config tweaks for batching

This reverts commit fa098c551a.

.github/workflows/gofmt.yml

@@ -14,7 +14,7 @@ jobs:
     runs-on: ubuntu-latest
     steps:
 
-    - uses: actions/checkout@v5
+    - uses: actions/checkout@v4
 
     - uses: actions/setup-go@v6
       with:

.github/workflows/release.yml

@@ -10,7 +10,7 @@ jobs:
     name: Build Linux/BSD All
     runs-on: ubuntu-latest
     steps:
-      - uses: actions/checkout@v5
+      - uses: actions/checkout@v4
 
       - uses: actions/setup-go@v6
         with:
@@ -24,7 +24,7 @@ jobs:
           mv build/*.tar.gz release
 
       - name: Upload artifacts
-        uses: actions/upload-artifact@v5
+        uses: actions/upload-artifact@v4
         with:
           name: linux-latest
           path: release
@@ -33,7 +33,7 @@ jobs:
     name: Build Windows
     runs-on: windows-latest
     steps:
-      - uses: actions/checkout@v5
+      - uses: actions/checkout@v4
 
       - uses: actions/setup-go@v6
         with:
@@ -55,7 +55,7 @@ jobs:
           mv dist\windows\wintun build\dist\windows\
 
       - name: Upload artifacts
-        uses: actions/upload-artifact@v5
+        uses: actions/upload-artifact@v4
         with:
           name: windows-latest
           path: build
@@ -66,7 +66,7 @@ jobs:
       HAS_SIGNING_CREDS: ${{ secrets.AC_USERNAME != '' }}
     runs-on: macos-latest
     steps:
-      - uses: actions/checkout@v5
+      - uses: actions/checkout@v4
 
       - uses: actions/setup-go@v6
         with:
@@ -104,7 +104,7 @@ jobs:
           fi
 
       - name: Upload artifacts
-        uses: actions/upload-artifact@v5
+        uses: actions/upload-artifact@v4
         with:
           name: darwin-latest
           path: ./release/*
@@ -124,11 +124,11 @@ jobs:
       # be overwritten
       - name: Checkout code
         if: ${{ env.HAS_DOCKER_CREDS == 'true' }}
-        uses: actions/checkout@v5
+        uses: actions/checkout@v4
 
       - name: Download artifacts
         if: ${{ env.HAS_DOCKER_CREDS == 'true' }}
-        uses: actions/download-artifact@v6
+        uses: actions/download-artifact@v4
         with:
           name: linux-latest
           path: artifacts
@@ -160,10 +160,10 @@ jobs:
     needs: [build-linux, build-darwin, build-windows]
     runs-on: ubuntu-latest
     steps:
-      - uses: actions/checkout@v5
+      - uses: actions/checkout@v4
 
       - name: Download artifacts
-        uses: actions/download-artifact@v6
+        uses: actions/download-artifact@v4
         with:
           path: artifacts
 

.github/workflows/smoke-extra.yml

@@ -20,7 +20,7 @@ jobs:
     runs-on: ubuntu-latest
     steps:
 
-    - uses: actions/checkout@v5
+    - uses: actions/checkout@v4
 
     - uses: actions/setup-go@v6
       with:

.github/workflows/smoke.yml

@@ -18,7 +18,7 @@ jobs:
     runs-on: ubuntu-latest
     steps:
 
-    - uses: actions/checkout@v5
+    - uses: actions/checkout@v4
 
     - uses: actions/setup-go@v6
       with:

.github/workflows/test.yml

@@ -18,7 +18,7 @@ jobs:
     runs-on: ubuntu-latest
     steps:
 
-    - uses: actions/checkout@v5
+    - uses: actions/checkout@v4
 
     - uses: actions/setup-go@v6
       with:
@@ -32,7 +32,7 @@ jobs:
       run: make vet
 
     - name: golangci-lint
-      uses: golangci/golangci-lint-action@v9
+      uses: golangci/golangci-lint-action@v8
       with:
         version: v2.5
 
@@ -45,7 +45,7 @@ jobs:
     - name: Build test mobile
       run: make build-test-mobile
 
-    - uses: actions/upload-artifact@v5
+    - uses: actions/upload-artifact@v4
       with:
         name: e2e packet flow linux-latest
         path: e2e/mermaid/linux-latest
@@ -56,7 +56,7 @@ jobs:
     runs-on: ubuntu-latest
     steps:
 
-    - uses: actions/checkout@v5
+    - uses: actions/checkout@v4
 
     - uses: actions/setup-go@v6
       with:
@@ -77,7 +77,7 @@ jobs:
     runs-on: ubuntu-latest
     steps:
 
-    - uses: actions/checkout@v5
+    - uses: actions/checkout@v4
 
     - uses: actions/setup-go@v6
       with:
@@ -98,7 +98,7 @@ jobs:
         os: [windows-latest, macos-latest]
     steps:
 
-    - uses: actions/checkout@v5
+    - uses: actions/checkout@v4
 
     - uses: actions/setup-go@v6
       with:
@@ -115,7 +115,7 @@ jobs:
       run: make vet
 
     - name: golangci-lint
-      uses: golangci/golangci-lint-action@v9
+      uses: golangci/golangci-lint-action@v8
       with:
         version: v2.5
 
@@ -125,7 +125,7 @@ jobs:
     - name: End 2 end
       run: make e2evv
 
-    - uses: actions/upload-artifact@v5
+    - uses: actions/upload-artifact@v4
       with:
         name: e2e packet flow ${{ matrix.os }}
         path: e2e/mermaid/${{ matrix.os }}

CHANGELOG.md

@@ -7,64 +7,12 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
 
 ## [Unreleased]
 
-## [1.10.0] - ????
-
-### Added
-
-- PKCS11 support for P256 keys when built with `pkcs11` tag (#1153)
-- ASN.1 based v2 nebula certificates with support for ipv6 and multiple ip addresses.
-  Certificates now have a unified interface for external implementations. (#1212, #1216, #1345)
-  **TODO: External documentation link!**
-- Add the ability to mark packets on linux to better target nebula packets in iptables/nftables. (#1331)
-- Add ECMP support for `unsafe_routes`. (#1332)
-
 ### Changed
 
 - `default_local_cidr_any` now defaults to false, meaning that any firewall rule
   intended to target an `unsafe_routes` entry must explicitly declare it via the
   `local_cidr` field. This is almost always the intended behavior. This flag is
-  deprecated and will be removed in a future release. (#1373)
-
-### Fixed
-
-- Fix moving a udp address from one vpn address to another in the `static_host_map`
-  which could cause rapid re-handshaking with an incorrect remote. (#1259)
-- Improve smoke tests in environments where the docker network is not the default. (#1347)
-
-## [1.9.7] - 2025-10-10
-
-### Security
-
-- Fix an issue where Nebula could incorrectly accept and process a packet from an erroneous source IP when the sender's
-  certificate is configured with unsafe_routes (cert v1/v2) or multiple IPs (cert v2). (#1494)
-
-### Changed
-
-- Disable sending `recv_error` messages when a packet is received outside the allowable counter window. (#1459)
-- Improve error messages and remove some unnecessary fatal conditions in the Windows and generic udp listener. (#1543)
-
-## [1.9.6] - 2025-7-15
-
-### Added
-
-- Support dropping inactive tunnels. This is disabled by default in this release but can be enabled with `tunnels.drop_inactive`. See example config for more details. (#1413)
-
-### Fixed
-
-- Fix Darwin freeze due to presence of some Network Extensions (#1426)
-- Ensure the same relay tunnel is always used when multiple relay tunnels are present (#1422)
-- Fix Windows freeze due to ICMP error handling (#1412)
-- Fix relay migration panic (#1403)
-
-## [1.9.5] - 2024-12-05
-
-### Added
-
-- Gracefully ignore v2 certificates. (#1282)
-
-### Fixed
-
-- Fix relays that refuse to re-establish after one of the remote tunnel pairs breaks. (#1277)
+  deprecated and will be removed in a future release.
 
 ## [1.9.4] - 2024-09-09
 
@@ -723,11 +671,7 @@ created.)
 
 - Initial public release.
 
-[Unreleased]: https://github.com/slackhq/nebula/compare/v1.10.0...HEAD
-[1.10.0]: https://github.com/slackhq/nebula/releases/tag/v1.10.0
-[1.9.7]: https://github.com/slackhq/nebula/releases/tag/v1.9.7
-[1.9.6]: https://github.com/slackhq/nebula/releases/tag/v1.9.6
-[1.9.5]: https://github.com/slackhq/nebula/releases/tag/v1.9.5
+[Unreleased]: https://github.com/slackhq/nebula/compare/v1.9.4...HEAD
 [1.9.4]: https://github.com/slackhq/nebula/releases/tag/v1.9.4
 [1.9.3]: https://github.com/slackhq/nebula/releases/tag/v1.9.3
 [1.9.2]: https://github.com/slackhq/nebula/releases/tag/v1.9.2

bits.go

@@ -5,17 +5,19 @@ import (
 	"github.com/sirupsen/logrus"
 )
 
+// TODO: Pretty sure this is just all sorts of racy now, we need it to be atomic
 type Bits struct {
 	length             uint64
 	current            uint64
 	bits               []bool
+	firstSeen          bool
 	lostCounter        metrics.Counter
 	dupeCounter        metrics.Counter
 	outOfWindowCounter metrics.Counter
 }
 
 func NewBits(bits uint64) *Bits {
-	b := &Bits{
+	return &Bits{
 		length:             bits,
 		bits:               make([]bool, bits, bits),
 		current:            0,
@@ -23,37 +25,34 @@ func NewBits(bits uint64) *Bits {
 		dupeCounter:        metrics.GetOrRegisterCounter("network.packets.duplicate", nil),
 		outOfWindowCounter: metrics.GetOrRegisterCounter("network.packets.out_of_window", nil),
 	}
-
-	// There is no counter value 0, mark it to avoid counting a lost packet later.
-	b.bits[0] = true
-	b.current = 0
-	return b
 }
 
-func (b *Bits) Check(l *logrus.Logger, i uint64) bool {
+func (b *Bits) Check(l logrus.FieldLogger, i uint64) bool {
 	// If i is the next number, return true.
-	if i > b.current {
+	if i > b.current || (i == 0 && b.firstSeen == false && b.current < b.length) {
 		return true
 	}
 
-	// If i is within the window, check if it's been set already.
-	if i > b.current-b.length || i < b.length && b.current < b.length {
+	// If i is within the window, check if it's been set already. The first window will fail this check
+	if i > b.current-b.length {
+		return !b.bits[i%b.length]
+	}
+
+	// If i is within the first window
+	if i < b.length {
 		return !b.bits[i%b.length]
 	}
 
 	// Not within the window
-	if l.Level >= logrus.DebugLevel {
-		l.Debugf("rejected a packet (top) %d %d\n", b.current, i)
-	}
+	l.Error("rejected a packet (top) %d %d\n", b.current, i)
 	return false
 }
 
 func (b *Bits) Update(l *logrus.Logger, i uint64) bool {
 	// If i is the next number, return true and update current.
 	if i == b.current+1 {
-		// Check if the oldest bit was lost since we are shifting the window by 1 and occupying it with this counter
-		// The very first window can only be tracked as lost once we are on the 2nd window or greater
-		if b.bits[i%b.length] == false && i > b.length {
+		// Report missed packets, we can only understand what was missed after the first window has been gone through
+		if i > b.length && b.bits[i%b.length] == false {
 			b.lostCounter.Inc(1)
 		}
 		b.bits[i%b.length] = true
@@ -61,32 +60,61 @@ func (b *Bits) Update(l *logrus.Logger, i uint64) bool {
 		return true
 	}
 
-	// If i is a jump, adjust the window, record lost, update current, and return true
-	if i > b.current {
-		lost := int64(0)
-		// Zero out the bits between the current and the new counter value, limited by the window size,
-		// since the window is shifting
-		for n := b.current + 1; n <= min(i, b.current+b.length); n++ {
-			if b.bits[n%b.length] == false && n > b.length {
-				lost++
-			}
+	// If i packet is greater than current but less than the maximum length of our bitmap,
+	// flip everything in between to false and move ahead.
+	if i > b.current && i < b.current+b.length {
+		// In between current and i need to be zero'd to allow those packets to come in later
+		for n := b.current + 1; n < i; n++ {
 			b.bits[n%b.length] = false
 		}
 
-		// Only record any skipped packets as a result of the window moving further than the window length
-		// Any loss within the new window will be accounted for in future calls
-		lost += max(0, int64(i-b.current-b.length))
+		b.bits[i%b.length] = true
+		b.current = i
+		//l.Debugf("missed %d packets between %d and %d\n", i-b.current, i, b.current)
+		return true
+	}
+
+	// If i is greater than the delta between current and the total length of our bitmap,
+	// just flip everything in the map and move ahead.
+	if i >= b.current+b.length {
+		// The current window loss will be accounted for later, only record the jump as loss up until then
+		lost := maxInt64(0, int64(i-b.current-b.length))
+		//TODO: explain this
+		if b.current == 0 {
+			lost++
+		}
+
+		for n := range b.bits {
+			// Don't want to count the first window as a loss
+			//TODO: this is likely wrong, we are wanting to track only the bit slots that we aren't going to track anymore and this is marking everything as missed
+			//if b.bits[n] == false {
+			//	lost++
+			//}
+			b.bits[n] = false
+		}
+
 		b.lostCounter.Inc(lost)
 
+		if l.Level >= logrus.DebugLevel {
+			l.WithField("receiveWindow", m{"accepted": true, "currentCounter": b.current, "incomingCounter": i, "reason": "window shifting"}).
+				Debug("Receive window")
+		}
 		b.bits[i%b.length] = true
 		b.current = i
 		return true
 	}
 
-	// If i is within the current window but below the current counter,
-	// Check to see if it's a duplicate
-	if i > b.current-b.length || i < b.length && b.current < b.length {
-		if b.current == i || b.bits[i%b.length] == true {
+	// Allow for the 0 packet to come in within the first window
+	if i == 0 && b.firstSeen == false && b.current < b.length {
+		b.firstSeen = true
+		b.bits[i%b.length] = true
+		return true
+	}
+
+	// If i is within the window of current minus length (the total pat window size),
+	// allow it and flip to true but to NOT change current. We also have to account for the first window
+	if ((b.current >= b.length && i > b.current-b.length) || (b.current < b.length && i < b.length)) && i <= b.current {
+		if b.current == i {
 			if l.Level >= logrus.DebugLevel {
 				l.WithField("receiveWindow", m{"accepted": false, "currentCounter": b.current, "incomingCounter": i, "reason": "duplicate"}).
 					Debug("Receive window")
@@ -95,8 +123,18 @@ func (b *Bits) Update(l *logrus.Logger, i uint64) bool {
 			return false
 		}
 
+		if b.bits[i%b.length] == true {
+			if l.Level >= logrus.DebugLevel {
+				l.WithField("receiveWindow", m{"accepted": false, "currentCounter": b.current, "incomingCounter": i, "reason": "old duplicate"}).
+					Debug("Receive window")
+			}
+			b.dupeCounter.Inc(1)
+			return false
+		}
+
 		b.bits[i%b.length] = true
 		return true
+
 	}
 
 	// In all other cases, fail and don't change current.
@@ -110,3 +148,11 @@ func (b *Bits) Update(l *logrus.Logger, i uint64) bool {
 	}
 	return false
 }
+
+func maxInt64(a, b int64) int64 {
+	if a > b {
+		return a
+	}
+
+	return b
+}

bits_test.go

@@ -15,41 +15,48 @@ func TestBits(t *testing.T) {
 	assert.Len(t, b.bits, 10)
 
 	// This is initialized to zero - receive one. This should work.
+
 	assert.True(t, b.Check(l, 1))
-	assert.True(t, b.Update(l, 1))
+	u := b.Update(l, 1)
+	assert.True(t, u)
 	assert.EqualValues(t, 1, b.current)
-	g := []bool{true, true, false, false, false, false, false, false, false, false}
+	g := []bool{false, true, false, false, false, false, false, false, false, false}
 	assert.Equal(t, g, b.bits)
 
 	// Receive two
 	assert.True(t, b.Check(l, 2))
-	assert.True(t, b.Update(l, 2))
+	u = b.Update(l, 2)
+	assert.True(t, u)
 	assert.EqualValues(t, 2, b.current)
-	g = []bool{true, true, true, false, false, false, false, false, false, false}
+	g = []bool{false, true, true, false, false, false, false, false, false, false}
 	assert.Equal(t, g, b.bits)
 
 	// Receive two again - it will fail
 	assert.False(t, b.Check(l, 2))
-	assert.False(t, b.Update(l, 2))
+	u = b.Update(l, 2)
+	assert.False(t, u)
 	assert.EqualValues(t, 2, b.current)
 
 	// Jump ahead to 15, which should clear everything and set the 6th element
 	assert.True(t, b.Check(l, 15))
-	assert.True(t, b.Update(l, 15))
+	u = b.Update(l, 15)
+	assert.True(t, u)
 	assert.EqualValues(t, 15, b.current)
 	g = []bool{false, false, false, false, false, true, false, false, false, false}
 	assert.Equal(t, g, b.bits)
 
 	// Mark 14, which is allowed because it is in the window
 	assert.True(t, b.Check(l, 14))
-	assert.True(t, b.Update(l, 14))
+	u = b.Update(l, 14)
+	assert.True(t, u)
 	assert.EqualValues(t, 15, b.current)
 	g = []bool{false, false, false, false, true, true, false, false, false, false}
 	assert.Equal(t, g, b.bits)
 
 	// Mark 5, which is not allowed because it is not in the window
 	assert.False(t, b.Check(l, 5))
-	assert.False(t, b.Update(l, 5))
+	u = b.Update(l, 5)
+	assert.False(t, u)
 	assert.EqualValues(t, 15, b.current)
 	g = []bool{false, false, false, false, true, true, false, false, false, false}
 	assert.Equal(t, g, b.bits)
@@ -62,29 +69,10 @@ func TestBits(t *testing.T) {
 
 	// Walk through a few windows in order
 	b = NewBits(10)
-	for i := uint64(1); i <= 100; i++ {
+	for i := uint64(0); i <= 100; i++ {
 		assert.True(t, b.Check(l, i), "Error while checking %v", i)
 		assert.True(t, b.Update(l, i), "Error while updating %v", i)
 	}
-
-	assert.False(t, b.Check(l, 1), "Out of window check")
-}
-
-func TestBitsLargeJumps(t *testing.T) {
-	l := test.NewLogger()
-	b := NewBits(10)
-	b.lostCounter.Clear()
-
-	b = NewBits(10)
-	b.lostCounter.Clear()
-	assert.True(t, b.Update(l, 55)) // We saw packet 55 and can still track 45,46,47,48,49,50,51,52,53,54
-	assert.Equal(t, int64(45), b.lostCounter.Count())
-
-	assert.True(t, b.Update(l, 100)) // We saw packet 55 and 100 and can still track 90,91,92,93,94,95,96,97,98,99
-	assert.Equal(t, int64(89), b.lostCounter.Count())
-
-	assert.True(t, b.Update(l, 200)) // We saw packet 55, 100, and 200 and can still track 190,191,192,193,194,195,196,197,198,199
-	assert.Equal(t, int64(188), b.lostCounter.Count())
 }
 
 func TestBitsDupeCounter(t *testing.T) {
@@ -136,7 +124,8 @@ func TestBitsOutOfWindowCounter(t *testing.T) {
 	assert.False(t, b.Update(l, 0))
 	assert.Equal(t, int64(1), b.outOfWindowCounter.Count())
 
-	assert.Equal(t, int64(19), b.lostCounter.Count()) // packet 0 wasn't lost
+	//tODO: make sure lostcounter doesn't increase in orderly increment
+	assert.Equal(t, int64(20), b.lostCounter.Count())
 	assert.Equal(t, int64(0), b.dupeCounter.Count())
 	assert.Equal(t, int64(1), b.outOfWindowCounter.Count())
 }
@@ -148,6 +137,8 @@ func TestBitsLostCounter(t *testing.T) {
 	b.dupeCounter.Clear()
 	b.outOfWindowCounter.Clear()
 
+	//assert.True(t, b.Update(0))
+	assert.True(t, b.Update(l, 0))
 	assert.True(t, b.Update(l, 20))
 	assert.True(t, b.Update(l, 21))
 	assert.True(t, b.Update(l, 22))
@@ -158,7 +149,7 @@ func TestBitsLostCounter(t *testing.T) {
 	assert.True(t, b.Update(l, 27))
 	assert.True(t, b.Update(l, 28))
 	assert.True(t, b.Update(l, 29))
-	assert.Equal(t, int64(19), b.lostCounter.Count()) // packet 0 wasn't lost
+	assert.Equal(t, int64(20), b.lostCounter.Count())
 	assert.Equal(t, int64(0), b.dupeCounter.Count())
 	assert.Equal(t, int64(0), b.outOfWindowCounter.Count())
 
@@ -167,6 +158,8 @@ func TestBitsLostCounter(t *testing.T) {
 	b.dupeCounter.Clear()
 	b.outOfWindowCounter.Clear()
 
+	assert.True(t, b.Update(l, 0))
+	assert.Equal(t, int64(0), b.lostCounter.Count())
 	assert.True(t, b.Update(l, 9))
 	assert.Equal(t, int64(0), b.lostCounter.Count())
 	// 10 will set 0 index, 0 was already set, no lost packets
@@ -221,62 +214,6 @@ func TestBitsLostCounter(t *testing.T) {
 	assert.Equal(t, int64(0), b.outOfWindowCounter.Count())
 }
 
-func TestBitsLostCounterIssue1(t *testing.T) {
-	l := test.NewLogger()
-	b := NewBits(10)
-	b.lostCounter.Clear()
-	b.dupeCounter.Clear()
-	b.outOfWindowCounter.Clear()
-
-	assert.True(t, b.Update(l, 4))
-	assert.Equal(t, int64(0), b.lostCounter.Count())
-	assert.True(t, b.Update(l, 1))
-	assert.Equal(t, int64(0), b.lostCounter.Count())
-	assert.True(t, b.Update(l, 9))
-	assert.Equal(t, int64(0), b.lostCounter.Count())
-	assert.True(t, b.Update(l, 2))
-	assert.Equal(t, int64(0), b.lostCounter.Count())
-	assert.True(t, b.Update(l, 3))
-	assert.Equal(t, int64(0), b.lostCounter.Count())
-	assert.True(t, b.Update(l, 5))
-	assert.Equal(t, int64(0), b.lostCounter.Count())
-	assert.True(t, b.Update(l, 6))
-	assert.Equal(t, int64(0), b.lostCounter.Count())
-	assert.True(t, b.Update(l, 7))
-	assert.Equal(t, int64(0), b.lostCounter.Count())
-	// assert.True(t, b.Update(l, 8))
-	assert.True(t, b.Update(l, 10))
-	assert.Equal(t, int64(0), b.lostCounter.Count())
-	assert.True(t, b.Update(l, 11))
-	assert.Equal(t, int64(0), b.lostCounter.Count())
-
-	assert.True(t, b.Update(l, 14))
-	assert.Equal(t, int64(0), b.lostCounter.Count())
-	// Issue seems to be here, we reset missing packet 8 to false here and don't increment the lost counter
-	assert.True(t, b.Update(l, 19))
-	assert.Equal(t, int64(1), b.lostCounter.Count())
-	assert.True(t, b.Update(l, 12))
-	assert.Equal(t, int64(1), b.lostCounter.Count())
-	assert.True(t, b.Update(l, 13))
-	assert.Equal(t, int64(1), b.lostCounter.Count())
-	assert.True(t, b.Update(l, 15))
-	assert.Equal(t, int64(1), b.lostCounter.Count())
-	assert.True(t, b.Update(l, 16))
-	assert.Equal(t, int64(1), b.lostCounter.Count())
-	assert.True(t, b.Update(l, 17))
-	assert.Equal(t, int64(1), b.lostCounter.Count())
-	assert.True(t, b.Update(l, 18))
-	assert.Equal(t, int64(1), b.lostCounter.Count())
-	assert.True(t, b.Update(l, 20))
-	assert.Equal(t, int64(1), b.lostCounter.Count())
-	assert.True(t, b.Update(l, 21))
-
-	// We missed packet 8 above
-	assert.Equal(t, int64(1), b.lostCounter.Count())
-	assert.Equal(t, int64(0), b.dupeCounter.Count())
-	assert.Equal(t, int64(0), b.outOfWindowCounter.Count())
-}
-
 func BenchmarkBits(b *testing.B) {
 	z := NewBits(10)
 	for n := 0; n < b.N; n++ {

cert/pem.go

@@ -1,8 +1,10 @@
 package cert
 
 import (
+	"encoding/hex"
 	"encoding/pem"
 	"fmt"
+	"time"
 
 	"golang.org/x/crypto/ed25519"
 )
@@ -138,6 +140,101 @@ func MarshalSigningPrivateKeyToPEM(curve Curve, b []byte) []byte {
 	}
 }
 
+// Backward compatibility functions for older API
+func MarshalX25519PublicKey(b []byte) []byte {
+	return MarshalPublicKeyToPEM(Curve_CURVE25519, b)
+}
+
+func MarshalX25519PrivateKey(b []byte) []byte {
+	return MarshalPrivateKeyToPEM(Curve_CURVE25519, b)
+}
+
+func MarshalPublicKey(curve Curve, b []byte) []byte {
+	return MarshalPublicKeyToPEM(curve, b)
+}
+
+func MarshalPrivateKey(curve Curve, b []byte) []byte {
+	return MarshalPrivateKeyToPEM(curve, b)
+}
+
+// NebulaCertificate is a compatibility wrapper for the old API
+type NebulaCertificate struct {
+	Details   NebulaCertificateDetails
+	Signature []byte
+	cert      Certificate
+}
+
+// NebulaCertificateDetails is a compatibility wrapper for certificate details
+type NebulaCertificateDetails struct {
+	Name      string
+	NotBefore time.Time
+	NotAfter  time.Time
+	PublicKey []byte
+	IsCA      bool
+	Issuer    []byte
+	Curve     Curve
+}
+
+// UnmarshalNebulaCertificateFromPEM provides backward compatibility with the old API
+func UnmarshalNebulaCertificateFromPEM(b []byte) (*NebulaCertificate, []byte, error) {
+	c, rest, err := UnmarshalCertificateFromPEM(b)
+	if err != nil {
+		return nil, rest, err
+	}
+
+	issuerBytes, err := func() ([]byte, error) {
+		issuer := c.Issuer()
+		if issuer == "" {
+			return nil, nil
+		}
+		decoded, err := hex.DecodeString(issuer)
+		if err != nil {
+			return nil, fmt.Errorf("failed to decode issuer fingerprint: %w", err)
+		}
+		return decoded, nil
+	}()
+	if err != nil {
+		return nil, rest, err
+	}
+
+	pubKey := c.PublicKey()
+	if pubKey != nil {
+		pubKey = append([]byte(nil), pubKey...)
+	}
+
+	sig := c.Signature()
+	if sig != nil {
+		sig = append([]byte(nil), sig...)
+	}
+
+	return &NebulaCertificate{
+		Details: NebulaCertificateDetails{
+			Name:      c.Name(),
+			NotBefore: c.NotBefore(),
+			NotAfter:  c.NotAfter(),
+			PublicKey: pubKey,
+			IsCA:      c.IsCA(),
+			Issuer:    issuerBytes,
+			Curve:     c.Curve(),
+		},
+		Signature: sig,
+		cert:      c,
+	}, rest, nil
+}
+
+// IssuerString returns the issuer in hex format for compatibility
+func (n *NebulaCertificate) IssuerString() string {
+	if n.Details.Issuer == nil {
+		return ""
+	}
+	return hex.EncodeToString(n.Details.Issuer)
+}
+
+// Certificate returns the underlying certificate (read-only)
+func (n *NebulaCertificate) Certificate() Certificate {
+	return n.cert
+}
+
 // UnmarshalPrivateKeyFromPEM will try to unmarshal the first pem block in a byte array, returning any non
 // consumed data or an error on failure
 func UnmarshalPrivateKeyFromPEM(b []byte) ([]byte, []byte, Curve, error) {

cert_test/cert.go

@@ -114,33 +114,6 @@ func NewTestCert(v cert.Version, curve cert.Curve, ca cert.Certificate, key []by
 	return c, pub, cert.MarshalPrivateKeyToPEM(curve, priv), pem
 }
 
-func NewTestCertDifferentVersion(c cert.Certificate, v cert.Version, ca cert.Certificate, key []byte) (cert.Certificate, []byte) {
-	nc := &cert.TBSCertificate{
-		Version:        v,
-		Curve:          c.Curve(),
-		Name:           c.Name(),
-		Networks:       c.Networks(),
-		UnsafeNetworks: c.UnsafeNetworks(),
-		Groups:         c.Groups(),
-		NotBefore:      time.Unix(c.NotBefore().Unix(), 0),
-		NotAfter:       time.Unix(c.NotAfter().Unix(), 0),
-		PublicKey:      c.PublicKey(),
-		IsCA:           false,
-	}
-
-	c, err := nc.Sign(ca, ca.Curve(), key)
-	if err != nil {
-		panic(err)
-	}
-
-	pem, err := c.MarshalPEM()
-	if err != nil {
-		panic(err)
-	}
-
-	return c, pem
-}
-
 func X25519Keypair() ([]byte, []byte) {
 	privkey := make([]byte, 32)
 	if _, err := io.ReadFull(rand.Reader, privkey); err != nil {

cmd/nebula-cert/ca.go

@@ -173,8 +173,6 @@ func ca(args []string, out io.Writer, errOut io.Writer, pr PasswordReader) error
 
 	var passphrase []byte
 	if !isP11 && *cf.encryption {
-		passphrase = []byte(os.Getenv("NEBULA_CA_PASSPHRASE"))
-		if len(passphrase) == 0 {
 		for i := 0; i < 5; i++ {
 			out.Write([]byte("Enter passphrase: "))
 			passphrase, err = pr.ReadPassword()
@@ -194,7 +192,6 @@ func ca(args []string, out io.Writer, errOut io.Writer, pr PasswordReader) error
 			return fmt.Errorf("no passphrase specified, remove -encrypt flag to write out-key in plaintext")
 		}
 	}
-	}
 
 	var curve cert.Curve
 	var pub, rawPriv []byte

cmd/nebula-cert/ca_test.go

@@ -171,17 +171,6 @@ func Test_ca(t *testing.T) {
 	assert.Equal(t, pwPromptOb, ob.String())
 	assert.Empty(t, eb.String())
 
-	// test encrypted key with passphrase environment variable
-	os.Remove(keyF.Name())
-	os.Remove(crtF.Name())
-	ob.Reset()
-	eb.Reset()
-	args = []string{"-version", "1", "-encrypt", "-name", "test", "-duration", "100m", "-groups", "1,2,3,4,5", "-out-crt", crtF.Name(), "-out-key", keyF.Name()}
-	os.Setenv("NEBULA_CA_PASSPHRASE", string(passphrase))
-	require.NoError(t, ca(args, ob, eb, testpw))
-	assert.Empty(t, eb.String())
-	os.Setenv("NEBULA_CA_PASSPHRASE", "")
-
 	// read encrypted key file and verify default params
 	rb, _ = os.ReadFile(keyF.Name())
 	k, _ := pem.Decode(rb)

cmd/nebula-cert/main.go

@@ -5,28 +5,10 @@ import (
 	"fmt"
 	"io"
 	"os"
-	"runtime/debug"
-	"strings"
 )
 
-// A version string that can be set with
-//
-//	-ldflags "-X main.Build=SOMEVERSION"
-//
-// at compile-time.
 var Build string
 
