Re: More Xen troubles in -current: booting DOMUs fails strangely

[Johan Ihrén <johani%johani.org@localhost>]

Hi John,

On Jul 22, 2011, at 00:16 , John Nemeth wrote:

> On Dec 11,  4:13pm, =?iso-8859-1?Q?Johan_Ihr=E9n?= wrote:
> }
> } This is a rather old box with an Intel Core 2 Quad CPU. This DOM0 runs
> } 5.1 and 5.99.43 with Xen 3.3.2nb1 just fine, and -current (5.99.55) also
> } boots nicely (apart from all the pmap_kenter_pa messages to the
> } console).
> } 
> } However, when upgrading the DOM0 to 5.99.55 (amd64) the DOMUs (i386pae,
> } not upgraded) no longer boot.
> } 
> } I.e. while running 5.99.43:
> } 
> } # xm create -c foo.conf
> } 
> } works fine, and connects me to the console of the booting DOMU "foo".
> } But upgrading the kernel to 5.99.55 from today (July 21) leads to the
> } same command hanging for a while and then the message:
> } 
> } Error: Device 0 (vif) could not be connected. Hotplug scripts not working.
> } Error: Domain 'foo' does not exist.
> 
>     What does your vif line look like?

Very relevant question, apologies for not providing any useful information 
about such.

tank# grep vif /usr/local/store/xen/lab/conf/master.i386pae
vif = [ 'mac=00:16:3e:12:00:ff, bridge=bridge0', 'bridge=bridge1' ]

I.e. fixed mac for the first xennet, dynamic for the second. Both bridges have 
been created, are up and running and have other DOM0 interfaces attached to 
them already:

tank# brconfig bridge0
bridge0: flags=41<UP,RUNNING>
        Configuration:
                priority 32768 hellotime 2 fwddelay 15 maxage 20
                ipfilter disabled flags 0x0
        Interfaces:
                tap0 flags=3<LEARNING,DISCOVER>
                        port 4 priority 128
        Address cache (max cache: 100, timeout: 1200):
tank# brconfig bridge1
bridge1: flags=41<UP,RUNNING>
        Configuration:
                priority 32768 hellotime 2 fwddelay 15 maxage 20
                ipfilter disabled flags 0x0
        Interfaces:
                wm0 flags=3<LEARNING,DISCOVER>
                        port 1 priority 128
        Address cache (max cache: 100, timeout: 1200):
                00:18:f3:01:f8:f9 wm0 1108 flags=0<>
                00:40:ca:28:8b:b8 wm0 1079 flags=0<>
                64:b9:e8:87:d9:7e wm0 1070 flags=0<>
                58:55:ca:f0:8b:db wm0 1064 flags=0<>
                00:16:cb:a9:cc:41 wm0 1062 flags=0<>

However, your question got me thinking and I changed the first line of 
/usr/pkg/etc/xen/scripts/vif-bridge to be '#!/bin/sh -ex' so I would see what 
happens. In the xendbackendd.log I get (among other things) the following:

+ 'iface=xvif2.0'
+ echo ifconfig 'xvif2.0' up
+ ifconfig 'xvif2.0' up
ifconfig: exec_matches: Device not configured

And later:

+ 'iface=xvif2.1'
+ echo ifconfig 'xvif2.1' up
+ ifconfig 'xvif2.1' up
ifconfig: exec_matches: Device not configured

I.e. both xennets/xvifs fail to get created. Now I have to admit that how the 
xvif interfaces are created and where is a mystery to me, but it would seem 
that this is where it fails. I.e. the problem is not connecting the xvif to the 
bridge, but rather the creation of the xvif in the first place.

I should also mention that while I usually run with my own kernels this is with 
a stock XEN3_DOM0 kernel from [repository of 
choice]/pub/NetBSD-daily/HEAD/201107210820Z.

Regards,

Johan