Re: After switching from "xm" to "xl" toolstack, can't get Guest networking to work.

[Jim Fehlig <jfehlig@suse.com>]

Ian Campbell wrote:
> On Tue, 2012-01-24 at 18:05 +0000, erin.balid@inoutbox.com wrote:
>   
>> Using the same log redirection you mentioned before
>>
>> edit /etc/xen/scripts/vif-bridge
>> 	#!/bin/bash
>> +       exec 1>>/tmp/vif-bridge.log
>> +       exec 2>&1
>> +       echo "`date`: Running $0 $*"
>> ...
>> +       ## TEST ##
>> +       echo "made it to 001"
>>
>> 	dir=$(dirname "$0")
>> 	. "$dir/vif-common.sh"
>>
>> +       ## TEST ##
>> +       echo "made it to 002"
>>
>>
>> 	domu=$(xenstore_read_default "$XENBUS_PATH/domain" "")
>> 	if [ -z "$domu" ]
>> 	then
>> 	    log debug "No device details in $XENBUS_PATH, exiting."
>>     
>
> So, presumably either the xenstore_read_default is failing or domu is
> zero length and the script is explicitly bailing here.
>
> However I do not see anything like this anywhere in the hotplug script
> shipped with Xen. Either this is a local modification or something done
> by your distribution, in which case you should contact them.
>   

Grrr, that is my patch here

http://old-list-archives.xen.org/archives/html/xen-devel/2011-10/msg01790.html

which thankfully was never applied upstream but did make its way into
openSUSE12.1.  In that thread, Ian suggested having the toolstack write
the literal tap device name to xenstore and then read it in the vif
script.  I have this on my todo list with very low priority, but seems I
should bump it now :-).

> What does this node contain under xend?
>   

The xend toolstack writes the domain name under this node but xl/libxl
doesn't, hence the failure Erin is seeing.

Erin, you can revert the patch mentioned above to fix your problem. 
I'll try to work on the proper fix soon.

Regards,
Jim