[asterisk-dev] Channel Read Queue locking

[Russell Bryant]

On 02/17/2010 02:32 PM, Damien Wedhorn wrote:
> I'm about to bring the readq locking patch
> (https://issues.asterisk.org/view.php?id=14998) up to date and put in on
> the review board. However a couple of issues before I do.
>
> Apparently it will fix/help with some chan_local issues
> (https://reviewboard.asterisk.org/r/219/). However, given the
> fundamental change, it could also lead to regressions elsewhere. Aragon
> has indicated that he expects this will be applied to all branches
> (based on feedback from tilghman's feedback.
>
> While this would normally just go to trunk, if it does fix bugs in
> releases, should we apply it to them and try to fix up anything else it
> breaks, or just apply it to trunk? Arguably, we can leave the chan_*
> locks as they are for current releases and just change the channels.c
> code which should limit possible issues, and only change the chan_*
> locking in trunk.

This is _way_ too invasive for me to be comfortable doing this is any 
release branch of Asterisk.

> The second thing is that the patch as it stands is only a half fix.
> Locking is still required to fiddle with the hangup stuff. I was going
> to modify the soft_hangup write stuff to be atomic and as such remove
> all of the requirements for channel locking before playing with readq.
> Should this idea be integrated into this patch or left for a subsequent
> patch?

I haven't looked at your patch in a while, so I don't remember the 
details of this.  However, based on this description, it sounds like 
something that should be addressed as a part of the patch and not as 
something separate.

-- 
Russell Bryant
Digium, Inc. | Engineering Manager, Open Source Software
445 Jan Davis Drive NW - Huntsville, AL 35806 - USA
www.digium.com -=- www.asterisk.org -=- blogs.asterisk.org