-func init() {
-	if Build == "" {
-		info, ok := debug.ReadBuildInfo()
-		if !ok {
-			return
-		}
-
-		Build = strings.TrimPrefix(info.Main.Version, "v")
-	}
-}
-
 type helpError struct {
 	s string
 }

cmd/nebula-cert/sign.go

@@ -116,10 +116,8 @@ func signCert(args []string, out io.Writer, errOut io.Writer, pr PasswordReader)
 		// naively attempt to decode the private key as though it is not encrypted
 		caKey, _, curve, err = cert.UnmarshalSigningPrivateKeyFromPEM(rawCAKey)
 		if errors.Is(err, cert.ErrPrivateKeyEncrypted) {
-			var passphrase []byte
-			passphrase = []byte(os.Getenv("NEBULA_CA_PASSPHRASE"))
-			if len(passphrase) == 0 {
 			// ask for a passphrase until we get one
+			var passphrase []byte
 			for i := 0; i < 5; i++ {
 				out.Write([]byte("Enter passphrase: "))
 				passphrase, err = pr.ReadPassword()
@@ -137,7 +135,7 @@ func signCert(args []string, out io.Writer, errOut io.Writer, pr PasswordReader)
 			if len(passphrase) == 0 {
 				return fmt.Errorf("cannot open encrypted ca-key without passphrase")
 			}
-			}
+
 			curve, caKey, _, err = cert.DecryptAndUnmarshalSigningPrivateKey(passphrase, rawCAKey)
 			if err != nil {
 				return fmt.Errorf("error while parsing encrypted ca-key: %s", err)

cmd/nebula-cert/sign_test.go

@@ -379,15 +379,6 @@ func Test_signCert(t *testing.T) {
 	assert.Equal(t, "Enter passphrase: ", ob.String())
 	assert.Empty(t, eb.String())
 
-	// test with the proper password in the environment
-	os.Remove(crtF.Name())
-	os.Remove(keyF.Name())
-	args = []string{"-version", "1", "-ca-crt", caCrtF.Name(), "-ca-key", caKeyF.Name(), "-name", "test", "-ip", "1.1.1.1/24", "-out-crt", crtF.Name(), "-out-key", keyF.Name(), "-duration", "100m", "-subnets", "10.1.1.1/32, ,   10.2.2.2/32   ,   ,  ,, 10.5.5.5/32", "-groups", "1,,   2    ,        ,,,3,4,5"}
-	os.Setenv("NEBULA_CA_PASSPHRASE", string(passphrase))
-	require.NoError(t, signCert(args, ob, eb, testpw))
-	assert.Empty(t, eb.String())
-	os.Setenv("NEBULA_CA_PASSPHRASE", "")
-
 	// test with the wrong password
 	ob.Reset()
 	eb.Reset()
@@ -398,17 +389,6 @@ func Test_signCert(t *testing.T) {
 	assert.Equal(t, "Enter passphrase: ", ob.String())
 	assert.Empty(t, eb.String())
 
-	// test with the wrong password in environment
-	ob.Reset()
-	eb.Reset()
-
-	os.Setenv("NEBULA_CA_PASSPHRASE", "invalid password")
-	args = []string{"-version", "1", "-ca-crt", caCrtF.Name(), "-ca-key", caKeyF.Name(), "-name", "test", "-ip", "1.1.1.1/24", "-out-crt", crtF.Name(), "-out-key", keyF.Name(), "-duration", "100m", "-subnets", "10.1.1.1/32, ,   10.2.2.2/32   ,   ,  ,, 10.5.5.5/32", "-groups", "1,,   2    ,        ,,,3,4,5"}
-	require.EqualError(t, signCert(args, ob, eb, nopw), "error while parsing encrypted ca-key: invalid passphrase or corrupt private key")
-	assert.Empty(t, ob.String())
-	assert.Empty(t, eb.String())
-	os.Setenv("NEBULA_CA_PASSPHRASE", "")
-
 	// test with the user not entering a password
 	ob.Reset()
 	eb.Reset()

cmd/nebula-service/main.go

@@ -4,8 +4,6 @@ import (
 	"flag"
 	"fmt"
 	"os"
-	"runtime/debug"
-	"strings"
 
 	"github.com/sirupsen/logrus"
 	"github.com/slackhq/nebula"
@@ -20,17 +18,6 @@ import (
 // at compile-time.
 var Build string
 
-func init() {
-	if Build == "" {
-		info, ok := debug.ReadBuildInfo()
-		if !ok {
-			return
-		}
-
-		Build = strings.TrimPrefix(info.Main.Version, "v")
-	}
-}
-
 func main() {
 	serviceFlag := flag.String("service", "", "Control the system service.")
 	configPath := flag.String("config", "", "Path to either a file or directory to load configuration from")
@@ -78,8 +65,16 @@ func main() {
 	}
 
 	if !*configTest {
-		ctrl.Start()
-		ctrl.ShutdownBlock()
+		wait, err := ctrl.Start()
+		if err != nil {
+			util.LogWithContextIfNeeded("Error while running", err, l)
+			os.Exit(1)
+		}
+
+		go ctrl.ShutdownBlock()
+		wait()
+
+		l.Info("Goodbye")
 	}
 
 	os.Exit(0)

cmd/nebula/main.go

@@ -3,9 +3,10 @@ package main
 import (
 	"flag"
 	"fmt"
+	"log"
+	"net/http"
+	_ "net/http/pprof"
 	"os"
-	"runtime/debug"
-	"strings"
 
 	"github.com/sirupsen/logrus"
 	"github.com/slackhq/nebula"
@@ -20,17 +21,6 @@ import (
 // at compile-time.
 var Build string
 
-func init() {
-	if Build == "" {
-		info, ok := debug.ReadBuildInfo()
-		if !ok {
-			return
-		}
-
-		Build = strings.TrimPrefix(info.Main.Version, "v")
-	}
-}
-
 func main() {
 	configPath := flag.String("config", "", "Path to either a file or directory to load configuration from")
 	configTest := flag.Bool("test", false, "Test the config and print the end result. Non zero exit indicates a faulty config")
@@ -71,10 +61,22 @@ func main() {
 		os.Exit(1)
 	}
 
+	go func() {
+		log.Println(http.ListenAndServe("0.0.0.0:6060", nil))
+	}()
+
 	if !*configTest {
-		ctrl.Start()
+		wait, err := ctrl.Start()
+		if err != nil {
+			util.LogWithContextIfNeeded("Error while running", err, l)
+			os.Exit(1)
+		}
+
+		go ctrl.ShutdownBlock()
 		notifyReady(l)
-		ctrl.ShutdownBlock()
+		wait()
+
+		l.Info("Goodbye")
 	}
 
 	os.Exit(0)

config/config.go

@@ -17,7 +17,7 @@ import (
 
 	"dario.cat/mergo"
 	"github.com/sirupsen/logrus"
-	"go.yaml.in/yaml/v3"
+	"gopkg.in/yaml.v3"
 )
 
 type C struct {

config/config_test.go

@@ -10,7 +10,7 @@ import (
 	"github.com/slackhq/nebula/test"
 	"github.com/stretchr/testify/assert"
 	"github.com/stretchr/testify/require"
-	"go.yaml.in/yaml/v3"
+	"gopkg.in/yaml.v3"
 )
 
 func TestConfig_Load(t *testing.T) {

connection_manager.go

@@ -354,6 +354,7 @@ func (cm *connectionManager) makeTrafficDecision(localIndex uint32, now time.Tim
 
 		if mainHostInfo {
 			decision = tryRehandshake
+
 		} else {
 			if cm.shouldSwapPrimary(hostinfo) {
 				decision = swapPrimary
@@ -460,10 +461,6 @@ func (cm *connectionManager) shouldSwapPrimary(current *HostInfo) bool {
 	}
 
 	crt := cm.intf.pki.getCertState().getCertificate(current.ConnectionState.myCert.Version())
-	if crt == nil {
-		//my cert was reloaded away. We should definitely swap from this tunnel
-		return true
-	}
 	// If this tunnel is using the latest certificate then we should swap it to primary for a bit and see if things
 	// settle down.
 	return bytes.Equal(current.ConnectionState.myCert.Signature(), crt.Signature())
@@ -478,34 +475,31 @@ func (cm *connectionManager) swapPrimary(current, primary *HostInfo) {
 	cm.hostMap.Unlock()
 }
 
-// isInvalidCertificate decides if we should destroy a tunnel.
-// returns true if pki.disconnect_invalid is true and the certificate is no longer valid.
-// Blocklisted certificates will skip the pki.disconnect_invalid check and return true.
+// isInvalidCertificate will check if we should destroy a tunnel if pki.disconnect_invalid is true and
+// the certificate is no longer valid. Block listed certificates will skip the pki.disconnect_invalid
+// check and return true.
 func (cm *connectionManager) isInvalidCertificate(now time.Time, hostinfo *HostInfo) bool {
 	remoteCert := hostinfo.GetCert()
 	if remoteCert == nil {
-		return false //don't tear down tunnels for handshakes in progress
+		return false
 	}
 
 	caPool := cm.intf.pki.GetCAPool()
 	err := caPool.VerifyCachedCertificate(now, remoteCert)
 	if err == nil {
-		return false //cert is still valid! yay!
-	} else if err == cert.ErrBlockListed { //avoiding errors.Is for speed
+		return false
+	}
+
+	if !cm.intf.disconnectInvalid.Load() && err != cert.ErrBlockListed {
 		// Block listed certificates should always be disconnected
-		hostinfo.logger(cm.l).WithError(err).
-			WithField("fingerprint", remoteCert.Fingerprint).
-			Info("Remote certificate is blocked, tearing down the tunnel")
-		return true
-	} else if cm.intf.disconnectInvalid.Load() {
+		return false
+	}
+
 	hostinfo.logger(cm.l).WithError(err).
 		WithField("fingerprint", remoteCert.Fingerprint).
 		Info("Remote certificate is no longer valid, tearing down the tunnel")
+
 	return true
-	} else {
-		//if we reach here, the cert is no longer valid, but we're configured to keep tunnels from now-invalid certs open
-		return false
-	}
 }
 
 func (cm *connectionManager) sendPunch(hostinfo *HostInfo) {
@@ -536,45 +530,15 @@ func (cm *connectionManager) sendPunch(hostinfo *HostInfo) {
 func (cm *connectionManager) tryRehandshake(hostinfo *HostInfo) {
 	cs := cm.intf.pki.getCertState()
 	curCrt := hostinfo.ConnectionState.myCert
-	curCrtVersion := curCrt.Version()
-	myCrt := cs.getCertificate(curCrtVersion)
-	if myCrt == nil {
-		cm.l.WithField("vpnAddrs", hostinfo.vpnAddrs).
-			WithField("version", curCrtVersion).
-			WithField("reason", "local certificate removed").
-			Info("Re-handshaking with remote")
-		cm.intf.handshakeManager.StartHandshake(hostinfo.vpnAddrs[0], nil)
+	myCrt := cs.getCertificate(curCrt.Version())
+	if curCrt.Version() >= cs.initiatingVersion && bytes.Equal(curCrt.Signature(), myCrt.Signature()) == true {
+		// The current tunnel is using the latest certificate and version, no need to rehandshake.
 		return
 	}
-	peerCrt := hostinfo.ConnectionState.peerCert
-	if peerCrt != nil && curCrtVersion < peerCrt.Certificate.Version() {
-		// if our certificate version is less than theirs, and we have a matching version available, rehandshake?
-		if cs.getCertificate(peerCrt.Certificate.Version()) != nil {
-			cm.l.WithField("vpnAddrs", hostinfo.vpnAddrs).
-				WithField("version", curCrtVersion).
-				WithField("peerVersion", peerCrt.Certificate.Version()).
-				WithField("reason", "local certificate version lower than peer, attempting to correct").
-				Info("Re-handshaking with remote")
-			cm.intf.handshakeManager.StartHandshake(hostinfo.vpnAddrs[0], func(hh *HandshakeHostInfo) {
-				hh.initiatingVersionOverride = peerCrt.Certificate.Version()
-			})
-			return
-		}
-	}
-	if !bytes.Equal(curCrt.Signature(), myCrt.Signature()) {
+
 	cm.l.WithField("vpnAddrs", hostinfo.vpnAddrs).
 		WithField("reason", "local certificate is not current").
 		Info("Re-handshaking with remote")
 
 	cm.intf.handshakeManager.StartHandshake(hostinfo.vpnAddrs[0], nil)
-		return
-	}
-	if curCrtVersion < cs.initiatingVersion {
-		cm.l.WithField("vpnAddrs", hostinfo.vpnAddrs).
-			WithField("reason", "current cert version < pki.initiatingVersion").
-			Info("Re-handshaking with remote")
-
-		cm.intf.handshakeManager.StartHandshake(hostinfo.vpnAddrs[0], nil)
-		return
-	}
 }

connection_state.go

@@ -13,6 +13,8 @@ import (
 	"github.com/slackhq/nebula/noiseutil"
 )
 
+// TODO: In a 5Gbps test, 1024 is not sufficient. With a 1400 MTU this is about 1.4Gbps of window, assuming full packets.
+// 4092 should be sufficient for 5Gbps
 const ReplayWindow = 1024
 
 type ConnectionState struct {

control.go

@@ -2,9 +2,11 @@ package nebula
 
 import (
 	"context"
+	"errors"
 	"net/netip"
 	"os"
 	"os/signal"
+	"sync"
 	"syscall"
 
 	"github.com/sirupsen/logrus"
@@ -13,6 +15,16 @@ import (
 	"github.com/slackhq/nebula/overlay"
 )
 
+type RunState int
+
+const (
+	Stopped  RunState = 0 // The control has yet to be started
+	Started  RunState = 1 // The control has been started
+	Stopping RunState = 2 // The control is stopping
+)
+
+var ErrAlreadyStarted = errors.New("nebula is already started")
+
 // Every interaction here needs to take extra care to copy memory and not return or use arguments "as is" when touching
 // core. This means copying IP objects, slices, de-referencing pointers and taking the actual value, etc
 
@@ -26,6 +38,9 @@ type controlHostLister interface {
 }
 
 type Control struct {
+	stateLock sync.Mutex
+	state     RunState
+
 	f                      *Interface
 	l                      *logrus.Logger
 	ctx                    context.Context
@@ -49,10 +64,21 @@ type ControlHostInfo struct {
 	CurrentRelaysThroughMe []netip.Addr     `json:"currentRelaysThroughMe"`
 }
 
-// Start actually runs nebula, this is a nonblocking call. To block use Control.ShutdownBlock()
-func (c *Control) Start() {
+// Start actually runs nebula, this is a nonblocking call.
+// The returned function can be used to wait for nebula to fully stop.
+func (c *Control) Start() (func(), error) {
+	c.stateLock.Lock()
+	if c.state != Stopped {
+		c.stateLock.Unlock()
+		return nil, ErrAlreadyStarted
+	}
+
 	// Activate the interface
-	c.f.activate()
+	err := c.f.activate()
+	if err != nil {
+		c.stateLock.Unlock()
+		return nil, err
+	}
 
 	// Call all the delayed funcs that waited patiently for the interface to be created.
 	if c.sshStart != nil {
@@ -72,15 +98,33 @@ func (c *Control) Start() {
 	}
 
 	// Start reading packets.
-	c.f.run()
+	c.state = Started
+	c.stateLock.Unlock()
+	return c.f.run(c.ctx)
+}
+
+func (c *Control) State() RunState {
+	c.stateLock.Lock()
+	defer c.stateLock.Unlock()
+	return c.state
 }
 
 func (c *Control) Context() context.Context {
 	return c.ctx
 }
 
-// Stop signals nebula to shutdown and close all tunnels, returns after the shutdown is complete
+// Stop is a non-blocking call that signals nebula to close all tunnels and shut down
 func (c *Control) Stop() {
+	c.stateLock.Lock()
+	if c.state != Started {
+		c.stateLock.Unlock()
+		// We are stopping or stopped already
+		return
+	}
+
+	c.state = Stopping
+	c.stateLock.Unlock()
+
 	// Stop the handshakeManager (and other services), to prevent new tunnels from
 	// being created while we're shutting them all down.
 	c.cancel()
@@ -89,7 +133,7 @@ func (c *Control) Stop() {
 	if err := c.f.Close(); err != nil {
 		c.l.WithError(err).Error("Close interface failed")
 	}
-	c.l.Info("Goodbye")
+	c.state = Stopped
 }
 
 // ShutdownBlock will listen for and block on term and interrupt signals, calling Control.Stop() once signalled

control_tester.go

@@ -174,10 +174,6 @@ func (c *Control) GetHostmap() *HostMap {
 	return c.f.hostMap
 }
 
-func (c *Control) GetF() *Interface {
-	return c.f
-}
-
 func (c *Control) GetCertState() *CertState {
 	return c.f.pki.getCertState()
 }

e2e/handshakes_test.go

@@ -20,7 +20,7 @@ import (
 	"github.com/slackhq/nebula/udp"
 	"github.com/stretchr/testify/assert"
 	"github.com/stretchr/testify/require"
-	"go.yaml.in/yaml/v3"
+	"gopkg.in/yaml.v3"
 )
 
 func BenchmarkHotPath(b *testing.B) {
@@ -97,41 +97,6 @@ func TestGoodHandshake(t *testing.T) {
 	theirControl.Stop()
 }
 
-func TestGoodHandshakeNoOverlap(t *testing.T) {
-	ca, _, caKey, _ := cert_test.NewTestCaCert(cert.Version2, cert.Curve_CURVE25519, time.Now(), time.Now().Add(10*time.Minute), nil, nil, []string{})
-	myControl, myVpnIpNet, myUdpAddr, _ := newSimpleServer(cert.Version2, ca, caKey, "me", "10.128.0.1/24", nil)
-	theirControl, theirVpnIpNet, theirUdpAddr, _ := newSimpleServer(cert.Version2, ca, caKey, "them", "2001::69/24", nil) //look ma, cross-stack!
-
-	// Put their info in our lighthouse
-	myControl.InjectLightHouseAddr(theirVpnIpNet[0].Addr(), theirUdpAddr)
-
-	// Start the servers
-	myControl.Start()
-	theirControl.Start()
-
-	empty := []byte{}
-	t.Log("do something to cause a handshake")
-	myControl.GetF().SendMessageToVpnAddr(header.Test, header.MessageNone, theirVpnIpNet[0].Addr(), empty, empty, empty)
-
-	t.Log("Have them consume my stage 0 packet. They have a tunnel now")
-	theirControl.InjectUDPPacket(myControl.GetFromUDP(true))
-
-	t.Log("Get their stage 1 packet")
-	stage1Packet := theirControl.GetFromUDP(true)
-
-	t.Log("Have me consume their stage 1 packet. I have a tunnel now")
-	myControl.InjectUDPPacket(stage1Packet)
-
-	t.Log("Wait until we see a test packet come through to make sure we give the tunnel time to complete")
-	myControl.WaitForType(header.Test, 0, theirControl)
-
-	t.Log("Make sure our host infos are correct")
-	assertHostInfoPair(t, myUdpAddr, theirUdpAddr, myVpnIpNet, theirVpnIpNet, myControl, theirControl)
-
-	myControl.Stop()
-	theirControl.Stop()
-}
-
 func TestWrongResponderHandshake(t *testing.T) {
 	ca, _, caKey, _ := cert_test.NewTestCaCert(cert.Version1, cert.Curve_CURVE25519, time.Now(), time.Now().Add(10*time.Minute), nil, nil, []string{})
 
@@ -499,35 +464,6 @@ func TestRelays(t *testing.T) {
 	r.RenderHostmaps("Final hostmaps", myControl, relayControl, theirControl)
 }
 
-func TestRelaysDontCareAboutIps(t *testing.T) {
-	ca, _, caKey, _ := cert_test.NewTestCaCert(cert.Version2, cert.Curve_CURVE25519, time.Now(), time.Now().Add(10*time.Minute), nil, nil, []string{})
-	myControl, myVpnIpNet, _, _ := newSimpleServer(cert.Version2, ca, caKey, "me     ", "10.128.0.1/24", m{"relay": m{"use_relays": true}})
-	relayControl, relayVpnIpNet, relayUdpAddr, _ := newSimpleServer(cert.Version2, ca, caKey, "relay  ", "2001::9999/24", m{"relay": m{"am_relay": true}})
-	theirControl, theirVpnIpNet, theirUdpAddr, _ := newSimpleServer(cert.Version2, ca, caKey, "them   ", "10.128.0.2/24", m{"relay": m{"use_relays": true}})
-
-	// Teach my how to get to the relay and that their can be reached via the relay
-	myControl.InjectLightHouseAddr(relayVpnIpNet[0].Addr(), relayUdpAddr)
-	myControl.InjectRelays(theirVpnIpNet[0].Addr(), []netip.Addr{relayVpnIpNet[0].Addr()})
-	relayControl.InjectLightHouseAddr(theirVpnIpNet[0].Addr(), theirUdpAddr)
-
-	// Build a router so we don't have to reason who gets which packet
-	r := router.NewR(t, myControl, relayControl, theirControl)
-	defer r.RenderFlow()
-
-	// Start the servers
-	myControl.Start()
-	relayControl.Start()
-	theirControl.Start()
-
-	t.Log("Trigger a handshake from me to them via the relay")
-	myControl.InjectTunUDPPacket(theirVpnIpNet[0].Addr(), 80, myVpnIpNet[0].Addr(), 80, []byte("Hi from me"))
-
-	p := r.RouteForAllUntilTxTun(theirControl)
-	r.Log("Assert the tunnel works")
-	assertUdpPacket(t, []byte("Hi from me"), p, myVpnIpNet[0].Addr(), theirVpnIpNet[0].Addr(), 80, 80)
-	r.RenderHostmaps("Final hostmaps", myControl, relayControl, theirControl)
-}
-
 func TestReestablishRelays(t *testing.T) {
 	ca, _, caKey, _ := cert_test.NewTestCaCert(cert.Version1, cert.Curve_CURVE25519, time.Now(), time.Now().Add(10*time.Minute), nil, nil, []string{})
 	myControl, myVpnIpNet, _, _ := newSimpleServer(cert.Version1, ca, caKey, "me     ", "10.128.0.1/24", m{"relay": m{"use_relays": true}})
@@ -1291,109 +1227,3 @@ func TestV2NonPrimaryWithLighthouse(t *testing.T) {
 	myControl.Stop()
 	theirControl.Stop()
 }
-
-func TestV2NonPrimaryWithOffNetLighthouse(t *testing.T) {
-	ca, _, caKey, _ := cert_test.NewTestCaCert(cert.Version2, cert.Curve_CURVE25519, time.Now(), time.Now().Add(10*time.Minute), nil, nil, []string{})
-	lhControl, lhVpnIpNet, lhUdpAddr, _ := newSimpleServer(cert.Version2, ca, caKey, "lh  ", "2001::1/64", m{"lighthouse": m{"am_lighthouse": true}})
-
-	o := m{
-		"static_host_map": m{
-			lhVpnIpNet[0].Addr().String(): []string{lhUdpAddr.String()},
-		},
-		"lighthouse": m{
-			"hosts": []string{lhVpnIpNet[0].Addr().String()},
-			"local_allow_list": m{
-				// Try and block our lighthouse updates from using the actual addresses assigned to this computer
-				// If we start discovering addresses the test router doesn't know about then test traffic cant flow
-				"10.0.0.0/24": true,
-				"::/0":        false,
-			},
-		},
-	}
-	myControl, myVpnIpNet, _, _ := newSimpleServer(cert.Version2, ca, caKey, "me  ", "10.128.0.2/24, ff::2/64", o)
-	theirControl, theirVpnIpNet, _, _ := newSimpleServer(cert.Version2, ca, caKey, "them", "10.128.0.3/24, ff::3/64", o)
-
-	// Build a router so we don't have to reason who gets which packet
-	r := router.NewR(t, lhControl, myControl, theirControl)
-	defer r.RenderFlow()
-
-	// Start the servers
-	lhControl.Start()
-	myControl.Start()
-	theirControl.Start()
-
-	t.Log("Stand up an ipv6 tunnel between me and them")
-	assert.True(t, myVpnIpNet[1].Addr().Is6())
-	assert.True(t, theirVpnIpNet[1].Addr().Is6())
-	assertTunnel(t, myVpnIpNet[1].Addr(), theirVpnIpNet[1].Addr(), myControl, theirControl, r)
-
-	lhControl.Stop()
-	myControl.Stop()
-	theirControl.Stop()
-}
-
-func TestGoodHandshakeUnsafeDest(t *testing.T) {
-	unsafePrefix := "192.168.6.0/24"
-	ca, _, caKey, _ := cert_test.NewTestCaCert(cert.Version2, cert.Curve_CURVE25519, time.Now(), time.Now().Add(10*time.Minute), nil, nil, []string{})
-	theirControl, theirVpnIpNet, theirUdpAddr, _ := newSimpleServerWithUdpAndUnsafeNetworks(cert.Version2, ca, caKey, "spooky", "10.128.0.2/24", netip.MustParseAddrPort("10.64.0.2:4242"), unsafePrefix, nil)
-	route := m{"route": unsafePrefix, "via": theirVpnIpNet[0].Addr().String()}
-	myCfg := m{
-		"tun": m{
-			"unsafe_routes": []m{route},
-		},
-	}
-	myControl, myVpnIpNet, myUdpAddr, myConfig := newSimpleServer(cert.Version2, ca, caKey, "me", "10.128.0.1/24", myCfg)
-	t.Logf("my config %v", myConfig)
-	// Put their info in our lighthouse
-	myControl.InjectLightHouseAddr(theirVpnIpNet[0].Addr(), theirUdpAddr)
-
-	spookyDest := netip.MustParseAddr("192.168.6.4")
-
-	// Start the servers
-	myControl.Start()
-	theirControl.Start()
-
-	t.Log("Send a udp packet through to begin standing up the tunnel, this should come out the other side")
-	myControl.InjectTunUDPPacket(spookyDest, 80, myVpnIpNet[0].Addr(), 80, []byte("Hi from me"))
-
-	t.Log("Have them consume my stage 0 packet. They have a tunnel now")
-	theirControl.InjectUDPPacket(myControl.GetFromUDP(true))
-
-	t.Log("Get their stage 1 packet so that we can play with it")
-	stage1Packet := theirControl.GetFromUDP(true)
-
-	t.Log("I consume a garbage packet with a proper nebula header for our tunnel")
-	// this should log a statement and get ignored, allowing the real handshake packet to complete the tunnel
-	badPacket := stage1Packet.Copy()
-	badPacket.Data = badPacket.Data[:len(badPacket.Data)-header.Len]
-	myControl.InjectUDPPacket(badPacket)
-
-	t.Log("Have me consume their real stage 1 packet. I have a tunnel now")
-	myControl.InjectUDPPacket(stage1Packet)
-
-	t.Log("Wait until we see my cached packet come through")
-	myControl.WaitForType(1, 0, theirControl)
-
-	t.Log("Make sure our host infos are correct")
-	assertHostInfoPair(t, myUdpAddr, theirUdpAddr, myVpnIpNet, theirVpnIpNet, myControl, theirControl)
-
-	t.Log("Get that cached packet and make sure it looks right")
-	myCachedPacket := theirControl.GetFromTun(true)
-	assertUdpPacket(t, []byte("Hi from me"), myCachedPacket, myVpnIpNet[0].Addr(), spookyDest, 80, 80)
-
-	//reply
-	theirControl.InjectTunUDPPacket(myVpnIpNet[0].Addr(), 80, spookyDest, 80, []byte("Hi from the spookyman"))
-	//wait for reply
-	theirControl.WaitForType(1, 0, myControl)
-	theirCachedPacket := myControl.GetFromTun(true)
-	assertUdpPacket(t, []byte("Hi from the spookyman"), theirCachedPacket, spookyDest, myVpnIpNet[0].Addr(), 80, 80)
-
-	t.Log("Do a bidirectional tunnel test")
-	r := router.NewR(t, myControl, theirControl)
-	defer r.RenderFlow()
-	assertTunnel(t, myVpnIpNet[0].Addr(), theirVpnIpNet[0].Addr(), myControl, theirControl, r)
-
-	r.RenderHostmaps("Final hostmaps", myControl, theirControl)
-	myControl.Stop()
-	theirControl.Stop()
-}

e2e/helpers_test.go

@@ -22,14 +22,15 @@ import (
 	"github.com/slackhq/nebula/config"
 	"github.com/slackhq/nebula/e2e/router"
 	"github.com/stretchr/testify/assert"
-	"github.com/stretchr/testify/require"
-	"go.yaml.in/yaml/v3"
+	"gopkg.in/yaml.v3"
 )
 
 type m = map[string]any
 
 // newSimpleServer creates a nebula instance with many assumptions
 func newSimpleServer(v cert.Version, caCrt cert.Certificate, caKey []byte, name string, sVpnNetworks string, overrides m) (*nebula.Control, []netip.Prefix, netip.AddrPort, *config.C) {
+	l := NewTestLogger()
+
 	var vpnNetworks []netip.Prefix
 	for _, sn := range strings.Split(sVpnNetworks, ",") {
 		vpnIpNet, err := netip.ParsePrefix(strings.TrimSpace(sn))
@@ -55,54 +56,7 @@ func newSimpleServer(v cert.Version, caCrt cert.Certificate, caKey []byte, name
 		budpIp[3] = 239
 		udpAddr = netip.AddrPortFrom(netip.AddrFrom16(budpIp), 4242)
 	}
-	return newSimpleServerWithUdp(v, caCrt, caKey, name, sVpnNetworks, udpAddr, overrides)
-}
-
-func newSimpleServerWithUdp(v cert.Version, caCrt cert.Certificate, caKey []byte, name string, sVpnNetworks string, udpAddr netip.AddrPort, overrides m) (*nebula.Control, []netip.Prefix, netip.AddrPort, *config.C) {
-	return newSimpleServerWithUdpAndUnsafeNetworks(v, caCrt, caKey, name, sVpnNetworks, udpAddr, "", overrides)
-}
-
-func newSimpleServerWithUdpAndUnsafeNetworks(v cert.Version, caCrt cert.Certificate, caKey []byte, name string, sVpnNetworks string, udpAddr netip.AddrPort, sUnsafeNetworks string, overrides m) (*nebula.Control, []netip.Prefix, netip.AddrPort, *config.C) {
-	l := NewTestLogger()
-
-	var vpnNetworks []netip.Prefix
-	for _, sn := range strings.Split(sVpnNetworks, ",") {
-		vpnIpNet, err := netip.ParsePrefix(strings.TrimSpace(sn))
-		if err != nil {
-			panic(err)
-		}
-		vpnNetworks = append(vpnNetworks, vpnIpNet)
-	}
-
-	if len(vpnNetworks) == 0 {
-		panic("no vpn networks")
-	}
-
-	firewallInbound := []m{{
-		"proto": "any",
-		"port":  "any",
-		"host":  "any",
-	}}
-
-	var unsafeNetworks []netip.Prefix
-	if sUnsafeNetworks != "" {
-		firewallInbound = []m{{
-			"proto":      "any",
-			"port":       "any",
-			"host":       "any",
-			"local_cidr": "0.0.0.0/0",
-		}}
-
-		for _, sn := range strings.Split(sUnsafeNetworks, ",") {
-			x, err := netip.ParsePrefix(strings.TrimSpace(sn))
-			if err != nil {
-				panic(err)
-			}
-			unsafeNetworks = append(unsafeNetworks, x)
-		}
-	}
-
-	_, _, myPrivKey, myPEM := cert_test.NewTestCert(v, cert.Curve_CURVE25519, caCrt, caKey, name, time.Now(), time.Now().Add(5*time.Minute), vpnNetworks, unsafeNetworks, []string{})
+	_, _, myPrivKey, myPEM := cert_test.NewTestCert(v, cert.Curve_CURVE25519, caCrt, caKey, name, time.Now(), time.Now().Add(5*time.Minute), vpnNetworks, nil, []string{})
 
 	caB, err := caCrt.MarshalPEM()
 	if err != nil {
@@ -122,7 +76,11 @@ func newSimpleServerWithUdpAndUnsafeNetworks(v cert.Version, caCrt cert.Certific
 				"port":  "any",
 				"host":  "any",
 			}},
-			"inbound": firewallInbound,
+			"inbound": []m{{
+				"proto": "any",
+				"port":  "any",
+				"host":  "any",
+			}},
 		},
 		//"handshakes": m{
 		//	"try_interval": "1s",
@@ -171,109 +129,6 @@ func newSimpleServerWithUdpAndUnsafeNetworks(v cert.Version, caCrt cert.Certific
 	return control, vpnNetworks, udpAddr, c
 }
 
