[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [Xen-users] DomU crashes with device_model_stubdomain_override=1 under Xen 4.5.0



W dniu 2015-02-18 o 14:43, Paul Durrant pisze:
-----Original Message-----
From: Ian Campbell
Sent: 18 February 2015 13:38
To: Kuba; Wei Liu; Paul Durrant
Cc: xen-users@xxxxxxxxxxxxx
Subject: Re: [Xen-users] DomU crashes with
device_model_stubdomain_override=1 under Xen 4.5.0

On Mon, 2015-02-16 at 22:33 +0100, Kuba wrote:

Wei/Paul, I suppose this is the ioreq server related regression which
was discovered and discussed on xen-devel recently?


Certainly sounds like it. The symptom was hvmloader() bailing because PCi 
enumeration got screwed up.

I'm still catching up on my email backlog from being away -- what's the
status of the fix?

I posted a fix (in Xen) to the list a couple of weeks back and Jan committed it 
on Feb 10th:

http://xenbits.xen.org/gitweb/?p=xen.git;a=commit;h=dd748d128d86996592afafea02e578cc7d4e6d42

I don't see it in the stable-4.5 branch yet though.

   Paul


Dear List,

I'm trying to bring up an HVM domain with

device_model_stubdomain_override=1

under Xen 4.5.0, but the domain always crashes just after creation.
Everything works flawlessly in Xen 4.4.x. Under 4.5.0 the domain starts
just fine if I change that line to "device_model_stubdomain_override=0".

I'm using stub domains because they allow me to use storage driver
domains.

# xl -vvv create consumer.conf &> create.log
# xl list
Name                        ID   Mem VCPUs      State   Time(s)
Domain-0                     0  6144     4     r-----      36.1
consumer                    10   255     1     ---sc-       0.0
consumer-dm                 11    32     1     -b----       0.1

There is no output in the VNC console.

The logs are from Xen compiled with debug ?= y, verbose ?= y and
crash_debug ?= y, but the same happens with non-debug build.

My consumer.conf:

name='consumer'
device_model_stubdomain_override=1
builder='hvm'
vcpus=1
memory=256
disk=[
#'file:/root/fbsd.iso,xvda,r,devtype=cdrom'
]
boot='d'
pae=1
nx=1
videoram=16
stdvga=1
sdl=0
vnc=1
vnclisten="0.0.0.0"
vncdisplay=31
localtime=1
xen_platform_pci=1
on_crash="preserve"

I've ran out of ideas and I'd be very grateful for any advice on how to
find the cause of this issue.

Best regards,
Kuba

I have applied Paul's patch to Xen 4.5.0 and it seems to resolve the issue:) Thank you!

Now, unfortunately, adding a vif crashes the stub domain:

# xl list
Name                                ID   Mem VCPUs      State   Time(s)
Domain-0                             0  6144     4     r-----      12.2
consumer                             3   239     1     ------       0.0
consumer-dm                          4    32     1     ---sc-       0.0

This happens with both standard Linux bridge and Open vSwitch using:
vif=['bridge=xenbr0']
or
vif=['script=vif-openvswitch,bridge=xenbr0']

Similar errors come up in both cases in qemu-dm-consumer.log:

************************ NETFRONT for device/vif/0 **********


net TX ring size 256
net RX ring size 256
backend at /local/domain/0/backend/vif/4/0
mac is 00:16:3e:4f:92:50
backend not avalable, state=5
TAP open failed
Could not initialize device 'tap'
close(0)
close(1)
close(2)
main returned 1
Do_exit called!
base is 0x5efa00 caller is 0x101c52
base is 0x5efa10 caller is 0xe4fe8
base is 0x5efa30 caller is 0xe5b09
base is 0x5efa60 caller is 0x102acb
base is 0x5efa80 caller is 0x8744
base is 0x5efe10 caller is 0xe5a7d
base is 0x5effe0 caller is 0x343b

I'd be very grateful for further assistance. Everything works fine with device_model_stubdomain_override=0 using both linux bridge and ovs.

Any help will be much appreciated.

Best regards,
Kuba


_______________________________________________
Xen-users mailing list
Xen-users@xxxxxxxxxxxxx
http://lists.xen.org/xen-users


 


Rackspace

Lists.xenproject.org is hosted with RackSpace, monitoring our
servers 24x7x365 and backed by RackSpace's Fanatical Support®.