[asterisk-dev] [Code Review] Resolve dnsmgr memory corruption caused by chan_iax2

Russell Bryant russell at digium.com
Fri Oct 1 15:25:23 CDT 2010


-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://reviewboard.asterisk.org/r/965/#review2804
-----------------------------------------------------------


I just found another bug I need to fix.  I didn't do a test that included outgoing registrations on the box with dnsmgr enabled.  This exact same bug exists for the iax2_registry structure.

- Russell


On 2010-10-01 15:11:41, Russell Bryant wrote:
> 
> -----------------------------------------------------------
> This is an automatically generated e-mail. To reply, visit:
> https://reviewboard.asterisk.org/r/965/
> -----------------------------------------------------------
> 
> (Updated 2010-10-01 15:11:41)
> 
> 
> Review request for Asterisk Developers.
> 
> 
> Summary
> -------
> 
> chan_iax2 registered a stack buffer with the dnsmgr, which stores it off to keep it up to date using periodic DNS lookups.  This is bad and led to bizarre failures.  
> 
> This was a regression in the merge of IPv6 support.  The dnsmgr API now takes an ast_sockaddr instead of a sockaddr_in structure.  To fix this, the address structure in the iax_peer structure needed to be changed to an ast_sockaddr.  The patch is pretty much fixing up all of the ripple effects of changing that part of the peer structure.
> 
> 
> This addresses bug 17902.
>     https://issues.asterisk.org/view.php?id=17902
> 
> 
> Diffs
> -----
> 
>   /branches/1.8/channels/chan_iax2.c 289543 
> 
> Diff: https://reviewboard.asterisk.org/r/965/diff
> 
> 
> Testing
> -------
> 
> I set up a few peers using dnsmgr in both IAX2 and SIP and verified that the addresses were populated properly.  I tested an IAX2 registration, including the code that pulls the registration data back out of astdb.  I also ran an IAX2 call.  I did all of this under valgrind and found no problems.
> 
> 
> Thanks,
> 
> Russell
> 
>




More information about the asterisk-dev mailing list