Aborted with sctp_timeout_handler: tmr->self corrupted, but tmr->self is null in the core dump about usrsctp HOT 6 OPEN

This has now happened three more times, with the same symptoms. Please let me know any further information you need.

from usrsctp.

How often does it happen? Do you have a way to reproduce this?

from usrsctp.

Sadly, I don't have a way to reproduce it reliably. It happens once every few days across our fleet of meet.jit.si production servers.

I have the core dumps so if there's any other information I can share that would be useful let me know.

See also #676 which is a rarer crash (I've only seen it once so far) but I suspect has the same root cause, and may be more revealing?

from usrsctp.

I've finally managed to extract a Java heap dump corresponding to this core dump, which lets me correlate my user-level objects and logs with the usrsctp objects. (For most of the crashes I'm running into a Java bug which is preventing this heap dump from being created.)

In this case it appears that the socket with the crashing timer received an SCTP packet just under 200 ms before the crash. Five other SCTP sockets received a packet in the interval between that packet receipt and the crash.

The crashing timer appears to be a SCTP_TIMER_TYPE_RECV timer.

from usrsctp.

I've finally managed to extract a Java heap dump corresponding to this core dump, which lets me correlate my user-level objects and logs with the usrsctp objects. (For most of the crashes I'm running into a Java bug which is preventing this heap dump from being created.)

In this case it appears that the socket with the crashing timer received an SCTP packet just under 200 ms before the crash. Five other SCTP sockets received a packet in the interval between that packet receipt and the crash.

The crashing timer appears to be a SCTP_TIMER_TYPE_RECV timer.

That is the delay ACK timer and normally expires at 200ms. I'm looking at a timer related problem where a socket is freed twice. Once I have a fix committed, you could try it, if it fixes also your issue. I'll let you know once I have solved the issue.

from usrsctp.

Any news on this?

from usrsctp.