|Summary:||ASTERISK-11860: [patch] SS7 support for chan_zap is missing clean-up code|
|Reporter:||Moises Silva (moy)||Labels:|
|Date Opened:||2008-04-15 20:28:23||Date Closed:||2011-06-07 14:07:54|
|Environment:||Attachments:||( 0) chan_zap.c-ss7-rev-114164.patch|
|Description:||I was adding code to chan_zap for MFR2 support and I took as example the SS7 stuff. I realized it seems to be missing proper thread and CLI clean up. Is not unregistering the CLI commands it register on load time and is not shutting down the threads with pthread_cancel. The attached patch address both issues.|
|Comments:||By: Moises Silva (moy) 2008-04-15 20:32:39|
Hum, just wondering, any particular reason I did not get Karma for my last feature submission: 11268? :-(
By: Moises Silva (moy) 2008-04-15 22:11:15
do you think we should be disabling cancelation during event handling?
By: Jason Parker (jparker) 2008-05-06 14:22:45
mattf, any thoughts?
By: Matthew Fredrickson (mattf) 2008-05-06 14:47:31
While this is a good thing IMHO, and a step in the right direction, there are deeper issues here with the unload that will fail not just for SS7 linksets but also for PRIs.
chan_zap isn't in a state where unload can be done safely with PRI and SS7 channels types (and probably analog channels as well), so if you are playing with this, you will find other problems (besides the very obvious things such as application unregistering, etc).
By: Moises Silva (moy) 2008-05-06 21:38:41
weeeell it was not my intention to open a can of worms, but ... indeed even unloading when only analog channels are present can cause leaks, since some memory seems to be allocated in do_monitor without using thread keys that can be destroyed when the thread is cancelled.
In anycase, I think at least this should be commited and then further analysis can be made for chan_zap unloading, I just looked at the code and that bug appeared in front of my eyes, not that I think is the only one on chan_zap :-)
By: Moises Silva (moy) 2008-06-19 09:12:01
if nothing will be done, probably you just should close this one and forget about it :)
By: Leif Madsen (lmadsen) 2008-11-21 09:11:08.000-0600
Suspending this issue per the reporter as there has been no movement on this bug in quite some time. Please feel free to reopen this issue should we have a way to move this forward.