Bug #2880
closed
OsmoMSC never releases MO call if MNCC doesn't respond to SETUP
Added by laforge over 6 years ago.
Updated almost 5 years ago.
Description
It seems like the call is never released in case the MNCC interface never responds to a MO SETUP.
See attached trace.
Terminating the MNCC handler doesn't seem to change this either, at leaset not on the CC signalling side. The log prints:
<0004> mncc_sock.c:85 MNCC Socket has LOST connection
<0001> gsm_04_08.c:191 Clearing all currently active transactions!!!
<0007> msc_mgcp.c:1049 (subscriber:IMSI:262420000000026) invalid mgcp context, call release failed.
<0001> gsm_04_08.c:1307 (ti 08 sub IMSI:262420000000026) new state INITIATED -> NULL
however, no attempt is made to send a CC release to the MS, nor is the SCCP connection cleared.
See MSC_Tests.TC_mo_setup_and_nothing
Files
- Related to Bug #3125: testcase for fixed "osmo-msc crashes while handling a call" added
There is a patch series coming to refactor the MSC's conn FSM to discard connections as soon as the use count reaches zero.
It is worth checking this error separately, i.e. make sure that an MNCC SETUP timeout indeed results in a use count decrement.
The FSM does not handle CC timeouts, that is still done the "old" way within our CC implementation.
Might be worth a thought to implement an actual CC FSM as well, as a child of the subscr_conn FSM, with timeouts for each state except the active call.
- Assignee changed from 4368 to neels
(I'm not that familiar with MNCC setups, maybe someone else could investigate this one?)
(I have since then become familiar with MNCC setups and can indeed investigate this one. Once I get to it.)
- Status changed from New to Resolved
TC_mo_setup_and_nothing passes.
I hope that means the issue is resolved.
Also available in: Atom
PDF