Re: [Xen-devel] [PATCH] tools/xenconsoled: Initialise pointers before trying to use them

[Andrew Cooper <andrew.cooper3@xxxxxxxxxx>]

On 07/03/13 17:15, Wei Liu wrote:
> On Thu, 2013-03-07 at 16:22 +0000, Andrew Cooper wrote:
>> On 07/03/13 16:20, Wei Liu wrote:
>>> On Thu, 2013-03-07 at 15:13 +0000, Andrew Cooper wrote:
>>>> This is a regression introduced by
>>>>
>>>> "Switch from select() to poll() in xenconsoled's IO loop."
>>>>   hg c/s 26405:7359c3122c5d
>>>>   git cc5434c933153c4b8812d1df901f8915c22830a8
>>>>
>>>> which results in reliable segfaults during VM power operations.
>>>>
>>>> Signed-off-by: Marcus Granado <marcus.granado@xxxxxxxxxx>
>>>> Signed-off-by: Andrew Cooper <andrew.cooper3@xxxxxxxxxx>
>>>>
>>> Good catch. Thanks.
>>>
>>>
>>> Wei.
>>>
>> Sadly, after fixing these segfaults, the code as currently is will cause
>> xenconsoled to exit gracefully as soon as you try and boot the 128th
>> domain.  We are currently investigating the issue.
>>
> Odd. So you were implying if you didn't fix this bug, you succeeded in
> booting up >128 guests?
>
> All the exit paths have "dolog" which outputs to stderr, so maybe run
> xenconsoled in foreground can help.
>
>
> Wei.

After fixing the segfault bugs, xenconsoled intermittently fails and
exits with:

Mar  7 16:45:49 localhost /usr/sbin/xenconsoled: Failure in poll
xs_handle: 3 (No such process)

The test case is attempting to sequentially boot 1000 PV VMs on the same
host.

So far, the common theme of failures appear to be at multiples of 128
VMs.  This is usually at the 128th VM, but also seen at the 384th VM.

~Andrew

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