1860072 - DTLS transport might be connected twice to DataChannelConnection

Status: RESOLVED FIXED

(Core :: WebRTC: Networking, defect, P2)

Description

[Victor Boivie]

User Agent: Mozilla/5.0 (Macintosh; Intel Mac OS X 10_15_7) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/118.0.0.0 Safari/537.36

Steps to reproduce:

• Go to http://meet.new

Actual results:

The DTLS transport is connected twice to the DataChannelConnection, which results in incoming DTLS packets (which are SCTP packets) are provided twice to usrsctp.

You will notice it mostly if doing a SCTP packet capture, seeing the incoming DATA are seen twice, and every outgoing SACK has the "duplicate TSN" field set.

This affects the congestion control algorithm and results in SACKs being sent for every DATA, instead of for every second data.

This is most likely triggered by Meet operating on the SDP. This doesn't always happen; Most simple examples (like https://tomashubelbauer.github.io/webrtc-data-channel-demo/) doesn't trigger it.

Expected results:

The SCTP transport is properly set up, and sent messages from the server are received once by the client's SCTP implementation.

Comment 1

[Byron Campen [:bwc]]

It is entirely possible that this is resulting in significant performance problems, particularly with peer connections with lots of renegotiations.

Comment 2

[Byron Campen [:bwc]]

Try runs:
https://treeherder.mozilla.org/jobs?repo=try&revision=2497f03c5a71b5eb957237c5ebab562fa92af214
https://treeherder.mozilla.org/jobs?repo=try&revision=d2e7515fa3f7dca2ce85831bc4c913d7134cf010
https://treeherder.mozilla.org/jobs?repo=try&revision=807c40471900e095468a13cff0e6991f6fdcf149

Comment 3

[Byron Campen [:bwc]]

Attachment: Bug 1860072: Avoid duplicate connect calls to the transport handler. r?mjf

Comment 4

[Byron Campen [:bwc]]

Try looks fine.

Comment 5

[Pulsebot]

Pushed by [e-mail]:
https://hg.mozilla.org/integration/autoland/rev/e3dad51c31b1
Avoid duplicate connect calls to the transport handler. r=mjf

Comment 6

[Cristina Horotan [:chorotan]]

https://hg.mozilla.org/mozilla-central/rev/e3dad51c31b1

Comment 7

[Brindusa Tot, Desktop QA]

Hi,
Is manual testing required here? If yes, can you provide us some details how to verify the SCTP transport?

Comment 8

[Victor Boivie]

I have done manual verification of that one as part of also verifying bug 1464914.

Comment 9

[Byron Campen [:bwc]]

Looks like we're already covered here.

Comment 10

[Timea Zsoldos [:zstimi/tzsoldos], Desktop QA]

Based on comment 8 I will set the status flag as verified on Firefox 121.