-// newServer creates a nebula instance with fewer assumptions
-func newServer(caCrt []cert.Certificate, certs []cert.Certificate, key []byte, overrides m) (*nebula.Control, []netip.Prefix, netip.AddrPort, *config.C) {
-	l := NewTestLogger()
-
-	vpnNetworks := certs[len(certs)-1].Networks()
-
-	var udpAddr netip.AddrPort
-	if vpnNetworks[0].Addr().Is4() {
-		budpIp := vpnNetworks[0].Addr().As4()
-		budpIp[1] -= 128
-		udpAddr = netip.AddrPortFrom(netip.AddrFrom4(budpIp), 4242)
-	} else {
-		budpIp := vpnNetworks[0].Addr().As16()
-		// beef for funsies
-		budpIp[2] = 190
-		budpIp[3] = 239
-		udpAddr = netip.AddrPortFrom(netip.AddrFrom16(budpIp), 4242)
-	}
-
-	caStr := ""
-	for _, ca := range caCrt {
-		x, err := ca.MarshalPEM()
-		if err != nil {
-			panic(err)
-		}
-		caStr += string(x)
-	}
-	certStr := ""
-	for _, c := range certs {
-		x, err := c.MarshalPEM()
-		if err != nil {
-			panic(err)
-		}
-		certStr += string(x)
-	}
-
-	mc := m{
-		"pki": m{
-			"ca":   caStr,
-			"cert": certStr,
-			"key":  string(key),
-		},
-		//"tun": m{"disabled": true},
-		"firewall": m{
-			"outbound": []m{{
-				"proto": "any",
-				"port":  "any",
-				"host":  "any",
-			}},
-			"inbound": []m{{
-				"proto": "any",
-				"port":  "any",
-				"host":  "any",
-			}},
-		},
-		//"handshakes": m{
-		//	"try_interval": "1s",
-		//},
-		"listen": m{
-			"host": udpAddr.Addr().String(),
-			"port": udpAddr.Port(),
-		},
-		"logging": m{
-			"timestamp_format": fmt.Sprintf("%v 15:04:05.000000", certs[0].Name()),
-			"level":            l.Level.String(),
-		},
-		"timers": m{
-			"pending_deletion_interval": 2,
-			"connection_alive_interval": 2,
-		},
-	}
-
-	if overrides != nil {
-		final := m{}
-		err := mergo.Merge(&final, overrides, mergo.WithAppendSlice)
-		if err != nil {
-			panic(err)
-		}
-		err = mergo.Merge(&final, mc, mergo.WithAppendSlice)
-		if err != nil {
-			panic(err)
-		}
-		mc = final
-	}
-
-	cb, err := yaml.Marshal(mc)
-	if err != nil {
-		panic(err)
-	}
-
-	c := config.NewC(l)
-	cStr := string(cb)
-	c.LoadString(cStr)
-
-	control, err := nebula.Main(c, false, "e2e-test", l, nil)
-
-	if err != nil {
-		panic(err)
-	}
-
-	return control, vpnNetworks, udpAddr, c
-}
-
 type doneCb func()
 
 func deadline(t *testing.T, seconds time.Duration) doneCb {
@@ -308,10 +163,10 @@ func assertHostInfoPair(t *testing.T, addrA, addrB netip.AddrPort, vpnNetsA, vpn
 	// Get both host infos
 	//TODO: CERT-V2 we may want to loop over each vpnAddr and assert all the things
 	hBinA := controlA.GetHostInfoByVpnAddr(vpnNetsB[0].Addr(), false)
-	require.NotNil(t, hBinA, "Host B was not found by vpnAddr in controlA")
+	assert.NotNil(t, hBinA, "Host B was not found by vpnAddr in controlA")
 
 	hAinB := controlB.GetHostInfoByVpnAddr(vpnNetsA[0].Addr(), false)
-	require.NotNil(t, hAinB, "Host A was not found by vpnAddr in controlB")
+	assert.NotNil(t, hAinB, "Host A was not found by vpnAddr in controlB")
 
 	// Check that both vpn and real addr are correct
 	assert.EqualValues(t, getAddrs(vpnNetsB), hBinA.VpnAddrs, "Host B VpnIp is wrong in control A")

e2e/tunnels_test.go

@@ -4,16 +4,12 @@
 package e2e
 
 import (
-	"fmt"
-	"net/netip"
 	"testing"
 	"time"
 
 	"github.com/slackhq/nebula/cert"
 	"github.com/slackhq/nebula/cert_test"
 	"github.com/slackhq/nebula/e2e/router"
-	"github.com/stretchr/testify/assert"
-	"gopkg.in/yaml.v3"
 )
 
 func TestDropInactiveTunnels(t *testing.T) {
@@ -59,309 +55,3 @@ func TestDropInactiveTunnels(t *testing.T) {
 	myControl.Stop()
 	theirControl.Stop()
 }
-
-func TestCertUpgrade(t *testing.T) {
-	// The goal of this test is to ensure the shortest inactivity timeout will close the tunnel on both sides
-	// under ideal conditions
-	ca, _, caKey, _ := cert_test.NewTestCaCert(cert.Version1, cert.Curve_CURVE25519, time.Now(), time.Now().Add(10*time.Minute), nil, nil, []string{})
-	caB, err := ca.MarshalPEM()
-	if err != nil {
-		panic(err)
-	}
-	ca2, _, caKey2, _ := cert_test.NewTestCaCert(cert.Version2, cert.Curve_CURVE25519, time.Now(), time.Now().Add(10*time.Minute), nil, nil, []string{})
-
-	ca2B, err := ca2.MarshalPEM()
-	if err != nil {
-		panic(err)
-	}
-	caStr := fmt.Sprintf("%s\n%s", caB, ca2B)
-
-	myCert, _, myPrivKey, _ := cert_test.NewTestCert(cert.Version1, cert.Curve_CURVE25519, ca, caKey, "me", time.Now(), time.Now().Add(5*time.Minute), []netip.Prefix{netip.MustParsePrefix("10.128.0.1/24")}, nil, []string{})
-	_, myCert2Pem := cert_test.NewTestCertDifferentVersion(myCert, cert.Version2, ca2, caKey2)
-
-	theirCert, _, theirPrivKey, _ := cert_test.NewTestCert(cert.Version1, cert.Curve_CURVE25519, ca, caKey, "them", time.Now(), time.Now().Add(5*time.Minute), []netip.Prefix{netip.MustParsePrefix("10.128.0.2/24")}, nil, []string{})
-	theirCert2, _ := cert_test.NewTestCertDifferentVersion(theirCert, cert.Version2, ca2, caKey2)
-
-	myControl, myVpnIpNet, myUdpAddr, myC := newServer([]cert.Certificate{ca, ca2}, []cert.Certificate{myCert}, myPrivKey, m{})
-	theirControl, theirVpnIpNet, theirUdpAddr, _ := newServer([]cert.Certificate{ca, ca2}, []cert.Certificate{theirCert, theirCert2}, theirPrivKey, m{})
-
-	// Share our underlay information
-	myControl.InjectLightHouseAddr(theirVpnIpNet[0].Addr(), theirUdpAddr)
-	theirControl.InjectLightHouseAddr(myVpnIpNet[0].Addr(), myUdpAddr)
-
-	// Start the servers
-	myControl.Start()
-	theirControl.Start()
-
-	r := router.NewR(t, myControl, theirControl)
-	defer r.RenderFlow()
-
-	r.Log("Assert the tunnel between me and them works")
-	assertTunnel(t, myVpnIpNet[0].Addr(), theirVpnIpNet[0].Addr(), myControl, theirControl, r)
-	r.Log("yay")
-	//todo ???
-	time.Sleep(1 * time.Second)
-	r.FlushAll()
-
-	mc := m{
-		"pki": m{
-			"ca":   caStr,
-			"cert": string(myCert2Pem),
-			"key":  string(myPrivKey),
-		},
-		//"tun": m{"disabled": true},
-		"firewall": myC.Settings["firewall"],
-		//"handshakes": m{
-		//	"try_interval": "1s",
-		//},
-		"listen":  myC.Settings["listen"],
-		"logging": myC.Settings["logging"],
-		"timers":  myC.Settings["timers"],
-	}
-
-	cb, err := yaml.Marshal(mc)
-	if err != nil {
-		panic(err)
-	}
-
-	r.Logf("reload new v2-only config")
-	err = myC.ReloadConfigString(string(cb))
-	assert.NoError(t, err)
-	r.Log("yay, spin until their sees it")
-	waitStart := time.Now()
-	for {
-		assertTunnel(t, myVpnIpNet[0].Addr(), theirVpnIpNet[0].Addr(), myControl, theirControl, r)
-		c := theirControl.GetHostInfoByVpnAddr(myVpnIpNet[0].Addr(), false)
-		if c == nil {
-			r.Log("nil")
-		} else {
-			version := c.Cert.Version()
-			r.Logf("version %d", version)
-			if version == cert.Version2 {
-				break
-			}
-		}
-		since := time.Since(waitStart)
-		if since > time.Second*10 {
-			t.Fatal("Cert should be new by now")
-		}
-		time.Sleep(time.Second)
-	}
-
-	r.RenderHostmaps("Final hostmaps", myControl, theirControl)
-
-	myControl.Stop()
-	theirControl.Stop()
-}
-
-func TestCertDowngrade(t *testing.T) {
-	// The goal of this test is to ensure the shortest inactivity timeout will close the tunnel on both sides
-	// under ideal conditions
-	ca, _, caKey, _ := cert_test.NewTestCaCert(cert.Version1, cert.Curve_CURVE25519, time.Now(), time.Now().Add(10*time.Minute), nil, nil, []string{})
-	caB, err := ca.MarshalPEM()
-	if err != nil {
-		panic(err)
-	}
-	ca2, _, caKey2, _ := cert_test.NewTestCaCert(cert.Version2, cert.Curve_CURVE25519, time.Now(), time.Now().Add(10*time.Minute), nil, nil, []string{})
-
-	ca2B, err := ca2.MarshalPEM()
-	if err != nil {
-		panic(err)
-	}
-	caStr := fmt.Sprintf("%s\n%s", caB, ca2B)
-
-	myCert, _, myPrivKey, myCertPem := cert_test.NewTestCert(cert.Version1, cert.Curve_CURVE25519, ca, caKey, "me", time.Now(), time.Now().Add(5*time.Minute), []netip.Prefix{netip.MustParsePrefix("10.128.0.1/24")}, nil, []string{})
-	myCert2, _ := cert_test.NewTestCertDifferentVersion(myCert, cert.Version2, ca2, caKey2)
-
-	theirCert, _, theirPrivKey, _ := cert_test.NewTestCert(cert.Version1, cert.Curve_CURVE25519, ca, caKey, "them", time.Now(), time.Now().Add(5*time.Minute), []netip.Prefix{netip.MustParsePrefix("10.128.0.2/24")}, nil, []string{})
-	theirCert2, _ := cert_test.NewTestCertDifferentVersion(theirCert, cert.Version2, ca2, caKey2)
-
-	myControl, myVpnIpNet, myUdpAddr, myC := newServer([]cert.Certificate{ca, ca2}, []cert.Certificate{myCert2}, myPrivKey, m{})
-	theirControl, theirVpnIpNet, theirUdpAddr, _ := newServer([]cert.Certificate{ca, ca2}, []cert.Certificate{theirCert, theirCert2}, theirPrivKey, m{})
-
-	// Share our underlay information
-	myControl.InjectLightHouseAddr(theirVpnIpNet[0].Addr(), theirUdpAddr)
-	theirControl.InjectLightHouseAddr(myVpnIpNet[0].Addr(), myUdpAddr)
-
-	// Start the servers
-	myControl.Start()
-	theirControl.Start()
-
-	r := router.NewR(t, myControl, theirControl)
-	defer r.RenderFlow()
-
-	r.Log("Assert the tunnel between me and them works")
-	//assertTunnel(t, theirVpnIpNet[0].Addr(), myVpnIpNet[0].Addr(), theirControl, myControl, r)
-	//r.Log("yay")
-	assertTunnel(t, myVpnIpNet[0].Addr(), theirVpnIpNet[0].Addr(), myControl, theirControl, r)
-	r.Log("yay")
-	//todo ???
-	time.Sleep(1 * time.Second)
-	r.FlushAll()
-
-	mc := m{
-		"pki": m{
-			"ca":   caStr,
-			"cert": string(myCertPem),
-			"key":  string(myPrivKey),
-		},
-		"firewall": myC.Settings["firewall"],
-		"listen":   myC.Settings["listen"],
-		"logging":  myC.Settings["logging"],
-		"timers":   myC.Settings["timers"],
-	}
-
-	cb, err := yaml.Marshal(mc)
-	if err != nil {
-		panic(err)
-	}
-
-	r.Logf("reload new v1-only config")
-	err = myC.ReloadConfigString(string(cb))
-	assert.NoError(t, err)
-	r.Log("yay, spin until their sees it")
-	waitStart := time.Now()
-	for {
-		assertTunnel(t, myVpnIpNet[0].Addr(), theirVpnIpNet[0].Addr(), myControl, theirControl, r)
-		c := theirControl.GetHostInfoByVpnAddr(myVpnIpNet[0].Addr(), false)
-		c2 := myControl.GetHostInfoByVpnAddr(theirVpnIpNet[0].Addr(), false)
-		if c == nil || c2 == nil {
-			r.Log("nil")
-		} else {
-			version := c.Cert.Version()
-			theirVersion := c2.Cert.Version()
-			r.Logf("version %d,%d", version, theirVersion)
-			if version == cert.Version1 {
-				break
-			}
-		}
-		since := time.Since(waitStart)
-		if since > time.Second*5 {
-			r.Log("it is unusual that the cert is not new yet, but not a failure yet")
-		}
-		if since > time.Second*10 {
-			r.Log("wtf")
-			t.Fatal("Cert should be new by now")
-		}
-		time.Sleep(time.Second)
-	}
-
-	r.RenderHostmaps("Final hostmaps", myControl, theirControl)
-
-	myControl.Stop()
-	theirControl.Stop()
-}
-
-func TestCertMismatchCorrection(t *testing.T) {
-	// The goal of this test is to ensure the shortest inactivity timeout will close the tunnel on both sides
-	// under ideal conditions
-	ca, _, caKey, _ := cert_test.NewTestCaCert(cert.Version1, cert.Curve_CURVE25519, time.Now(), time.Now().Add(10*time.Minute), nil, nil, []string{})
-	ca2, _, caKey2, _ := cert_test.NewTestCaCert(cert.Version2, cert.Curve_CURVE25519, time.Now(), time.Now().Add(10*time.Minute), nil, nil, []string{})
-
-	myCert, _, myPrivKey, _ := cert_test.NewTestCert(cert.Version1, cert.Curve_CURVE25519, ca, caKey, "me", time.Now(), time.Now().Add(5*time.Minute), []netip.Prefix{netip.MustParsePrefix("10.128.0.1/24")}, nil, []string{})
-	myCert2, _ := cert_test.NewTestCertDifferentVersion(myCert, cert.Version2, ca2, caKey2)
-
-	theirCert, _, theirPrivKey, _ := cert_test.NewTestCert(cert.Version1, cert.Curve_CURVE25519, ca, caKey, "them", time.Now(), time.Now().Add(5*time.Minute), []netip.Prefix{netip.MustParsePrefix("10.128.0.2/24")}, nil, []string{})
-	theirCert2, _ := cert_test.NewTestCertDifferentVersion(theirCert, cert.Version2, ca2, caKey2)
-
-	myControl, myVpnIpNet, myUdpAddr, _ := newServer([]cert.Certificate{ca, ca2}, []cert.Certificate{myCert2}, myPrivKey, m{})
-	theirControl, theirVpnIpNet, theirUdpAddr, _ := newServer([]cert.Certificate{ca, ca2}, []cert.Certificate{theirCert, theirCert2}, theirPrivKey, m{})
-
-	// Share our underlay information
-	myControl.InjectLightHouseAddr(theirVpnIpNet[0].Addr(), theirUdpAddr)
-	theirControl.InjectLightHouseAddr(myVpnIpNet[0].Addr(), myUdpAddr)
-
-	// Start the servers
-	myControl.Start()
-	theirControl.Start()
-
-	r := router.NewR(t, myControl, theirControl)
-	defer r.RenderFlow()
-
-	r.Log("Assert the tunnel between me and them works")
-	//assertTunnel(t, theirVpnIpNet[0].Addr(), myVpnIpNet[0].Addr(), theirControl, myControl, r)
-	//r.Log("yay")
-	assertTunnel(t, myVpnIpNet[0].Addr(), theirVpnIpNet[0].Addr(), myControl, theirControl, r)
-	r.Log("yay")
-	//todo ???
-	time.Sleep(1 * time.Second)
-	r.FlushAll()
-
-	waitStart := time.Now()
-	for {
-		assertTunnel(t, myVpnIpNet[0].Addr(), theirVpnIpNet[0].Addr(), myControl, theirControl, r)
-		c := theirControl.GetHostInfoByVpnAddr(myVpnIpNet[0].Addr(), false)
-		c2 := myControl.GetHostInfoByVpnAddr(theirVpnIpNet[0].Addr(), false)
-		if c == nil || c2 == nil {
-			r.Log("nil")
-		} else {
-			version := c.Cert.Version()
-			theirVersion := c2.Cert.Version()
-			r.Logf("version %d,%d", version, theirVersion)
-			if version == theirVersion {
-				break
-			}
-		}
-		since := time.Since(waitStart)
-		if since > time.Second*5 {
-			r.Log("wtf")
-		}
-		if since > time.Second*10 {
-			r.Log("wtf")
-			t.Fatal("Cert should be new by now")
-		}
-		time.Sleep(time.Second)
-	}
-
-	r.RenderHostmaps("Final hostmaps", myControl, theirControl)
-
-	myControl.Stop()
-	theirControl.Stop()
-}
-
-func TestCrossStackRelaysWork(t *testing.T) {
-	ca, _, caKey, _ := cert_test.NewTestCaCert(cert.Version2, cert.Curve_CURVE25519, time.Now(), time.Now().Add(10*time.Minute), nil, nil, []string{})
-	myControl, myVpnIpNet, _, _ := newSimpleServer(cert.Version2, ca, caKey, "me     ", "10.128.0.1/24,fc00::1/64", m{"relay": m{"use_relays": true}})
-	relayControl, relayVpnIpNet, relayUdpAddr, _ := newSimpleServer(cert.Version2, ca, caKey, "relay  ", "10.128.0.128/24,fc00::128/64", m{"relay": m{"am_relay": true}})
-	theirUdp := netip.MustParseAddrPort("10.0.0.2:4242")
-	theirControl, theirVpnIpNet, theirUdpAddr, _ := newSimpleServerWithUdp(cert.Version2, ca, caKey, "them   ", "fc00::2/64", theirUdp, m{"relay": m{"use_relays": true}})
-
-	//myVpnV4 := myVpnIpNet[0]
-	myVpnV6 := myVpnIpNet[1]
-	relayVpnV4 := relayVpnIpNet[0]
-	relayVpnV6 := relayVpnIpNet[1]
-	theirVpnV6 := theirVpnIpNet[0]
-
-	// Teach my how to get to the relay and that their can be reached via the relay
-	myControl.InjectLightHouseAddr(relayVpnV4.Addr(), relayUdpAddr)
-	myControl.InjectLightHouseAddr(relayVpnV6.Addr(), relayUdpAddr)
-	myControl.InjectRelays(theirVpnV6.Addr(), []netip.Addr{relayVpnV6.Addr()})
-	relayControl.InjectLightHouseAddr(theirVpnV6.Addr(), theirUdpAddr)
-
-	// Build a router so we don't have to reason who gets which packet
-	r := router.NewR(t, myControl, relayControl, theirControl)
-	defer r.RenderFlow()
-
-	// Start the servers
-	myControl.Start()
-	relayControl.Start()
-	theirControl.Start()
-
-	t.Log("Trigger a handshake from me to them via the relay")
-	myControl.InjectTunUDPPacket(theirVpnV6.Addr(), 80, myVpnV6.Addr(), 80, []byte("Hi from me"))
-
-	p := r.RouteForAllUntilTxTun(theirControl)
-	r.Log("Assert the tunnel works")
-	assertUdpPacket(t, []byte("Hi from me"), p, myVpnV6.Addr(), theirVpnV6.Addr(), 80, 80)
-
-	t.Log("reply?")
-	theirControl.InjectTunUDPPacket(myVpnV6.Addr(), 80, theirVpnV6.Addr(), 80, []byte("Hi from them"))
-	p = r.RouteForAllUntilTxTun(myControl)
-	assertUdpPacket(t, []byte("Hi from them"), p, theirVpnV6.Addr(), myVpnV6.Addr(), 80, 80)
-
-	r.RenderHostmaps("Final hostmaps", myControl, relayControl, theirControl)
-	//t.Log("finish up")
-	//myControl.Stop()
-	//theirControl.Stop()
-	//relayControl.Stop()
-}

examples/config.yml

@@ -132,6 +132,13 @@ listen:
   # Sets the max number of packets to pull from the kernel for each syscall (under systems that support recvmmsg)
   # default is 64, does not support reload
   #batch: 64
+
+# Control batching between UDP and TUN pipelines
+#batch:
+#  inbound_size: 32        # packets to queue from UDP before handing to workers
+#  outbound_size: 32       # packets to queue from TUN before handing to workers
+#  flush_interval: 50us    # flush partially filled batches after this duration
+#  max_outstanding: 1028   # batches buffered per routine on each channel
   # Configure socket buffers for the udp side (outside), leave unset to use the system defaults. Values will be doubled by the kernel
   # Default is net.core.rmem_default and net.core.wmem_default (/proc/sys/net/core/rmem_default and /proc/sys/net/core/rmem_default)
   # Maximum is limited by memory in the system, SO_RCVBUFFORCE and SO_SNDBUFFORCE is used to avoid having to raise the system wide

firewall.go

@@ -417,10 +417,8 @@ func AddFirewallRulesFromConfig(l *logrus.Logger, inbound bool, c *config.C, fw
 	return nil
 }
 
-var ErrUnknownNetworkType = errors.New("unknown network type")
-var ErrPeerRejected = errors.New("remote address is not within a network that we handle")
-var ErrInvalidRemoteIP = errors.New("remote address is not in remote certificate networks")
-var ErrInvalidLocalIP = errors.New("local address is not in list of handled local addresses")
+var ErrInvalidRemoteIP = errors.New("remote IP is not in remote certificate subnets")
+var ErrInvalidLocalIP = errors.New("local IP is not in list of handled local IPs")
 var ErrNoMatchingRule = errors.New("no matching rule in firewall table")
 
 // Drop returns an error if the packet should be dropped, explaining why. It
@@ -431,31 +429,18 @@ func (f *Firewall) Drop(fp firewall.Packet, incoming bool, h *HostInfo, caPool *
 		return nil
 	}
 
-	// Make sure remote address matches nebula certificate, and determine how to treat it
-	if h.networks == nil {
-		// Simple case: Certificate has one address and no unsafe networks
-		if h.vpnAddrs[0] != fp.RemoteAddr {
+	// Make sure remote address matches nebula certificate
+	if h.networks != nil {
+		if !h.networks.Contains(fp.RemoteAddr) {
 			f.metrics(incoming).droppedRemoteAddr.Inc(1)
 			return ErrInvalidRemoteIP
 		}
 	} else {
-		nwType, ok := h.networks.Lookup(fp.RemoteAddr)
-		if !ok {
+		// Simple case: Certificate has one address and no unsafe networks
+		if h.vpnAddrs[0] != fp.RemoteAddr {
 			f.metrics(incoming).droppedRemoteAddr.Inc(1)
 			return ErrInvalidRemoteIP
 		}
-		switch nwType {
-		case NetworkTypeVPN:
-			break // nothing special
-		case NetworkTypeVPNPeer:
-			f.metrics(incoming).droppedRemoteAddr.Inc(1)
-			return ErrPeerRejected // reject for now, one day this may have different FW rules
-		case NetworkTypeUnsafe:
-			break // nothing special, one day this may have different FW rules
-		default:
-			f.metrics(incoming).droppedRemoteAddr.Inc(1)
-			return ErrUnknownNetworkType //should never happen
-		}
 	}
 
 	// Make sure we are supposed to be handling this local ip address

firewall_test.go

@@ -8,8 +8,6 @@ import (
 	"testing"
 	"time"
 
-	"github.com/gaissmai/bart"
-	"github.com/sirupsen/logrus"
 	"github.com/slackhq/nebula/cert"
 	"github.com/slackhq/nebula/config"
 	"github.com/slackhq/nebula/firewall"
@@ -106,13 +104,13 @@ func TestFirewall_AddRule(t *testing.T) {
 	fw = NewFirewall(l, time.Second, time.Minute, time.Hour, c)
 	require.NoError(t, fw.AddRule(false, firewall.ProtoAny, 1, 1, []string{}, "", netip.Prefix{}, ti, "", ""))
 	assert.NotNil(t, fw.OutRules.AnyProto[1].Any.Any)
-	ok = fw.OutRules.AnyProto[1].Any.Any.LocalCIDR.Get(ti)
+	_, ok = fw.OutRules.AnyProto[1].Any.Any.LocalCIDR.Get(ti)
 	assert.True(t, ok)
 
 	fw = NewFirewall(l, time.Second, time.Minute, time.Hour, c)
 	require.NoError(t, fw.AddRule(false, firewall.ProtoAny, 1, 1, []string{}, "", netip.Prefix{}, ti6, "", ""))
 	assert.NotNil(t, fw.OutRules.AnyProto[1].Any.Any)
-	ok = fw.OutRules.AnyProto[1].Any.Any.LocalCIDR.Get(ti6)
+	_, ok = fw.OutRules.AnyProto[1].Any.Any.LocalCIDR.Get(ti6)
 	assert.True(t, ok)
 
 	fw = NewFirewall(l, time.Second, time.Minute, time.Hour, c)
@@ -151,8 +149,7 @@ func TestFirewall_Drop(t *testing.T) {
 	l := test.NewLogger()
 	ob := &bytes.Buffer{}
 	l.SetOutput(ob)
-	myVpnNetworksTable := new(bart.Lite)
-	myVpnNetworksTable.Insert(netip.MustParsePrefix("1.1.1.1/8"))
+
 	p := firewall.Packet{
 		LocalAddr:  netip.MustParseAddr("1.2.3.4"),
 		RemoteAddr: netip.MustParseAddr("1.2.3.4"),
@@ -177,7 +174,7 @@ func TestFirewall_Drop(t *testing.T) {
 		},
 		vpnAddrs: []netip.Addr{netip.MustParseAddr("1.2.3.4")},
 	}
-	h.buildNetworks(myVpnNetworksTable, &c)
+	h.buildNetworks(c.networks, c.unsafeNetworks)
 
 	fw := NewFirewall(l, time.Second, time.Minute, time.Hour, &c)
 	require.NoError(t, fw.AddRule(true, firewall.ProtoAny, 0, 0, []string{"any"}, "", netip.Prefix{}, netip.Prefix{}, "", ""))
@@ -229,9 +226,6 @@ func TestFirewall_DropV6(t *testing.T) {
 	ob := &bytes.Buffer{}
 	l.SetOutput(ob)
 
-	myVpnNetworksTable := new(bart.Lite)
-	myVpnNetworksTable.Insert(netip.MustParsePrefix("fd00::/7"))
-
 	p := firewall.Packet{
 		LocalAddr:  netip.MustParseAddr("fd12::34"),
 		RemoteAddr: netip.MustParseAddr("fd12::34"),
@@ -256,7 +250,7 @@ func TestFirewall_DropV6(t *testing.T) {
 		},
 		vpnAddrs: []netip.Addr{netip.MustParseAddr("fd12::34")},
 	}
-	h.buildNetworks(myVpnNetworksTable, &c)
+	h.buildNetworks(c.networks, c.unsafeNetworks)
 
 	fw := NewFirewall(l, time.Second, time.Minute, time.Hour, &c)
 	require.NoError(t, fw.AddRule(true, firewall.ProtoAny, 0, 0, []string{"any"}, "", netip.Prefix{}, netip.Prefix{}, "", ""))
@@ -398,7 +392,7 @@ func BenchmarkFirewallTable_match(b *testing.B) {
 		c := &cert.CachedCertificate{
 			Certificate: &dummyCert{
 				name:     "nope",
-				networks: []netip.Prefix{netip.MustParsePrefix("fd99::99/128")},
+				networks: []netip.Prefix{netip.MustParsePrefix("fd99:99/128")},
 			},
 			InvertedGroups: map[string]struct{}{"nope": {}},
 		}
@@ -459,8 +453,6 @@ func TestFirewall_Drop2(t *testing.T) {
 	l := test.NewLogger()
 	ob := &bytes.Buffer{}
 	l.SetOutput(ob)
-	myVpnNetworksTable := new(bart.Lite)
-	myVpnNetworksTable.Insert(netip.MustParsePrefix("1.1.1.1/8"))
 
 	p := firewall.Packet{
 		LocalAddr:  netip.MustParseAddr("1.2.3.4"),
@@ -486,7 +478,7 @@ func TestFirewall_Drop2(t *testing.T) {
 		},
 		vpnAddrs: []netip.Addr{network.Addr()},
 	}
-	h.buildNetworks(myVpnNetworksTable, c.Certificate)
+	h.buildNetworks(c.Certificate.Networks(), c.Certificate.UnsafeNetworks())
 
 	c1 := cert.CachedCertificate{
 		Certificate: &dummyCert{
@@ -501,7 +493,7 @@ func TestFirewall_Drop2(t *testing.T) {
 			peerCert: &c1,
 		},
 	}
-	h1.buildNetworks(myVpnNetworksTable, c1.Certificate)
+	h1.buildNetworks(c1.Certificate.Networks(), c1.Certificate.UnsafeNetworks())
 
 	fw := NewFirewall(l, time.Second, time.Minute, time.Hour, c.Certificate)
 	require.NoError(t, fw.AddRule(true, firewall.ProtoAny, 0, 0, []string{"default-group", "test-group"}, "", netip.Prefix{}, netip.Prefix{}, "", ""))
@@ -518,8 +510,6 @@ func TestFirewall_Drop3(t *testing.T) {
 	l := test.NewLogger()
 	ob := &bytes.Buffer{}
 	l.SetOutput(ob)
-	myVpnNetworksTable := new(bart.Lite)
-	myVpnNetworksTable.Insert(netip.MustParsePrefix("1.1.1.1/8"))
 
 	p := firewall.Packet{
 		LocalAddr:  netip.MustParseAddr("1.2.3.4"),
@@ -551,7 +541,7 @@ func TestFirewall_Drop3(t *testing.T) {
 		},
 		vpnAddrs: []netip.Addr{network.Addr()},
 	}
-	h1.buildNetworks(myVpnNetworksTable, c1.Certificate)
+	h1.buildNetworks(c1.Certificate.Networks(), c1.Certificate.UnsafeNetworks())
 
 	c2 := cert.CachedCertificate{
 		Certificate: &dummyCert{
@@ -566,7 +556,7 @@ func TestFirewall_Drop3(t *testing.T) {
 		},
 		vpnAddrs: []netip.Addr{network.Addr()},
 	}
-	h2.buildNetworks(myVpnNetworksTable, c2.Certificate)
+	h2.buildNetworks(c2.Certificate.Networks(), c2.Certificate.UnsafeNetworks())
 
 	c3 := cert.CachedCertificate{
 		Certificate: &dummyCert{
@@ -581,7 +571,7 @@ func TestFirewall_Drop3(t *testing.T) {
 		},
 		vpnAddrs: []netip.Addr{network.Addr()},
 	}
-	h3.buildNetworks(myVpnNetworksTable, c3.Certificate)
+	h3.buildNetworks(c3.Certificate.Networks(), c3.Certificate.UnsafeNetworks())
 
 	fw := NewFirewall(l, time.Second, time.Minute, time.Hour, c.Certificate)
 	require.NoError(t, fw.AddRule(true, firewall.ProtoAny, 1, 1, []string{}, "host1", netip.Prefix{}, netip.Prefix{}, "", ""))
@@ -607,8 +597,6 @@ func TestFirewall_Drop3V6(t *testing.T) {
 	l := test.NewLogger()
 	ob := &bytes.Buffer{}
 	l.SetOutput(ob)
-	myVpnNetworksTable := new(bart.Lite)
-	myVpnNetworksTable.Insert(netip.MustParsePrefix("fd00::/7"))
 
 	p := firewall.Packet{
 		LocalAddr:  netip.MustParseAddr("fd12::34"),
@@ -632,7 +620,7 @@ func TestFirewall_Drop3V6(t *testing.T) {
 		},
 		vpnAddrs: []netip.Addr{network.Addr()},
 	}
-	h.buildNetworks(myVpnNetworksTable, c.Certificate)
+	h.buildNetworks(c.Certificate.Networks(), c.Certificate.UnsafeNetworks())
 
