Compare commits

..

10 Commits

Author SHA1 Message Date
JackDoan a8a4090831 drop runtime.LockOSThread() on listenOut 2026-04-14 12:59:03 -05:00
JackDoan f4c8dfdfd2 fix vet oops 2026-04-14 12:32:24 -05:00
JackDoan 4c4faa3fac switch back to raw writes 2026-04-14 11:43:06 -05:00
JackDoan f03f753baa fix 2026-04-14 11:27:57 -05:00
JackDoan 28d2b47164 udp_linux: wrap socket operations with syscall.RawConn for clean teardown 2026-04-13 14:30:27 -05:00
Jack Doan 51308b845b connection-track ICMP traffic (#1602)
gofmt / Run gofmt (push) Failing after 2s
smoke-extra / Run extra smoke tests (push) Failing after 3s
smoke / Run multi node smoke test (push) Failing after 2s
Build and test / Build all and test on ubuntu-linux (push) Failing after 2s
Build and test / Build and test on linux with boringcrypto (push) Failing after 2s
Build and test / Build and test on linux with pkcs11 (push) Failing after 2s
Build and test / Build and test on macos-latest (push) Has been cancelled
Build and test / Build and test on windows-latest (push) Has been cancelled
* connection-track ICMP and ICMPv6 traffic

* icmpv6 only has identifier on echo
2026-02-18 23:19:37 -06:00
Wade Simmons 422fc2ad1e go fix (#1608)
gofmt / Run gofmt (push) Failing after 3s
smoke-extra / Run extra smoke tests (push) Failing after 2s
smoke / Run multi node smoke test (push) Failing after 2s
Build and test / Build all and test on ubuntu-linux (push) Failing after 2s
Build and test / Build and test on linux with boringcrypto (push) Failing after 3s
Build and test / Build and test on linux with pkcs11 (push) Failing after 2s
Build and test / Build and test on macos-latest (push) Has been cancelled
Build and test / Build and test on windows-latest (push) Has been cancelled
2026-02-17 11:42:14 -05:00
Wade Simmons e8bb874e14 smoke-extra: try AMD-V workaround (#1610)
gofmt / Run gofmt (push) Failing after 3s
smoke-extra / Run extra smoke tests (push) Failing after 2s
smoke / Run multi node smoke test (push) Failing after 2s
Build and test / Build all and test on ubuntu-linux (push) Failing after 3s
Build and test / Build and test on linux with boringcrypto (push) Failing after 2s
Build and test / Build and test on linux with pkcs11 (push) Failing after 2s
Build and test / Build and test on macos-latest (push) Has been cancelled
Build and test / Build and test on windows-latest (push) Has been cancelled
* smoke-extra: try AMD-V workaround

- https://github.com/slackhq/nebula/actions/runs/21995850645/job/63555492676?pr=1602
- https://github.com/actions/runner-images/issues/13202
- https://github.com/cri-o/packaging/pull/306/changes
2026-02-13 12:55:19 -06:00
Jack Doan 353ad1f271 firewall: icmp no longer requires a port spec (#1609) 2026-02-13 11:10:40 -06:00
Jack Doan f573e8a266 Merge commit from fork
gofmt / Run gofmt (push) Failing after 3s
smoke-extra / Run extra smoke tests (push) Failing after 2s
smoke / Run multi node smoke test (push) Failing after 3s
Build and test / Build all and test on ubuntu-linux (push) Failing after 2s
Build and test / Build and test on linux with boringcrypto (push) Failing after 3s
Build and test / Build and test on linux with pkcs11 (push) Failing after 2s
Build and test / Build and test on macos-latest (push) Has been cancelled
Build and test / Build and test on windows-latest (push) Has been cancelled
Newly signed P256 based certificates will have their signature clamped to the low-s form.

