[SR-Users] Keep-Alive in dialog "freeing a free fragment"

[Daniel-Constantin Mierla]

Hello,

the respective patch was done to fix generating CDRs on local BYE. Maybe
cnxcc is locking slots as well, I am not familiar with the module at all.

Perhaps the best is to change the mutexes to be re-entrant, to avoid
such conflicts between dialog and upper layer modules relying on it. I
will push a patch for it.

Cheers,
Daniel

On 16/06/15 13:59, Dirk Teurlings - SIGNET B.V. wrote:
> Hello,
>
> Following up on this issue, though KA seems to work fine now. We're
> running into another problem. We use a fork of CNXCC that terminates
> all calls when the user doesn't have enough credit. But with these
> patches that doesn't work anymore and the current thread even locks up
> completely.
>
>
> We've tracked it down to these changes:
>
> diff --git a/modules/dialog/dlg_req_within.c
> b/modules/dialog/dlg_req_within.c
> index 206d13e..5668c3c 100644
> --- a/modules/dialog/dlg_req_within.c
> +++ b/modules/dialog/dlg_req_within.c
> @@ -214,7 +214,7 @@ void bye_reply_cb(struct cell* t, int type, struct
> tmcb_params* ps){
>                         unref++;
>                 }
>                 /* dialog terminated (BYE) */
> -               run_dlg_callbacks( DLGCB_TERMINATED, dlg, ps->req,
> ps->rpl, DLG_DIR_NONE, 0);
> +               run_dlg_callbacks( DLGCB_TERMINATED_CONFIRMED, dlg,
> ps->req, ps->rpl, DLG_DIR_NONE, 0);
>
>                 LM_DBG("first final reply\n");
>                 /* derefering the dialog */
> @@ -521,6 +525,9 @@ int dlg_bye_all(struct dlg_cell *dlg, str *hdrs)
>         str all_hdrs = { 0, 0 };
>         int ret;
>
> +       /* run dialog terminated callbacks */
> +       run_dlg_callbacks( DLGCB_TERMINATED, dlg, NULL, NULL,
> DLG_DIR_NONE, 0);
> +
>         if ((build_extra_hdr(dlg, hdrs, &all_hdrs)) != 0)
>         {
>                 LM_ERR("failed to build dlg headers\n");
>
>
>
>
>
> This is a diff between 4.2.3 and 4.2.5, once reverted, everthing works
> fine.
>
> Can you tell me the reason for these additions? What is the expected
> difference in behaviour?
>
>
> Cheers,
> Dirk
>
>
> On 12-05-15 10:49, Daniel-Constantin Mierla wrote:
>> Hello,
>>
>> master is still the code to be 4.3, we will make a dedicated branch
>> sometime soon.
>>
>> The patch will be backported to 4.2 as well. Thanks for helping to
>> troubleshoot and testing.
>>
>> Cheers,
>> Daniel
>>
>
> _______________________________________________
> SIP Express Router (SER) and Kamailio (OpenSER) - sr-users mailing list
> sr-users at lists.sip-router.org
> http://lists.sip-router.org/cgi-bin/mailman/listinfo/sr-users

-- 
Daniel-Constantin Mierla
http://twitter.com/#!/miconda - http://www.linkedin.com/in/miconda
Book: SIP Routing With Kamailio - http://www.asipto.com