[Freeipa-devel] [PATCH 0023] topology plugins sigsev when adding a managed host

thierry bordaz tbordaz at redhat.com
Fri Jun 10 16:00:13 UTC 2016



On 06/10/2016 05:56 PM, Ludwig Krispenz wrote:
>
> On 06/10/2016 05:41 PM, thierry bordaz wrote:
>>
>>
>> On 06/10/2016 05:23 PM, Ludwig Krispenz wrote:
>>>
>>> On 06/10/2016 04:44 PM, thierry bordaz wrote:
>>>> Hi Ludwig,
>>>>
>>>> I agree with you there is no path to add a host with an empty hostname.
>>>> You fix looks valid but I would prefer a log in FATAL rather in PLUGIN.
>>> yes, of course that was my intention, copy paste :-)
>>>>
>>>> Also I wonder if a reason of empty hostname could be a 
>>>> slapi_ch_free on it but with the host remaining in the list.
>>>> Looking at 
>>>> https://git.fedorahosted.org/cgit/freeipa.git/tree/daemons/ipa-slapi-plugins/topology/topology_cfg.c#n852,
>>>> I wonder if the two lines 852 and 853 could lead to this situation.
>>> but this frees the complete replica structure tconf and in the 
>>> caller tconf is set to null, so should never be used again.
>>
>> Yes you are right, it can not conduct to empty hostname into tconf.
>> However I think it can leak because host will be set to 0 at the end 
>> of the first iteration.
> yes, line 852 and 853 should be switched
>>
>> An other possibility is that in
>> https://git.fedorahosted.org/cgit/freeipa.git/tree/daemons/ipa-slapi-plugins/topology/topology_util.c#n1449|
>> ipa_topo_util_init_hosts, 'cn' has an empty value. newhost is not 
>> null but empty, so we may create an empty hostname.
>> |

|You are definitely right there is no path that can lead to NULL 
hostname :-(

If we are unable to get a reproducible test case,
A crazy idea would be to do some internal rotating logging in a buffer 
on each events (add/del host, add/del segments...),
when the incoherency is detected, then dump the buffer into the error 
logs so that we would have more data to find the RC.


|
>> ||
>>>>
>>>> thanks
>>>> thierry
>>>>
>>>>
>>>> On 06/10/2016 12:36 PM, Ludwig Krispenz wrote:
>>>>> Hi,
>>>>> the attached patch will prevent the crash reported in ticket #5928.
>>>>>
>>>>> So far I do not understand how this situation can occur, there is 
>>>>> no reproducer yet. I do not really like this fix as it hides a 
>>>>> probable corrupted data structure and would prefer to find the 
>>>>> root cause.
>>>>>
>>>>> But please review it, so we can commit it if there is no progress 
>>>>> on the root cause.
>>>>>
>>>>> Ludwig
>>>>>
>>>>>
>>>>>
>>>>
>>>
>>> -- 
>>> Red Hat GmbH,http://www.de.redhat.com/, Registered seat: Grasbrunn,
>>> Commercial register: Amtsgericht Muenchen, HRB 153243,
>>> Managing Directors: Charles Cachera, Michael Cunningham, Michael O'Neill, Eric Shander
>>
>
> -- 
> Red Hat GmbH,http://www.de.redhat.com/, Registered seat: Grasbrunn,
> Commercial register: Amtsgericht Muenchen, HRB 153243,
> Managing Directors: Charles Cachera, Michael Cunningham, Michael O'Neill, Eric Shander

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://listman.redhat.com/archives/freeipa-devel/attachments/20160610/a136e4b0/attachment.htm>


More information about the Freeipa-devel mailing list