 	// Test a remote address match
 	fw := NewFirewall(l, time.Second, time.Minute, time.Hour, c.Certificate)
@@ -645,8 +633,6 @@ func TestFirewall_DropConntrackReload(t *testing.T) {
 	l := test.NewLogger()
 	ob := &bytes.Buffer{}
 	l.SetOutput(ob)
-	myVpnNetworksTable := new(bart.Lite)
-	myVpnNetworksTable.Insert(netip.MustParsePrefix("1.1.1.1/8"))
 
 	p := firewall.Packet{
 		LocalAddr:  netip.MustParseAddr("1.2.3.4"),
@@ -673,7 +659,7 @@ func TestFirewall_DropConntrackReload(t *testing.T) {
 		},
 		vpnAddrs: []netip.Addr{network.Addr()},
 	}
-	h.buildNetworks(myVpnNetworksTable, c.Certificate)
+	h.buildNetworks(c.Certificate.Networks(), c.Certificate.UnsafeNetworks())
 
 	fw := NewFirewall(l, time.Second, time.Minute, time.Hour, c.Certificate)
 	require.NoError(t, fw.AddRule(true, firewall.ProtoAny, 0, 0, []string{"any"}, "", netip.Prefix{}, netip.Prefix{}, "", ""))
@@ -706,52 +692,6 @@ func TestFirewall_DropConntrackReload(t *testing.T) {
 	assert.Equal(t, fw.Drop(p, false, &h, cp, nil), ErrNoMatchingRule)
 }
 
-func TestFirewall_DropIPSpoofing(t *testing.T) {
-	l := test.NewLogger()
-	ob := &bytes.Buffer{}
-	l.SetOutput(ob)
-	myVpnNetworksTable := new(bart.Lite)
-	myVpnNetworksTable.Insert(netip.MustParsePrefix("192.0.2.1/24"))
-
-	c := cert.CachedCertificate{
-		Certificate: &dummyCert{
-			name:     "host-owner",
-			networks: []netip.Prefix{netip.MustParsePrefix("192.0.2.1/24")},
-		},
-	}
-
-	c1 := cert.CachedCertificate{
-		Certificate: &dummyCert{
-			name:           "host",
-			networks:       []netip.Prefix{netip.MustParsePrefix("192.0.2.2/24")},
-			unsafeNetworks: []netip.Prefix{netip.MustParsePrefix("198.51.100.0/24")},
-		},
-	}
-	h1 := HostInfo{
-		ConnectionState: &ConnectionState{
-			peerCert: &c1,
-		},
-		vpnAddrs: []netip.Addr{c1.Certificate.Networks()[0].Addr()},
-	}
-	h1.buildNetworks(myVpnNetworksTable, c1.Certificate)
-
-	fw := NewFirewall(l, time.Second, time.Minute, time.Hour, c.Certificate)
-
-	require.NoError(t, fw.AddRule(true, firewall.ProtoAny, 1, 1, []string{}, "", netip.Prefix{}, netip.Prefix{}, "", ""))
-	cp := cert.NewCAPool()
-
-	// Packet spoofed by `c1`. Note that the remote addr is not a valid one.
-	p := firewall.Packet{
-		LocalAddr:  netip.MustParseAddr("192.0.2.1"),
-		RemoteAddr: netip.MustParseAddr("192.0.2.3"),
-		LocalPort:  1,
-		RemotePort: 1,
-		Protocol:   firewall.ProtoUDP,
-		Fragment:   false,
-	}
-	assert.Equal(t, fw.Drop(p, true, &h1, cp, nil), ErrInvalidRemoteIP)
-}
-
 func BenchmarkLookup(b *testing.B) {
 	ml := func(m map[string]struct{}, a [][]string) {
 		for n := 0; n < b.N; n++ {
@@ -1063,171 +1003,6 @@ func TestFirewall_convertRule(t *testing.T) {
 	assert.Equal(t, "group1", r.Group)
 }
 
-type testcase struct {
-	h   *HostInfo
-	p   firewall.Packet
-	c   cert.Certificate
-	err error
-}
-
-func (c *testcase) Test(t *testing.T, fw *Firewall) {
-	t.Helper()
-	cp := cert.NewCAPool()
-	resetConntrack(fw)
-	err := fw.Drop(c.p, true, c.h, cp, nil)
-	if c.err == nil {
-		require.NoError(t, err, "failed to not drop remote address %s", c.p.RemoteAddr)
-	} else {
-		require.ErrorIs(t, c.err, err, "failed to drop remote address %s", c.p.RemoteAddr)
-	}
-}
-
-func buildTestCase(setup testsetup, err error, theirPrefixes ...netip.Prefix) testcase {
-	c1 := dummyCert{
-		name:     "host1",
-		networks: theirPrefixes,
-		groups:   []string{"default-group"},
-		issuer:   "signer-shasum",
-	}
-	h := HostInfo{
-		ConnectionState: &ConnectionState{
-			peerCert: &cert.CachedCertificate{
-				Certificate:    &c1,
-				InvertedGroups: map[string]struct{}{"default-group": {}},
-			},
-		},
-		vpnAddrs: make([]netip.Addr, len(theirPrefixes)),
-	}
-	for i := range theirPrefixes {
-		h.vpnAddrs[i] = theirPrefixes[i].Addr()
-	}
-	h.buildNetworks(setup.myVpnNetworksTable, &c1)
-	p := firewall.Packet{
-		LocalAddr:  setup.c.Networks()[0].Addr(), //todo?
-		RemoteAddr: theirPrefixes[0].Addr(),
-		LocalPort:  10,
-		RemotePort: 90,
-		Protocol:   firewall.ProtoUDP,
-		Fragment:   false,
-	}
-	return testcase{
-		h:   &h,
-		p:   p,
-		c:   &c1,
-		err: err,
-	}
-}
-
-type testsetup struct {
-	c                  dummyCert
-	myVpnNetworksTable *bart.Lite
-	fw                 *Firewall
-}
-
-func newSetup(t *testing.T, l *logrus.Logger, myPrefixes ...netip.Prefix) testsetup {
-	c := dummyCert{
-		name:     "me",
-		networks: myPrefixes,
-		groups:   []string{"default-group"},
-		issuer:   "signer-shasum",
-	}
-
-	return newSetupFromCert(t, l, c)
-}
-
-func newSetupFromCert(t *testing.T, l *logrus.Logger, c dummyCert) testsetup {
-	myVpnNetworksTable := new(bart.Lite)
-	for _, prefix := range c.Networks() {
-		myVpnNetworksTable.Insert(prefix)
-	}
-	fw := NewFirewall(l, time.Second, time.Minute, time.Hour, &c)
-	require.NoError(t, fw.AddRule(true, firewall.ProtoAny, 0, 0, []string{"any"}, "", netip.Prefix{}, netip.Prefix{}, "", ""))
-
-	return testsetup{
-		c:                  c,
-		fw:                 fw,
-		myVpnNetworksTable: myVpnNetworksTable,
-	}
-}
-
-func TestFirewall_Drop_EnforceIPMatch(t *testing.T) {
-	t.Parallel()
-	l := test.NewLogger()
-	ob := &bytes.Buffer{}
-	l.SetOutput(ob)
-
-	myPrefix := netip.MustParsePrefix("1.1.1.1/8")
-	// for now, it's okay that these are all "incoming", the logic this test tries to check doesn't care about in/out
-	t.Run("allow inbound all matching", func(t *testing.T) {
-		t.Parallel()
-		setup := newSetup(t, l, myPrefix)
-		tc := buildTestCase(setup, nil, netip.MustParsePrefix("1.2.3.4/24"))
-		tc.Test(t, setup.fw)
-	})
-	t.Run("allow inbound local matching", func(t *testing.T) {
-		t.Parallel()
-		setup := newSetup(t, l, myPrefix)
-		tc := buildTestCase(setup, ErrInvalidLocalIP, netip.MustParsePrefix("1.2.3.4/24"))
-		tc.p.LocalAddr = netip.MustParseAddr("1.2.3.8")
-		tc.Test(t, setup.fw)
-	})
-	t.Run("block inbound remote mismatched", func(t *testing.T) {
-		t.Parallel()
-		setup := newSetup(t, l, myPrefix)
-		tc := buildTestCase(setup, ErrInvalidRemoteIP, netip.MustParsePrefix("1.2.3.4/24"))
-		tc.p.RemoteAddr = netip.MustParseAddr("9.9.9.9")
-		tc.Test(t, setup.fw)
-	})
-	t.Run("Block a vpn peer packet", func(t *testing.T) {
-		t.Parallel()
-		setup := newSetup(t, l, myPrefix)
-		tc := buildTestCase(setup, ErrPeerRejected, netip.MustParsePrefix("2.2.2.2/24"))
-		tc.Test(t, setup.fw)
-	})
-	twoPrefixes := []netip.Prefix{
-		netip.MustParsePrefix("1.2.3.4/24"), netip.MustParsePrefix("2.2.2.2/24"),
-	}
-	t.Run("allow inbound one matching", func(t *testing.T) {
-		t.Parallel()
-		setup := newSetup(t, l, myPrefix)
-		tc := buildTestCase(setup, nil, twoPrefixes...)
-		tc.Test(t, setup.fw)
-	})
-	t.Run("block inbound multimismatch", func(t *testing.T) {
-		t.Parallel()
-		setup := newSetup(t, l, myPrefix)
-		tc := buildTestCase(setup, ErrInvalidRemoteIP, twoPrefixes...)
-		tc.p.RemoteAddr = netip.MustParseAddr("9.9.9.9")
-		tc.Test(t, setup.fw)
-	})
-	t.Run("allow inbound 2nd one matching", func(t *testing.T) {
-		t.Parallel()
-		setup2 := newSetup(t, l, netip.MustParsePrefix("2.2.2.1/24"))
-		tc := buildTestCase(setup2, nil, twoPrefixes...)
-		tc.p.RemoteAddr = twoPrefixes[1].Addr()
-		tc.Test(t, setup2.fw)
-	})
-	t.Run("allow inbound unsafe route", func(t *testing.T) {
-		t.Parallel()
-		unsafePrefix := netip.MustParsePrefix("192.168.0.0/24")
-		c := dummyCert{
-			name:           "me",
-			networks:       []netip.Prefix{myPrefix},
-			unsafeNetworks: []netip.Prefix{unsafePrefix},
-			groups:         []string{"default-group"},
-			issuer:         "signer-shasum",
-		}
-		unsafeSetup := newSetupFromCert(t, l, c)
-		tc := buildTestCase(unsafeSetup, nil, twoPrefixes...)
-		tc.p.LocalAddr = netip.MustParseAddr("192.168.0.3")
-		tc.err = ErrNoMatchingRule
-		tc.Test(t, unsafeSetup.fw) //should hit firewall and bounce off
-		require.NoError(t, unsafeSetup.fw.AddRule(true, firewall.ProtoAny, 0, 0, []string{"any"}, "", netip.Prefix{}, unsafePrefix, "", ""))
-		tc.err = nil
-		tc.Test(t, unsafeSetup.fw) //should pass
-	})
-}
-
 type addRuleCall struct {
 	incoming  bool
 	proto     uint8

go.mod

@@ -8,31 +8,30 @@ require (
 	github.com/armon/go-radix v1.0.0
 	github.com/cyberdelia/go-metrics-graphite v0.0.0-20161219230853-39f87cc3b432
 	github.com/flynn/noise v1.1.0
-	github.com/gaissmai/bart v0.26.0
+	github.com/gaissmai/bart v0.20.4
 	github.com/gogo/protobuf v1.3.2
 	github.com/google/gopacket v1.1.19
-	github.com/kardianos/service v1.2.4
-	github.com/miekg/dns v1.1.68
+	github.com/kardianos/service v1.2.2
+	github.com/miekg/dns v1.1.65
 	github.com/miekg/pkcs11 v1.1.2-0.20231115102856-9078ad6b9d4b
 	github.com/nbrownus/go-metrics-prometheus v0.0.0-20210712211119-974a6260965f
-	github.com/prometheus/client_golang v1.23.2
+	github.com/prometheus/client_golang v1.22.0
 	github.com/rcrowley/go-metrics v0.0.0-20201227073835-cf1acfcdf475
 	github.com/sirupsen/logrus v1.9.3
 	github.com/skip2/go-qrcode v0.0.0-20200617195104-da1b6568686e
 	github.com/stefanberger/go-pkcs11uri v0.0.0-20230803200340-78284954bff6
-	github.com/stretchr/testify v1.11.1
+	github.com/stretchr/testify v1.10.0
 	github.com/vishvananda/netlink v1.3.1
-	go.yaml.in/yaml/v3 v3.0.4
-	golang.org/x/crypto v0.44.0
+	golang.org/x/crypto v0.37.0
 	golang.org/x/exp v0.0.0-20230725093048-515e97ebf090
-	golang.org/x/net v0.46.0
-	golang.org/x/sync v0.18.0
-	golang.org/x/sys v0.38.0
-	golang.org/x/term v0.37.0
+	golang.org/x/net v0.39.0
+	golang.org/x/sync v0.13.0
+	golang.org/x/sys v0.32.0
+	golang.org/x/term v0.31.0
 	golang.zx2c4.com/wintun v0.0.0-20230126152724-0fa3db229ce2
 	golang.zx2c4.com/wireguard v0.0.0-20230325221338-052af4a8072b
 	golang.zx2c4.com/wireguard/windows v0.5.3
-	google.golang.org/protobuf v1.36.10
+	google.golang.org/protobuf v1.36.6
 	gopkg.in/yaml.v3 v3.0.1
 	gvisor.dev/gvisor v0.0.0-20240423190808-9d7a357edefe
 )
@@ -44,12 +43,11 @@ require (
 	github.com/google/btree v1.1.2 // indirect
 	github.com/munnerz/goautoneg v0.0.0-20191010083416-a7dc8b61c822 // indirect
 	github.com/pmezard/go-difflib v1.0.0 // indirect
-	github.com/prometheus/client_model v0.6.2 // indirect
-	github.com/prometheus/common v0.66.1 // indirect
-	github.com/prometheus/procfs v0.16.1 // indirect
+	github.com/prometheus/client_model v0.6.1 // indirect
+	github.com/prometheus/common v0.62.0 // indirect
+	github.com/prometheus/procfs v0.15.1 // indirect
 	github.com/vishvananda/netns v0.0.5 // indirect
-	go.yaml.in/yaml/v2 v2.4.2 // indirect
-	golang.org/x/mod v0.24.0 // indirect
+	golang.org/x/mod v0.23.0 // indirect
 	golang.org/x/time v0.5.0 // indirect
-	golang.org/x/tools v0.33.0 // indirect
+	golang.org/x/tools v0.30.0 // indirect
 )

go.sum

@@ -24,8 +24,8 @@ github.com/davecgh/go-spew v1.1.1 h1:vj9j/u1bqnvCEfJOwUhtlOARqs3+rkHYY13jYWTU97c
 github.com/davecgh/go-spew v1.1.1/go.mod h1:J7Y8YcW2NihsgmVo/mv3lAwl/skON4iLHjSsI+c5H38=
 github.com/flynn/noise v1.1.0 h1:KjPQoQCEFdZDiP03phOvGi11+SVVhBG2wOWAorLsstg=
 github.com/flynn/noise v1.1.0/go.mod h1:xbMo+0i6+IGbYdJhF31t2eR1BIU0CYc12+BNAKwUTag=
-github.com/gaissmai/bart v0.26.0 h1:xOZ57E9hJLBiQaSyeZa9wgWhGuzfGACgqp4BE77OkO0=
-github.com/gaissmai/bart v0.26.0/go.mod h1:GREWQfTLRWz/c5FTOsIw+KkscuFkIV5t8Rp7Nd1Td5c=
+github.com/gaissmai/bart v0.20.4 h1:Ik47r1fy3jRVU+1eYzKSW3ho2UgBVTVnUS8O993584U=
+github.com/gaissmai/bart v0.20.4/go.mod h1:cEed+ge8dalcbpi8wtS9x9m2hn/fNJH5suhdGQOHnYk=
 github.com/go-kit/kit v0.8.0/go.mod h1:xBxKIO96dXMWWy0MnWVtmwkA9/13aqxPnvrjFYMA2as=
 github.com/go-kit/kit v0.9.0/go.mod h1:xBxKIO96dXMWWy0MnWVtmwkA9/13aqxPnvrjFYMA2as=
 github.com/go-kit/log v0.1.0/go.mod h1:zbhenjAZHb184qTLMA9ZjW7ThYL0H2mk7Q6pNt4vbaY=
@@ -64,8 +64,8 @@ github.com/json-iterator/go v1.1.10/go.mod h1:KdQUCv79m/52Kvf8AW2vK1V8akMuk1QjK/
 github.com/json-iterator/go v1.1.11/go.mod h1:KdQUCv79m/52Kvf8AW2vK1V8akMuk1QjK/uOdHXbAo4=
 github.com/julienschmidt/httprouter v1.2.0/go.mod h1:SYymIcj16QtmaHHD7aYtjjsJG7VTCxuUUipMqKk8s4w=
 github.com/julienschmidt/httprouter v1.3.0/go.mod h1:JR6WtHb+2LUe8TCKY3cZOxFyyO8IZAc4RVcycCCAKdM=
-github.com/kardianos/service v1.2.4 h1:XNlGtZOYNx2u91urOdg/Kfmc+gfmuIo1Dd3rEi2OgBk=
-github.com/kardianos/service v1.2.4/go.mod h1:E4V9ufUuY82F7Ztlu1eN9VXWIQxg8NoLQlmFe0MtrXc=
+github.com/kardianos/service v1.2.2 h1:ZvePhAHfvo0A7Mftk/tEzqEZ7Q4lgnR8sGz4xu1YX60=
+github.com/kardianos/service v1.2.2/go.mod h1:CIMRFEJVL+0DS1a3Nx06NaMn4Dz63Ng6O7dl0qH0zVM=
 github.com/kisielk/errcheck v1.5.0/go.mod h1:pFxgyoBC7bSaBwPgfKdkLd5X25qrDl4LWUI2bnpBCr8=
 github.com/kisielk/gotool v1.0.0/go.mod h1:XhKaO+MFFWcvkIS/tQcRk01m1F5IRFswLeQ+oQHNcck=
 github.com/klauspost/compress v1.18.0 h1:c/Cqfb0r+Yi+JtIEq73FWXVkRonBlf0CRNYc8Zttxdo=
@@ -83,8 +83,8 @@ github.com/kr/text v0.1.0/go.mod h1:4Jbv+DJW3UT/LiOwJeYQe1efqtUx/iVham/4vfdArNI=
 github.com/kylelemons/godebug v1.1.0 h1:RPNrshWIDI6G2gRW9EHilWtl7Z6Sb1BR0xunSBf0SNc=
 github.com/kylelemons/godebug v1.1.0/go.mod h1:9/0rRGxNHcop5bhtWyNeEfOS8JIWk580+fNqagV/RAw=
 github.com/matttproud/golang_protobuf_extensions v1.0.1/go.mod h1:D8He9yQNgCq6Z5Ld7szi9bcBfOoFv/3dc6xSMkL2PC0=
-github.com/miekg/dns v1.1.68 h1:jsSRkNozw7G/mnmXULynzMNIsgY2dHC8LO6U6Ij2JEA=
-github.com/miekg/dns v1.1.68/go.mod h1:fujopn7TB3Pu3JM69XaawiU0wqjpL9/8xGop5UrTPps=
+github.com/miekg/dns v1.1.65 h1:0+tIPHzUW0GCge7IiK3guGP57VAw7hoPDfApjkMD1Fc=
+github.com/miekg/dns v1.1.65/go.mod h1:Dzw9769uoKVaLuODMDZz9M6ynFU6Em65csPuoi8G0ck=
 github.com/miekg/pkcs11 v1.1.2-0.20231115102856-9078ad6b9d4b h1:J/AzCvg5z0Hn1rqZUJjpbzALUmkKX0Zwbc/i4fw7Sfk=
 github.com/miekg/pkcs11 v1.1.2-0.20231115102856-9078ad6b9d4b/go.mod h1:XsNlhZGX73bx86s2hdc/FuaLm2CPZJemRLMA+WTFxgs=
 github.com/modern-go/concurrent v0.0.0-20180228061459-e0a39a4cb421/go.mod h1:6dJC0mAP4ikYIbvyc7fijjWJddQyLn8Ig3JB5CqoB9Q=
@@ -106,24 +106,24 @@ github.com/prometheus/client_golang v0.9.1/go.mod h1:7SWBe2y4D6OKWSNQJUaRYU/AaXP
 github.com/prometheus/client_golang v1.0.0/go.mod h1:db9x61etRT2tGnBNRi70OPL5FsnadC4Ky3P0J6CfImo=
 github.com/prometheus/client_golang v1.7.1/go.mod h1:PY5Wy2awLA44sXw4AOSfFBetzPP4j5+D6mVACh+pe2M=
 github.com/prometheus/client_golang v1.11.0/go.mod h1:Z6t4BnS23TR94PD6BsDNk8yVqroYurpAkEiz0P2BEV0=
-github.com/prometheus/client_golang v1.23.2 h1:Je96obch5RDVy3FDMndoUsjAhG5Edi49h0RJWRi/o0o=
-github.com/prometheus/client_golang v1.23.2/go.mod h1:Tb1a6LWHB3/SPIzCoaDXI4I8UHKeFTEQ1YCr+0Gyqmg=
+github.com/prometheus/client_golang v1.22.0 h1:rb93p9lokFEsctTys46VnV1kLCDpVZ0a/Y92Vm0Zc6Q=
+github.com/prometheus/client_golang v1.22.0/go.mod h1:R7ljNsLXhuQXYZYtw6GAE9AZg8Y7vEW5scdCXrWRXC0=
 github.com/prometheus/client_model v0.0.0-20180712105110-5c3871d89910/go.mod h1:MbSGuTsp3dbXC40dX6PRTWyKYBIrTGTE9sqQNg2J8bo=
 github.com/prometheus/client_model v0.0.0-20190129233127-fd36f4220a90/go.mod h1:xMI15A0UPsDsEKsMN9yxemIoYk6Tm2C1GtYGdfGttqA=
 github.com/prometheus/client_model v0.2.0/go.mod h1:xMI15A0UPsDsEKsMN9yxemIoYk6Tm2C1GtYGdfGttqA=
-github.com/prometheus/client_model v0.6.2 h1:oBsgwpGs7iVziMvrGhE53c/GrLUsZdHnqNwqPLxwZyk=
-github.com/prometheus/client_model v0.6.2/go.mod h1:y3m2F6Gdpfy6Ut/GBsUqTWZqCUvMVzSfMLjcu6wAwpE=
+github.com/prometheus/client_model v0.6.1 h1:ZKSh/rekM+n3CeS952MLRAdFwIKqeY8b62p8ais2e9E=
+github.com/prometheus/client_model v0.6.1/go.mod h1:OrxVMOVHjw3lKMa8+x6HeMGkHMQyHDk9E3jmP2AmGiY=
 github.com/prometheus/common v0.4.1/go.mod h1:TNfzLD0ON7rHzMJeJkieUDPYmFC7Snx/y86RQel1bk4=
 github.com/prometheus/common v0.10.0/go.mod h1:Tlit/dnDKsSWFlCLTWaA1cyBgKHSMdTB80sz/V91rCo=
 github.com/prometheus/common v0.26.0/go.mod h1:M7rCNAaPfAosfx8veZJCuw84e35h3Cfd9VFqTh1DIvc=
-github.com/prometheus/common v0.66.1 h1:h5E0h5/Y8niHc5DlaLlWLArTQI7tMrsfQjHV+d9ZoGs=
-github.com/prometheus/common v0.66.1/go.mod h1:gcaUsgf3KfRSwHY4dIMXLPV0K/Wg1oZ8+SbZk/HH/dA=
+github.com/prometheus/common v0.62.0 h1:xasJaQlnWAeyHdUBeGjXmutelfJHWMRr+Fg4QszZ2Io=
+github.com/prometheus/common v0.62.0/go.mod h1:vyBcEuLSvWos9B1+CyL7JZ2up+uFzXhkqml0W5zIY1I=
 github.com/prometheus/procfs v0.0.0-20181005140218-185b4288413d/go.mod h1:c3At6R/oaqEKCNdg8wHV1ftS6bRYblBhIjjI8uT2IGk=
 github.com/prometheus/procfs v0.0.2/go.mod h1:TjEm7ze935MbeOT/UhFTIMYKhuLP4wbCsTZCD3I8kEA=
 github.com/prometheus/procfs v0.1.3/go.mod h1:lV6e/gmhEcM9IjHGsFOCxxuZ+z1YqCvr4OA4YeYWdaU=
 github.com/prometheus/procfs v0.6.0/go.mod h1:cz+aTbrPOrUb4q7XlbU9ygM+/jj0fzG6c1xBZuNvfVA=
-github.com/prometheus/procfs v0.16.1 h1:hZ15bTNuirocR6u0JZ6BAHHmwS1p8B4P6MRqxtzMyRg=
-github.com/prometheus/procfs v0.16.1/go.mod h1:teAbpZRB1iIAJYREa1LsoWUXykVXA1KlTmWl8x/U+Is=
+github.com/prometheus/procfs v0.15.1 h1:YagwOFzUgYfKKHX6Dr+sHT7km/hxC76UB0learggepc=
+github.com/prometheus/procfs v0.15.1/go.mod h1:fB45yRUv8NstnjriLhBQLuOUt+WW4BsoGhij/e3PBqk=
 github.com/rcrowley/go-metrics v0.0.0-20201227073835-cf1acfcdf475 h1:N/ElC8H3+5XpJzTSTfLsJV/mx9Q9g7kxmchpfZyxgzM=
 github.com/rcrowley/go-metrics v0.0.0-20201227073835-cf1acfcdf475/go.mod h1:bCqnVzQkZxMG4s8nGwiZ5l3QUCyqpo9Y+/ZMZ9VjZe4=
 github.com/rogpeppe/go-internal v1.10.0 h1:TMyTOH3F/DB16zRVcYyreMH6GnZZrwQVAoYjRBZyWFQ=
@@ -143,35 +143,29 @@ github.com/stretchr/testify v1.2.2/go.mod h1:a8OnRcib4nhh0OaRAV+Yts87kKdq0PP7pXf
 github.com/stretchr/testify v1.3.0/go.mod h1:M5WIy9Dh21IEIfnGCwXGc5bZfKNJtfHm1UVUgZn+9EI=
 github.com/stretchr/testify v1.4.0/go.mod h1:j7eGeouHqKxXV5pUuKE4zz7dFj8WfuZ+81PSLYec5m4=
 github.com/stretchr/testify v1.7.0/go.mod h1:6Fq8oRcR53rry900zMqJjRRixrwX3KX962/h/Wwjteg=
-github.com/stretchr/testify v1.11.1 h1:7s2iGBzp5EwR7/aIZr8ao5+dra3wiQyKjjFuvgVKu7U=
-github.com/stretchr/testify v1.11.1/go.mod h1:wZwfW3scLgRK+23gO65QZefKpKQRnfz6sD981Nm4B6U=
+github.com/stretchr/testify v1.10.0 h1:Xv5erBjTwe/5IxqUQTdXv5kgmIvbHo3QQyRwhJsOfJA=
+github.com/stretchr/testify v1.10.0/go.mod h1:r2ic/lqez/lEtzL7wO/rwa5dbSLXVDPFyf8C91i36aY=
 github.com/vishvananda/netlink v1.3.1 h1:3AEMt62VKqz90r0tmNhog0r/PpWKmrEShJU0wJW6bV0=
 github.com/vishvananda/netlink v1.3.1/go.mod h1:ARtKouGSTGchR8aMwmkzC0qiNPrrWO5JS/XMVl45+b4=
 github.com/vishvananda/netns v0.0.5 h1:DfiHV+j8bA32MFM7bfEunvT8IAqQ/NzSJHtcmW5zdEY=
 github.com/vishvananda/netns v0.0.5/go.mod h1:SpkAiCQRtJ6TvvxPnOSyH3BMl6unz3xZlaprSwhNNJM=
 github.com/yuin/goldmark v1.1.27/go.mod h1:3hX8gzYuyVAZsxl0MRgGTJEmQBFcNTphYh9decYSb74=
 github.com/yuin/goldmark v1.2.1/go.mod h1:3hX8gzYuyVAZsxl0MRgGTJEmQBFcNTphYh9decYSb74=
-go.uber.org/goleak v1.3.0 h1:2K3zAYmnTNqV73imy9J1T3WC+gmCePx2hEGkimedGto=
-go.uber.org/goleak v1.3.0/go.mod h1:CoHD4mav9JJNrW/WLlf7HGZPjdw8EucARQHekz1X6bE=
-go.yaml.in/yaml/v2 v2.4.2 h1:DzmwEr2rDGHl7lsFgAHxmNz/1NlQ7xLIrlN2h5d1eGI=
-go.yaml.in/yaml/v2 v2.4.2/go.mod h1:081UH+NErpNdqlCXm3TtEran0rJZGxAYx9hb/ELlsPU=
-go.yaml.in/yaml/v3 v3.0.4 h1:tfq32ie2Jv2UxXFdLJdh3jXuOzWiL1fo0bu/FbuKpbc=
-go.yaml.in/yaml/v3 v3.0.4/go.mod h1:DhzuOOF2ATzADvBadXxruRBLzYTpT36CKvDb3+aBEFg=
 golang.org/x/crypto v0.0.0-20180904163835-0709b304e793/go.mod h1:6SG95UA2DQfeDnfUPMdvaQW0Q7yPrPDi9nlGo2tz2b4=
 golang.org/x/crypto v0.0.0-20190308221718-c2843e01d9a2/go.mod h1:djNgcEr1/C05ACkg1iLfiJU5Ep61QUkGW8qpdssI0+w=
 golang.org/x/crypto v0.0.0-20191011191535-87dc89f01550/go.mod h1:yigFU9vqHzYiE8UmvKecakEJjdnWj3jj499lnFckfCI=
 golang.org/x/crypto v0.0.0-20200622213623-75b288015ac9/go.mod h1:LzIPMQfyMNhhGPhUkYOs5KpL4U8rLKemX1yGLhDgUto=
 golang.org/x/crypto v0.0.0-20210322153248-0c34fe9e7dc2/go.mod h1:T9bdIzuCu7OtxOm1hfPfRQxPLYneinmdGuTeoZ9dtd4=
-golang.org/x/crypto v0.44.0 h1:A97SsFvM3AIwEEmTBiaxPPTYpDC47w720rdiiUvgoAU=
-golang.org/x/crypto v0.44.0/go.mod h1:013i+Nw79BMiQiMsOPcVCB5ZIJbYkerPrGnOa00tvmc=
+golang.org/x/crypto v0.37.0 h1:kJNSjF/Xp7kU0iB2Z+9viTPMW4EqqsrywMXLJOOsXSE=
+golang.org/x/crypto v0.37.0/go.mod h1:vg+k43peMZ0pUMhYmVAWysMK35e6ioLh3wB8ZCAfbVc=
 golang.org/x/exp v0.0.0-20230725093048-515e97ebf090 h1:Di6/M8l0O2lCLc6VVRWhgCiApHV8MnQurBnFSHsQtNY=
 golang.org/x/exp v0.0.0-20230725093048-515e97ebf090/go.mod h1:FXUEEKJgO7OQYeo8N01OfiKP8RXMtf6e8aTskBGqWdc=
 golang.org/x/lint v0.0.0-20200302205851-738671d3881b/go.mod h1:3xt1FjdF8hUf6vQPIChWIBhFzV8gjjsPE/fR3IyQdNY=
 golang.org/x/mod v0.1.1-0.20191105210325-c90efee705ee/go.mod h1:QqPTAvyqsEbceGzBzNggFXnrqF1CaUcvgkdR5Ot7KZg=
 golang.org/x/mod v0.2.0/go.mod h1:s0Qsj1ACt9ePp/hMypM3fl4fZqREWJwdYDEqhRiZZUA=
 golang.org/x/mod v0.3.0/go.mod h1:s0Qsj1ACt9ePp/hMypM3fl4fZqREWJwdYDEqhRiZZUA=
-golang.org/x/mod v0.24.0 h1:ZfthKaKaT4NrhGVZHO1/WDTwGES4De8KtWO0SIbNJMU=
-golang.org/x/mod v0.24.0/go.mod h1:IXM97Txy2VM4PJ3gI61r1YEk/gAj6zAHN3AdZt6S9Ww=
+golang.org/x/mod v0.23.0 h1:Zb7khfcRGKk+kqfxFaP5tZqCnDZMjC5VtUBs87Hr6QM=
+golang.org/x/mod v0.23.0/go.mod h1:6SkKJ3Xj0I0BrPOZoBy3bdMptDDU9oJrpohJ3eWZ1fY=
 golang.org/x/net v0.0.0-20180724234803-3673e40ba225/go.mod h1:mL1N/T3taQHkDXs73rZJwtUhF3w3ftmwwsq0BUmARs4=
 golang.org/x/net v0.0.0-20181114220301-adae6a3d119a/go.mod h1:mL1N/T3taQHkDXs73rZJwtUhF3w3ftmwwsq0BUmARs4=
 golang.org/x/net v0.0.0-20190108225652-1e06a53dbb7e/go.mod h1:mL1N/T3taQHkDXs73rZJwtUhF3w3ftmwwsq0BUmARs4=
@@ -182,8 +176,8 @@ golang.org/x/net v0.0.0-20200226121028-0de0cce0169b/go.mod h1:z5CRVTTTmAJ677TzLL
 golang.org/x/net v0.0.0-20200625001655-4c5254603344/go.mod h1:/O7V0waA8r7cgGh81Ro3o1hOxt32SMVPicZroKQ2sZA=
 golang.org/x/net v0.0.0-20201021035429-f5854403a974/go.mod h1:sp8m0HH+o8qH0wwXwYZr8TS3Oi6o0r6Gce1SSxlDquU=
 golang.org/x/net v0.0.0-20210226172049-e18ecbb05110/go.mod h1:m0MpNAwzfU5UDzcl9v0D8zg8gWTRqZa9RBIspLL5mdg=
-golang.org/x/net v0.46.0 h1:giFlY12I07fugqwPuWJi68oOnpfqFnJIJzaIIm2JVV4=
-golang.org/x/net v0.46.0/go.mod h1:Q9BGdFy1y4nkUwiLvT5qtyhAnEHgnQ/zd8PfU6nc210=
+golang.org/x/net v0.39.0 h1:ZCu7HMWDxpXpaiKdhzIfaltL9Lp31x/3fCP11bc6/fY=
+golang.org/x/net v0.39.0/go.mod h1:X7NRbYVEA+ewNkCNyJ513WmMdQ3BineSwVtN2zD/d+E=
 golang.org/x/oauth2 v0.0.0-20190226205417-e64efc72b421/go.mod h1:gOpvHmFTYa4IltrdGE7lF6nIHvwfUNPOp7c8zoXwtLw=
 golang.org/x/sync v0.0.0-20181108010431-42b317875d0f/go.mod h1:RxMgew5VJxzue5/jJTE5uejpjVlOe/izrB70Jof72aM=
 golang.org/x/sync v0.0.0-20181221193216-37e7f081c4d4/go.mod h1:RxMgew5VJxzue5/jJTE5uejpjVlOe/izrB70Jof72aM=
@@ -191,8 +185,8 @@ golang.org/x/sync v0.0.0-20190423024810-112230192c58/go.mod h1:RxMgew5VJxzue5/jJ
 golang.org/x/sync v0.0.0-20190911185100-cd5d95a43a6e/go.mod h1:RxMgew5VJxzue5/jJTE5uejpjVlOe/izrB70Jof72aM=
 golang.org/x/sync v0.0.0-20201020160332-67f06af15bc9/go.mod h1:RxMgew5VJxzue5/jJTE5uejpjVlOe/izrB70Jof72aM=
 golang.org/x/sync v0.0.0-20201207232520-09787c993a3a/go.mod h1:RxMgew5VJxzue5/jJTE5uejpjVlOe/izrB70Jof72aM=
-golang.org/x/sync v0.18.0 h1:kr88TuHDroi+UVf+0hZnirlk8o8T+4MrK6mr60WkH/I=
-golang.org/x/sync v0.18.0/go.mod h1:9KTHXmSnoGruLpwFjVSX0lNNA75CykiMECbovNTZqGI=
+golang.org/x/sync v0.13.0 h1:AauUjRAJ9OSnvULf/ARrrVywoJDy0YS2AwQ98I37610=
+golang.org/x/sync v0.13.0/go.mod h1:1dzgHSNfp02xaA81J2MS99Qcpr2w7fw1gpm99rleRqA=
 golang.org/x/sys v0.0.0-20180905080454-ebe1bf3edb33/go.mod h1:STP8DvDyc/dI5b8T5hshtkjS+E42TnysNCUPdjciGhY=
 golang.org/x/sys v0.0.0-20181116152217-5ac8a444bdc5/go.mod h1:STP8DvDyc/dI5b8T5hshtkjS+E42TnysNCUPdjciGhY=
 golang.org/x/sys v0.0.0-20190215142949-d0b11bdaac8a/go.mod h1:STP8DvDyc/dI5b8T5hshtkjS+E42TnysNCUPdjciGhY=
@@ -203,17 +197,18 @@ golang.org/x/sys v0.0.0-20200323222414-85ca7c5b95cd/go.mod h1:h1NjWce9XRLGQEsW7w
 golang.org/x/sys v0.0.0-20200615200032-f1bc736245b1/go.mod h1:h1NjWce9XRLGQEsW7wpKNCjG9DtNlClVuFLEZdDNbEs=
 golang.org/x/sys v0.0.0-20200625212154-ddb9806d33ae/go.mod h1:h1NjWce9XRLGQEsW7wpKNCjG9DtNlClVuFLEZdDNbEs=
 golang.org/x/sys v0.0.0-20200930185726-fdedc70b468f/go.mod h1:h1NjWce9XRLGQEsW7wpKNCjG9DtNlClVuFLEZdDNbEs=
+golang.org/x/sys v0.0.0-20201015000850-e3ed0017c211/go.mod h1:h1NjWce9XRLGQEsW7wpKNCjG9DtNlClVuFLEZdDNbEs=
 golang.org/x/sys v0.0.0-20201119102817-f84b799fce68/go.mod h1:h1NjWce9XRLGQEsW7wpKNCjG9DtNlClVuFLEZdDNbEs=
 golang.org/x/sys v0.0.0-20210124154548-22da62e12c0c/go.mod h1:h1NjWce9XRLGQEsW7wpKNCjG9DtNlClVuFLEZdDNbEs=
 golang.org/x/sys v0.0.0-20210603081109-ebe580a85c40/go.mod h1:oPkhp1MJrh7nUepCBck5+mAzfO9JrbApNNgaTdGDITg=
 golang.org/x/sys v0.0.0-20220715151400-c0bba94af5f8/go.mod h1:oPkhp1MJrh7nUepCBck5+mAzfO9JrbApNNgaTdGDITg=
 golang.org/x/sys v0.2.0/go.mod h1:oPkhp1MJrh7nUepCBck5+mAzfO9JrbApNNgaTdGDITg=
 golang.org/x/sys v0.10.0/go.mod h1:oPkhp1MJrh7nUepCBck5+mAzfO9JrbApNNgaTdGDITg=
-golang.org/x/sys v0.38.0 h1:3yZWxaJjBmCWXqhN1qh02AkOnCQ1poK6oF+a7xWL6Gc=
-golang.org/x/sys v0.38.0/go.mod h1:OgkHotnGiDImocRcuBABYBEXf8A9a87e/uXjp9XT3ks=
+golang.org/x/sys v0.32.0 h1:s77OFDvIQeibCmezSnk/q6iAfkdiQaJi4VzroCFrN20=
+golang.org/x/sys v0.32.0/go.mod h1:BJP2sWEmIv4KK5OTEluFJCKSidICx8ciO85XgH3Ak8k=
 golang.org/x/term v0.0.0-20201126162022-7de9c90e9dd1/go.mod h1:bj7SfCRtBDWHUb9snDiAeCFNEtKQo2Wmx5Cou7ajbmo=
-golang.org/x/term v0.37.0 h1:8EGAD0qCmHYZg6J17DvsMy9/wJ7/D/4pV/wfnld5lTU=
-golang.org/x/term v0.37.0/go.mod h1:5pB4lxRNYYVZuTLmy8oR2BH8dflOR+IbTYFD8fi3254=
+golang.org/x/term v0.31.0 h1:erwDkOK1Msy6offm1mOgvspSkslFnIGsFnxOKoufg3o=
+golang.org/x/term v0.31.0/go.mod h1:R4BeIy7D95HzImkxGkTW1UQTtP54tio2RyHz7PwK0aw=
 golang.org/x/text v0.3.0/go.mod h1:NqM8EUOU14njkJ3fqMW+pc6Ldnwhi/IjpwHt7yyuwOQ=
 golang.org/x/text v0.3.2/go.mod h1:bEr9sfX3Q8Zfm5fL9x+3itogRgK3+ptLWKqgva+5dAk=
 golang.org/x/text v0.3.3/go.mod h1:5Zoc/QRtKVWzQhOtBMvqHzDpF6irO9z98xDceosuGiQ=
@@ -224,8 +219,8 @@ golang.org/x/tools v0.0.0-20191119224855-298f0cb1881e/go.mod h1:b+2E5dAYhXwXZwtn
 golang.org/x/tools v0.0.0-20200130002326-2f3ba24bd6e7/go.mod h1:TB2adYChydJhpapKDTa4BR/hXlZSLoq2Wpct/0txZ28=
 golang.org/x/tools v0.0.0-20200619180055-7c47624df98f/go.mod h1:EkVYQZoAsY45+roYkvgYkIh4xh/qjgUK9TdY2XT94GE=
 golang.org/x/tools v0.0.0-20210106214847-113979e3529a/go.mod h1:emZCQorbCU4vsT4fOWvOPXz4eW1wZW4PmDk9uLelYpA=
-golang.org/x/tools v0.33.0 h1:4qz2S3zmRxbGIhDIAgjxvFutSvH5EfnsYrRBj0UI0bc=
-golang.org/x/tools v0.33.0/go.mod h1:CIJMaWEY88juyUfo7UbgPqbC8rU2OqfAV1h2Qp0oMYI=
+golang.org/x/tools v0.30.0 h1:BgcpHewrV5AUp2G9MebG4XPFI1E2W41zU1SaqVA9vJY=
+golang.org/x/tools v0.30.0/go.mod h1:c347cR/OJfw5TI+GfX7RUPNMdDRRbjvYTS0jPyvsVtY=
 golang.org/x/xerrors v0.0.0-20190717185122-a985d3407aa7/go.mod h1:I/5z698sn9Ka8TeJc9MKroUUfqBBauWjQqLJ2OPfmY0=
 golang.org/x/xerrors v0.0.0-20191011141410-1b5146add898/go.mod h1:I/5z698sn9Ka8TeJc9MKroUUfqBBauWjQqLJ2OPfmY0=
 golang.org/x/xerrors v0.0.0-20191204190536-9bdfabe68543/go.mod h1:I/5z698sn9Ka8TeJc9MKroUUfqBBauWjQqLJ2OPfmY0=
@@ -244,8 +239,8 @@ google.golang.org/protobuf v1.20.1-0.20200309200217-e05f789c0967/go.mod h1:A+miE
 google.golang.org/protobuf v1.21.0/go.mod h1:47Nbq4nVaFHyn7ilMalzfO3qCViNmqZ2kzikPIcrTAo=
 google.golang.org/protobuf v1.23.0/go.mod h1:EGpADcykh3NcUnDUJcl1+ZksZNG86OlYog2l/sGQquU=
 google.golang.org/protobuf v1.26.0-rc.1/go.mod h1:jlhhOSvTdKEhbULTjvd4ARK9grFBp09yW+WbY/TyQbw=
-google.golang.org/protobuf v1.36.10 h1:AYd7cD/uASjIL6Q9LiTjz8JLcrh/88q5UObnmY3aOOE=
-google.golang.org/protobuf v1.36.10/go.mod h1:HTf+CrKn2C3g5S8VImy6tdcUvCska2kB7j23XfzDpco=
+google.golang.org/protobuf v1.36.6 h1:z1NpPI8ku2WgiWnf+t9wTPsn6eP1L7ksHUlkfLvd9xY=
+google.golang.org/protobuf v1.36.6/go.mod h1:jduwjTPXsFjZGTmRluh+L6NjiWu7pchiJ2/5YcXBHnY=
 gopkg.in/alecthomas/kingpin.v2 v2.2.6/go.mod h1:FMv+mEhP44yOT+4EoQTLFTRgOQ1FBLkstjWtayDeSgw=
 gopkg.in/check.v1 v0.0.0-20161208181325-20d25e280405/go.mod h1:Co6ibVJAznAaIkqp8huTwlJQCZ016jof/cbN4VW5Yz0=
 gopkg.in/check.v1 v1.0.0-20190902080502-41f04d3bba15/go.mod h1:Co6ibVJAznAaIkqp8huTwlJQCZ016jof/cbN4VW5Yz0=

