Re: [Xen-devel] Hotplug scripts not working... xen/ia64 domU stoppedworking

[Ewan Mellor <ewan@xxxxxxxxxxxxx>]

On Tue, Dec 06, 2005 at 12:58:11PM -0500, Dave Thompson (davetho) wrote:

> Ewan,
> 
> I found some other udev messages in the syslog that don't look good.
> 
> Dec  6 09:36:33 davetho-intel udevd[6015]: init_udevd_socket: bind
> failed, Address already in use
> Dec  6 09:36:33 davetho-intel udevd[6018]: init_udevd_socket: bind
> failed, Address already in use
> Dec  6 09:36:33 davetho-intel udevd[6021]: init_udevd_socket: bind
> failed, Address already in use

> Dec  6 09:36:33 davetho-intel kernel: udevd[6012] trap stack segment
> rip:404967 rsp:7fffff89e7a0 error:0

This message states that your udevd process has fallen off the end of its
stack in some way (and crashed, of course).  I think the other messages come
from multiple instances of udevsend all starting udevd having found that there
isn't one alive, and then racing with each other.  Those that lose the race
fail to bind the socket.  This seems like a udev bug to me, see

http://www.mail-archive.com/lfs-book@xxxxxxxxxxxxxxxxxxxx/msg04694.html

for example.  However, this is really only a secondary problem -- the primary
problem is that udevd has crashed.

It would help to know what was at the rip specified in the error message
(objdump -d /sbin/hotplug) and even better would be if you could start udevd
under gdb, and get a backtrace and the register contents when it crashes.

> Does the udev daemon rely on something in the kernel that either I
> haven't included in the .config or perhaps isn't available in 2.6.12?

udevd relies on a good number of things being turned on in the kernel, but I
would have thought that you would have got a more definitive error if that was
what you had done.

Ewan.

_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxxxxxxxx
http://lists.xensource.com/xen-devel