Checkov, Andrew wrote:
> Hello all,
>
> I suggest to think about reversed logic of T.38 processing - IMHO it will
> be better to _disable_ it by default and _enable_ with some variable like
> T38_ENABLE for selected peers or sessions.
>
> Currently we are in process of switching our customers
> from Asterisk-based server to CW-based and we realized that there a
> lot of problems with different CPE gateways: some of them just disconnect
> outgoung call when receive INVITE with T.38, some of them sends
> totally wrong UDPL packets - e.g. yesterday one of our CW-servers was ruined
> by some unknown gateway which results with constant segfault in the following
> function:
>
> #0  0x00455ae1 in opbx_udptl_read (udptl=0x0) at udptl.c:360
> 360             if (seq_no > s->rx_seq_no)
>
> I will try to findout which exact model was the killer and which
> packets it sends, but currently I have to disable UDPTL on this server
> at all till the end of investigation.
> BTW, is their any way to get this data from core dump?
>
> With T38_ENABLE it will be possible to enable T.38 only to _selected_
> gateways which a) really need it   b) has really working T.38
> implementation.
>
> IMHO using session variable is very flexible way because we
> can control T.38 processing in both ways - through SetVar in SIP peer
> definition and through dialplan.
>
> Andrew Checkov
>   
If you can segfault the software there may be a security vulnerability 
as well as a segfault nuisance. We need to sort this out. I thought we 
had chased all these things out of the software.

Steve

_______________________________________________
Callweaver-users mailing list
[email protected]
http://lists.callweaver.org/mailman/listinfo/callweaver-users

Reply via email to