handshake_ix.go

@@ -2,6 +2,7 @@ package nebula
 
 import (
 	"net/netip"
+	"slices"
 	"time"
 
 	"github.com/flynn/noise"
@@ -22,19 +23,15 @@ func ixHandshakeStage0(f *Interface, hh *HandshakeHostInfo) bool {
 		return false
 	}
 
+	// If we're connecting to a v6 address we must use a v2 cert
 	cs := f.pki.getCertState()
 	v := cs.initiatingVersion
-	if hh.initiatingVersionOverride != cert.VersionPre1 {
-		v = hh.initiatingVersionOverride
-	} else if v < cert.Version2 {
-		// If we're connecting to a v6 address we should encourage use of a V2 cert
 	for _, a := range hh.hostinfo.vpnAddrs {
 		if a.Is6() {
 			v = cert.Version2
 			break
 		}
 	}
-	}
 
 	crt := cs.getCertificate(v)
 	if crt == nil {
@@ -51,7 +48,6 @@ func ixHandshakeStage0(f *Interface, hh *HandshakeHostInfo) bool {
 			WithField("handshake", m{"stage": 0, "style": "ix_psk0"}).
 			WithField("certVersion", v).
 			Error("Unable to handshake with host because no certificate handshake bytes is available")
-		return false
 	}
 
 	ci, err := NewConnectionState(f.l, cs, crt, true, noise.HandshakeIX)
@@ -107,7 +103,6 @@ func ixHandshakeStage1(f *Interface, addr netip.AddrPort, via *ViaSender, packet
 			WithField("handshake", m{"stage": 0, "style": "ix_psk0"}).
 			WithField("certVersion", cs.initiatingVersion).
 			Error("Unable to handshake with host because no certificate is available")
-		return
 	}
 
 	ci, err := NewConnectionState(f.l, cs, crt, false, noise.HandshakeIX)
@@ -148,8 +143,8 @@ func ixHandshakeStage1(f *Interface, addr netip.AddrPort, via *ViaSender, packet
 
 	remoteCert, err := f.pki.GetCAPool().VerifyCertificate(time.Now(), rc)
 	if err != nil {
-		fp, fperr := rc.Fingerprint()
-		if fperr != nil {
+		fp, err := rc.Fingerprint()
+		if err != nil {
 			fp = "<error generating certificate fingerprint>"
 		}
 
@@ -168,19 +163,16 @@ func ixHandshakeStage1(f *Interface, addr netip.AddrPort, via *ViaSender, packet
 
 	if remoteCert.Certificate.Version() != ci.myCert.Version() {
 		// We started off using the wrong certificate version, lets see if we can match the version that was sent to us
-		myCertOtherVersion := cs.getCertificate(remoteCert.Certificate.Version())
-		if myCertOtherVersion == nil {
-			if f.l.Level >= logrus.DebugLevel {
-				f.l.WithError(err).WithFields(m{
-					"udpAddr":   addr,
-					"handshake": m{"stage": 1, "style": "ix_psk0"},
-					"cert":      remoteCert,
-				}).Debug("Might be unable to handshake with host due to missing certificate version")
+		rc := cs.getCertificate(remoteCert.Certificate.Version())
+		if rc == nil {
+			f.l.WithError(err).WithField("udpAddr", addr).
+				WithField("handshake", m{"stage": 1, "style": "ix_psk0"}).WithField("cert", remoteCert).
+				Info("Unable to handshake with host due to missing certificate version")
+			return
 		}
-		} else {
+
 		// Record the certificate we are actually using
-			ci.myCert = myCertOtherVersion
-		}
+		ci.myCert = rc
 	}
 
 	if len(remoteCert.Certificate.Networks()) == 0 {
@@ -191,17 +183,17 @@ func ixHandshakeStage1(f *Interface, addr netip.AddrPort, via *ViaSender, packet
 		return
 	}
 
+	var vpnAddrs []netip.Addr
+	var filteredNetworks []netip.Prefix
 	certName := remoteCert.Certificate.Name()
 	certVersion := remoteCert.Certificate.Version()
 	fingerprint := remoteCert.Fingerprint
 	issuer := remoteCert.Certificate.Issuer()
-	vpnNetworks := remoteCert.Certificate.Networks()
 
-	anyVpnAddrsInCommon := false
-	vpnAddrs := make([]netip.Addr, len(vpnNetworks))
-	for i, network := range vpnNetworks {
-		if f.myVpnAddrsTable.Contains(network.Addr()) {
-			f.l.WithField("vpnNetworks", vpnNetworks).WithField("udpAddr", addr).
+	for _, network := range remoteCert.Certificate.Networks() {
+		vpnAddr := network.Addr()
+		if f.myVpnAddrsTable.Contains(vpnAddr) {
+			f.l.WithField("vpnAddr", vpnAddr).WithField("udpAddr", addr).
 				WithField("certName", certName).
 				WithField("certVersion", certVersion).
 				WithField("fingerprint", fingerprint).
@@ -209,10 +201,24 @@ func ixHandshakeStage1(f *Interface, addr netip.AddrPort, via *ViaSender, packet
 				WithField("handshake", m{"stage": 1, "style": "ix_psk0"}).Error("Refusing to handshake with myself")
 			return
 		}
-		vpnAddrs[i] = network.Addr()
-		if f.myVpnNetworksTable.Contains(network.Addr()) {
-			anyVpnAddrsInCommon = true
+
+		// vpnAddrs outside our vpn networks are of no use to us, filter them out
+		if !f.myVpnNetworksTable.Contains(vpnAddr) {
+			continue
 		}
+
+		filteredNetworks = append(filteredNetworks, network)
+		vpnAddrs = append(vpnAddrs, vpnAddr)
+	}
+
+	if len(vpnAddrs) == 0 {
+		f.l.WithError(err).WithField("udpAddr", addr).
+			WithField("certName", certName).
+			WithField("certVersion", certVersion).
+			WithField("fingerprint", fingerprint).
+			WithField("issuer", issuer).
+			WithField("handshake", m{"stage": 1, "style": "ix_psk0"}).Error("No usable vpn addresses from host, refusing handshake")
+		return
 	}
 
 	if addr.IsValid() {
@@ -249,30 +255,26 @@ func ixHandshakeStage1(f *Interface, addr netip.AddrPort, via *ViaSender, packet
 		},
 	}
 
-	msgRxL := f.l.WithFields(m{
-		"vpnAddrs":       vpnAddrs,
-		"udpAddr":        addr,
-		"certName":       certName,
-		"certVersion":    certVersion,
-		"fingerprint":    fingerprint,
-		"issuer":         issuer,
-		"initiatorIndex": hs.Details.InitiatorIndex,
-		"responderIndex": hs.Details.ResponderIndex,
-		"remoteIndex":    h.RemoteIndex,
-		"handshake":      m{"stage": 1, "style": "ix_psk0"},
-	})
-
-	if anyVpnAddrsInCommon {
-		msgRxL.Info("Handshake message received")
-	} else {
-		//todo warn if not lighthouse or relay?
-		msgRxL.Info("Handshake message received, but no vpnNetworks in common.")
-	}
+	f.l.WithField("vpnAddrs", vpnAddrs).WithField("udpAddr", addr).
+		WithField("certName", certName).
+		WithField("certVersion", certVersion).
+		WithField("fingerprint", fingerprint).
+		WithField("issuer", issuer).
+		WithField("initiatorIndex", hs.Details.InitiatorIndex).WithField("responderIndex", hs.Details.ResponderIndex).
+		WithField("remoteIndex", h.RemoteIndex).WithField("handshake", m{"stage": 1, "style": "ix_psk0"}).
+		Info("Handshake message received")
 
 	hs.Details.ResponderIndex = myIndex
 	hs.Details.Cert = cs.getHandshakeBytes(ci.myCert.Version())
 	if hs.Details.Cert == nil {
-		msgRxL.WithField("myCertVersion", ci.myCert.Version()).
+		f.l.WithField("vpnAddrs", vpnAddrs).WithField("udpAddr", addr).
+			WithField("certName", certName).
+			WithField("certVersion", certVersion).
+			WithField("fingerprint", fingerprint).
+			WithField("issuer", issuer).
+			WithField("initiatorIndex", hs.Details.InitiatorIndex).WithField("responderIndex", hs.Details.ResponderIndex).
+			WithField("remoteIndex", h.RemoteIndex).WithField("handshake", m{"stage": 1, "style": "ix_psk0"}).
+			WithField("certVersion", ci.myCert.Version()).
 			Error("Unable to handshake with host because no certificate handshake bytes is available")
 		return
 	}
@@ -330,7 +332,7 @@ func ixHandshakeStage1(f *Interface, addr netip.AddrPort, via *ViaSender, packet
 
 	hostinfo.remotes = f.lightHouse.QueryCache(vpnAddrs)
 	hostinfo.SetRemote(addr)
-	hostinfo.buildNetworks(f.myVpnNetworksTable, remoteCert.Certificate)
+	hostinfo.buildNetworks(filteredNetworks, remoteCert.Certificate.UnsafeNetworks())
 
 	existing, err := f.handshakeManager.CheckAndComplete(hostinfo, 0, f)
 	if err != nil {
@@ -571,22 +573,31 @@ func ixHandshakeStage2(f *Interface, addr netip.AddrPort, via *ViaSender, hh *Ha
 		hostinfo.relayState.InsertRelayTo(via.relayHI.vpnAddrs[0])
 	}
 
-	correctHostResponded := false
-	anyVpnAddrsInCommon := false
-	vpnAddrs := make([]netip.Addr, len(vpnNetworks))
-	for i, network := range vpnNetworks {
-		vpnAddrs[i] = network.Addr()
-		if f.myVpnNetworksTable.Contains(network.Addr()) {
-			anyVpnAddrsInCommon = true
+	var vpnAddrs []netip.Addr
+	var filteredNetworks []netip.Prefix
+	for _, network := range vpnNetworks {
+		// vpnAddrs outside our vpn networks are of no use to us, filter them out
+		vpnAddr := network.Addr()
+		if !f.myVpnNetworksTable.Contains(vpnAddr) {
+			continue
 		}
-		if hostinfo.vpnAddrs[0] == network.Addr() {
-			// todo is it more correct to see if any of hostinfo.vpnAddrs are in the cert? it should have len==1, but one day it might not?
-			correctHostResponded = true
+
+		filteredNetworks = append(filteredNetworks, network)
+		vpnAddrs = append(vpnAddrs, vpnAddr)
 	}
+
+	if len(vpnAddrs) == 0 {
+		f.l.WithError(err).WithField("udpAddr", addr).
+			WithField("certName", certName).
+			WithField("certVersion", certVersion).
+			WithField("fingerprint", fingerprint).
+			WithField("issuer", issuer).
+			WithField("handshake", m{"stage": 2, "style": "ix_psk0"}).Error("No usable vpn addresses from host, refusing handshake")
+		return true
 	}
 
 	// Ensure the right host responded
-	if !correctHostResponded {
+	if !slices.Contains(vpnAddrs, hostinfo.vpnAddrs[0]) {
 		f.l.WithField("intendedVpnAddrs", hostinfo.vpnAddrs).WithField("haveVpnNetworks", vpnNetworks).
 			WithField("udpAddr", addr).
 			WithField("certName", certName).
@@ -598,7 +609,6 @@ func ixHandshakeStage2(f *Interface, addr netip.AddrPort, via *ViaSender, hh *Ha
 		f.handshakeManager.DeleteHostInfo(hostinfo)
 
 		// Create a new hostinfo/handshake for the intended vpn ip
-		//TODO is hostinfo.vpnAddrs[0] always the address to use?
 		f.handshakeManager.StartHandshake(hostinfo.vpnAddrs[0], func(newHH *HandshakeHostInfo) {
 			// Block the current used address
 			newHH.hostinfo.remotes = hostinfo.remotes
@@ -625,7 +635,7 @@ func ixHandshakeStage2(f *Interface, addr netip.AddrPort, via *ViaSender, hh *Ha
 	ci.window.Update(f.l, 2)
 
 	duration := time.Since(hh.startTime).Nanoseconds()
-	msgRxL := f.l.WithField("vpnAddrs", vpnAddrs).WithField("udpAddr", addr).
+	f.l.WithField("vpnAddrs", vpnAddrs).WithField("udpAddr", addr).
 		WithField("certName", certName).
 		WithField("certVersion", certVersion).
 		WithField("fingerprint", fingerprint).
@@ -633,17 +643,12 @@ func ixHandshakeStage2(f *Interface, addr netip.AddrPort, via *ViaSender, hh *Ha
 		WithField("initiatorIndex", hs.Details.InitiatorIndex).WithField("responderIndex", hs.Details.ResponderIndex).
 		WithField("remoteIndex", h.RemoteIndex).WithField("handshake", m{"stage": 2, "style": "ix_psk0"}).
 		WithField("durationNs", duration).
-		WithField("sentCachedPackets", len(hh.packetStore))
-	if anyVpnAddrsInCommon {
-		msgRxL.Info("Handshake message received")
-	} else {
-		//todo warn if not lighthouse or relay?
-		msgRxL.Info("Handshake message received, but no vpnNetworks in common.")
-	}
+		WithField("sentCachedPackets", len(hh.packetStore)).
+		Info("Handshake message received")
 
 	// Build up the radix for the firewall if we have subnets in the cert
 	hostinfo.vpnAddrs = vpnAddrs
-	hostinfo.buildNetworks(f.myVpnNetworksTable, remoteCert.Certificate)
+	hostinfo.buildNetworks(filteredNetworks, remoteCert.Certificate.UnsafeNetworks())
 
 	// Complete our handshake and update metrics, this will replace any existing tunnels for the vpnAddrs here
 	f.handshakeManager.Complete(hostinfo, f)

handshake_manager.go

@@ -70,7 +70,6 @@ type HandshakeHostInfo struct {
 
 	startTime   time.Time        // Time that we first started trying with this handshake
 	ready       bool             // Is the handshake ready
-	initiatingVersionOverride cert.Version     // Should we use a non-default cert version for this handshake?
 	counter     int64            // How many attempts have we made so far
 	lastRemotes []netip.AddrPort // Remotes that we sent to during the previous attempt
 	packetStore []*cachedPacket  // A set of packets to be transmitted once the handshake completes
@@ -269,12 +268,12 @@ func (hm *HandshakeManager) handleOutbound(vpnIp netip.Addr, lighthouseTriggered
 		hostinfo.logger(hm.l).WithField("relays", hostinfo.remotes.relays).Info("Attempt to relay through hosts")
 		// Send a RelayRequest to all known Relay IP's
 		for _, relay := range hostinfo.remotes.relays {
-			// Don't relay through the host I'm trying to connect to
+			// Don't relay to myself
 			if relay == vpnIp {
 				continue
 			}
 
-			// Don't relay to myself
+			// Don't relay through the host I'm trying to connect to
 			if hm.f.myVpnAddrsTable.Contains(relay) {
 				continue
 			}

hostmap.go

@@ -212,18 +212,6 @@ func (rs *RelayState) InsertRelay(ip netip.Addr, idx uint32, r *Relay) {
 	rs.relayForByIdx[idx] = r
 }
 
-type NetworkType uint8
-
-const (
-	NetworkTypeUnknown NetworkType = iota
-	// NetworkTypeVPN is a network that overlaps one or more of the vpnNetworks in our certificate
-	NetworkTypeVPN
-	// NetworkTypeVPNPeer is a network that does not overlap one of our networks
-	NetworkTypeVPNPeer
-	// NetworkTypeUnsafe is a network from Certificate.UnsafeNetworks()
-	NetworkTypeUnsafe
-)
-
 type HostInfo struct {
 	remote          netip.AddrPort
 	remotes         *RemoteList
@@ -237,8 +225,8 @@ type HostInfo struct {
 	// vpn networks but were removed because they are not usable
 	vpnAddrs []netip.Addr
 
-	// networks is a combination of specific vpn addresses (not prefixes!) and full unsafe networks assigned to this host.
-	networks   *bart.Table[NetworkType]
+	// networks are both all vpn and unsafe networks assigned to this host
+	networks   *bart.Lite
 	relayState RelayState
 
 	// HandshakePacket records the packets used to create this hostinfo
@@ -742,26 +730,19 @@ func (i *HostInfo) SetRemoteIfPreferred(hm *HostMap, newRemote netip.AddrPort) b
 	return false
 }
 
-// buildNetworks fills in the networks field of HostInfo. It accepts a cert.Certificate so you never ever mix the network types up.
-func (i *HostInfo) buildNetworks(myVpnNetworksTable *bart.Lite, c cert.Certificate) {
-	if len(c.Networks()) == 1 && len(c.UnsafeNetworks()) == 0 {
-		if myVpnNetworksTable.Contains(c.Networks()[0].Addr()) {
-			return // Simple case, no BART needed
-		}
+func (i *HostInfo) buildNetworks(networks, unsafeNetworks []netip.Prefix) {
+	if len(networks) == 1 && len(unsafeNetworks) == 0 {
+		// Simple case, no CIDRTree needed
+		return
 	}
 
-	i.networks = new(bart.Table[NetworkType])
-	for _, network := range c.Networks() {
-		nprefix := netip.PrefixFrom(network.Addr(), network.Addr().BitLen())
-		if myVpnNetworksTable.Contains(network.Addr()) {
-			i.networks.Insert(nprefix, NetworkTypeVPN)
-		} else {
-			i.networks.Insert(nprefix, NetworkTypeVPNPeer)
-		}
+	i.networks = new(bart.Lite)
+	for _, network := range networks {
+		i.networks.Insert(network)
 	}
 
-	for _, network := range c.UnsafeNetworks() {
-		i.networks.Insert(network, NetworkTypeUnsafe)
+	for _, network := range unsafeNetworks {
+		i.networks.Insert(network)
 	}
 }
 

inside.go

@@ -120,10 +120,9 @@ func (f *Interface) rejectOutside(packet []byte, ci *ConnectionState, hostinfo *
 	f.sendNoMetrics(header.Message, 0, ci, hostinfo, netip.AddrPort{}, out, nb, packet, q)
 }
 
-// Handshake will attempt to initiate a tunnel with the provided vpn address. This is a no-op if the tunnel is already established or being established
-// it does not check if it is within our vpn networks!
+// Handshake will attempt to initiate a tunnel with the provided vpn address if it is within our vpn networks. This is a no-op if the tunnel is already established or being established
 func (f *Interface) Handshake(vpnAddr netip.Addr) {
-	f.handshakeManager.GetOrHandshake(vpnAddr, nil)
+	f.getOrHandshakeNoRouting(vpnAddr, nil)
 }
 
 // getOrHandshakeNoRouting returns nil if the vpnAddr is not routable.
@@ -139,6 +138,7 @@ func (f *Interface) getOrHandshakeNoRouting(vpnAddr netip.Addr, cacheCallback fu
 // getOrHandshakeConsiderRouting will try to find the HostInfo to handle this packet, starting a handshake if necessary.
 // If the 2nd return var is false then the hostinfo is not ready to be used in a tunnel.
 func (f *Interface) getOrHandshakeConsiderRouting(fwPacket *firewall.Packet, cacheCallback func(*HandshakeHostInfo)) (*HostInfo, bool) {
+
 	destinationAddr := fwPacket.RemoteAddr
 
 	hostinfo, ready := f.getOrHandshakeNoRouting(destinationAddr, cacheCallback)
@@ -231,10 +231,9 @@ func (f *Interface) sendMessageNow(t header.MessageType, st header.MessageSubTyp
 	f.sendNoMetrics(header.Message, st, hostinfo.ConnectionState, hostinfo, netip.AddrPort{}, p, nb, out, 0)
 }
 
