Re: [PATCH 00/13] Valgrind patches for hypercalls

[Andrew Cooper <andrew.cooper3@citrix.com>]

On 02/12/13 10:49, Ian Campbell wrote:
> On Mon, 2013-12-02 at 10:46 +0000, Andrew Cooper wrote:
>> On 02/12/13 10:37, Ian Campbell wrote:
>>> On Fri, 2013-11-29 at 11:45 +0000, Ian Campbell wrote:
>>>> On Thu, 2013-11-28 at 18:56 +0000, Andrew Cooper wrote:
>>>>> This set of patches teaches valgrind about new hypercalls.
>>>>>
>>>>> Valgrind can now completely inspect xc_domain_save()/restore()
>>>>>
>>>>> Signed-off-by: Andrew Cooper <andrew.cooper3@citrix.com>
>>>>> CC: Ian Campbell <Ian.Campbell@citrix.com>
>>>> Thanks, these look OK to me on a quick skim and build so I've sent them
>>>> upstream to: https://bugs.kde.org/show_bug.cgi?id=328205
>>> Which has just resulted in them being applied, thanks to Bart Van
>>> Assche.
>>>
>>> Thanks,
>>> Ian.
>>>
>>>
>> Wow - that was quick going.
> Yes!
>
>> I wonder how well valgrind now does with xl create...
> It used to work for me, that's what I was doing when I first implemented
> this stuff.
>
> IIRC it doesn't play especially well with the deaemonisation aspect for
> some reason, but you can keep it in the foreground and/or disable the
> monitoring bit (both on the cmdline) and it works ok, for at least bog
> standard PV and HVM guests.
>
> That was then of course.
>
> Ian.
>

Sadly, its the daemonisation aspect which is the interesting one to look
at.  Memory leaks in the daemon are far more critical than in the
short-lived runs.

~Andrew