[asterisk-dev] [Code Review] Convert device state callbacks to ao2 objects to fix a deadlock.

[Jeff Peeler]

> On 2011-01-10 02:42:10, schmidts wrote:
> > /branches/1.6.2/main/pbx.c, line 3929
> > <https://reviewboard.asterisk.org/r/1072/diff/1/?file=14863#file14863line3929>
> >
> >     sorry if this is a stupid question but why do you set the refcounter for this container one down?
> >     
> >     you dont use the same container to iterate through so IMHO you dont need to reset the refcounter.
> >     
> >     the iterator is reinitialised below this row for the callback container of this hint so i dont think you need this.
> >     
> >     be honest if i am wrong ;)

You are correct, removed!


> On 2011-01-10 02:42:10, schmidts wrote:
> > /branches/1.6.2/main/pbx.c, line 4132
> > <https://reviewboard.asterisk.org/r/1072/diff/1/?file=14863#file14863line4132>
> >
> >     IMHO you dont need to allocate this container with 563 buckets, cause you will in fact only need a link list. 
> >     The normal callback container should have buckets, even if they are useless cause all general callbacks have id=0. 
> >     maybe you could change this cause with this patch general and linked callbacks are split up. 
> >     the id=0 was only used to find these general callbacks. 
> >     but again the general callbacks and also the linked only need a link list ;)

I think I'll just keep them separate for now and reduce its size.


- Jeff


-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://reviewboard.asterisk.org/r/1072/#review3086
-----------------------------------------------------------


On 2011-01-07 15:06:45, Jeff Peeler wrote:
> 
> -----------------------------------------------------------
> This is an automatically generated e-mail. To reply, visit:
> https://reviewboard.asterisk.org/r/1072/
> -----------------------------------------------------------
> 
> (Updated 2011-01-07 15:06:45)
> 
> 
> Review request for Asterisk Developers.
> 
> 
> Summary
> -------
> 
> Lock scenario:
> Thread 1
>  holds ast_rdlock_contexts &conlock
>  holds handle_statechange hints
>  holds handle_statechange hint
>   waiting for cb_extensionstate
>    Locked Here: chan_sip.c line 7428 (find_call)
> Thread 2
>  holds handle_request_do &netlock
>  holds find_call sip_pvt_ptr
>   waiting for ast_rdlock_contexts &conlock
>    Locked Here: pbx.c line 9911 (ast_rdlock_contexts)
> 
> Chan_sip has an established locking order of locking the sip_pvt and then getting the context lock. So the as stated by the summary, the operations in thread 2 have been modified to no longer require the context lock.
> 
> 
> This addresses bug 18310.
>     https://issues.asterisk.org/view.php?id=18310
> 
> 
> Diffs
> -----
> 
>   /branches/1.6.2/main/pbx.c 301088 
> 
> Diff: https://reviewboard.asterisk.org/r/1072/diff
> 
> 
> Testing
> -------
> 
> one47 has reported on the issue that it has been going well so far (for 18 hours):
> https://issues.asterisk.org/view.php?id=18310#130297
> 
> 
> Thanks,
> 
> Jeff
> 
>

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.digium.com/pipermail/asterisk-dev/attachments/20110112/5d877c81/attachment.htm>