Update CHANGELOG.md
2026-02-06 14:26:51 -05:00
59 changed files with 914 additions and 1596 deletions
+3
View File
@@ -30,6 +30,9 @@ jobs:
- name: add hashicorp source - name: add hashicorp source
run: wget -O- https://apt.releases.hashicorp.com/gpg | gpg --dearmor | sudo tee /usr/share/keyrings/hashicorp-archive-keyring.gpg && echo "deb [signed-by=/usr/share/keyrings/hashicorp-archive-keyring.gpg] https://apt.releases.hashicorp.com $(lsb_release -cs) main" | sudo tee /etc/apt/sources.list.d/hashicorp.list run: wget -O- https://apt.releases.hashicorp.com/gpg | gpg --dearmor | sudo tee /usr/share/keyrings/hashicorp-archive-keyring.gpg && echo "deb [signed-by=/usr/share/keyrings/hashicorp-archive-keyring.gpg] https://apt.releases.hashicorp.com $(lsb_release -cs) main" | sudo tee /etc/apt/sources.list.d/hashicorp.list
- name: workaround AMD-V issue # https://github.com/cri-o/packaging/pull/306
run: sudo rmmod kvm_amd
- name: install vagrant - name: install vagrant
run: sudo apt-get update && sudo apt-get install -y vagrant virtualbox run: sudo apt-get update && sudo apt-get install -y vagrant virtualbox
+10 -9
View File
@@ -37,17 +37,18 @@ docker run --name host4 --device /dev/net/tun:/dev/net/tun --cap-add NET_ADMIN -
sleep 1 sleep 1
# grab tcpdump pcaps for debugging # grab tcpdump pcaps for debugging
docker exec lighthouse1 tcpdump -i nebula1 -q -w - -U 2>logs/lighthouse1.inside.log >logs/lighthouse1.inside.pcap & docker exec lighthouse1 tcpdump -i tun0 -q -w - -U 2>logs/lighthouse1.inside.log >logs/lighthouse1.inside.pcap &
docker exec lighthouse1 tcpdump -i eth0 -q -w - -U 2>logs/lighthouse1.outside.log >logs/lighthouse1.outside.pcap & docker exec lighthouse1 tcpdump -i eth0 -q -w - -U 2>logs/lighthouse1.outside.log >logs/lighthouse1.outside.pcap &
docker exec host2 tcpdump -i nebula1 -q -w - -U 2>logs/host2.inside.log >logs/host2.inside.pcap & docker exec host2 tcpdump -i tun0 -q -w - -U 2>logs/host2.inside.log >logs/host2.inside.pcap &
docker exec host2 tcpdump -i eth0 -q -w - -U 2>logs/host2.outside.log >logs/host2.outside.pcap & docker exec host2 tcpdump -i eth0 -q -w - -U 2>logs/host2.outside.log >logs/host2.outside.pcap &
docker exec host3 tcpdump -i nebula1 -q -w - -U 2>logs/host3.inside.log >logs/host3.inside.pcap & docker exec host3 tcpdump -i tun0 -q -w - -U 2>logs/host3.inside.log >logs/host3.inside.pcap &
docker exec host3 tcpdump -i eth0 -q -w - -U 2>logs/host3.outside.log >logs/host3.outside.pcap & docker exec host3 tcpdump -i eth0 -q -w - -U 2>logs/host3.outside.log >logs/host3.outside.pcap &
docker exec host4 tcpdump -i nebula1 -q -w - -U 2>logs/host4.inside.log >logs/host4.inside.pcap & docker exec host4 tcpdump -i tun0 -q -w - -U 2>logs/host4.inside.log >logs/host4.inside.pcap &
docker exec host4 tcpdump -i eth0 -q -w - -U 2>logs/host4.outside.log >logs/host4.outside.pcap & docker exec host4 tcpdump -i eth0 -q -w - -U 2>logs/host4.outside.log >logs/host4.outside.pcap &
docker exec host2 ncat -nklv 0.0.0.0 2000 & docker exec host2 ncat -nklv 0.0.0.0 2000 &
docker exec host3 ncat -nklv 0.0.0.0 2000 & docker exec host3 ncat -nklv 0.0.0.0 2000 &
docker exec host4 ncat -nkluv 0.0.0.0 4000 &
docker exec host2 ncat -e '/usr/bin/echo host2' -nkluv 0.0.0.0 3000 & docker exec host2 ncat -e '/usr/bin/echo host2' -nkluv 0.0.0.0 3000 &
docker exec host3 ncat -e '/usr/bin/echo host3' -nkluv 0.0.0.0 3000 & docker exec host3 ncat -e '/usr/bin/echo host3' -nkluv 0.0.0.0 3000 &
@@ -119,11 +120,11 @@ echo
echo " *** Testing conntrack" echo " *** Testing conntrack"
echo echo
set -x set -x
# host2 can ping host3 now that host3 pinged it first
docker exec host2 ping -c1 192.168.100.3 # host2 speaking to host4 on UDP 4000 should allow it to reply, when firewall rules would normally not permit this
# host4 can ping host2 once conntrack established docker exec host2 sh -c "/usr/bin/echo host2 | ncat -nuv 192.168.100.4 4000"
docker exec host2 ping -c1 192.168.100.4 docker exec host2 ncat -e '/usr/bin/echo helloagainfromhost2' -nkluv 0.0.0.0 4000 &
docker exec host4 ping -c1 192.168.100.2 docker exec host4 sh -c "/usr/bin/echo host4 | ncat -nuv 192.168.100.2 4000"
docker exec host4 sh -c 'kill 1' docker exec host4 sh -c 'kill 1'
docker exec host3 sh -c 'kill 1' docker exec host3 sh -c 'kill 1'
+15 -1
View File
@@ -7,6 +7,19 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
## [Unreleased] ## [Unreleased]
## [1.10.3] - 2026-02-06
### Security
- Fix an issue where blocklist bypass is possible when using curve P256 since the signature can have 2 valid representations.
