|Summary:||ASTERISK-16161: [patch] Deadlock in chan_local.c causes crash.|
|Reporter:||Barry GATEFAIT (bwg)||Labels:|
|Date Opened:||2010-05-27 18:10:28||Date Closed:||2011-06-07 14:00:32|
|Environment:||Attachments:||( 0) gdb_BWG_28mai10.txt|
( 1) localchan_deadlock.diff.txt
( 2) locks_BWG_28mai10.txt
( 3) Pb_locks_29mai10.txt
A mutual exclusion occured in chan_local.c causing a system crash.
Have attached a asterisk CLI trace, gdb trace and lock trace resulting.
|Comments:||By: Barry GATEFAIT (bwg) 2010-05-27 21:18:49|
After further research, the locks occured in (chan_local.c).
By: Leif Madsen (lmadsen) 2010-06-02 10:01:14
Thanks for the debug information. Acknowledging this issue.
By: Leif Madsen (lmadsen) 2010-06-03 11:09:01
There have been deadlock fixes since the revision you state (in chan_local), so please test a newer revision -- ideally the latest revision.
Also, which branch are you testing this against? You haven't stated that in your bug report.
By: Barry GATEFAIT (bwg) 2010-06-06 17:05:22
I'm working on SVN-TRUNK.
following your advise, now testing latest revision (267714) with local patches ...
Will give news on the outcome.
By: Alec Davis (alecdavis) 2010-06-12 03:29:51
uploaded localchan_deadlock.diff.txt against Asterisk SVN-trunk-r270042M
The deadlock avoidance method is simlar to chan_dahdi.
This should hopefully avoid the deadlock pairs you reported in 'locks_BWG_28mai..'
Please provide details how to force this deadlock. Dialplan etc.
By: Tilghman Lesher (tilghman) 2010-07-22 11:22:37
This shouldn't be necessary, since the ordained channel locking order is used: channel first, then pvt. It should only be necessary to do deadlock avoidance when the reverse locking order is used.
By: Tilghman Lesher (tilghman) 2010-07-29 16:16:02
bwg: have the problems gone away since you updated to a later copy of trunk?
By: Barry GATEFAIT (bwg) 2010-07-29 17:46:00
We, indeed, updated to SVN-trunk-r273055 and since no deadlock occured.
Thank you for your advise ! Cheers.