Race condition for chunkmap access in payloadQueue

[cohosh]

Your environment.

• Version: v1.7.5
• Browser: N/A (Snowflake)
• Other Information - see Tor ticket #33211

What did you do?

There are more details in the ticket, but we found our application going into a CPU-intensive infinite loop. Profiling pointed us towards the markAllToRetransmit function in payload_queue.go: https://github.com/pion/sctp/blob/master/payload_queue.go#L163

I'm not sure whether the infinite loop behaviour we saw was due to this race condition but it seems plausible.

[enobufs]

@cohosh Thanks for your report and the PR.
I recently fixed a bug (#109) that was causing 100% CPU usage which was fixed in v1.7.5. Looking at the Tor ticket #33211, at the very bottom, I see the identical logs reported in #109 and I am wondering if that has already been solved...

You reported above that your environment was pion/sctp v1.7.5 which is supposed to have the fix. Could you please double-check if the Snowflake version experiencing the problem was actually using pion/sctp v1.7.5? (The last tag for pion/webrtc is v2.2.0 which still uses pion/sctp v1.7.4).

[Sean-Der]

Hey @cohosh @enobufs!

I tagged v2.2.1 I am working on profiling stuff (ICE performance) and coincidentally tagged it.

thanks!

[cohosh]

Hey @enobufs I think you're right. I wasn't actually able to reproduce the bug myself using the latest version. I'm guessing this is the issue. Thanks for that!

[enobufs]

@cohosh Let me close this case. (If it happens with pion/[email protected], please reopen this.)