Both fingerprint representations will be tested against the blocklist.
Any newly issued P256 based certificates will have their signature clamped to the low-s form.
Nebula will assert the low-s signature form when validating certificates in a future version. [GHSA-69x3-g4r3-p962](https://github.com/slackhq/nebula/security/advisories/GHSA-69x3-g4r3-p962)
### Changed
- Improve error reporting if nebula fails to start due to a tun device naming issue. (#1588)
## [1.10.2] - 2026-01-21 ## [1.10.2] - 2026-01-21
### Fixed ### Fixed
@@ -775,7 +788,8 @@ created.)
- Initial public release. - Initial public release.
[Unreleased]: https://github.com/slackhq/nebula/compare/v1.10.2...HEAD [Unreleased]: https://github.com/slackhq/nebula/compare/v1.10.3...HEAD
[1.10.3]: https://github.com/slackhq/nebula/releases/tag/v1.10.3
[1.10.2]: https://github.com/slackhq/nebula/releases/tag/v1.10.2 [1.10.2]: https://github.com/slackhq/nebula/releases/tag/v1.10.2
[1.10.1]: https://github.com/slackhq/nebula/releases/tag/v1.10.1 [1.10.1]: https://github.com/slackhq/nebula/releases/tag/v1.10.1
[1.10.0]: https://github.com/slackhq/nebula/releases/tag/v1.10.0 [1.10.0]: https://github.com/slackhq/nebula/releases/tag/v1.10.0
-1
View File
@@ -1,5 +1,4 @@
//go:build boringcrypto //go:build boringcrypto
// +build boringcrypto
package nebula package nebula
+18
View File
@@ -141,10 +141,23 @@ func (ncp *CAPool) VerifyCertificate(now time.Time, c Certificate) (*CachedCerti
return nil, err return nil, err
} }
// Pre nebula v1.10.3 could generate signatures in either high or low s form and validation
// of signatures allowed for either. Nebula v1.10.3 and beyond clamps signature generation to low-s form
// but validation still allows for either. Since a change in the signature bytes affects the fingerprint, we
// need to test both forms until such a time comes that we enforce low-s form on signature validation.
fp2, err := CalculateAlternateFingerprint(c)
if err != nil {
return nil, fmt.Errorf("could not calculate alternate fingerprint to verify: %w", err)
}
if fp2 != "" && ncp.IsBlocklisted(fp2) {
return nil, ErrBlockListed
}
cc := CachedCertificate{ cc := CachedCertificate{
Certificate: c, Certificate: c,
InvertedGroups: make(map[string]struct{}), InvertedGroups: make(map[string]struct{}),
Fingerprint: fp, Fingerprint: fp,
fingerprint2: fp2,
signerFingerprint: signer.Fingerprint, signerFingerprint: signer.Fingerprint,
} }
@@ -158,6 +171,11 @@ func (ncp *CAPool) VerifyCertificate(now time.Time, c Certificate) (*CachedCerti
// VerifyCachedCertificate is the same as VerifyCertificate other than it operates on a pre-verified structure and // VerifyCachedCertificate is the same as VerifyCertificate other than it operates on a pre-verified structure and
// is a cheaper operation to perform as a result. // is a cheaper operation to perform as a result.
