endpoint address latching issue

[achalkov]

Hello!
Case:
sip schema: a - opensips v.2.1 (with rtpproxy) - b
media schema: a - rtpproxy (commit 7c27f22) - b
call schema:
step 1. opensips -> INVITE -> B (real IP in SDP)
step 2. B -> 200 -> opensips (real IP in SDP(1))
step 3. B (from real IP(1)) -> RTP -> rtpproxy
After step 3 session establishing successfully and RTP goes to both sides.

step 4. B -> Re-INVITE -> opensips (local IP in SDP(2))
step 5. opensips -> 200 -> B (same IP as in step 1 in SDP)
step 6. B (from real IP(3)) -> RTP -> rtpproxy
After step 5 rtpproxy pre-fills local address (2), which opensips tells him via UR command, but on step 6 when RTP-packets arrives from real IP (3) - rtpproxy didn't latch new real address and ignore all RTP from it.

Earlier i tried to reproduce issue on older version of rtpproxy (commit f9d2b03) and it successfully latched IP (3) after step 6, so i think something went wrong with address latching mechanics between commits f9d2b03 and 7c27f22

Issue always reproducing on 7c27f22 and never reproducing on f9d2b03 (tested on more than 10 instances of each of both versions)

[sobomax]

Thanks for a detailed debug! We are looking into it.

[sobomax]

@achalkov Artem, please try attached patch and let us know if it helps with the issue in question. Thanks!
47.diff.txt

[achalkov]

@sobomax Thanks for fast workaround. I'll try check how rtpproxy works with this patch. Unfortunally, it might take some time (few days or a week). I'll let you know when i will have some results.

[achalkov]

Sorry for long wait, i had to make sure that the build with patch is stable. Unfortunately, i could not reproduce initial case (originally it was on customer's UAC, which not available for us now) , but the build seems stable. So, seems like all is ok now.