[Asterisk-code-review] chan iax2: Prevent deadlock between hangup and sending lagrq... (asterisk[master])

[Richard Mudgett]

Richard Mudgett has posted comments on this change.

Change subject: chan_iax2: Prevent deadlock between hangup and sending lagrq/ping
......................................................................


Patch Set 11:

(2 comments)

https://gerrit.asterisk.org/#/c/169/11/channels/chan_iax2.c
File channels/chan_iax2.c:

Line 1706: 	/* Mark pingid as invalid scheduler id. */
         : 	iaxs[callno]->pingid = -1;
> I understood that after send_ping return (and before __send_ping start to e
I've been going back and forth on setting pingid to -1 in send_ping() and in __send_ping().

Moving the setting of pingid to __send_ping() eliminates the locking in send_ping() and practically eliminates the deadlock problem in iax2_destroy_helper() and in make_trunk().  However, there is now the potential for DEV_MODE to assert on a stale id in make_trunk() when it attempts to delete the scheduled ping callback.

Leaving the setting of pingid in send_ping() keeps the deadlock potential in make_trunk() and with no assert potential on a stale pingid.

I think the iaxsl[] locking with make_trunk() to eliminate its deadlock potential needs to be explored.


Line 2076: 	if (ast_sched_find_data(sched, sched_id)) {
This check can fail if callno is zero.  A simple way around the problem is to encode the send_ping/send_lagrq's data parameter as callno + 1 so data will never be NULL.


-- 
To view, visit https://gerrit.asterisk.org/169
To unsubscribe, visit https://gerrit.asterisk.org/settings

Gerrit-MessageType: comment
Gerrit-Change-Id: I03bec1fc8faacb89630269e935fa667c6d6c080c
Gerrit-PatchSet: 11
Gerrit-Project: asterisk
Gerrit-Branch: master
Gerrit-Owner: Y Ateya <y.ateya at starkbits.com>
Gerrit-Reviewer: Ashley Sanders <asanders at digium.com>
Gerrit-Reviewer: George Joseph <george.joseph at fairview5.com>
Gerrit-Reviewer: Joshua Colp <jcolp at digium.com>
Gerrit-Reviewer: Matt Jordan <mjordan at digium.com>
Gerrit-Reviewer: Richard Mudgett <rmudgett at digium.com>
Gerrit-Reviewer: Y Ateya <y.ateya at starkbits.com>
Gerrit-HasComments: Yes