[Freeipa-devel] [PATCH] 370-371 ipa-kdb: remove memory leaks

[Petr Spacek]

On 14.2.2013 10:07, Sumit Bose wrote:
> On Tue, Feb 12, 2013 at 04:42:52PM +0100, Martin Kosek wrote:
>> On 02/12/2013 04:26 PM, Simo Sorce wrote:
>>> On Tue, 2013-02-12 at 16:14 +0100, Martin Kosek wrote:
>>>> Explained in the commit description - this may not be super-critical, I just
>>>> followed info in ldap_search_ext() man page:
>>>>
>>>> ...
>>>>
>>>>         Note that res parameter  of  ldap_search_ext_s()  and  ldap_search_s()
>>>> should  be  freed  with
>>>>         ldap_msgfree() regardless of return value of these functions.
>>>> ...
>>>
>>> OK.
>>>
>>>>> This snippet seem to change the logic.
>>>>>
>>>>> Before the condition was !ipadb_need_retry() and no you change it to
>>>>> ipadb_need_retry() so it looks reversed, was this on purpose ?
>>>>
>>>> As noted in commit description, this check was wrong and it was causing the
>>>> function to _always_ retry at least once because ipadb_need_retry returns true
>>>> when we need to retry and not 0 -> thus no "!" is needed.
>>>
>>> I do not like much 'sneaking' this kind of change in the same patch that
>>> fixes memory leaks, so if you want to plit in 2 patches it would be
>>> nice, but I am of course ok with the change.
>>
>> I agree with that. Lets do it right as we have not pushed it yet. I moved this
>> change to a separate patch (attached).
>
> I didn't found any issues while testing, so ACK.
>
> When I run kinit in a endless loop I still see a constant increase in
> the resident memory size, but it looks like it is not related to the
> ipadb code. I will investigate further.
>
> bye,
> Sumit

KDC has built-in in-memory "lookaside cache", see 
http://mailman.mit.edu/pipermail/krbdev/2012-June/010985.html

You can recompile KDC without lookaside cache or run test for more than 120 
seconds. Memory consumption should be roughly stable (if request rate is 
stable all the time).

-- 
Petr^2 Spacek