounds like the problem I have been experiencing. > > > > > > > > If you examine your packet capture carefully on the failed > > > connections, you will probably see some transport layer checksum errors > > > (tcp/udp.) > > > > > > > > > > > > If you want a routed setup, a workaround is to build gre tunnels > > > over the top of the Dom0 vifX.0 <-> guest eth0 links. > > > > > > > > The new style bridged setup works, with the apparent limitation that > > > Dom0 can now only be attached to one bridge group. > > > > > > > > > > > > If I remember correctly, vif0.0 should be bridged with vifX.0. veth0 > > > should be assigned an appropriate address for your network. > > > > Dom0 veth0 is the equivalent of eth0 in a guest, vif0.0 being the > > > matching end of the virtual link. > > > > > > > > > > > > Outside network access should work if you bridge eth0 in the same > > > bridge group. Have fun if you want to run a firewall... > > > > > > > > Tim:> > > > <snip> > > > Hmmm . . . I''ll take a stab in the dark at this. I''ve seen those types > > > of errors with checksum offloading. Someone on this list in reply to a > > > problem I was having mentioned that unstable implements checksum > > > offloading. What happens if you disable it? - John > > > -- > > > John A. Sullivan III > > > Open Source Development Corporation > > > +1 207-985-7880 > > > jsullivan@opensourcedevel.com > > > > > > If you would like to participate in the development of an open source > > > enterprise class network security management system, please visit > > > http://iscs.sourceforge.net > > > > > > > > >_______________________________________________ Xen-users mailing list Xen-users@lists.xensource.com http://lists.xensource.com/xen-users
Ramiro Brito Willmersdorf
2005-Sep-12 14:02 UTC
Re: [Xen-users] Xen on production enviroment
Hi Tim, Yes, I guess so. The card which I have problems with is, coincidentally or not, also tg3 based. I cannot switch off checksum offloading on it, I get: Operation not supported from ethtool. Em Sunday 11 September 2005 22:07, Tim Durack escreveu:> Same thing doesn''t work on a tg3 based system. Perhaps a driver bug? > > Might explain why Ramiro is having problems too. > > Tim:>-- Ramiro Brito Willmersdorf Dep. Engenharia Mecânica/UFPE ramiro@willmersdorf.net tel: +55 81 2126-8231e239 _______________________________________________ Xen-users mailing list Xen-users@lists.xensource.com http://lists.xensource.com/xen-users
Tim Durack wrote:> Same thing doesn''t work on a tg3 based system. Perhaps a driver bug? > > Might explain why Ramiro is having problems too.Didn''t think the tg3 supported disabling the checksum offload. thanks, Nivedita> Tim:> > > On 9/7/05, Tim Durack <tdurack@gmail.com> wrote: > >>Okay, I''ll eat my words (just not my hat yet): >> >>guest> ethtool -K eth0 tx off >> >>fixed the checksum problem (at least on my system with an e100 driver.) >> >>I''ll try it again on my tg3 based system. I will eat my hat if it works >>there... >> >>Apologies if I mislead anyone._______________________________________________ Xen-users mailing list Xen-users@lists.xensource.com http://lists.xensource.com/xen-users
tdurack@zeta:~$ sudo ethtool -k eth0 Offload parameters for eth0: rx-checksumming: on tx-checksumming: on scatter-gather: off tcp segmentation offload: off tdurack@zeta:~$ sudo ethtool -K eth0 tx off tdurack@zeta:~$ sudo ethtool -k eth0 Offload parameters for eth0: rx-checksumming: on tx-checksumming: off scatter-gather: off tcp segmentation offload: off Ethtool works, what it''s really doing is a different matter. Tim:> On 9/13/05, Nivedita Singhvi <niv@us.ibm.com> wrote:> > Tim Durack wrote: > > Same thing doesn''t work on a tg3 based system. Perhaps a driver bug? > > > > Might explain why Ramiro is having problems too. > > > Didn''t think the tg3 supported disabling the checksum offload. > > thanks, > Nivedita > > > Tim:> > > > > On 9/7/05, Tim Durack <tdurack@gmail.com> wrote: > > > >>Okay, I''ll eat my words (just not my hat yet): > >> > >>guest> ethtool -K eth0 tx off > >> > >>fixed the checksum problem (at least on my system with an e100 driver.) > >> > >>I''ll try it again on my tg3 based system. I will eat my hat if it works > >>there... > >> > >>Apologies if I mislead anyone. > >_______________________________________________ Xen-users mailing list Xen-users@lists.xensource.com http://lists.xensource.com/xen-users