-// SendMessageToVpnAddr handles real addr:port lookup and sends to the current best known address for vpnAddr.
-// This function ignores myVpnNetworksTable, and will always attempt to treat the address as a vpnAddr
+// SendMessageToVpnAddr handles real addr:port lookup and sends to the current best known address for vpnAddr
 func (f *Interface) SendMessageToVpnAddr(t header.MessageType, st header.MessageSubType, vpnAddr netip.Addr, p, nb, out []byte) {
-	hostInfo, ready := f.handshakeManager.GetOrHandshake(vpnAddr, func(hh *HandshakeHostInfo) {
+	hostInfo, ready := f.getOrHandshakeNoRouting(vpnAddr, func(hh *HandshakeHostInfo) {
 		hh.cachePacket(f.l, t, st, p, f.SendMessageToHostInfo, f.cachedPacketMetrics)
 	})
 

interface.go

@@ -6,8 +6,8 @@ import (
 	"fmt"
 	"io"
 	"net/netip"
-	"os"
 	"runtime"
+	"sync"
 	"sync/atomic"
 	"time"
 
@@ -18,10 +18,18 @@ import (
 	"github.com/slackhq/nebula/firewall"
 	"github.com/slackhq/nebula/header"
 	"github.com/slackhq/nebula/overlay"
+	"github.com/slackhq/nebula/packet"
 	"github.com/slackhq/nebula/udp"
 )
 
-const mtu = 9001
+const (
+	mtu = 9001
+
+	inboundBatchSizeDefault      = 32
+	outboundBatchSizeDefault     = 32
+	batchFlushIntervalDefault    = 50 * time.Microsecond
+	maxOutstandingBatchesDefault = 1028
+)
 
 type InterfaceConfig struct {
 	HostMap            *HostMap
@@ -47,9 +55,17 @@ type InterfaceConfig struct {
 	reQueryWait     time.Duration
 
 	ConntrackCacheTimeout time.Duration
+	BatchConfig           BatchConfig
 	l                     *logrus.Logger
 }
 
+type BatchConfig struct {
+	InboundBatchSize      int
+	OutboundBatchSize     int
+	FlushInterval         time.Duration
+	MaxOutstandingPerChan int
+}
+
 type Interface struct {
 	hostMap               *HostMap
 	outside               udp.Conn
@@ -87,12 +103,95 @@ type Interface struct {
 
 	writers []udp.Conn
 	readers []io.ReadWriteCloser
+	wg      sync.WaitGroup
 
 	metricHandshakes    metrics.Histogram
 	messageMetrics      *MessageMetrics
 	cachedPacketMetrics *cachedPacketMetrics
 
 	l *logrus.Logger
+
+	inPool  sync.Pool
+	inbound []chan *packetBatch
+
+	outPool  sync.Pool
+	outbound []chan *outboundBatch
+
+	packetBatchPool   sync.Pool
+	outboundBatchPool sync.Pool
+
+	inboundBatchSize      int
+	outboundBatchSize     int
+	batchFlushInterval    time.Duration
+	maxOutstandingPerChan int
+}
+
+type packetBatch struct {
+	packets []*packet.Packet
+}
+
+func newPacketBatch(capacity int) *packetBatch {
+	return &packetBatch{
+		packets: make([]*packet.Packet, 0, capacity),
+	}
+}
+
+func (b *packetBatch) add(p *packet.Packet) {
+	b.packets = append(b.packets, p)
+}
+
+func (b *packetBatch) reset() {
+	for i := range b.packets {
+		b.packets[i] = nil
+	}
+	b.packets = b.packets[:0]
+}
+
+func (f *Interface) getPacketBatch() *packetBatch {
+	if v := f.packetBatchPool.Get(); v != nil {
+		b := v.(*packetBatch)
+		b.reset()
+		return b
+	}
+	return newPacketBatch(f.inboundBatchSize)
+}
+
+func (f *Interface) releasePacketBatch(b *packetBatch) {
+	b.reset()
+	f.packetBatchPool.Put(b)
+}
+
+type outboundBatch struct {
+	payloads []*[]byte
+}
+
+func newOutboundBatch(capacity int) *outboundBatch {
+	return &outboundBatch{payloads: make([]*[]byte, 0, capacity)}
+}
+
+func (b *outboundBatch) add(buf *[]byte) {
+	b.payloads = append(b.payloads, buf)
+}
+
+func (b *outboundBatch) reset() {
+	for i := range b.payloads {
+		b.payloads[i] = nil
+	}
+	b.payloads = b.payloads[:0]
+}
+
+func (f *Interface) getOutboundBatch() *outboundBatch {
+	if v := f.outboundBatchPool.Get(); v != nil {
+		b := v.(*outboundBatch)
+		b.reset()
+		return b
+	}
+	return newOutboundBatch(f.outboundBatchSize)
+}
+
+func (f *Interface) releaseOutboundBatch(b *outboundBatch) {
+	b.reset()
+	f.outboundBatchPool.Put(b)
 }
 
 type EncWriter interface {
@@ -162,6 +261,20 @@ func NewInterface(ctx context.Context, c *InterfaceConfig) (*Interface, error) {
 	}
 
 	cs := c.pki.getCertState()
+
+	bc := c.BatchConfig
+	if bc.InboundBatchSize <= 0 {
+		bc.InboundBatchSize = inboundBatchSizeDefault
+	}
+	if bc.OutboundBatchSize <= 0 {
+		bc.OutboundBatchSize = outboundBatchSizeDefault
+	}
+	if bc.FlushInterval <= 0 {
+		bc.FlushInterval = batchFlushIntervalDefault
+	}
+	if bc.MaxOutstandingPerChan <= 0 {
+		bc.MaxOutstandingPerChan = maxOutstandingBatchesDefault
+	}
 	ifce := &Interface{
 		pki:                   c.pki,
 		hostMap:               c.HostMap,
@@ -194,9 +307,39 @@ func NewInterface(ctx context.Context, c *InterfaceConfig) (*Interface, error) {
 			dropped: metrics.GetOrRegisterCounter("hostinfo.cached_packets.dropped", nil),
 		},
 
+		inbound:  make([]chan *packetBatch, c.routines),
+		outbound: make([]chan *outboundBatch, c.routines),
+
 		l: c.l,
+
+		inboundBatchSize:      bc.InboundBatchSize,
+		outboundBatchSize:     bc.OutboundBatchSize,
+		batchFlushInterval:    bc.FlushInterval,
+		maxOutstandingPerChan: bc.MaxOutstandingPerChan,
 	}
 
+	for i := 0; i < c.routines; i++ {
+		ifce.inbound[i] = make(chan *packetBatch, ifce.maxOutstandingPerChan)
+		ifce.outbound[i] = make(chan *outboundBatch, ifce.maxOutstandingPerChan)
+	}
+
+	ifce.inPool = sync.Pool{New: func() any {
+		return packet.New()
+	}}
+
+	ifce.outPool = sync.Pool{New: func() any {
+		t := make([]byte, mtu)
+		return &t
+	}}
+
+	ifce.packetBatchPool = sync.Pool{New: func() any {
+		return newPacketBatch(ifce.inboundBatchSize)
+	}}
+
+	ifce.outboundBatchPool = sync.Pool{New: func() any {
+		return newOutboundBatch(ifce.outboundBatchSize)
+	}}
+
 	ifce.tryPromoteEvery.Store(c.tryPromoteEvery)
 	ifce.reQueryEvery.Store(c.reQueryEvery)
 	ifce.reQueryWait.Store(int64(c.reQueryWait))
@@ -209,7 +352,7 @@ func NewInterface(ctx context.Context, c *InterfaceConfig) (*Interface, error) {
 // activate creates the interface on the host. After the interface is created, any
 // other services that want to bind listeners to its IP may do so successfully. However,
 // the interface isn't going to process anything until run() is called.
-func (f *Interface) activate() {
+func (f *Interface) activate() error {
 	// actually turn on tun dev
 
 	addr, err := f.outside.LocalAddr()
@@ -230,33 +373,44 @@ func (f *Interface) activate() {
 		if i > 0 {
 			reader, err = f.inside.NewMultiQueueReader()
 			if err != nil {
-				f.l.Fatal(err)
+				return err
 			}
 		}
 		f.readers[i] = reader
 	}
 
-	if err := f.inside.Activate(); err != nil {
+	if err = f.inside.Activate(); err != nil {
 		f.inside.Close()
-		f.l.Fatal(err)
+		return err
 	}
+
+	return nil
 }
 
-func (f *Interface) run() {
-	// Launch n queues to read packets from udp
+func (f *Interface) run(c context.Context) (func(), error) {
 	for i := 0; i < f.routines; i++ {
+		// Launch n queues to read packets from udp
+		f.wg.Add(1)
 		go f.listenOut(i)
-	}
 
 		// Launch n queues to read packets from tun dev
-	for i := 0; i < f.routines; i++ {
+		f.wg.Add(1)
 		go f.listenIn(f.readers[i], i)
+
+		// Launch n queues to read packets from tun dev
+		f.wg.Add(1)
+		go f.workerIn(i, c)
+
+		// Launch n queues to read packets from tun dev
+		f.wg.Add(1)
+		go f.workerOut(i, c)
 	}
+
+	return f.wg.Wait, nil
 }
 
 func (f *Interface) listenOut(i int) {
 	runtime.LockOSThread()
-
 	var li udp.Conn
 	if i > 0 {
 		li = f.writers[i]
@@ -264,41 +418,142 @@ func (f *Interface) listenOut(i int) {
 		li = f.outside
 	}
 
-	ctCache := firewall.NewConntrackCacheTicker(f.conntrackCacheTimeout)
-	lhh := f.lightHouse.NewRequestHandler()
-	plaintext := make([]byte, udp.MTU)
-	h := &header.H{}
-	fwPacket := &firewall.Packet{}
-	nb := make([]byte, 12, 12)
+	batch := f.getPacketBatch()
+	lastFlush := time.Now()
 
-	li.ListenOut(func(fromUdpAddr netip.AddrPort, payload []byte) {
-		f.readOutsidePackets(fromUdpAddr, nil, plaintext[:0], payload, h, fwPacket, lhh, nb, i, ctCache.Get(f.l))
+	flush := func(force bool) {
+		if len(batch.packets) == 0 {
+			if force {
+				f.releasePacketBatch(batch)
+			}
+			return
+		}
+
+		f.inbound[i] <- batch
+		batch = f.getPacketBatch()
+		lastFlush = time.Now()
+	}
+
+	err := li.ListenOut(func(fromUdpAddr netip.AddrPort, payload []byte) {
+		p := f.inPool.Get().(*packet.Packet)
+		p.Payload = p.Payload[:mtu]
+		copy(p.Payload, payload)
+		p.Payload = p.Payload[:len(payload)]
+		p.Addr = fromUdpAddr
+		batch.add(p)
+
+		if len(batch.packets) >= f.inboundBatchSize || time.Since(lastFlush) >= f.batchFlushInterval {
+			flush(false)
+		}
 	})
+
+	if len(batch.packets) > 0 {
+		f.inbound[i] <- batch
+	} else {
+		f.releasePacketBatch(batch)
+	}
+
+	if err != nil && !f.closed.Load() {
+		f.l.WithError(err).Error("Error while reading packet inbound packet, closing")
+		//TODO: Trigger Control to close
+	}
+
+	f.l.Debugf("underlay reader %v is done", i)
+	f.wg.Done()
 }
 
 func (f *Interface) listenIn(reader io.ReadWriteCloser, i int) {
 	runtime.LockOSThread()
 
-	packet := make([]byte, mtu)
-	out := make([]byte, mtu)
-	fwPacket := &firewall.Packet{}
-	nb := make([]byte, 12, 12)
+	batch := f.getOutboundBatch()
+	lastFlush := time.Now()
 
-	conntrackCache := firewall.NewConntrackCacheTicker(f.conntrackCacheTimeout)
-
-	for {
-		n, err := reader.Read(packet)
-		if err != nil {
-			if errors.Is(err, os.ErrClosed) && f.closed.Load() {
+	flush := func(force bool) {
+		if len(batch.payloads) == 0 {
+			if force {
+				f.releaseOutboundBatch(batch)
+			}
 			return
 		}
 
-			f.l.WithError(err).Error("Error while reading outbound packet")
-			// This only seems to happen when something fatal happens to the fd, so exit.
-			os.Exit(2)
+		f.outbound[i] <- batch
+		batch = f.getOutboundBatch()
+		lastFlush = time.Now()
 	}
 
-		f.consumeInsidePacket(packet[:n], fwPacket, nb, out, i, conntrackCache.Get(f.l))
+	for {
+		p := f.outPool.Get().(*[]byte)
+		*p = (*p)[:mtu]
+		n, err := reader.Read(*p)
+		if err != nil {
+			if !f.closed.Load() {
+				f.l.WithError(err).Error("Error while reading outbound packet, closing")
+				//TODO: Trigger Control to close
+			}
+			break
+		}
+
+		*p = (*p)[:n]
+		batch.add(p)
+
+		if len(batch.payloads) >= f.outboundBatchSize || time.Since(lastFlush) >= f.batchFlushInterval {
+			flush(false)
+		}
+	}
+
+	if len(batch.payloads) > 0 {
+		f.outbound[i] <- batch
+	} else {
+		f.releaseOutboundBatch(batch)
+	}
+
+	f.l.Debugf("overlay reader %v is done", i)
+	f.wg.Done()
+}
+
+func (f *Interface) workerIn(i int, ctx context.Context) {
+	lhh := f.lightHouse.NewRequestHandler()
+	conntrackCache := firewall.NewConntrackCacheTicker(f.conntrackCacheTimeout)
+	fwPacket2 := &firewall.Packet{}
+	nb2 := make([]byte, 12, 12)
+	result2 := make([]byte, mtu)
+	h := &header.H{}
+
+	for {
+		select {
+		case batch := <-f.inbound[i]:
+			for _, p := range batch.packets {
+				f.readOutsidePackets(p.Addr, nil, result2[:0], p.Payload, h, fwPacket2, lhh, nb2, i, conntrackCache.Get(f.l))
+				p.Payload = p.Payload[:mtu]
+				f.inPool.Put(p)
+			}
+			f.releasePacketBatch(batch)
+		case <-ctx.Done():
+			f.wg.Done()
+			return
+		}
+	}
+}
+
+func (f *Interface) workerOut(i int, ctx context.Context) {
+	conntrackCache := firewall.NewConntrackCacheTicker(f.conntrackCacheTimeout)
+	fwPacket1 := &firewall.Packet{}
+	nb1 := make([]byte, 12, 12)
+	result1 := make([]byte, mtu)
+
+	for {
+		select {
+		case batch := <-f.outbound[i]:
+			for _, data := range batch.payloads {
+				f.consumeInsidePacket(*data, fwPacket1, nb1, result1, i, conntrackCache.Get(f.l))
+				*data = (*data)[:mtu]
+				f.outPool.Put(data)
+			}
+			f.releaseOutboundBatch(batch)
+		case <-ctx.Done():
+			f.wg.Done()
+			return
+		}
 	}
 }
 
@@ -451,6 +706,7 @@ func (f *Interface) GetCertState() *CertState {
 func (f *Interface) Close() error {
 	f.closed.Store(true)
 
+	// Release the udp readers
 	for _, u := range f.writers {
 		err := u.Close()
 		if err != nil {
@@ -458,6 +714,13 @@ func (f *Interface) Close() error {
 		}
 	}
 
-	// Release the tun device
-	return f.inside.Close()
+	// Release the tun readers
+	for _, u := range f.readers {
+		err := u.Close()
+		if err != nil {
+			f.l.WithError(err).Error("Error while closing tun device")
+		}
+	}
+
+	return nil
 }

lighthouse.go

@@ -360,8 +360,7 @@ func (lh *LightHouse) parseLighthouses(c *config.C) ([]netip.Addr, error) {
 		}
 
 		if !lh.myVpnNetworksTable.Contains(addr) {
-			lh.l.WithFields(m{"vpnAddr": addr, "networks": lh.myVpnNetworks}).
-				Warn("lighthouse host is not within our networks, lighthouse functionality will work but layer 3 network traffic to the lighthouse will not")
+			return nil, util.NewContextualError("lighthouse host is not in our networks, invalid", m{"vpnAddr": addr, "networks": lh.myVpnNetworks}, nil)
 		}
 		out[i] = addr
 	}
@@ -432,8 +431,7 @@ func (lh *LightHouse) loadStaticMap(c *config.C, staticList map[netip.Addr]struc
 		}
 
 		if !lh.myVpnNetworksTable.Contains(vpnAddr) {
-			lh.l.WithFields(m{"vpnAddr": vpnAddr, "networks": lh.myVpnNetworks, "entry": i + 1}).
-				Warn("static_host_map key is not within our networks, layer 3 network traffic to this host will not work")
+			return util.NewContextualError("static_host_map key is not in our network, invalid", m{"vpnAddr": vpnAddr, "networks": lh.myVpnNetworks, "entry": i + 1}, nil)
 		}
 
 		vals, ok := v.([]any)
@@ -1339,19 +1337,12 @@ func (lhh *LightHouseHandler) handleHostPunchNotification(n *NebulaMeta, fromVpn
 		}
 	}
 
-	remoteAllowList := lhh.lh.GetRemoteAllowList()
 	for _, a := range n.Details.V4AddrPorts {
-		b := protoV4AddrPortToNetAddrPort(a)
-		if remoteAllowList.Allow(detailsVpnAddr, b.Addr()) {
-			punch(b, detailsVpnAddr)
-		}
+		punch(protoV4AddrPortToNetAddrPort(a), detailsVpnAddr)
 	}
 
 	for _, a := range n.Details.V6AddrPorts {
-		b := protoV6AddrPortToNetAddrPort(a)
-		if remoteAllowList.Allow(detailsVpnAddr, b.Addr()) {
-			punch(b, detailsVpnAddr)
-		}
+		punch(protoV6AddrPortToNetAddrPort(a), detailsVpnAddr)
 	}
 
 	// This sends a nebula test packet to the host trying to contact us. In the case

lighthouse_test.go

@@ -14,7 +14,7 @@ import (
 	"github.com/slackhq/nebula/test"
 	"github.com/stretchr/testify/assert"
 	"github.com/stretchr/testify/require"
-	"go.yaml.in/yaml/v3"
+	"gopkg.in/yaml.v3"
 )
 
 func TestOldIPv4Only(t *testing.T) {

main.go

@@ -5,8 +5,6 @@ import (
 	"fmt"
 	"net"
 	"net/netip"
-	"runtime/debug"
-	"strings"
 	"time"
 
 	"github.com/sirupsen/logrus"
@@ -15,7 +13,7 @@ import (
 	"github.com/slackhq/nebula/sshd"
 	"github.com/slackhq/nebula/udp"
 	"github.com/slackhq/nebula/util"
-	"go.yaml.in/yaml/v3"
+	"gopkg.in/yaml.v3"
 )
 
 type m = map[string]any
@@ -29,10 +27,6 @@ func Main(c *config.C, configTest bool, buildVersion string, logger *logrus.Logg
 		}
 	}()
 
-	if buildVersion == "" {
-		buildVersion = moduleVersion()
-	}
-
 	l := logger
 	l.Formatter = &logrus.TextFormatter{
 		FullTimestamp: true,
@@ -81,8 +75,7 @@ func Main(c *config.C, configTest bool, buildVersion string, logger *logrus.Logg
 	if c.GetBool("sshd.enabled", false) {
 		sshStart, err = configSSH(l, ssh, c)
 		if err != nil {
-			l.WithError(err).Warn("Failed to configure sshd, ssh debugging will not be available")
-			sshStart = nil
+			return nil, util.ContextualizeIfNeeded("Error while configuring the sshd", err)
 		}
 	}
 
@@ -228,6 +221,13 @@ func Main(c *config.C, configTest bool, buildVersion string, logger *logrus.Logg
 		}
 	}
 
+	batchCfg := BatchConfig{
+		InboundBatchSize:      c.GetInt("batch.inbound_size", inboundBatchSizeDefault),
+		OutboundBatchSize:     c.GetInt("batch.outbound_size", outboundBatchSizeDefault),
+		FlushInterval:         c.GetDuration("batch.flush_interval", batchFlushIntervalDefault),
+		MaxOutstandingPerChan: c.GetInt("batch.max_outstanding", maxOutstandingBatchesDefault),
+	}
+
 	ifConfig := &InterfaceConfig{
 		HostMap:               hostMap,
 		Inside:                tun,
@@ -249,6 +249,7 @@ func Main(c *config.C, configTest bool, buildVersion string, logger *logrus.Logg
 		relayManager:          NewRelayManager(ctx, l, hostMap, c),
 		punchy:                punchy,
 		ConntrackCacheTimeout: conntrackCacheTimeout,
+		BatchConfig:           batchCfg,
 		l:                     l,
 	}
 
@@ -291,29 +292,14 @@ func Main(c *config.C, configTest bool, buildVersion string, logger *logrus.Logg
 	}
 
 	return &Control{
-		ifce,
-		l,
-		ctx,
-		cancel,
-		sshStart,
-		statsStart,
-		dnsStart,
-		lightHouse.StartUpdateWorker,
-		connManager.Start,
+		f:                      ifce,
+		l:                      l,
+		ctx:                    ctx,
+		cancel:                 cancel,
+		sshStart:               sshStart,
+		statsStart:             statsStart,
+		dnsStart:               dnsStart,
+		lighthouseStart:        lightHouse.StartUpdateWorker,
+		connectionManagerStart: connManager.Start,
 	}, nil
 }
-
-func moduleVersion() string {
-	info, ok := debug.ReadBuildInfo()
-	if !ok {
-		return ""
-	}
-
-	for _, dep := range info.Deps {
-		if dep.Path == "github.com/slackhq/nebula" {
-			return strings.TrimPrefix(dep.Version, "v")
-		}
-	}
-
-	return ""
-}

outside.go

@@ -29,7 +29,7 @@ func (f *Interface) readOutsidePackets(ip netip.AddrPort, via *ViaSender, out []
 		return
 	}
 
-	//l.Error("in packet ", header, packet[HeaderLen:])
+	//f.l.Error("in packet ", h)
 	if ip.IsValid() {
 		if f.myVpnNetworksTable.Contains(ip.Addr()) {
 			if f.l.Level >= logrus.DebugLevel {
@@ -245,6 +245,7 @@ func (f *Interface) handleHostRoaming(hostinfo *HostInfo, udpAddr netip.AddrPort
 			return
 		}
 
+		//TODO: Seems we have a bunch of stuff racing here, since we don't have a lock on hostinfo anymore we announce roaming in bursts
 		hostinfo.logger(f.l).WithField("udpAddr", hostinfo.remote).WithField("newAddr", udpAddr).
 			Info("Host roamed to new udp ip/port.")
 		hostinfo.lastRoam = time.Now()
@@ -470,7 +471,7 @@ func (f *Interface) decryptToTun(hostinfo *HostInfo, messageCounter uint64, out
 
 	out, err = hostinfo.ConnectionState.dKey.DecryptDanger(out, packet[:header.Len], packet[header.Len:], messageCounter, nb)
 	if err != nil {
-		hostinfo.logger(f.l).WithError(err).Error("Failed to decrypt packet")
+		hostinfo.logger(f.l).WithError(err).WithField("fwPacket", fwPacket).Error("Failed to decrypt packet")
 		return false
 	}
 

overlay/tun.go

@@ -1,7 +1,6 @@
 package overlay
 
 import (
-	"fmt"
 	"net"
 	"net/netip"
 
@@ -82,41 +81,3 @@ func prefixToMask(prefix netip.Prefix) netip.Addr {
 	addr, _ := netip.AddrFromSlice(net.CIDRMask(prefix.Bits(), pLen))
 	return addr
 }
-
-func flipBytes(b []byte) []byte {
-	for i := 0; i < len(b); i++ {
-		b[i] ^= 0xFF
-	}
-	return b
-}
-func orBytes(a []byte, b []byte) []byte {
-	ret := make([]byte, len(a))
-	for i := 0; i < len(a); i++ {
-		ret[i] = a[i] | b[i]
-	}
-	return ret
-}
-
-func getBroadcast(cidr netip.Prefix) netip.Addr {
-	broadcast, _ := netip.AddrFromSlice(
-		orBytes(
-			cidr.Addr().AsSlice(),
-			flipBytes(prefixToMask(cidr).AsSlice()),
-		),
-	)
-	return broadcast
-}
-
-func selectGateway(dest netip.Prefix, gateways []netip.Prefix) (netip.Prefix, error) {
-	for _, gateway := range gateways {
-		if dest.Addr().Is4() && gateway.Addr().Is4() {
-			return gateway, nil
-		}
-
-		if dest.Addr().Is6() && gateway.Addr().Is6() {
-			return gateway, nil
-		}
-	}
-
-	return netip.Prefix{}, fmt.Errorf("no gateway found for %v in the list of vpn networks", dest)
-}

overlay/tun_darwin.go

@@ -294,6 +294,7 @@ func (t *tun) activate6(network netip.Prefix) error {
 			Vltime: 0xffffffff,
 			Pltime: 0xffffffff,
 		},
+		//TODO: CERT-V2 should we disable DAD (duplicate address detection) and mark this as a secured address?
 		Flags: _IN6_IFF_NODAD,
 	}
 

overlay/tun_freebsd.go

@@ -501,6 +501,30 @@ func (t *tun) deviceBytes() (o [16]byte) {
 	return
 }
 
+func flipBytes(b []byte) []byte {
+	for i := 0; i < len(b); i++ {
+		b[i] ^= 0xFF
+	}
+	return b
+}
+func orBytes(a []byte, b []byte) []byte {
+	ret := make([]byte, len(a))
+	for i := 0; i < len(a); i++ {
+		ret[i] = a[i] | b[i]
+	}
+	return ret
+}
+
+func getBroadcast(cidr netip.Prefix) netip.Addr {
+	broadcast, _ := netip.AddrFromSlice(
+		orBytes(
+			cidr.Addr().AsSlice(),
+			flipBytes(prefixToMask(cidr).AsSlice()),
+		),
+	)
+	return broadcast
+}
+
 func addRoute(prefix netip.Prefix, gateway netroute.Addr) error {
 	sock, err := unix.Socket(unix.AF_ROUTE, unix.SOCK_RAW, unix.AF_UNSPEC)
 	if err != nil {

overlay/tun_netbsd.go

@@ -4,12 +4,13 @@
 package overlay
 
 import (
-	"errors"
 	"fmt"
 	"io"
 	"net/netip"
 	"os"
+	"os/exec"
 	"regexp"
+	"strconv"
 	"sync/atomic"
 	"syscall"
 	"unsafe"
@@ -19,42 +20,11 @@ import (
 	"github.com/slackhq/nebula/config"
 	"github.com/slackhq/nebula/routing"
 	"github.com/slackhq/nebula/util"
-	netroute "golang.org/x/net/route"
-	"golang.org/x/sys/unix"
 )
 
-const (
-	SIOCAIFADDR_IN6 = 0x8080696b
-	TUNSIFHEAD      = 0x80047442
-	TUNSIFMODE      = 0x80047458
-)
-
-type ifreqAlias4 struct {
-	Name     [unix.IFNAMSIZ]byte
-	Addr     unix.RawSockaddrInet4
-	DstAddr  unix.RawSockaddrInet4
-	MaskAddr unix.RawSockaddrInet4
-}
-
-type ifreqAlias6 struct {
-	Name       [unix.IFNAMSIZ]byte
-	Addr       unix.RawSockaddrInet6
-	DstAddr    unix.RawSockaddrInet6
-	PrefixMask unix.RawSockaddrInet6
-	Flags      uint32
-	Lifetime   addrLifetime
-}
-
-type ifreq struct {
-	Name [unix.IFNAMSIZ]byte
-	data int
-}
-
-type addrLifetime struct {
-	Expire    uint64
-	Preferred uint64
-	Vltime    uint32
-	Pltime    uint32
+type ifreqDestroy struct {
+	Name [16]byte
+	pad  [16]byte
 }
 
 type tun struct {
@@ -64,18 +34,40 @@ type tun struct {
 	Routes      atomic.Pointer[[]Route]
 	routeTree   atomic.Pointer[bart.Table[routing.Gateways]]
 	l           *logrus.Logger
-	f           *os.File
-	fd          int
+
+	io.ReadWriteCloser
 }
 
-var deviceNameRE = regexp.MustCompile(`^tun[0-9]+$`)
+func (t *tun) Close() error {
+	if t.ReadWriteCloser != nil {
+		if err := t.ReadWriteCloser.Close(); err != nil {
+			return err
+		}
+
+		s, err := syscall.Socket(syscall.AF_INET, syscall.SOCK_DGRAM, syscall.IPPROTO_IP)
+		if err != nil {
+			return err
+		}
+		defer syscall.Close(s)
+
+		ifreq := ifreqDestroy{Name: t.deviceBytes()}
+
+		err = ioctl(uintptr(s), syscall.SIOCIFDESTROY, uintptr(unsafe.Pointer(&ifreq)))
+
+		return err
+	}
+	return nil
+}
 
 func newTunFromFd(_ *config.C, _ *logrus.Logger, _ int, _ []netip.Prefix) (*tun, error) {
 	return nil, fmt.Errorf("newTunFromFd not supported in NetBSD")
 }
 
+var deviceNameRE = regexp.MustCompile(`^tun[0-9]+$`)
+
 func newTun(c *config.C, l *logrus.Logger, vpnNetworks []netip.Prefix, _ bool) (*tun, error) {
 	// Try to open tun device
+	var file *os.File
 	var err error
 	deviceName := c.GetString("tun.dev", "")
 	if deviceName == "" {
@@ -85,19 +77,13 @@ func newTun(c *config.C, l *logrus.Logger, vpnNetworks []netip.Prefix, _ bool) (
 		return nil, fmt.Errorf("a device name in the format of /dev/tunN must be specified")
 	}
 
-	fd, err := unix.Open("/dev/"+deviceName, os.O_RDWR, 0)
+	file, err = os.OpenFile("/dev/"+deviceName, os.O_RDWR, 0)
 	if err != nil {
 		return nil, err
 	}
 
-	err = unix.SetNonblock(fd, true)
-	if err != nil {
-		l.WithError(err).Warn("Failed to set the tun device as nonblocking")
-	}
-
 	t := &tun{
-		f:           os.NewFile(uintptr(fd), ""),
-		fd:          fd,
+		ReadWriteCloser: file,
 		Device:          deviceName,
 		vpnNetworks:     vpnNetworks,
 		MTU:             c.GetInt("tun.mtu", DefaultMTU),
@@ -119,225 +105,40 @@ func newTun(c *config.C, l *logrus.Logger, vpnNetworks []netip.Prefix, _ bool) (
 	return t, nil
 }
 
