Re: Unable to create VM with nic device on Arndale

["Roger Pau Monné" <roger.pau@citrix.com>]

On 18/01/14 03:35, Dennis Lan (dlan) wrote:
> On Fri, Jan 17, 2014 at 9:31 PM, Roger Pau Monné <roger.pau@citrix.com> wrote:
>> On 17/01/14 12:59, Ian Campbell wrote:
>>> On Fri, 2014-01-17 at 19:43 +0800, Dennis Lan (dlan) wrote:
>>>> On Fri, Jan 17, 2014 at 7:01 PM, Ian Campbell <Ian.Campbell@citrix.com> wrote:
>>>>
>>>>
>>>>> vif-bridge and the common scripts which it includes would be a good
>>>>> start. Just an echo at the top to confirm that the script is running
>>>>> would be useful.
>>>>>
>>>>> I used to do "exec 1>/tmp/hotplug.log 2>&1" at the top to aid debugging
>>>>> when these scripts were launched by udev, but now that libxl runs them
>>>>> you may find that the debug from the script comes out on stdout/err of
>>>>> the xl create command so perhaps that isn't needed any more.
>>>>>
>>>>>> headless here.
>>>>>
>>>>> That shouldn't matter, you are looking for output from userspace
>>>>> scripts, not kernel or hypervisor logs.
>>>>>
>>>>> Ian.
>>>>>
>>>>
>>>> Hi Ian
>>>> I suspect for 4.4.0, the network devices even was not detected.
>>>> this is output from 4.3.1, notes follow lines.
>>>>
>>>> libxl: debug: libxl_device.c:959:device_hotplug: calling hotplug
>>>> script: /etc/xen/scripts/vif-bridge online
>>>> dlan: vif-bridge start
>>>> dlan: vif-common start
>>>>
>>>> dlan: vif-bridge start -> output from vif-bridge script
>>>> dlan: vif-common start -> output from vif-common.sh script
>>>
>>> So these are the 4.3 logs? Have you tried 4.4 and found that it doesn't
>>> produce the same output?
>>>
>>> (please can you try and set the text type to "preformatted" for the logs
>>> -- having them wrapped makes them very hard to read).
>>>
>>> The lack of
>>> libxl: debug: libxl_device.c:959:device_hotplug: calling hotplug script: /etc/xen/scripts/vif-bridge online
>>> in your original logs is a bit concerning.
>>>
>>> Roger -- any ideas?
>>
>> My first guess would be that libxl__get_domid failed, however I'm not
>> able to reproduce this. I'm attaching a patch to add an error message
>> if libxl__get_domid fails, and also prevent the removal of xenstore
>> entries so we can see what's going on. Dennis/Eugene, could you try the
>> attached patch and send the output of xl -vvv create <...> and
>> xenstore-ls -fp after the failed creation?
>>
>> ---
>> diff --git a/tools/libxl/libxl_create.c b/tools/libxl/libxl_create.c
>> index a604cd8..03f9fe9 100644
>> --- a/tools/libxl/libxl_create.c
>> +++ b/tools/libxl/libxl_create.c
>> @@ -1296,6 +1296,9 @@ static void domcreate_complete(libxl__egc *egc,
>>          rc = xc_flask_relabel_domain(CTX->xch, dcs->guest_domid, d_config->b_info.exec_ssidref);
>>
>>      if (rc) {
>> +        LOG(ERROR, "domain creation failed, not doing removal of xs entries");
>> +        dcs->callback(egc, dcs, rc, -1);
>> +        return;
>>          if (dcs->guest_domid) {
>>              dcs->dds.ao = ao;
>>              dcs->dds.domid = dcs->guest_domid;
>> diff --git a/tools/libxl/libxl_device.c b/tools/libxl/libxl_device.c
>> index ba7d100..56d8162 100644
>> --- a/tools/libxl/libxl_device.c
>> +++ b/tools/libxl/libxl_device.c
>> @@ -965,7 +965,10 @@ static void device_hotplug(libxl__egc *egc, libxl__ao_device *aodev)
>>       * hotplug scripts
>>       */
>>      rc = libxl__get_domid(gc, &domid);
>> -    if (rc) goto out;
>> +    if (rc) {
>> +        LOG(ERROR, "unable to get domain id, error: %d", rc);
>> +        goto out;
>> +    }
>>      if (aodev->dev->backend_domid != domid) {
>>          if (aodev->action != LIBXL__DEVICE_ACTION_REMOVE)
>>              goto out;
>>
> with this patch applied, I got following err. or see attached file for more info
> 
> ofire configs # xl create -c test1_stable
> Parsing config from test1_stable
> libxl: error: libxl_device.c:969:device_hotplug: unable to get domain
> id, error: -3
> libxl: error: libxl_device.c:969:device_hotplug: unable to get domain
> id, error: -3
> libxl: error: libxl_create.c:1206:domcreate_attach_vtpms: unable to
> add nic devices
> libxl: error: libxl_create.c:1279:domcreate_complete: domain creation
> failed, not doing removal of xs entries
> 

Hello,

Thanks for the log, could you please post the output of xenstore-ls -fp
after a failed domain creation?

My first guess is that your xencommons init script is outdated, could
you check if your xencommons init script has the following line:

113                 ${BINDIR}/xenstore-write "/local/domain/0/domid" 0

Roger.