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

[Xen-devel] Re: No shutdown of domU: xenbus_dev_shutdown: device/console/0: Initialising != Connected, skipping


  • To: Sander Eikelenboom <linux@xxxxxxxxxxxxxx>
  • From: Keir Fraser <keir@xxxxxxx>
  • Date: Wed, 17 Nov 2010 15:36:11 +0000
  • Cc: Ian Campbell <Ian.Campbell@xxxxxxxxxxxxx>, Jeremy Fitzhardinge <jeremy@xxxxxxxx>, "Xen-devel@xxxxxxxxxxxxxxxxxxx" <Xen-devel@xxxxxxxxxxxxxxxxxxx>, Ian Jackson <Ian.Jackson@xxxxxxxxxxxxx>
  • Delivery-date: Wed, 17 Nov 2010 07:36:51 -0800
  • Domainkey-signature: a=rsa-sha1; c=nofws; d=gmail.com; s=gamma; h=sender:user-agent:date:subject:from:to:cc:message-id:thread-topic :thread-index:in-reply-to:mime-version:content-type :content-transfer-encoding; b=JcHhjWBz7FO9kXegXoL/FLH+6/fDrTgUHPEqjdU4Mhg6x8ZTwtu5toExCywLiQ+Clv U7OgbA8sL0eLE4OmKbLvPB7mD5iDMOYvbVNEoK9XwQloJMEGo2knDHntdXsqb7O5Qdom v24If+G1O2TdLKWV0C0liE/Rv/Vs0CsB2Y1SU=
  • List-id: Xen developer discussion <xen-devel.lists.xensource.com>
  • Thread-index: AcuGbSkkl+0ZRSCezUqcMJ4LqJ/cWw==
  • Thread-topic: No shutdown of domU: xenbus_dev_shutdown: device/console/0: Initialising != Connected, skipping

Our 2.6.18 tree explicitly ignores console devices in
drivers/xen/xenbus/xenbus_probe.c:xenbus_probe_frontend(). More modern
kernels should be doing similar -- I think it's a bug for guest's xenbus
driver to attempt to manage console devices. One of our kernel guys will be
able to say more, no doubt.

 K.

On 17/11/2010 12:28, "Sander Eikelenboom" <linux@xxxxxxxxxxxxxx> wrote:

> Ahh the difference between console on the other devices seems to be "XENBUS:
> Device with no driver: device/console/0"
> That makes it states "initializing" for ever ...
> It then probably goes wrong with the "out:" path in xenbus_dev_shutdown()
> doing the put_device ?
> 
> I only fail to see why it seems to cause a problem with multiple vpcu's
> assigned and not with just one.
> 
> --
> Sander
> 
> 
> Wednesday, November 17, 2010, 1:06:04 PM, you wrote:
> 
>> Consoles do not have a connection handshake. If there is a state field in
>> xenstore, it is only unused detritus written by the toolstack
>> (xend/libxl/whatever).
> 
>>  -- Keir
> 
>> On 17/11/2010 11:58, "Sander Eikelenboom" <linux@xxxxxxxxxxxxxx> wrote:
> 
>>> Hmm .. i haven't received any response, is there anyone who could point me
>>> to
>>> the functions involved in communicating the state of the console from
>>> "initializing" to "connected" ?
>>> That way i could at some additional printk's to find out why the state of
>>> domU
>>> consoles stays "1" instead of "4" in xenstore.
>>> 
>>> --
>>> Sander
>>> 
>>> Friday, November 12, 2010, 6:30:10 PM, you wrote:
>>> 
>>>> I'm encountering the following problem:
>>> 
>>>> When trying to shutdown a PV domU with more than 1 vpcu, the shutdown
>>>> fails.
>>>> The domain keeps running with 100% cpu, and it's still possible to get the
>>>> console of this domU with xm console.
>>>> When only 1 vcpu is assigned the domain does shutdown.
>>> 
>>>> Last lines of the PV domU console:
>>> 
>>>> Debian GNU/Linux 5.0 tv hvc0
>>> 
>>>> INIT: Switching to runlevel: 0
>>>> INIT: Sending processes the TERM signal
>>>> Stopping web server: apache2 ... waiting .
>>>> Stopping MythTV server: mythbackend No /usr/bin/mythbackend found running;
>>>> none killed.
>>>> .
>>>> Stopping MTA: exim4_listener.
>>>> Stopping rsync daemon: rsync.
>>>> Stopping MySQL database server: mysqld.
>>>> Saving the system clock.
>>>> Cannot access the Hardware Clock via any known method.
>>>> Use the --debug option to see the details of our search for an access
>>>> method.
>>>> Stopping enhanced syslogd: rsyslogd.
>>>> Asking all remaining processes to terminate...done.
>>>> All processes ended within 2 seconds....done.
>>>> Deconfiguring network interfaces...done.
>>>> Cleaning up ifupdown....
>>>> Deactivating swap...done.
>>>> Unmounting local filesystems...done.
>>>> Will now halt.
>>>> [ 4336.046876] md: stopping all md devices.
>>>> [ 4337.047171] xenbus_dev_shutdown:  trying shutdown of device/console/0:
>>>> Initialising
>>>> [ 4337.047194] xenbus_dev_shutdown: device/console/0: Initialising !=
>>>> Connected, skipping
>>>> [ 4337.047200] xenbus_dev_shutdown: result of shutdown of device/console/0:
>>>> Initialising
>>>> [ 4337.047205] xenbus_dev_shutdown:  trying shutdown of device/vif/0:
>>>> Connected
>>>> [ 4337.110869] xenbus_dev_shutdown: result of shutdown of device/vif/0:
>>>> Closed
>>>> [ 4337.110883] xenbus_dev_shutdown:  trying shutdown of device/vbd/51714:
>>>> Connected
>>>> [ 4337.161975] xenbus_dev_shutdown: result of shutdown of device/vbd/51714:
>>>> Closed
>>>> [ 4337.161989] xenbus_dev_shutdown:  trying shutdown of device/vbd/51713:
>>>> Connected
>>>> [ 4337.217136] xenbus_dev_shutdown: result of shutdown of device/vbd/51713:
>>>> Closed
>>>> [ 4337.217167] System halted.
>>> 
>>> 
>>>> But when using xenstore-ls .. i see that for every domain (but 1 and
>>>> multiple
>>>> vcpu's):
>>>>     - All devices have state=4
>>>>     - Except all backend = "/local/domain/0/backend/console/*/0" entries,
>>>> those have state=1
>>>>     - Although xenstore is saying the state is initializing .. xm console
>>>> works perfectly for all domains.
>>>>     - Perhaps this also explains the high event/0 load in dom0, related to
>>>> tty and xenconsoled ?
>>> 
>>>> DomU kernels vary from debian 2.6.26-xen kernels, to 2.6.37-rc1 mainline.
>>>> Xen-unstable-tip and xen-next-2.6.32 dom0
>>> 
>>> 
>>>> --
>>>> Sander
>>> 
>>> 
> 
> 
> 
> 



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


 


Rackspace

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