[asterisk-bugs] [JIRA] (ASTERISK-28472) Asterisk Crash

Kevin Harwell (JIRA) noreply at issues.asterisk.org
Wed Jul 10 18:29:47 CDT 2019


    [ https://issues.asterisk.org/jira/browse/ASTERISK-28472?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=247564#comment-247564 ] 

Kevin Harwell commented on ASTERISK-28472:
------------------------------------------

Thanks for the added info. srtp related problems, and crashes can be quite challenging to figure out unfortunately. 
{quote}
> 1. Can you think of anything that changed just before noticing the crash? For instance, had you upgraded Asterisk, or libsrtp?

The before was Asterisk 13.25 with chan_sip, so the changes are quite substantial and probably not related 
{quote}
Ha yeah quite the substantial change. You are correct that's not gonna help us here.
{quote}
> what kind of endpoints/phones are used in the crash scenario?

Could be anything really - our customers can connect whatever SIP compatible device they want. I can try to figure out which exact models were in use at the time of the crash if you think it'll be helpful.
{quote}
I'll leave whether or not to track down that info up to you (it may or may not be helpful). As mentioned srtp problems can be a pain to figure out. Any little data point might help with a solution and/or workaround. In this case as the crash seems to be occurring in the "unprotect" function it's possible the endpoint is sending a bad packet at times, which may cause the crash. If you knew it was the same endpoint involved each time it might point us in a direction to try (for instance, upgrade phone's firmware, avoid using those types of endpoints if possible, easier to packet capture only that endpoint, etc...).
{quote}
Enabling level 5 debug might be a tall order - this is a pretty busy production server during peak hours.
{quote}
No worries quite understandable. In any case I'd double check the log on each crash. Logging is done in a separate thread so it's possible more logging could be written prior to a crash vs another crash. It's all about the timing. So perhaps in one instance something might turn up.
{quote}
> Lastly, if you can, install srtp with debug symbols.

Yes... though I'm unsure how to go about doing that. Can you provide instructions on how I'd go about doing so? I've compiled libsrtp from source manually.
{quote}
hrm, how did you configure and build it? Something like the following?
{noformat}
$ ./configure --enable-openssl
$ make shared_library
{noformat}
If so try something like the following then (or similar - the main thing is adding the -g3 to CFLAGS):
{noformat}
$ CFLAGS='-fPIC -Wall -g3 -funroll-loops' ./configure --enable-openssl
$ make shared_library
{noformat}
One more thing. You could try upgrading your version of srtp. 1.5.4 to something higher. I believe 1.5.4 is the minimum version we support. We should also allow higher versions as well.

> Asterisk Crash
> --------------
>
>                 Key: ASTERISK-28472
>                 URL: https://issues.asterisk.org/jira/browse/ASTERISK-28472
>             Project: Asterisk
>          Issue Type: Bug
>      Security Level: None
>          Components: General
>    Affects Versions: 16.4.0
>         Environment: Amazon Linux 2
>            Reporter: Jonas Swiatek
>            Assignee: Unassigned
>            Severity: Critical
>              Labels: pjsip
>         Attachments: core-2019-07-10T17-36-23+0000-brief.txt, core-2019-07-10T17-36-23+0000-full.txt, core-2019-07-10T17-36-23+0000-locks.txt, core-2019-07-10T17-36-23+0000-thread1.txt, core-brief.txt, core-full.txt, core-locks.txt, core-thread1.txt, extensions.conf, pjsip.conf
>
>
> We're seeing this crash every now and then - last time was around 3 weeks ago, on Asterisk 16.3, though we've upgraded to 16.4 since then, and enabled BETTER_BACKTRACES and DONT_OPTIMIZED to try and get to the bottom of it.
> I'm at a loss for what might be causing it. When this server crashed it was serving 6 channels, and the only exotic thing being used by any of them were ARI, but I don't believe any of the active channels were actually in the Stasis application.
> I'm not really qualified to read the crash dump, but there is some mention of libsrtp in one of the files, so I've included which version of that we're using (1.5.4, as recommended on the wiki).
> Any insight would be greatly appreciated, and I can provide whatever else information might be valuable.



--
This message was sent by Atlassian JIRA
(v6.2#6252)



More information about the asterisk-bugs mailing list