Implement 9260 ack and rtx timers by JoTurk · Pull Request #518 · pion/sctp

JoTurk · 2026-10-01T00:17:57Z

Description

Folds: #403 and #429

on top of them it actually integrates the timers, makes the backoff reset actually work, fixes SACK
ordering, stops T3 after validation simplifies the code and removes unused helpers that were added
in the original PRs.

Reference issue

closes #428

Folds: #403 #429 on top of them it actually integerates the timers, makes the backoff reset actually work, fixes SACK ordering, stops T3 after validation simplifes the code and removes unused helpers that were added in the original PRs. closes: #428 Co-authored-by: philipch07 <philipch07@gmail.com>

codecov · 2026-10-01T00:20:36Z

Codecov Report

❌ Patch coverage is 97.22222% with 1 line in your changes missing coverage. Please review.
✅ Project coverage is 89.47%. Comparing base (e1fbf8b) to head (a5a8fe0).
⚠️ Report is 1 commits behind head on main.

Files with missing lines	Patch %	Lines
association.go	95.23%	1 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #518      +/-   ##
==========================================
+ Coverage   89.04%   89.47%   +0.43%     
==========================================
  Files          56       56              
  Lines        4756     4770      +14     
==========================================
+ Hits         4235     4268      +33     
+ Misses        521      502      -19

Flag	Coverage Δ
go	`89.47% <97.22%> (+0.43%)`	⬆️

Flags with carried forward coverage won't be shown. Click here to find out more.

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

Ibn-Butt-Toota · 2026-10-01T00:22:49Z

+// measureRTT applies RFC 9260 section 6.3.1 C4/C5: sample once per round trip
+// and exclude retransmitted DATA (Karn's algorithm). The caller holds the lock.
+func (a *Association) measureRTT(c *chunkPayloadData, now time.Time) {
+	if c.nSent != 1 || sna32LT(c.tsn, a.minTSN2MeasureRTT) {
+		return
+	}
+	a.minTSN2MeasureRTT = a.myNextTSN
+	rtt := now.Sub(c.since)
+	srtt := a.rtoMgr.setNewRTT(rtt.Seconds() * 1000)
+	a.srtt.Store(srtt)
+	// Keep a windowed minimum so RACK can adapt when the path RTT changes.
+	a.rack.rackMinRTTWnd.Push(now, rtt)
+	a.log.Tracef("[%s] SACK: measured-rtt=%f srtt=%f new-rto=%f",
+		a.name, rtt.Seconds()*1000, srtt, a.rtoMgr.getRTO())
+}


This is really nice, we needed to pull this out a long time ago. Maybe it can even be in a helper for the v2.

oh i didn't write this lol, this was just refactored from main from two places from processSelectiveAck, I just made it into a helper :)

sctp/association.go

Lines 2972 to 2992 in e1fbf8b

// C5) Karn's algorithm: RTT measurements MUST NOT be made using

// packets that were retransmitted (and thus for which it is

// ambiguous whether the reply was for the first instance of the

// chunk or for a later instance)

if sna32GTE(chunkPayload.tsn, a.minTSN2MeasureRTT) {

// Only original transmissions for classic RTT measurement (Karn's rule)

if chunkPayload.nSent == 1 {

a.minTSN2MeasureRTT = a.myNextTSN

rtt := now.Sub(chunkPayload.since).Seconds() * 1000.0

srtt := a.rtoMgr.setNewRTT(rtt)

a.srtt.Store(srtt)

// use a window to determine minRtt instead of a global min

// as the RTT can fluctuate, which can cause problems if going from a

// high RTT to a low RTT.

a.rack.rackMinRTTWnd.Push(now, now.Sub(chunkPayload.since))

a.log.Tracef("[%s] SACK: measured-rtt=%f srtt=%f new-rto=%f",

a.name, rtt, srtt, a.rtoMgr.getRTO())

}

}

Ibn-Butt-Toota · 2026-10-01T00:26:11Z

+	if a.ackState != ackStateDelay {
+		return
+	}


Whoops looks like we forgot to wire this up. Thanks for catching it!

JoTurk requested a review from Ibn-Butt-Toota October 1, 2026 00:27

Ibn-Butt-Toota approved these changes Oct 1, 2026

View reviewed changes

JoTurk merged commit a5a8fe0 into main Oct 1, 2026
20 checks passed

JoTurk deleted the jo/rfc9260-timers branch October 1, 2026 00:36

Navigation Menu

Implement 9260 ack and rtx timers - #518

Implement 9260 ack and rtx timers#518
JoTurk merged 1 commit into
mainfrom
jo/rfc9260-timers

JoTurk commented Oct 1, 2026

Uh oh!

codecov Bot commented Oct 1, 2026 •

edited

Loading

Uh oh!

Ibn-Butt-Toota Oct 1, 2026

Uh oh!

JoTurk Oct 1, 2026

Uh oh!

Ibn-Butt-Toota Oct 1, 2026

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

	// C5) Karn's algorithm: RTT measurements MUST NOT be made using
	// packets that were retransmitted (and thus for which it is
	// ambiguous whether the reply was for the first instance of the
	// chunk or for a later instance)
	if sna32GTE(chunkPayload.tsn, a.minTSN2MeasureRTT) {
	// Only original transmissions for classic RTT measurement (Karn's rule)
	if chunkPayload.nSent == 1 {
	a.minTSN2MeasureRTT = a.myNextTSN
	rtt := now.Sub(chunkPayload.since).Seconds() * 1000.0
	srtt := a.rtoMgr.setNewRTT(rtt)
	a.srtt.Store(srtt)

	// use a window to determine minRtt instead of a global min
	// as the RTT can fluctuate, which can cause problems if going from a
	// high RTT to a low RTT.
	a.rack.rackMinRTTWnd.Push(now, now.Sub(chunkPayload.since))

	a.log.Tracef("[%s] SACK: measured-rtt=%f srtt=%f new-rto=%f",
	a.name, rtt, srtt, a.rtoMgr.getRTO())
	}
	}

Conversation

JoTurk commented Oct 1, 2026

Description

Reference issue

Uh oh!

codecov Bot commented Oct 1, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Codecov Report

Uh oh!

Ibn-Butt-Toota Oct 1, 2026

Choose a reason for hiding this comment

Uh oh!

JoTurk Oct 1, 2026

Choose a reason for hiding this comment

Uh oh!

Ibn-Butt-Toota Oct 1, 2026

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

codecov Bot commented Oct 1, 2026 •

edited

Loading