I have a problem with networking under XEN, I hope someone can help me
out. The problem is that after some time (a day or so) the network
suddenly stops working, and it takes some waiting and/or a reboot to fix
I'm running XEN 3.1.0-rc7 on an Intel Core2 (x86_64). It's a Fedora 8
system, with the following packages:
o Xen version 3.1.0-rc7-2950.fc8 (kojibuilder@(none)) (gcc version 4.1.2
20070925 (Red Hat 4.1.2-32)) Tue Oct 23 12:21:08 EDT 2007
The error message the kernel gives me is the following:
Nov 16 14:38:31 xen2 kernel: NETDEV WATCHDOG: peth0: transmit timed out
Nov 16 14:38:31 xen2 kernel: sky2 peth0: tx timeout
Nov 16 14:38:31 xen2 kernel: sky2 peth0: disabling interface
Nov 16 14:38:31 xen2 kernel: sky2 peth0: enabling interface
Nov 16 14:38:31 xen2 kernel: sky2 peth0: ram buffer 48K
Nov 16 14:38:31 xen2 kernel: eth0: port 1(peth0) entering disabled state
When I googled around, searching for these error messages, I found three
previous e-mails to this xen-users list, reporting exactly the same
problem. Unfortunately none of these messages seem to have received any
reply? At least I wasn't able to find any.
Below I pasted the previous mails from Markus Goldstein (december 2006),
Brock Palen (january 2007) and Ian Tobin (august 2007) for more
information on this subject.
I hope someone has any idea what could be done to solve this problem, or
if this issue is already taken care of by someone, or any other insight
that might help.
On december 22, 2006, Markus Goldstein wrote: "Debian Etch and nvidia
chipset trouble (Kernel bug)"
> Hi all,
> I have a problem running Xen 3.0.3-1 on Debian Etch (amd64).
> Packages installed:
> linux-image-2.6.18-3-xen-amd64 2.6.18-7
> linux-modules-2.6.18-3-xen-amd64 2.6.18-7
> xen-hypervisor-3.0.3-1-amd64 3.0.3-0-2
> xen-ioemu-3.0.3-1 3.0.3-0-2
> xen-linux-system-2.6.18-3-xen-amd64 2.6.18-7
> I have a nvidia chipset and I am using the onboard gigabit ethernet
> From time to time, the networking hangs and gives the output
> Dec 21 19:09:40 xen kernel: NETDEV WATCHDOG: peth0: transmit timed out
> Dec 21 19:09:40 xen kernel: peth0: Got tx_timeout. irq: 00000000
> Dec 21 19:09:40 xen kernel: peth0: Ring at 4923c000: next 25635708
> nic 25635452
> Dec 21 19:09:40 xen kernel: peth0: Dumping tx registers
> (full output below)
> After rebooting the machine, I get a Kernel Bug:
> Dec 21 19:53:18 xen kernel: Kernel BUG at
> (full output below)
> After waiting a couple of hours and then rebooting the machine,
> everything works fine again for a certain time until the net hangs
> I am not quite sure, what causes this and how to debug.
> Any help is really appreciated.
On january 7, 2007, Brock Palen wrote: "more xen network problems"
> > Hello again, I put in pci network cards:
> > National Semiconductor Corporation DP83820
> Ok i have made progress, the problem listed below does go away, when
> using a old 3com pci card. So it looks like the ns83820 module has
> issues with Linux bridging. Is there a Wiki page for working
> networking? And hardware? The system works fine now (Dell PowerEdge
> 440SC) Just the internal networking is broken with xen, so you will
> need to add your own working networking.
> Other than that Is there a way to tell xend when it starts and
> creates a bridge to use eth1 and not eth0 to create the bridge? eth0
> (the ns83820) will be used as a crossover between teh two boxes for
> drbd. It works just fine if you dont create a bridge on that device.
> > They get addresses and make a /dev/eth0 allowing network access. I
> > had no luck making the bcm57xx work.
> > I now have a new problem, When I turned on xend networking no-
> > longer works, the output from the 'route' command is very slow to
> > appear. and i see the following in the logs, The system this is
> > replacing is a old xen-2.07 box so im not familiar with peth.
> > Jan 7 17:33:11 xen1 kernel: NETDEV WATCHDOG: peth0: transmit timed
> > out Jan 7 17:33:11 xen1 kernel: peth0: tx_timeout: tx_done_idx=10
> > free_idx=1 cmdsts=8000002a Jan 7 17:33:11 xen1 kernel: peth0: after:
> > tx_done_idx=10 free_idx=1 cmdsts=8000002a Jan 7 17:33:12 xen1
> > kernel: peth0: tx_timeout: tx_done_idx=10 free_idx=1 cmdsts=8000002a
> > Jan 7 17:33:12 xen1 kernel: peth0: after: tx_done_idx=10 free_idx=1
> > cmdsts=8000002a Jan 7 17:33:14 xen1 kernel: peth0: tx_timeout:
> > tx_done_idx=10 free_idx=1 cmdsts=8000002a Jan 7 17:33:14 xen1
> > kernel: peth0: after: tx_done_idx=10 free_idx=1 cmdsts=8000002a Jan
> > 7 17:33:16 xen1 kernel: peth0: tx_timeout: tx_done_idx=10 free_idx=1
> > cmdsts=8000002a Jan 7 17:33:16 xen1 kernel: peth0: after:
> > tx_done_idx=10 free_idx=1 cmdsts=8000002a
> > Its filling up my logs and filling up dmesg. I found some mentions
> > of this when i googled the archives but no solutions. Anyone have
> > any ideas?
> > Brock Palen
On august 16, 2007, Ian Tobin wrote: "annoying network problem"
> We are having an issue with XEN network where by after some time of
> the server being booted we start seeing these messages in the syslog
> kernel: NETDEV WATCHDOG: peth0: transmit timed out
> then all networking stops responding and the only way to solve it
> is to reboot the server.
> I have looked up and down on the web and some have mentioned putting
> pci=noacpi in the grub boot file but this has no affect.
> The network card is Compaq Computer Corporation Netelligent 10/100
> TX PCI
> Has anyone got any suggestion or work arounds for this?
> Any help is much appreciated
Xen-users mailing list