Re: [Xen-devel] Possible memory leak in qemu-dm (qemu-dm swapping 20GB+, adding 2gb+ per day)

[Andrew Cooper <andrew.cooper3@xxxxxxxxxx>]

On 25/03/14 11:03, Ian Campbell wrote:
> On Tue, 2014-03-25 at 21:48 +1100, Steven Haigh wrote:
>> On 25/03/14 21:28, Ian Campbell wrote:
>>> On Tue, 2014-03-25 at 09:09 +0200, Pasi KÃrkkÃinen wrote:
>>>> On Tue, Mar 25, 2014 at 01:08:00PM +1100, Steven Haigh wrote:
>>>>> Continuing from:
>>>>> http://lists.xenproject.org/archives/html/xen-devel/2013-11/msg03961.html
>>>>>
>>>>> Hi guys,
>>>>>
>>>>> Continuing on from this thread, has any progress been made on this?
>>>>>
>>>>> I have had a report from a user of my packages with the same problem:
>>>>>   http://xen.crc.id.au/bugs/view.php?id=25
>>>>>
>>>>> He has been able to reproduce this in a reliable manner.
>>>>>
>>>> I thought the fix was committed to all the maintained qemu-traditional 
>>>> branches..
>>> I think so too.
>>>
>>> But if not then given a reliable repro I think the advice to try it
>>> under valgrind (which AIUI can now traces qemus thanks to Andrew Coopers
>>> work) still holds as a useful next step.
>> Is there any guide on how to do this to gather said info? Documentation?
>> Implementation? Interpretation?
> http://blog.xen.org/index.php/2013/01/18/using-valgrind-to-debug-xen-toolstacks/
>  has some info on running valgrind on the toolstack, I think this should 
> extend to processes launched by the toolstack such as qemu, so it might be as 
> easy as following that.
>
> Otherwise Andrew might have some more concrete advise but I think the
> approach I would take is to create a wrapper script which does
> "valgrind /path/to/qemu $@" and then use that via the
> device_model_override directive in the domain config.
>
> Ian.
>
>

I have never used valgrind in combination with xl and qemu before, but
the intercepting it in a Xapi environment is mostly similar.

Something like:

#!/bin/sh
valgrind --log-file="/path/to/logs/qemu-%p-valgrind.log" /path/to/qemu "$@"

should work fine.

You will need the latest valgrind, and the patchset of 7 from
1393858404-15220-1-git-send-email-andrew.cooper3@xxxxxxxxxx as they are
still pending acceptance upstream.

At some point soon I will need to do some more patches for the new
SYSCTL and pending DOMCTL interface bumps new in unstable, but that wont
affect you if you are on a released version of Xen.

Finally, I have not yet tried qemu-upstream, so there might still be
some missing hypercalls, but qemu-traditional should work fine.

~Andrew

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