On 12/14/2017 05:53 PM, Jim Fehlig wrote:
On 12/14/2017 12:15 AM, Michal Privoznik wrote:
> On 12/13/2017 10:41 PM, Jim Fehlig wrote:
>> A Xen domain0 is better described as a persistent domain. Mark it
>> as such during intialization.
>>
>> Signed-off-by: Jim Fehlig <jfehlig(a)suse.com>
>> ---
>>
>> I noticed this while debugging a libvirt-guests issue. The list_guests()
>> function filters domain0 with
>>
>> echo "$list" | grep -v 00000000-0000-0000-0000-000000000000
>>
>> If domain0 is the only item in $list, the grep returns 1, causing
>> a failure of the stop operation when action is suspend. This
>> patch fixes the libvirt-guests issue, but I can also send a patch
>> to improve the filter if desired. E.g.
>>
>> echo "$list" | sed
"s/00000000-0000-0000-0000-000000000000//g"
>>
>> src/libxl/libxl_driver.c | 1 +
>> 1 file changed, 1 insertion(+)
>>
>> diff --git a/src/libxl/libxl_driver.c b/src/libxl/libxl_driver.c
>> index 40328a6cb..79e29ce07 100644
>> --- a/src/libxl/libxl_driver.c
>> +++ b/src/libxl/libxl_driver.c
>> @@ -609,6 +609,7 @@ libxlAddDom0(libxlDriverPrivatePtr driver)
>> def = NULL;
>> + vm->persistent = 1;
>> virDomainObjSetState(vm, VIR_DOMAIN_RUNNING,
>> VIR_DOMAIN_RUNNING_BOOTED);
>> if (virDomainDefSetVcpusMax(vm->def, d_info.vcpu_max_id + 1,
>> driver->xmlopt))
>> goto cleanup;
>>
>
> I think this is orthogonal to libvirt-guests fix. dom0 should be marked
> as persistent because it is persistent. I mean, even though we don't
> store its config anywhere, we can't really get rid of it (udefine &&
> destroy).
Right, that's why I didn't mention libvirt-guests in the commit message.
Is it ok to push this patch and send another for the libvirt-guests
improvement?
Oh right. Sorry, I forgot to write it explicitly: ACK.
Michal