On 14/04/15 14:50, Tomasz Bursztyka wrote:
> Hi Slava,
>
> This looks really suspicious. Such functions are always used to
> agglomerate BSS into networks.
> Which networks are then pushed to the wifi plugins through callbacks.
> Some of these exist because of different behaviors of wpa_supplicant
> and its different versions (from 0.7.3 If I remember well).
>
> This code path is used during wifi operation, so a leak should have
> been found long ago.
> What are the steps followed to get such leak?


The log file I was analyzing is over 3 million lines.
interface_network_added is invoked 246 times but there are only 22
leaks. I spent a few hours trying to identify these 22 special cases out
of 246, but no luck so far.

In the meantime I don't understand what interface_network_added() is
doing in the first place. If I read the code correctly, it allocates
GSupplicantNetwork, requests its properties and if everything goes well,
the network object ends up in merge_network() where it simply gets
deallocated. If something doesn't go well (those 22 cases I guess) it
doesn't make it to merge_network() and the network object doesn't get
deallocated.

Am I missing something? If this code isn't doing anything useful then
perhaps it makes more sense to remove it than to debug it.


> What patches do you have on top?
>

Nothing important in supplicant.c, if anything at all. We are a bit
behind though, but I checked the latest master and didn't notice any
relevant changes there.

Regards,

-Slava


> Tomasz
> _______________________________________________
> connman mailing list
> [email protected]
> https://lists.connman.net/mailman/listinfo/connman
>


_______________________________________________
connman mailing list
[email protected]
https://lists.connman.net/mailman/listinfo/connman

Reply via email to