| Summary: | ASTERISK-25163: Deadlock in chan_sip between reload of sip peer container and MWI Stasis callback | ||
| Reporter: | Dmitriy Serov (Demon) | Labels: | |
| Date Opened: | 2015-06-16 03:13:03 | Date Closed: | 2015-06-21 09:19:51 |
| Priority: | Major | Regression? | Yes |
| Status: | Closed/Complete | Components: | Channels/chan_sip/General |
| Versions: | 13.4.0 | Frequency of Occurrence | Frequent |
| Related Issues: | |||
| Environment: | Attachments: | ( 0) 2015_06_18.backtrace-threads.txt ( 1) backtrace-threads.txt | |
| Description: | Asterisk periodic hangs.
UDP Recv-Q greatly exceeds zero. No errors in log (like DNS error, function getaddr). used cdr_mysql, peers in text config, chan_sip & res_pjsip on different ports. | ||
| Comments: | By: Dmitriy Serov (Demon) 2015-06-16 03:14:08.979-0500 Hangs backtrace By: Dmitriy Serov (Demon) 2015-06-17 06:57:27.735-0500 The reason may be a high workload mariadb. For example, in the case of the db backup every night. Although, during day occurs, but much less frequently. After abandoning the use of res_pjsip (but the module is still loaded), the situation began to occur less frequently. By: Dmitriy Serov (Demon) 2015-06-18 03:24:21.916-0500 just one more backtrace By: Matt Jordan (mjordan) 2015-06-18 16:12:44.485-0500 This actually is a deadlock in {{chan_sip}}, due to an inversion in locking between the Stasis MWI callback and the act of reloading. By: Joshua C. Colp (jcolp) 2015-06-20 19:50:32.877-0500 I've put a change up at https://gerrit.asterisk.org/673 which should resolve this issue. | ||