-func (t *tun) Close() error {
-	if t.f != nil {
-		if err := t.f.Close(); err != nil {
-			return fmt.Errorf("error closing tun file: %w", err)
-		}
-
-		// t.f.Close should have handled it for us but let's be extra sure
-		_ = unix.Close(t.fd)
-
-		s, err := syscall.Socket(syscall.AF_INET, syscall.SOCK_DGRAM, syscall.IPPROTO_IP)
-		if err != nil {
-			return err
-		}
-		defer syscall.Close(s)
-
-		ifr := ifreq{Name: t.deviceBytes()}
-		err = ioctl(uintptr(s), syscall.SIOCIFDESTROY, uintptr(unsafe.Pointer(&ifr)))
-		return err
-	}
-	return nil
-}
-
-func (t *tun) Read(to []byte) (int, error) {
-	rc, err := t.f.SyscallConn()
-	if err != nil {
-		return 0, fmt.Errorf("failed to get syscall conn for tun: %w", err)
-	}
-
-	var errno syscall.Errno
-	var n uintptr
-	err = rc.Read(func(fd uintptr) bool {
-		// first 4 bytes is protocol family, in network byte order
-		head := [4]byte{}
-		iovecs := []syscall.Iovec{
-			{&head[0], 4},
-			{&to[0], uint64(len(to))},
-		}
-
-		n, _, errno = syscall.Syscall(syscall.SYS_READV, fd, uintptr(unsafe.Pointer(&iovecs[0])), uintptr(2))
-		if errno.Temporary() {
-			// We got an EAGAIN, EINTR, or EWOULDBLOCK, go again
-			return false
-		}
-		return true
-	})
-	if err != nil {
-		if err == syscall.EBADF || err.Error() == "use of closed file" {
-			// Go doesn't export poll.ErrFileClosing but happily reports it to us so here we are
-			// https://github.com/golang/go/blob/master/src/internal/poll/fd_poll_runtime.go#L121
-			return 0, os.ErrClosed
-		}
-		return 0, fmt.Errorf("failed to make read call for tun: %w", err)
-	}
-
-	if errno != 0 {
-		return 0, fmt.Errorf("failed to make inner read call for tun: %w", errno)
-	}
-
-	// fix bytes read number to exclude header
-	bytesRead := int(n)
-	if bytesRead < 0 {
-		return bytesRead, nil
-	} else if bytesRead < 4 {
-		return 0, nil
-	} else {
-		return bytesRead - 4, nil
-	}
-}
-
-// Write is only valid for single threaded use
-func (t *tun) Write(from []byte) (int, error) {
-	if len(from) <= 1 {
-		return 0, syscall.EIO
-	}
-
-	ipVer := from[0] >> 4
-	var head [4]byte
-	// first 4 bytes is protocol family, in network byte order
-	if ipVer == 4 {
-		head[3] = syscall.AF_INET
-	} else if ipVer == 6 {
-		head[3] = syscall.AF_INET6
-	} else {
-		return 0, fmt.Errorf("unable to determine IP version from packet")
-	}
-
-	rc, err := t.f.SyscallConn()
-	if err != nil {
-		return 0, err
-	}
-
-	var errno syscall.Errno
-	var n uintptr
-	err = rc.Write(func(fd uintptr) bool {
-		iovecs := []syscall.Iovec{
-			{&head[0], 4},
-			{&from[0], uint64(len(from))},
-		}
-
-		n, _, errno = syscall.Syscall(syscall.SYS_WRITEV, fd, uintptr(unsafe.Pointer(&iovecs[0])), uintptr(2))
-		// According to NetBSD documentation for TUN, writes will only return errors in which
-		// this packet will never be delivered so just go on living life.
-		return true
-	})
-	if err != nil {
-		return 0, err
-	}
-
-	if errno != 0 {
-		return 0, errno
-	}
-
-	return int(n) - 4, err
-}
-
 func (t *tun) addIp(cidr netip.Prefix) error {
-	if cidr.Addr().Is4() {
-		var req ifreqAlias4
-		req.Name = t.deviceBytes()
-		req.Addr = unix.RawSockaddrInet4{
-			Len:    unix.SizeofSockaddrInet4,
-			Family: unix.AF_INET,
-			Addr:   cidr.Addr().As4(),
-		}
-		req.DstAddr = unix.RawSockaddrInet4{
-			Len:    unix.SizeofSockaddrInet4,
-			Family: unix.AF_INET,
-			Addr:   cidr.Addr().As4(),
-		}
-		req.MaskAddr = unix.RawSockaddrInet4{
-			Len:    unix.SizeofSockaddrInet4,
-			Family: unix.AF_INET,
-			Addr:   prefixToMask(cidr).As4(),
+	var err error
+
+	// TODO use syscalls instead of exec.Command
+	cmd := exec.Command("/sbin/ifconfig", t.Device, cidr.String(), cidr.Addr().String())
+	t.l.Debug("command: ", cmd.String())
+	if err = cmd.Run(); err != nil {
+		return fmt.Errorf("failed to run 'ifconfig': %s", err)
 	}
 
-		s, err := unix.Socket(unix.AF_INET, unix.SOCK_DGRAM, unix.IPPROTO_IP)
-		if err != nil {
-			return err
-		}
-		defer syscall.Close(s)
-
-		if err := ioctl(uintptr(s), unix.SIOCAIFADDR, uintptr(unsafe.Pointer(&req))); err != nil {
-			return fmt.Errorf("failed to set tun address %s: %s", cidr.Addr(), err)
+	cmd = exec.Command("/sbin/route", "-n", "add", "-net", cidr.String(), cidr.Addr().String())
+	t.l.Debug("command: ", cmd.String())
+	if err = cmd.Run(); err != nil {
+		return fmt.Errorf("failed to run 'route add': %s", err)
 	}
 
-		return nil
-	}
-
-	if cidr.Addr().Is6() {
-		var req ifreqAlias6
-		req.Name = t.deviceBytes()
-		req.Addr = unix.RawSockaddrInet6{
-			Len:    unix.SizeofSockaddrInet6,
-			Family: unix.AF_INET6,
-			Addr:   cidr.Addr().As16(),
-		}
-		req.PrefixMask = unix.RawSockaddrInet6{
-			Len:    unix.SizeofSockaddrInet6,
-			Family: unix.AF_INET6,
-			Addr:   prefixToMask(cidr).As16(),
-		}
-		req.Lifetime = addrLifetime{
-			Vltime: 0xffffffff,
-			Pltime: 0xffffffff,
-		}
-
-		s, err := unix.Socket(unix.AF_INET6, unix.SOCK_DGRAM, unix.IPPROTO_IP)
-		if err != nil {
-			return err
-		}
-		defer syscall.Close(s)
-
-		if err := ioctl(uintptr(s), SIOCAIFADDR_IN6, uintptr(unsafe.Pointer(&req))); err != nil {
-			return fmt.Errorf("failed to set tun address %s: %s", cidr.Addr().String(), err)
-		}
-		return nil
-	}
-
-	return fmt.Errorf("unknown address type %v", cidr)
-}
-
-func (t *tun) Activate() error {
-	mode := int32(unix.IFF_BROADCAST)
-	err := ioctl(uintptr(t.fd), TUNSIFMODE, uintptr(unsafe.Pointer(&mode)))
-	if err != nil {
-		return fmt.Errorf("failed to set tun device mode: %w", err)
-	}
-
-	v := 1
-	err = ioctl(uintptr(t.fd), TUNSIFHEAD, uintptr(unsafe.Pointer(&v)))
-	if err != nil {
-		return fmt.Errorf("failed to set tun device head: %w", err)
-	}
-
-	err = t.doIoctlByName(unix.SIOCSIFMTU, uint32(t.MTU))
-	if err != nil {
-		return fmt.Errorf("failed to set tun mtu: %w", err)
-	}
-
-	for i := range t.vpnNetworks {
-		err = t.addIp(t.vpnNetworks[i])
-		if err != nil {
-			return err
-		}
+	cmd = exec.Command("/sbin/ifconfig", t.Device, "mtu", strconv.Itoa(t.MTU))
+	t.l.Debug("command: ", cmd.String())
+	if err = cmd.Run(); err != nil {
+		return fmt.Errorf("failed to run 'ifconfig': %s", err)
 	}
 
+	// Unsafe path routes
 	return t.addRoutes(false)
 }
 
-func (t *tun) doIoctlByName(ctl uintptr, value uint32) error {
-	s, err := unix.Socket(unix.AF_INET, unix.SOCK_DGRAM, unix.IPPROTO_IP)
+func (t *tun) Activate() error {
+	for i := range t.vpnNetworks {
+		err := t.addIp(t.vpnNetworks[i])
 		if err != nil {
 			return err
 		}
-	defer syscall.Close(s)
-
-	ir := ifreq{Name: t.deviceBytes(), data: int(value)}
-	err = ioctl(uintptr(s), ctl, uintptr(unsafe.Pointer(&ir)))
-	return err
+	}
+	return nil
 }
 
 func (t *tun) reload(c *config.C, initial bool) error {
@@ -396,23 +197,21 @@ func (t *tun) NewMultiQueueReader() (io.ReadWriteCloser, error) {
 
 func (t *tun) addRoutes(logErrors bool) error {
 	routes := *t.Routes.Load()
-
 	for _, r := range routes {
 		if len(r.Via) == 0 || !r.Install {
 			// We don't allow route MTUs so only install routes with a via
 			continue
 		}
 
-		err := addRoute(r.Cidr, t.vpnNetworks)
-		if err != nil {
-			retErr := util.NewContextualError("Failed to add route", map[string]any{"route": r}, err)
+		cmd := exec.Command("/sbin/route", "-n", "add", "-net", r.Cidr.String(), t.vpnNetworks[0].Addr().String())
+		t.l.Debug("command: ", cmd.String())
+		if err := cmd.Run(); err != nil {
+			retErr := util.NewContextualError("failed to run 'route add' for unsafe_route", map[string]any{"route": r}, err)
 			if logErrors {
 				retErr.Log(t.l)
 			} else {
 				return retErr
 			}
-		} else {
-			t.l.WithField("route", r).Info("Added route")
 		}
 	}
 
@@ -425,8 +224,10 @@ func (t *tun) removeRoutes(routes []Route) error {
 			continue
 		}
 
-		err := delRoute(r.Cidr, t.vpnNetworks)
-		if err != nil {
+		//TODO: CERT-V2 is this right?
+		cmd := exec.Command("/sbin/route", "-n", "delete", "-net", r.Cidr.String(), t.vpnNetworks[0].Addr().String())
+		t.l.Debug("command: ", cmd.String())
+		if err := cmd.Run(); err != nil {
 			t.l.WithError(err).WithField("route", r).Error("Failed to remove route")
 		} else {
 			t.l.WithField("route", r).Info("Removed route")
@@ -441,109 +242,3 @@ func (t *tun) deviceBytes() (o [16]byte) {
 	}
 	return
 }
-
-func addRoute(prefix netip.Prefix, gateways []netip.Prefix) error {
-	sock, err := unix.Socket(unix.AF_ROUTE, unix.SOCK_RAW, unix.AF_UNSPEC)
-	if err != nil {
-		return fmt.Errorf("unable to create AF_ROUTE socket: %v", err)
-	}
-	defer unix.Close(sock)
-
-	route := &netroute.RouteMessage{
-		Version: unix.RTM_VERSION,
-		Type:    unix.RTM_ADD,
-		Flags:   unix.RTF_UP | unix.RTF_GATEWAY,
-		Seq:     1,
-	}
-
-	if prefix.Addr().Is4() {
-		gw, err := selectGateway(prefix, gateways)
-		if err != nil {
-			return err
-		}
-		route.Addrs = []netroute.Addr{
-			unix.RTAX_DST:     &netroute.Inet4Addr{IP: prefix.Masked().Addr().As4()},
-			unix.RTAX_NETMASK: &netroute.Inet4Addr{IP: prefixToMask(prefix).As4()},
-			unix.RTAX_GATEWAY: &netroute.Inet4Addr{IP: gw.Addr().As4()},
-		}
-	} else {
-		gw, err := selectGateway(prefix, gateways)
-		if err != nil {
-			return err
-		}
-		route.Addrs = []netroute.Addr{
-			unix.RTAX_DST:     &netroute.Inet6Addr{IP: prefix.Masked().Addr().As16()},
-			unix.RTAX_NETMASK: &netroute.Inet6Addr{IP: prefixToMask(prefix).As16()},
-			unix.RTAX_GATEWAY: &netroute.Inet6Addr{IP: gw.Addr().As16()},
-		}
-	}
-
-	data, err := route.Marshal()
-	if err != nil {
-		return fmt.Errorf("failed to create route.RouteMessage: %w", err)
-	}
-
-	_, err = unix.Write(sock, data[:])
-	if err != nil {
-		if errors.Is(err, unix.EEXIST) {
-			// Try to do a change
-			route.Type = unix.RTM_CHANGE
-			data, err = route.Marshal()
-			if err != nil {
-				return fmt.Errorf("failed to create route.RouteMessage for change: %w", err)
-			}
-			_, err = unix.Write(sock, data[:])
-			return err
-		}
-		return fmt.Errorf("failed to write route.RouteMessage to socket: %w", err)
-	}
-
-	return nil
-}
-
-func delRoute(prefix netip.Prefix, gateways []netip.Prefix) error {
-	sock, err := unix.Socket(unix.AF_ROUTE, unix.SOCK_RAW, unix.AF_UNSPEC)
-	if err != nil {
-		return fmt.Errorf("unable to create AF_ROUTE socket: %v", err)
-	}
-	defer unix.Close(sock)
-
-	route := netroute.RouteMessage{
-		Version: unix.RTM_VERSION,
-		Type:    unix.RTM_DELETE,
-		Seq:     1,
-	}
-
-	if prefix.Addr().Is4() {
-		gw, err := selectGateway(prefix, gateways)
-		if err != nil {
-			return err
-		}
-		route.Addrs = []netroute.Addr{
-			unix.RTAX_DST:     &netroute.Inet4Addr{IP: prefix.Masked().Addr().As4()},
-			unix.RTAX_NETMASK: &netroute.Inet4Addr{IP: prefixToMask(prefix).As4()},
-			unix.RTAX_GATEWAY: &netroute.Inet4Addr{IP: gw.Addr().As4()},
-		}
-	} else {
-		gw, err := selectGateway(prefix, gateways)
-		if err != nil {
-			return err
-		}
-		route.Addrs = []netroute.Addr{
-			unix.RTAX_DST:     &netroute.Inet6Addr{IP: prefix.Masked().Addr().As16()},
-			unix.RTAX_NETMASK: &netroute.Inet6Addr{IP: prefixToMask(prefix).As16()},
-			unix.RTAX_GATEWAY: &netroute.Inet6Addr{IP: gw.Addr().As16()},
-		}
-	}
-
-	data, err := route.Marshal()
-	if err != nil {
-		return fmt.Errorf("failed to create route.RouteMessage: %w", err)
-	}
-	_, err = unix.Write(sock, data[:])
-	if err != nil {
-		return fmt.Errorf("failed to write route.RouteMessage to socket: %w", err)
-	}
-
-	return nil
-}

overlay/tun_openbsd.go

@@ -4,50 +4,23 @@
 package overlay
 
 import (
-	"errors"
 	"fmt"
 	"io"
 	"net/netip"
 	"os"
+	"os/exec"
 	"regexp"
+	"strconv"
 	"sync/atomic"
 	"syscall"
-	"unsafe"
 
 	"github.com/gaissmai/bart"
 	"github.com/sirupsen/logrus"
 	"github.com/slackhq/nebula/config"
 	"github.com/slackhq/nebula/routing"
 	"github.com/slackhq/nebula/util"
-	netroute "golang.org/x/net/route"
-	"golang.org/x/sys/unix"
 )
 
-const (
-	SIOCAIFADDR_IN6 = 0x8080691a
-)
-
-type ifreqAlias4 struct {
-	Name     [unix.IFNAMSIZ]byte
-	Addr     unix.RawSockaddrInet4
-	DstAddr  unix.RawSockaddrInet4
-	MaskAddr unix.RawSockaddrInet4
-}
-
-type ifreqAlias6 struct {
-	Name       [unix.IFNAMSIZ]byte
-	Addr       unix.RawSockaddrInet6
-	DstAddr    unix.RawSockaddrInet6
-	PrefixMask unix.RawSockaddrInet6
-	Flags      uint32
-	Lifetime   [2]uint32
-}
-
-type ifreq struct {
-	Name [unix.IFNAMSIZ]byte
-	data int
-}
-
 type tun struct {
 	Device      string
 	vpnNetworks []netip.Prefix
@@ -55,42 +28,44 @@ type tun struct {
 	Routes      atomic.Pointer[[]Route]
 	routeTree   atomic.Pointer[bart.Table[routing.Gateways]]
 	l           *logrus.Logger
-	f           *os.File
-	fd          int
+
+	io.ReadWriteCloser
+
 	// cache out buffer since we need to prepend 4 bytes for tun metadata
 	out []byte
 }
 
-var deviceNameRE = regexp.MustCompile(`^tun[0-9]+$`)
+func (t *tun) Close() error {
+	if t.ReadWriteCloser != nil {
+		return t.ReadWriteCloser.Close()
+	}
 
-func newTunFromFd(_ *config.C, _ *logrus.Logger, _ int, _ []netip.Prefix) (*tun, error) {
-	return nil, fmt.Errorf("newTunFromFd not supported in openbsd")
+	return nil
 }
 
+func newTunFromFd(_ *config.C, _ *logrus.Logger, _ int, _ []netip.Prefix) (*tun, error) {
+	return nil, fmt.Errorf("newTunFromFd not supported in OpenBSD")
+}
+
+var deviceNameRE = regexp.MustCompile(`^tun[0-9]+$`)
+
 func newTun(c *config.C, l *logrus.Logger, vpnNetworks []netip.Prefix, _ bool) (*tun, error) {
-	// Try to open tun device
-	var err error
 	deviceName := c.GetString("tun.dev", "")
 	if deviceName == "" {
-		return nil, fmt.Errorf("a device name in the format of /dev/tunN must be specified")
-	}
-	if !deviceNameRE.MatchString(deviceName) {
-		return nil, fmt.Errorf("a device name in the format of /dev/tunN must be specified")
+		return nil, fmt.Errorf("a device name in the format of tunN must be specified")
 	}
 
-	fd, err := unix.Open("/dev/"+deviceName, os.O_RDWR, 0)
+	if !deviceNameRE.MatchString(deviceName) {
+		return nil, fmt.Errorf("a device name in the format of tunN must be specified")
+	}
+
+	file, err := os.OpenFile("/dev/"+deviceName, os.O_RDWR, 0)
 	if err != nil {
 		return nil, err
 	}
 
-	err = unix.SetNonblock(fd, true)
-	if err != nil {
-		l.WithError(err).Warn("Failed to set the tun device as nonblocking")
-	}
-
 	t := &tun{
-		f:           os.NewFile(uintptr(fd), ""),
-		fd:          fd,
+		ReadWriteCloser: file,
 		Device:          deviceName,
 		vpnNetworks:     vpnNetworks,
 		MTU:             c.GetInt("tun.mtu", DefaultMTU),
@@ -112,154 +87,6 @@ func newTun(c *config.C, l *logrus.Logger, vpnNetworks []netip.Prefix, _ bool) (
 	return t, nil
 }
 
-func (t *tun) Close() error {
-	if t.f != nil {
-		if err := t.f.Close(); err != nil {
-			return fmt.Errorf("error closing tun file: %w", err)
-		}
-
-		// t.f.Close should have handled it for us but let's be extra sure
-		_ = unix.Close(t.fd)
-	}
-	return nil
-}
-
-func (t *tun) Read(to []byte) (int, error) {
-	buf := make([]byte, len(to)+4)
-
-	n, err := t.f.Read(buf)
-
-	copy(to, buf[4:])
-	return n - 4, err
-}
-
-// Write is only valid for single threaded use
-func (t *tun) Write(from []byte) (int, error) {
-	buf := t.out
-	if cap(buf) < len(from)+4 {
-		buf = make([]byte, len(from)+4)
-		t.out = buf
-	}
-	buf = buf[:len(from)+4]
-
-	if len(from) == 0 {
-		return 0, syscall.EIO
-	}
-
-	// Determine the IP Family for the NULL L2 Header
-	ipVer := from[0] >> 4
-	if ipVer == 4 {
-		buf[3] = syscall.AF_INET
-	} else if ipVer == 6 {
-		buf[3] = syscall.AF_INET6
-	} else {
-		return 0, fmt.Errorf("unable to determine IP version from packet")
-	}
-
-	copy(buf[4:], from)
-
-	n, err := t.f.Write(buf)
-	return n - 4, err
-}
-
-func (t *tun) addIp(cidr netip.Prefix) error {
-	if cidr.Addr().Is4() {
-		var req ifreqAlias4
-		req.Name = t.deviceBytes()
-		req.Addr = unix.RawSockaddrInet4{
-			Len:    unix.SizeofSockaddrInet4,
-			Family: unix.AF_INET,
-			Addr:   cidr.Addr().As4(),
-		}
-		req.DstAddr = unix.RawSockaddrInet4{
-			Len:    unix.SizeofSockaddrInet4,
-			Family: unix.AF_INET,
-			Addr:   cidr.Addr().As4(),
-		}
-		req.MaskAddr = unix.RawSockaddrInet4{
-			Len:    unix.SizeofSockaddrInet4,
-			Family: unix.AF_INET,
-			Addr:   prefixToMask(cidr).As4(),
-		}
-
-		s, err := unix.Socket(unix.AF_INET, unix.SOCK_DGRAM, unix.IPPROTO_IP)
-		if err != nil {
-			return err
-		}
-		defer syscall.Close(s)
-
-		if err := ioctl(uintptr(s), unix.SIOCAIFADDR, uintptr(unsafe.Pointer(&req))); err != nil {
-			return fmt.Errorf("failed to set tun address %s: %s", cidr.Addr(), err)
-		}
-
-		err = addRoute(cidr, t.vpnNetworks)
-		if err != nil {
-			return fmt.Errorf("failed to set route for vpn network %v: %w", cidr, err)
-		}
-
-		return nil
-	}
-
-	if cidr.Addr().Is6() {
-		var req ifreqAlias6
-		req.Name = t.deviceBytes()
-		req.Addr = unix.RawSockaddrInet6{
-			Len:    unix.SizeofSockaddrInet6,
-			Family: unix.AF_INET6,
-			Addr:   cidr.Addr().As16(),
-		}
-		req.PrefixMask = unix.RawSockaddrInet6{
-			Len:    unix.SizeofSockaddrInet6,
-			Family: unix.AF_INET6,
-			Addr:   prefixToMask(cidr).As16(),
-		}
-		req.Lifetime[0] = 0xffffffff
-		req.Lifetime[1] = 0xffffffff
-
-		s, err := unix.Socket(unix.AF_INET6, unix.SOCK_DGRAM, unix.IPPROTO_IP)
-		if err != nil {
-			return err
-		}
-		defer syscall.Close(s)
-
-		if err := ioctl(uintptr(s), SIOCAIFADDR_IN6, uintptr(unsafe.Pointer(&req))); err != nil {
-			return fmt.Errorf("failed to set tun address %s: %s", cidr.Addr().String(), err)
-		}
-
-		return nil
-	}
-
-	return fmt.Errorf("unknown address type %v", cidr)
-}
-
-func (t *tun) Activate() error {
-	err := t.doIoctlByName(unix.SIOCSIFMTU, uint32(t.MTU))
-	if err != nil {
-		return fmt.Errorf("failed to set tun mtu: %w", err)
-	}
-
-	for i := range t.vpnNetworks {
-		err = t.addIp(t.vpnNetworks[i])
-		if err != nil {
-			return err
-		}
-	}
-
-	return t.addRoutes(false)
-}
-
-func (t *tun) doIoctlByName(ctl uintptr, value uint32) error {
-	s, err := unix.Socket(unix.AF_INET, unix.SOCK_DGRAM, unix.IPPROTO_IP)
-	if err != nil {
-		return err
-	}
-	defer syscall.Close(s)
-
-	ir := ifreq{Name: t.deviceBytes(), data: int(value)}
-	err = ioctl(uintptr(s), ctl, uintptr(unsafe.Pointer(&ir)))
-	return err
-}
-
 func (t *tun) reload(c *config.C, initial bool) error {
 	change, routes, err := getAllRoutesFromConfig(c, t.vpnNetworks, initial)
 	if err != nil {
@@ -297,42 +124,63 @@ func (t *tun) reload(c *config.C, initial bool) error {
 	return nil
 }
 
+func (t *tun) addIp(cidr netip.Prefix) error {
+	var err error
+	// TODO use syscalls instead of exec.Command
+	cmd := exec.Command("/sbin/ifconfig", t.Device, cidr.String(), cidr.Addr().String())
+	t.l.Debug("command: ", cmd.String())
+	if err = cmd.Run(); err != nil {
+		return fmt.Errorf("failed to run 'ifconfig': %s", err)
+	}
+
+	cmd = exec.Command("/sbin/ifconfig", t.Device, "mtu", strconv.Itoa(t.MTU))
+	t.l.Debug("command: ", cmd.String())
+	if err = cmd.Run(); err != nil {
+		return fmt.Errorf("failed to run 'ifconfig': %s", err)
+	}
+
+	cmd = exec.Command("/sbin/route", "-n", "add", "-inet", cidr.String(), cidr.Addr().String())
+	t.l.Debug("command: ", cmd.String())
+	if err = cmd.Run(); err != nil {
+		return fmt.Errorf("failed to run 'route add': %s", err)
+	}
+
+	// Unsafe path routes
+	return t.addRoutes(false)
+}
+
+func (t *tun) Activate() error {
+	for i := range t.vpnNetworks {
+		err := t.addIp(t.vpnNetworks[i])
+		if err != nil {
+			return err
+		}
+	}
+	return nil
+}
+
 func (t *tun) RoutesFor(ip netip.Addr) routing.Gateways {
 	r, _ := t.routeTree.Load().Lookup(ip)
 	return r
 }
 
-func (t *tun) Networks() []netip.Prefix {
-	return t.vpnNetworks
-}
-
-func (t *tun) Name() string {
-	return t.Device
-}
-
-func (t *tun) NewMultiQueueReader() (io.ReadWriteCloser, error) {
-	return nil, fmt.Errorf("TODO: multiqueue not implemented for openbsd")
-}
-
 func (t *tun) addRoutes(logErrors bool) error {
 	routes := *t.Routes.Load()
-
 	for _, r := range routes {
 		if len(r.Via) == 0 || !r.Install {
 			// We don't allow route MTUs so only install routes with a via
 			continue
 		}
-
-		err := addRoute(r.Cidr, t.vpnNetworks)
-		if err != nil {
-			retErr := util.NewContextualError("Failed to add route", map[string]any{"route": r}, err)
+		//TODO: CERT-V2 is this right?
+		cmd := exec.Command("/sbin/route", "-n", "add", "-inet", r.Cidr.String(), t.vpnNetworks[0].Addr().String())
+		t.l.Debug("command: ", cmd.String())
+		if err := cmd.Run(); err != nil {
+			retErr := util.NewContextualError("failed to run 'route add' for unsafe_route", map[string]any{"route": r}, err)
 			if logErrors {
 				retErr.Log(t.l)
 			} else {
 				return retErr
 			}
-		} else {
-			t.l.WithField("route", r).Info("Added route")
 		}
 	}
 
@@ -344,9 +192,10 @@ func (t *tun) removeRoutes(routes []Route) error {
 		if !r.Install {
 			continue
 		}
-
-		err := delRoute(r.Cidr, t.vpnNetworks)
-		if err != nil {
+		//TODO: CERT-V2 is this right?
+		cmd := exec.Command("/sbin/route", "-n", "delete", "-inet", r.Cidr.String(), t.vpnNetworks[0].Addr().String())
+		t.l.Debug("command: ", cmd.String())
+		if err := cmd.Run(); err != nil {
 			t.l.WithError(err).WithField("route", r).Error("Failed to remove route")
 		} else {
 			t.l.WithField("route", r).Info("Removed route")
@@ -355,115 +204,52 @@ func (t *tun) removeRoutes(routes []Route) error {
 	return nil
 }
 
-func (t *tun) deviceBytes() (o [16]byte) {
-	for i, c := range t.Device {
-		o[i] = byte(c)
-	}
-	return
+func (t *tun) Networks() []netip.Prefix {
+	return t.vpnNetworks
 }
 
-func addRoute(prefix netip.Prefix, gateways []netip.Prefix) error {
-	sock, err := unix.Socket(unix.AF_ROUTE, unix.SOCK_RAW, unix.AF_UNSPEC)
-	if err != nil {
-		return fmt.Errorf("unable to create AF_ROUTE socket: %v", err)
-	}
-	defer unix.Close(sock)
+func (t *tun) Name() string {
+	return t.Device
+}
 
-	route := &netroute.RouteMessage{
-		Version: unix.RTM_VERSION,
-		Type:    unix.RTM_ADD,
-		Flags:   unix.RTF_UP | unix.RTF_GATEWAY,
-		Seq:     1,
+func (t *tun) NewMultiQueueReader() (io.ReadWriteCloser, error) {
+	return nil, fmt.Errorf("TODO: multiqueue not implemented for freebsd")
+}
+
+func (t *tun) Read(to []byte) (int, error) {
+	buf := make([]byte, len(to)+4)
+
+	n, err := t.ReadWriteCloser.Read(buf)
+
+	copy(to, buf[4:])
+	return n - 4, err
+}
+
+// Write is only valid for single threaded use
+func (t *tun) Write(from []byte) (int, error) {
+	buf := t.out
+	if cap(buf) < len(from)+4 {
+		buf = make([]byte, len(from)+4)
+		t.out = buf
+	}
+	buf = buf[:len(from)+4]
+
+	if len(from) == 0 {
+		return 0, syscall.EIO
 	}
 
-	if prefix.Addr().Is4() {
-		gw, err := selectGateway(prefix, gateways)
-		if err != nil {
-			return err
-		}
-		route.Addrs = []netroute.Addr{
-			unix.RTAX_DST:     &netroute.Inet4Addr{IP: prefix.Masked().Addr().As4()},
-			unix.RTAX_NETMASK: &netroute.Inet4Addr{IP: prefixToMask(prefix).As4()},
-			unix.RTAX_GATEWAY: &netroute.Inet4Addr{IP: gw.Addr().As4()},
-		}
+	// Determine the IP Family for the NULL L2 Header
+	ipVer := from[0] >> 4
+	if ipVer == 4 {
+		buf[3] = syscall.AF_INET
+	} else if ipVer == 6 {
+		buf[3] = syscall.AF_INET6
 	} else {
-		gw, err := selectGateway(prefix, gateways)
-		if err != nil {
-			return err
-		}
-		route.Addrs = []netroute.Addr{
-			unix.RTAX_DST:     &netroute.Inet6Addr{IP: prefix.Masked().Addr().As16()},
-			unix.RTAX_NETMASK: &netroute.Inet6Addr{IP: prefixToMask(prefix).As16()},
-			unix.RTAX_GATEWAY: &netroute.Inet6Addr{IP: gw.Addr().As16()},
-		}
+		return 0, fmt.Errorf("unable to determine IP version from packet")
 	}
 
-	data, err := route.Marshal()
-	if err != nil {
-		return fmt.Errorf("failed to create route.RouteMessage: %w", err)
-	}
+	copy(buf[4:], from)
 
