[sr-dev] Possible memory leakage in ndb_redis module

[Vicente Hernando]

Hello,

On 05/03/2012 01:29 PM, Daniel-Constantin Mierla wrote:
> Hello,
>
> thanks for the patch -- when the module is destroyed, practically is the shut down of kamailio, when the entire memory used is freed (pkg and shm) as well as the one from the system by the system manager. Anyhow, for a clean shut down, is good to have it.
>
> One thing is not clear, where the function redisc_free_reply(str *name) is used? The patch seems just to define it.
The patch is not finished, only to show a possible solution idea.

Right now, if a user performs:

redis_cmd("srvN", "GET foo", "r");

a redisc_reply_t structure is created and assigned to name "r".  If user 
performs again:
redis_cmd("srvN", "GET bar", "r");

then old redisk_reply_t.rplRedis is freed with freeReplyObject and 
hiredis creates a new redisReply for rplRedis field.
No problem here, then when kamailio closes that last structure would be 
freed.

However, if user would perform:
redis_cmd("srvN", "GET foo", "r1");
redis_cmd("srvN", "GET foo", "r2");
redis_cmd("srvN", "GET foo", "r3");
...

those structures would not be freed from memory.

In my opinion if a user creates a lot of different replies, it should be 
his responsibility to free them using e.g:
redis_free("r1")
redis_free("r2")

Most users will not need this. I am planning exporting ndb_redis 
functions in an API, and use them heavily in another new module.

Regards,
Vicente.
>
> Cheers,
> Daniel
>
>
> On 5/2/12 9:38 PM, Vicente Hernando wrote:
>> Ops, sorry, I send a wrong patch, that one creates an API for 
>> ndb_redis module.
>>
>> The patch attached now is the correct one.
>>
>> Regards,
>> Vicente.
>>
>>
>>
>> On 05/02/2012 09:33 PM, Vicente Hernando wrote:
>>> Hello,
>>>
>>> looking at ndb_redis module source code I see replies are not freed 
>>> when module is destroyed.
>>>
>>> Also a reply is only freed when another one with same name is 
>>> performed. This could cause problems if an user would ask for
>>> a lot of different named replies.
>>>
>>> IMHO freeing reply list when module is destroyed and creating a new 
>>> redisc_free_reply function would solve this.
>>>
>>>
>>> I attach a patch that compiles but not tested to show the idea.
>>>
>>> If you agree I can perform the task, I suppose redisc_free_reply 
>>> function should also be exported in cmd_export_t module structure.
>>>
>>>
>>> This solution would be compatible with old ndb redis module 
>>> configuration.
>>>
>>>
>>>
>>> Regards,
>>> Vicente.
>>>
>>>
>>>
>>> _______________________________________________
>>> sr-dev mailing list
>>> sr-dev at lists.sip-router.org
>>> http://lists.sip-router.org/cgi-bin/mailman/listinfo/sr-dev
>>
>>
>>
>> _______________________________________________
>> sr-dev mailing list
>> sr-dev at lists.sip-router.org
>> http://lists.sip-router.org/cgi-bin/mailman/listinfo/sr-dev
>
> -- 
> Daniel-Constantin Mierla -http://www.asipto.com
> http://twitter.com/#!/miconda  -http://www.linkedin.com/in/miconda

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.sip-router.org/pipermail/sr-dev/attachments/20120503/06d4411f/attachment.htm>