-	_, err = unix.Write(sock, data[:])
-	if err != nil {
-		if errors.Is(err, unix.EEXIST) {
-			// Try to do a change
-			route.Type = unix.RTM_CHANGE
-			data, err = route.Marshal()
-			if err != nil {
-				return fmt.Errorf("failed to create route.RouteMessage for change: %w", err)
-			}
-			_, err = unix.Write(sock, data[:])
-			return err
-		}
-		return fmt.Errorf("failed to write route.RouteMessage to socket: %w", err)
-	}
-
-	return nil
-}
-
-func delRoute(prefix netip.Prefix, gateways []netip.Prefix) error {
-	sock, err := unix.Socket(unix.AF_ROUTE, unix.SOCK_RAW, unix.AF_UNSPEC)
-	if err != nil {
-		return fmt.Errorf("unable to create AF_ROUTE socket: %v", err)
-	}
-	defer unix.Close(sock)
-
-	route := netroute.RouteMessage{
-		Version: unix.RTM_VERSION,
-		Type:    unix.RTM_DELETE,
-		Seq:     1,
-	}
-
-	if prefix.Addr().Is4() {
-		gw, err := selectGateway(prefix, gateways)
-		if err != nil {
-			return err
-		}
-		route.Addrs = []netroute.Addr{
-			unix.RTAX_DST:     &netroute.Inet4Addr{IP: prefix.Masked().Addr().As4()},
-			unix.RTAX_NETMASK: &netroute.Inet4Addr{IP: prefixToMask(prefix).As4()},
-			unix.RTAX_GATEWAY: &netroute.Inet4Addr{IP: gw.Addr().As4()},
-		}
-	} else {
-		gw, err := selectGateway(prefix, gateways)
-		if err != nil {
-			return err
-		}
-		route.Addrs = []netroute.Addr{
-			unix.RTAX_DST:     &netroute.Inet6Addr{IP: prefix.Masked().Addr().As16()},
-			unix.RTAX_NETMASK: &netroute.Inet6Addr{IP: prefixToMask(prefix).As16()},
-			unix.RTAX_GATEWAY: &netroute.Inet6Addr{IP: gw.Addr().As16()},
-		}
-	}
-
-	data, err := route.Marshal()
-	if err != nil {
-		return fmt.Errorf("failed to create route.RouteMessage: %w", err)
-	}
-	_, err = unix.Write(sock, data[:])
-	if err != nil {
-		return fmt.Errorf("failed to write route.RouteMessage to socket: %w", err)
-	}
-
-	return nil
+	n, err := t.ReadWriteCloser.Write(buf)
+	return n - 4, err
 }

packet/packet.go

@@ -0,0 +1,12 @@
+package packet
+
+import "net/netip"
+
+type Packet struct {
+	Payload []byte
+	Addr    netip.AddrPort
+}
+
+func New() *Packet {
+	return &Packet{Payload: make([]byte, 9001)}
+}

pki.go

@@ -100,36 +100,41 @@ func (p *PKI) reloadCerts(c *config.C, initial bool) *util.ContextualError {
 		currentState := p.cs.Load()
 		if newState.v1Cert != nil {
 			if currentState.v1Cert == nil {
-				//adding certs is fine, actually. Networks-in-common confirmed in newCertState().
-			} else {
+				return util.NewContextualError("v1 certificate was added, restart required", nil, err)
+			}
+
 			// did IP in cert change? if so, don't set
 			if !slices.Equal(currentState.v1Cert.Networks(), newState.v1Cert.Networks()) {
 				return util.NewContextualError(
 					"Networks in new cert was different from old",
-						m{"new_networks": newState.v1Cert.Networks(), "old_networks": currentState.v1Cert.Networks(), "cert_version": cert.Version1},
+					m{"new_networks": newState.v1Cert.Networks(), "old_networks": currentState.v1Cert.Networks()},
 					nil,
 				)
 			}
 
 			if currentState.v1Cert.Curve() != newState.v1Cert.Curve() {
 				return util.NewContextualError(
-						"Curve in new v1 cert was different from old",
-						m{"new_curve": newState.v1Cert.Curve(), "old_curve": currentState.v1Cert.Curve(), "cert_version": cert.Version1},
+					"Curve in new cert was different from old",
+					m{"new_curve": newState.v1Cert.Curve(), "old_curve": currentState.v1Cert.Curve()},
 					nil,
 				)
 			}
-			}
+
+		} else if currentState.v1Cert != nil {
+			//TODO: CERT-V2 we should be able to tear this down
+			return util.NewContextualError("v1 certificate was removed, restart required", nil, err)
 		}
 
 		if newState.v2Cert != nil {
 			if currentState.v2Cert == nil {
-				//adding certs is fine, actually
-			} else {
+				return util.NewContextualError("v2 certificate was added, restart required", nil, err)
+			}
+
 			// did IP in cert change? if so, don't set
 			if !slices.Equal(currentState.v2Cert.Networks(), newState.v2Cert.Networks()) {
 				return util.NewContextualError(
 					"Networks in new cert was different from old",
-						m{"new_networks": newState.v2Cert.Networks(), "old_networks": currentState.v2Cert.Networks(), "cert_version": cert.Version2},
+					m{"new_networks": newState.v2Cert.Networks(), "old_networks": currentState.v2Cert.Networks()},
 					nil,
 				)
 			}
@@ -137,25 +142,13 @@ func (p *PKI) reloadCerts(c *config.C, initial bool) *util.ContextualError {
 			if currentState.v2Cert.Curve() != newState.v2Cert.Curve() {
 				return util.NewContextualError(
 					"Curve in new cert was different from old",
-						m{"new_curve": newState.v2Cert.Curve(), "old_curve": currentState.v2Cert.Curve(), "cert_version": cert.Version2},
+					m{"new_curve": newState.v2Cert.Curve(), "old_curve": currentState.v2Cert.Curve()},
 					nil,
 				)
 			}
-			}
 
 		} else if currentState.v2Cert != nil {
-			//newState.v1Cert is non-nil bc empty certstates aren't permitted
-			if newState.v1Cert == nil {
-				return util.NewContextualError("v1 and v2 certs are nil, this should be impossible", nil, err)
-			}
-			//if we're going to v1-only, we need to make sure we didn't orphan any v2-cert vpnaddrs
-			if !slices.Equal(currentState.v2Cert.Networks(), newState.v1Cert.Networks()) {
-				return util.NewContextualError(
-					"Removing a V2 cert is not permitted unless it has identical networks to the new V1 cert",
-					m{"new_v1_networks": newState.v1Cert.Networks(), "old_v2_networks": currentState.v2Cert.Networks()},
-					nil,
-				)
-			}
+			return util.NewContextualError("v2 certificate was removed, restart required", nil, err)
 		}
 
 		// Cipher cant be hot swapped so just leave it at what it was before
@@ -523,14 +516,10 @@ func loadCAPoolFromConfig(l *logrus.Logger, c *config.C) (*cert.CAPool, error) {
 		return nil, fmt.Errorf("error while adding CA certificate to CA trust store: %s", err)
 	}
 
-	bl := c.GetStringSlice("pki.blocklist", []string{})
-	if len(bl) > 0 {
-		for _, fp := range bl {
+	for _, fp := range c.GetStringSlice("pki.blocklist", []string{}) {
+		l.WithField("fingerprint", fp).Info("Blocklisting cert")
 		caPool.BlocklistFingerprint(fp)
 	}
 
-		l.WithField("fingerprintCount", len(bl)).Info("Blocklisted certificates")
-	}
-
 	return caPool, nil
 }

service/service.go

@@ -44,7 +44,10 @@ type Service struct {
 }
 
 func New(control *nebula.Control) (*Service, error) {
-	control.Start()
+	wait, err := control.Start()
+	if err != nil {
+		return nil, err
+	}
 
 	ctx := control.Context()
 	eg, ctx := errgroup.WithContext(ctx)
@@ -141,6 +144,12 @@ func New(control *nebula.Control) (*Service, error) {
 		}
 	})
 
+	// Add the nebula wait function to the group
+	eg.Go(func() error {
+		wait()
+		return nil
+	})
+
 	return &s, nil
 }
 

service/service_test.go

@@ -16,8 +16,8 @@ import (
 	"github.com/slackhq/nebula/cert_test"
 	"github.com/slackhq/nebula/config"
 	"github.com/slackhq/nebula/overlay"
-	"go.yaml.in/yaml/v3"
 	"golang.org/x/sync/errgroup"
+	"gopkg.in/yaml.v3"
 )
 
 type m = map[string]any

udp/conn.go

@@ -16,7 +16,7 @@ type EncReader func(
 type Conn interface {
 	Rebind() error
 	LocalAddr() (netip.AddrPort, error)
-	ListenOut(r EncReader)
+	ListenOut(r EncReader) error
 	WriteTo(b []byte, addr netip.AddrPort) error
 	ReloadConfig(c *config.C)
 	Close() error
@@ -30,8 +30,8 @@ func (NoopConn) Rebind() error {
 func (NoopConn) LocalAddr() (netip.AddrPort, error) {
 	return netip.AddrPort{}, nil
 }
-func (NoopConn) ListenOut(_ EncReader) {
-	return
+func (NoopConn) ListenOut(_ EncReader) error {
+	return nil
 }
 func (NoopConn) WriteTo(_ []byte, _ netip.AddrPort) error {
 	return nil

udp/udp_darwin.go

@@ -165,7 +165,7 @@ func NewUDPStatsEmitter(udpConns []Conn) func() {
 	return func() {}
 }
 
-func (u *StdConn) ListenOut(r EncReader) {
+func (u *StdConn) ListenOut(r EncReader) error {
 	buffer := make([]byte, MTU)
 
 	for {
@@ -174,14 +174,17 @@ func (u *StdConn) ListenOut(r EncReader) {
 		if err != nil {
 			if errors.Is(err, net.ErrClosed) {
 				u.l.WithError(err).Debug("udp socket is closed, exiting read loop")
-				return
+				return err
 			}
 
 			u.l.WithError(err).Error("unexpected udp socket receive error")
+			continue
 		}
 
 		r(netip.AddrPortFrom(rua.Addr().Unmap(), rua.Port()), buffer[:n])
 	}
+
+	return nil
 }
 
 func (u *StdConn) Rebind() error {

udp/udp_generic.go

@@ -71,15 +71,14 @@ type rawMessage struct {
 	Len uint32
 }
 
-func (u *GenericConn) ListenOut(r EncReader) {
+func (u *GenericConn) ListenOut(r EncReader) error {
 	buffer := make([]byte, MTU)
 
 	for {
 		// Just read one packet at a time
 		n, rua, err := u.ReadFromUDPAddrPort(buffer)
 		if err != nil {
-			u.l.WithError(err).Debug("udp socket is closed, exiting read loop")
-			return
+			return err
 		}
 
 		r(netip.AddrPortFrom(rua.Addr().Unmap(), rua.Port()), buffer[:n])

udp/udp_linux.go

@@ -5,10 +5,13 @@ package udp
 
 import (
 	"encoding/binary"
+	"errors"
 	"fmt"
 	"net"
 	"net/netip"
+	"sync"
 	"syscall"
+	"time"
 	"unsafe"
 
 	"github.com/rcrowley/go-metrics"
@@ -17,19 +20,40 @@ import (
 	"golang.org/x/sys/unix"
 )
 
+var readTimeout = unix.NsecToTimeval(int64(time.Millisecond * 500))
+
+const (
+	defaultGSOMaxSegments    = 8
+	defaultGSOFlushTimeout   = 150 * time.Microsecond
+	defaultGROReadBufferSize = MTU * defaultGSOMaxSegments
+	maxGSOBatchBytes         = 0xFFFF
+)
+
+var (
+	errGSOFallback = errors.New("udp gso fallback")
+	errGSODisabled = errors.New("udp gso disabled")
+)
+
 type StdConn struct {
 	sysFd int
 	isV4  bool
 	l     *logrus.Logger
 	batch int
-}
 
-func maybeIPV4(ip net.IP) (net.IP, bool) {
-	ip4 := ip.To4()
-	if ip4 != nil {
-		return ip4, true
-	}
-	return ip, false
+	enableGRO bool
+	enableGSO bool
+
+	gsoMu           sync.Mutex
+	gsoBuf          []byte
+	gsoAddr         netip.AddrPort
+	gsoSegSize      int
+	gsoSegments     int
+	gsoMaxSegments  int
+	gsoMaxBytes     int
+	gsoFlushTimeout time.Duration
+	gsoTimer        *time.Timer
+
+	groBufSize int
 }
 
 func NewListener(l *logrus.Logger, ip netip.Addr, port int, multi bool, batch int) (Conn, error) {
@@ -55,6 +79,11 @@ func NewListener(l *logrus.Logger, ip netip.Addr, port int, multi bool, batch in
 		}
 	}
 
+	// Set a read timeout
+	if err = unix.SetsockoptTimeval(fd, unix.SOL_SOCKET, unix.SO_RCVTIMEO, &readTimeout); err != nil {
+		return nil, fmt.Errorf("unable to set SO_RCVTIMEO: %s", err)
+	}
+
 	var sa unix.Sockaddr
 	if ip.Is4() {
 		sa4 := &unix.SockaddrInet4{Port: port}
@@ -69,7 +98,16 @@ func NewListener(l *logrus.Logger, ip netip.Addr, port int, multi bool, batch in
 		return nil, fmt.Errorf("unable to bind to socket: %s", err)
 	}
 
-	return &StdConn{sysFd: fd, isV4: ip.Is4(), l: l, batch: batch}, err
+	return &StdConn{
+		sysFd:           fd,
+		isV4:            ip.Is4(),
+		l:               l,
+		batch:           batch,
+		gsoMaxSegments:  defaultGSOMaxSegments,
+		gsoMaxBytes:     MTU * defaultGSOMaxSegments,
+		gsoFlushTimeout: defaultGSOFlushTimeout,
+		groBufSize:      MTU,
+	}, err
 }
 
 func (u *StdConn) Rebind() error {
@@ -118,20 +156,46 @@ func (u *StdConn) LocalAddr() (netip.AddrPort, error) {
 	}
 }
 
-func (u *StdConn) ListenOut(r EncReader) {
-	var ip netip.Addr
+func (u *StdConn) ListenOut(r EncReader) error {
+	var (
+		ip       netip.Addr
+		controls [][]byte
+	)
 
-	msgs, buffers, names := u.PrepareRawMessages(u.batch)
+	bufSize := u.readBufferSize()
+	msgs, buffers, names := u.PrepareRawMessages(u.batch, bufSize)
 	read := u.ReadMulti
 	if u.batch == 1 {
 		read = u.ReadSingle
 	}
 
 	for {
+		desired := u.readBufferSize()
+		if len(buffers) == 0 || cap(buffers[0]) < desired {
+			msgs, buffers, names = u.PrepareRawMessages(u.batch, desired)
+			controls = nil
+		}
+
+		if u.enableGRO {
+			if controls == nil {
+				controls = make([][]byte, len(msgs))
+				for i := range controls {
+					controls[i] = make([]byte, unix.CmsgSpace(4))
+				}
+			}
+			for i := range msgs {
+				setRawMessageControl(&msgs[i], controls[i])
+			}
+		} else if controls != nil {
+			for i := range msgs {
+				setRawMessageControl(&msgs[i], nil)
+			}
+			controls = nil
+		}
+
 		n, err := read(msgs)
 		if err != nil {
-			u.l.WithError(err).Debug("udp socket is closed, exiting read loop")
-			return
+			return err
 		}
 
 		for i := 0; i < n; i++ {
@@ -141,9 +205,80 @@ func (u *StdConn) ListenOut(r EncReader) {
 			} else {
 				ip, _ = netip.AddrFromSlice(names[i][8:24])
 			}
-			r(netip.AddrPortFrom(ip.Unmap(), binary.BigEndian.Uint16(names[i][2:4])), buffers[i][:msgs[i].Len])
+			addr := netip.AddrPortFrom(ip.Unmap(), binary.BigEndian.Uint16(names[i][2:4]))
+			payload := buffers[i][:msgs[i].Len]
+
+			if u.enableGRO && u.l.IsLevelEnabled(logrus.DebugLevel) {
+				ctrlLen := getRawMessageControlLen(&msgs[i])
+				msgFlags := getRawMessageFlags(&msgs[i])
+				u.l.WithFields(logrus.Fields{
+					"tag":         "gro-debug",
+					"stage":       "recv",
+					"payload_len": len(payload),
+					"ctrl_len":    ctrlLen,
+					"msg_flags":   msgFlags,
+				}).Debug("gro batch data")
+				if controls != nil && ctrlLen > 0 {
+					maxDump := ctrlLen
+					if maxDump > 16 {
+						maxDump = 16
+					}
+					u.l.WithFields(logrus.Fields{
+						"tag":         "gro-debug",
+						"stage":       "control-bytes",
+						"control_hex": fmt.Sprintf("%x", controls[i][:maxDump]),
+						"datalen":     ctrlLen,
+					}).Debug("gro control dump")
 				}
 			}
+
+			sawControl := false
+			if controls != nil {
+				if ctrlLen := getRawMessageControlLen(&msgs[i]); ctrlLen > 0 {
+					if segSize, segCount := parseGROControl(controls[i][:ctrlLen]); segSize > 0 {
+						sawControl = true
+						if u.l.IsLevelEnabled(logrus.DebugLevel) {
+							u.l.WithFields(logrus.Fields{
+								"tag":        "gro-debug",
+								"stage":      "control",
+								"seg_size":   segSize,
+								"seg_count":  segCount,
+								"payloadLen": len(payload),
+							}).Debug("gro control parsed")
+						}
+						segSize = normalizeGROSegSize(segSize, segCount, len(payload))
+						if segSize > 0 && segSize < len(payload) {
+							if u.emitGROSegments(r, addr, payload, segSize) {
+								continue
+							}
+						}
+					}
+				}
+			}
+
+			if u.enableGRO && len(payload) > MTU {
+				if !sawControl && u.l.IsLevelEnabled(logrus.DebugLevel) {
+					u.l.WithFields(logrus.Fields{
+						"tag":         "gro-debug",
+						"stage":       "fallback",
+						"payload_len": len(payload),
+					}).Debug("gro control missing; splitting payload by MTU")
+				}
+				if u.emitGROSegments(r, addr, payload, MTU) {
+					continue
+				}
+			}
+
+			r(addr, payload)
+		}
+	}
+}
+
+func (u *StdConn) readBufferSize() int {
+	if u.enableGRO && u.groBufSize > MTU {
+		return u.groBufSize
+	}
+	return MTU
 }
 
 func (u *StdConn) ReadSingle(msgs []rawMessage) (int, error) {
@@ -159,6 +294,9 @@ func (u *StdConn) ReadSingle(msgs []rawMessage) (int, error) {
 		)
 
 		if err != 0 {
+			if err == unix.EAGAIN || err == unix.EINTR {
+				continue
+			}
 			return 0, &net.OpError{Op: "recvmsg", Err: err}
 		}
 
@@ -180,6 +318,9 @@ func (u *StdConn) ReadMulti(msgs []rawMessage) (int, error) {
 		)
 
 		if err != 0 {
+			if err == unix.EAGAIN || err == unix.EINTR {
+				continue
+			}
 			return 0, &net.OpError{Op: "recvmmsg", Err: err}
 		}
 
@@ -188,6 +329,14 @@ func (u *StdConn) ReadMulti(msgs []rawMessage) (int, error) {
 }
 
 func (u *StdConn) WriteTo(b []byte, ip netip.AddrPort) error {
+	if u.enableGSO && ip.IsValid() {
+		if err := u.queueGSOPacket(b, ip); err == nil {
+			return nil
+		} else if !errors.Is(err, errGSOFallback) {
+			return err
+		}
+	}
+
 	if u.isV4 {
 		return u.writeTo4(b, ip)
 	}
@@ -221,7 +370,7 @@ func (u *StdConn) writeTo6(b []byte, ip netip.AddrPort) error {
 
 func (u *StdConn) writeTo4(b []byte, ip netip.AddrPort) error {
 	if !ip.Addr().Is4() {
-		return ErrInvalidIPv6RemoteForSocket
+		return fmt.Errorf("Listener is IPv4, but writing to IPv6 remote")
 	}
 
 	var rsa unix.RawSockaddrInet4
@@ -294,6 +443,94 @@ func (u *StdConn) ReloadConfig(c *config.C) {
 			u.l.WithError(err).Error("Failed to set listen.so_mark")
 		}
 	}
+
+	u.configureGRO(c)
+	u.configureGSO(c)
+}
+
+func (u *StdConn) configureGRO(c *config.C) {
+	if c == nil {
+		return
+	}
+
+	enable := c.GetBool("listen.enable_gro", false)
+	if enable == u.enableGRO {
+		if enable {
+			if size := c.GetInt("listen.gro_read_buffer", 0); size > 0 {
+				u.setGROBufferSize(size)
+			}
+		}
+		return
+	}
+
+	if enable {
+		if err := unix.SetsockoptInt(u.sysFd, unix.SOL_UDP, unix.UDP_GRO, 1); err != nil {
+			u.l.WithError(err).Warn("Failed to enable UDP GRO")
+			return
+		}
+		u.enableGRO = true
+		u.setGROBufferSize(c.GetInt("listen.gro_read_buffer", defaultGROReadBufferSize))
+		u.l.WithField("buffer_size", u.groBufSize).Info("UDP GRO enabled")
+		return
+	}
+
+	if err := unix.SetsockoptInt(u.sysFd, unix.SOL_UDP, unix.UDP_GRO, 0); err != nil && err != unix.ENOPROTOOPT {
+		u.l.WithError(err).Warn("Failed to disable UDP GRO")
+	}
+	u.enableGRO = false
+	u.groBufSize = MTU
+}
+
+func (u *StdConn) configureGSO(c *config.C) {
+	enable := c.GetBool("listen.enable_gso", false)
+	if !enable {
+		u.disableGSO()
+	} else {
+		u.enableGSO = true
+	}
+
+	segments := c.GetInt("listen.gso_max_segments", defaultGSOMaxSegments)
+	if segments < 1 {
+		segments = 1
+	}
+	u.gsoMaxSegments = segments
+
+	maxBytes := c.GetInt("listen.gso_max_bytes", 0)
+	if maxBytes <= 0 {
+		maxBytes = MTU * segments
+	}
+	if maxBytes > maxGSOBatchBytes {
+		u.l.WithField("requested", maxBytes).Warn("listen.gso_max_bytes larger than UDP limit; clamping")
+		maxBytes = maxGSOBatchBytes
+	}
+	u.gsoMaxBytes = maxBytes
+
+	timeout := c.GetDuration("listen.gso_flush_timeout", defaultGSOFlushTimeout)
+	if timeout < 0 {
+		timeout = 0
+	}
+	u.gsoFlushTimeout = timeout
+}
+
+func (u *StdConn) setGROBufferSize(size int) {
+	if size < MTU {
+		size = defaultGROReadBufferSize
+	}
+	if size > maxGSOBatchBytes {
+		size = maxGSOBatchBytes
+	}
+	u.groBufSize = size
+}
+
+func (u *StdConn) disableGSO() {
+	u.gsoMu.Lock()
+	defer u.gsoMu.Unlock()
+	u.enableGSO = false
+	_ = u.flushGSOlocked()
+	u.gsoBuf = nil
+	u.gsoSegments = 0
+	u.gsoSegSize = 0
+	u.stopGSOTimerLocked()
 }
 
 func (u *StdConn) getMemInfo(meminfo *[unix.SK_MEMINFO_VARS]uint32) error {
@@ -305,7 +542,239 @@ func (u *StdConn) getMemInfo(meminfo *[unix.SK_MEMINFO_VARS]uint32) error {
 	return nil
 }
 
+func (u *StdConn) queueGSOPacket(b []byte, addr netip.AddrPort) error {
+	if len(b) == 0 {
+		return nil
+	}
+
+	u.gsoMu.Lock()
+	defer u.gsoMu.Unlock()
+
+	if !u.enableGSO || !addr.IsValid() || len(b) > u.gsoMaxBytes {
+		if err := u.flushGSOlocked(); err != nil {
+			return err
+		}
+		return errGSOFallback
+	}
+
+	if u.gsoSegments == 0 {
+		if cap(u.gsoBuf) < u.gsoMaxBytes {
+			u.gsoBuf = make([]byte, 0, u.gsoMaxBytes)
+		}
+		u.gsoAddr = addr
+		u.gsoSegSize = len(b)
+	} else if addr != u.gsoAddr || len(b) != u.gsoSegSize {
+		if err := u.flushGSOlocked(); err != nil {
+			return err
+		}
+		if cap(u.gsoBuf) < u.gsoMaxBytes {
+			u.gsoBuf = make([]byte, 0, u.gsoMaxBytes)
+		}
+		u.gsoAddr = addr
+		u.gsoSegSize = len(b)
+	}
+
+	if len(u.gsoBuf)+len(b) > u.gsoMaxBytes {
+		if err := u.flushGSOlocked(); err != nil {
+			return err
+		}
+		if cap(u.gsoBuf) < u.gsoMaxBytes {
+			u.gsoBuf = make([]byte, 0, u.gsoMaxBytes)
+		}
+		u.gsoAddr = addr
+		u.gsoSegSize = len(b)
+	}
+
+	u.gsoBuf = append(u.gsoBuf, b...)
+	u.gsoSegments++
+
+	if u.gsoSegments >= u.gsoMaxSegments || u.gsoFlushTimeout <= 0 {
+		return u.flushGSOlocked()
+	}
+
+	u.scheduleGSOFlushLocked()
+	return nil
+}
+
+func (u *StdConn) flushGSOlocked() error {
+	if u.gsoSegments == 0 {
+		u.stopGSOTimerLocked()
+		return nil
+	}
+
+	payload := append([]byte(nil), u.gsoBuf...)
+	addr := u.gsoAddr
+	segSize := u.gsoSegSize
+
+	u.gsoBuf = u.gsoBuf[:0]
+	u.gsoSegments = 0
+	u.gsoSegSize = 0
+	u.stopGSOTimerLocked()
+
+	if segSize <= 0 {
+		return errGSOFallback
+	}
+
+	err := u.sendSegmented(payload, addr, segSize)
+	if errors.Is(err, errGSODisabled) {
+		u.l.WithField("addr", addr).Warn("UDP GSO disabled by kernel, falling back to sendto")
+		u.enableGSO = false
+		return u.sendSegmentsIndividually(payload, addr, segSize)
+	}
+
+	return err
+}
+
+func (u *StdConn) sendSegmented(payload []byte, addr netip.AddrPort, segSize int) error {
+	if len(payload) == 0 {
+		return nil
+	}
+
+	control := make([]byte, unix.CmsgSpace(2))
+	hdr := (*unix.Cmsghdr)(unsafe.Pointer(&control[0]))
+	hdr.Level = unix.SOL_UDP
+	hdr.Type = unix.UDP_SEGMENT
+	setCmsgLen(hdr, unix.CmsgLen(2))
+	binary.NativeEndian.PutUint16(control[unix.CmsgLen(0):unix.CmsgLen(0)+2], uint16(segSize))
+
+	var sa unix.Sockaddr
+	if addr.Addr().Is4() {
+		var sa4 unix.SockaddrInet4
+		sa4.Port = int(addr.Port())
+		sa4.Addr = addr.Addr().As4()
+		sa = &sa4
+	} else {
+		var sa6 unix.SockaddrInet6
+		sa6.Port = int(addr.Port())
+		sa6.Addr = addr.Addr().As16()
+		sa = &sa6
+	}
+
+	if _, err := unix.SendmsgN(u.sysFd, payload, control, sa, 0); err != nil {
+		if errno, ok := err.(syscall.Errno); ok && (errno == unix.EINVAL || errno == unix.ENOTSUP || errno == unix.EOPNOTSUPP) {
+			return errGSODisabled
+		}
+		return &net.OpError{Op: "sendmsg", Err: err}
+	}
+	return nil
+}
+
+func (u *StdConn) sendSegmentsIndividually(buf []byte, addr netip.AddrPort, segSize int) error {
+	if segSize <= 0 {
+		return errGSOFallback
+	}
+
+	for offset := 0; offset < len(buf); offset += segSize {
+		end := offset + segSize
+		if end > len(buf) {
+			end = len(buf)
+		}
+		var err error
+		if u.isV4 {
+			err = u.writeTo4(buf[offset:end], addr)
+		} else {
+			err = u.writeTo6(buf[offset:end], addr)
+		}
+		if err != nil {
+			return err
+		}
+	}
+	return nil
+}
+
+func (u *StdConn) scheduleGSOFlushLocked() {
+	if u.gsoTimer == nil {
+		u.gsoTimer = time.AfterFunc(u.gsoFlushTimeout, u.gsoFlushTimer)
+		return
+	}
+	u.gsoTimer.Reset(u.gsoFlushTimeout)
+}
+
+func (u *StdConn) stopGSOTimerLocked() {
+	if u.gsoTimer != nil {
+		u.gsoTimer.Stop()
+		u.gsoTimer = nil
+	}
+}
+
+func (u *StdConn) gsoFlushTimer() {
+	u.gsoMu.Lock()
+	defer u.gsoMu.Unlock()
+	_ = u.flushGSOlocked()
+}
+
+func parseGROControl(control []byte) (int, int) {
+	if len(control) == 0 {
+		return 0, 0
+	}
+
+	cmsgs, err := unix.ParseSocketControlMessage(control)
+	if err != nil {
+		return 0, 0
+	}
+
+	for _, c := range cmsgs {
+		if c.Header.Level == unix.SOL_UDP && c.Header.Type == unix.UDP_GRO && len(c.Data) >= 2 {
+			segSize := int(binary.NativeEndian.Uint16(c.Data[:2]))
+			segCount := 0
+			if len(c.Data) >= 4 {
+				segCount = int(binary.NativeEndian.Uint16(c.Data[2:4]))
+			}
+			return segSize, segCount
+		}
+	}
+
+	return 0, 0
+}
+
+func (u *StdConn) emitGROSegments(r EncReader, addr netip.AddrPort, payload []byte, segSize int) bool {
+	if segSize <= 0 {
+		return false
+	}
+
+	for offset := 0; offset < len(payload); offset += segSize {
+		end := offset + segSize
+		if end > len(payload) {
+			end = len(payload)
+		}
+		segment := make([]byte, end-offset)
+		copy(segment, payload[offset:end])
+		r(addr, segment)
+	}
+	return true
+}
+
+func normalizeGROSegSize(segSize, segCount, total int) int {
+	if segSize <= 0 || total <= 0 {
+		return segSize
+	}
+
+	if segSize > total && segCount > 0 {
+		segSize = total / segCount
+		if segSize == 0 {
+			segSize = total
+		}
+	}
+
+	if segCount <= 1 && segSize > 0 && total > segSize {
+		calculated := total / segSize
+		if calculated <= 1 {
+			calculated = (total + segSize - 1) / segSize
+		}
+		if calculated > 1 {
+			segCount = calculated
+		}
+	}
+
+	if segSize > MTU {
+		return MTU
+	}
+
+	return segSize
+}
+
 func (u *StdConn) Close() error {
+	u.disableGSO()
 	return syscall.Close(u.sysFd)
 }
 

udp/udp_linux_32.go

@@ -30,13 +30,16 @@ type rawMessage struct {
 	Len uint32
 }
 
-func (u *StdConn) PrepareRawMessages(n int) ([]rawMessage, [][]byte, [][]byte) {
+func (u *StdConn) PrepareRawMessages(n int, bufSize int) ([]rawMessage, [][]byte, [][]byte) {
+	if bufSize <= 0 {
+		bufSize = MTU
+	}
 	msgs := make([]rawMessage, n)
 	buffers := make([][]byte, n)
 	names := make([][]byte, n)
 
 	for i := range msgs {
-		buffers[i] = make([]byte, MTU)
+		buffers[i] = make([]byte, bufSize)
 		names[i] = make([]byte, unix.SizeofSockaddrInet6)
 
 		vs := []iovec{
@@ -52,3 +55,25 @@ func (u *StdConn) PrepareRawMessages(n int) ([]rawMessage, [][]byte, [][]byte) {
 
 	return msgs, buffers, names
 }
+
+func setRawMessageControl(msg *rawMessage, buf []byte) {
+	if len(buf) == 0 {
+		msg.Hdr.Control = nil
+		msg.Hdr.Controllen = 0
+		return
+	}
+	msg.Hdr.Control = &buf[0]
+	msg.Hdr.Controllen = uint32(len(buf))
+}
+
+func getRawMessageControlLen(msg *rawMessage) int {
+	return int(msg.Hdr.Controllen)
+}
+
+func getRawMessageFlags(msg *rawMessage) int {
+	return int(msg.Hdr.Flags)
+}
+
+func setCmsgLen(h *unix.Cmsghdr, l int) {
+	h.Len = uint32(l)
+}

udp/udp_linux_64.go

@@ -33,13 +33,16 @@ type rawMessage struct {
 	Pad0 [4]byte
 }
 
-func (u *StdConn) PrepareRawMessages(n int) ([]rawMessage, [][]byte, [][]byte) {
+func (u *StdConn) PrepareRawMessages(n int, bufSize int) ([]rawMessage, [][]byte, [][]byte) {
+	if bufSize <= 0 {
+		bufSize = MTU
+	}
 	msgs := make([]rawMessage, n)
 	buffers := make([][]byte, n)
 	names := make([][]byte, n)
 
 	for i := range msgs {
-		buffers[i] = make([]byte, MTU)
+		buffers[i] = make([]byte, bufSize)
 		names[i] = make([]byte, unix.SizeofSockaddrInet6)
 
 		vs := []iovec{
@@ -55,3 +58,25 @@ func (u *StdConn) PrepareRawMessages(n int) ([]rawMessage, [][]byte, [][]byte) {
 
 	return msgs, buffers, names
 }
+
+func setRawMessageControl(msg *rawMessage, buf []byte) {
+	if len(buf) == 0 {
+		msg.Hdr.Control = nil
+		msg.Hdr.Controllen = 0
+		return
+	}
+	msg.Hdr.Control = &buf[0]
+	msg.Hdr.Controllen = uint64(len(buf))
+}
+
+func getRawMessageControlLen(msg *rawMessage) int {
+	return int(msg.Hdr.Controllen)
+}
+
+func getRawMessageFlags(msg *rawMessage) int {
+	return int(msg.Hdr.Flags)
+}
+
+func setCmsgLen(h *unix.Cmsghdr, l int) {
+	h.Len = uint64(l)
+}

udp/udp_rio_windows.go

@@ -134,7 +134,7 @@ func (u *RIOConn) bind(sa windows.Sockaddr) error {
 	return nil
 }
 
-func (u *RIOConn) ListenOut(r EncReader) {
+func (u *RIOConn) ListenOut(r EncReader) error {
 	buffer := make([]byte, MTU)
 
 	for {