func (ncp *CAPool) VerifyCachedCertificate(now time.Time, c *CachedCertificate) error { func (ncp *CAPool) VerifyCachedCertificate(now time.Time, c *CachedCertificate) error {
// Check any available alternate fingerprint forms for this certificate, re P256 high-s/low-s
if c.fingerprint2 != "" && ncp.IsBlocklisted(c.fingerprint2) {
return ErrBlockListed
}
_, err := ncp.verify(c.Certificate, now, c.Fingerprint, c.signerFingerprint) _, err := ncp.verify(c.Certificate, now, c.Fingerprint, c.signerFingerprint)
return err return err
} }
+43 -4
View File
@@ -5,6 +5,7 @@ import (
"testing" "testing"
"time" "time"
"github.com/slackhq/nebula/cert/p256"
"github.com/stretchr/testify/assert" "github.com/stretchr/testify/assert"
"github.com/stretchr/testify/require" "github.com/stretchr/testify/require"
) )
@@ -170,6 +171,15 @@ func TestCertificateV1_VerifyP256(t *testing.T) {
_, err = caPool.VerifyCertificate(time.Now(), c) _, err = caPool.VerifyCertificate(time.Now(), c)
require.EqualError(t, err, "certificate is in the block list") require.EqualError(t, err, "certificate is in the block list")
// Create a copy of the cert and swap to the alternate form for the signature
nc := c.Copy()
b, err := p256.Swap(c.Signature())
require.NoError(t, err)
require.NoError(t, nc.(*certificateV1).setSignature(b))
_, err = caPool.VerifyCertificate(time.Now(), nc)
require.EqualError(t, err, "certificate is in the block list")
caPool.ResetCertBlocklist() caPool.ResetCertBlocklist()
_, err = caPool.VerifyCertificate(time.Now(), c) _, err = caPool.VerifyCertificate(time.Now(), c)
require.NoError(t, err) require.NoError(t, err)
@@ -187,7 +197,7 @@ func TestCertificateV1_VerifyP256(t *testing.T) {
require.NoError(t, err) require.NoError(t, err)
caPool = NewCAPool() caPool = NewCAPool()
b, err := caPool.AddCAFromPEM(caPem) b, err = caPool.AddCAFromPEM(caPem)
require.NoError(t, err) require.NoError(t, err)
assert.Empty(t, b) assert.Empty(t, b)
@@ -196,7 +206,17 @@ func TestCertificateV1_VerifyP256(t *testing.T) {
}) })
c, _, _, _ = NewTestCert(Version1, Curve_P256, ca, caKey, "test", time.Now(), time.Now().Add(5*time.Minute), nil, nil, []string{"test1"}) c, _, _, _ = NewTestCert(Version1, Curve_P256, ca, caKey, "test", time.Now(), time.Now().Add(5*time.Minute), nil, nil, []string{"test1"})
_, err = caPool.VerifyCertificate(time.Now(), c) cc, err := caPool.VerifyCertificate(time.Now(), c)
require.NoError(t, err)
// Reset the blocklist and block the alternate form fingerprint
caPool.ResetCertBlocklist()
caPool.BlocklistFingerprint(cc.fingerprint2)
err = caPool.VerifyCachedCertificate(time.Now(), cc)
require.EqualError(t, err, "certificate is in the block list")
caPool.ResetCertBlocklist()
err = caPool.VerifyCachedCertificate(time.Now(), cc)
require.NoError(t, err) require.NoError(t, err)
} }
@@ -394,6 +414,15 @@ func TestCertificateV2_VerifyP256(t *testing.T) {
_, err = caPool.VerifyCertificate(time.Now(), c) _, err = caPool.VerifyCertificate(time.Now(), c)
require.EqualError(t, err, "certificate is in the block list") require.EqualError(t, err, "certificate is in the block list")
// Create a copy of the cert and swap to the alternate form for the signature
nc := c.Copy()
b, err := p256.Swap(c.Signature())
require.NoError(t, err)
require.NoError(t, nc.(*certificateV2).setSignature(b))
_, err = caPool.VerifyCertificate(time.Now(), nc)
require.EqualError(t, err, "certificate is in the block list")
caPool.ResetCertBlocklist() caPool.ResetCertBlocklist()
_, err = caPool.VerifyCertificate(time.Now(), c) _, err = caPool.VerifyCertificate(time.Now(), c)
require.NoError(t, err) require.NoError(t, err)
@@ -411,7 +440,7 @@ func TestCertificateV2_VerifyP256(t *testing.T) {
require.NoError(t, err) require.NoError(t, err)
caPool = NewCAPool() caPool = NewCAPool()
b, err := caPool.AddCAFromPEM(caPem) b, err = caPool.AddCAFromPEM(caPem)
require.NoError(t, err) require.NoError(t, err)
assert.Empty(t, b) assert.Empty(t, b)
@@ -420,7 +449,17 @@ func TestCertificateV2_VerifyP256(t *testing.T) {
}) })
c, _, _, _ = NewTestCert(Version2, Curve_P256, ca, caKey, "test", time.Now(), time.Now().Add(5*time.Minute), nil, nil, []string{"test1"}) c, _, _, _ = NewTestCert(Version2, Curve_P256, ca, caKey, "test", time.Now(), time.Now().Add(5*time.Minute), nil, nil, []string{"test1"})
_, err = caPool.VerifyCertificate(time.Now(), c) cc, err := caPool.VerifyCertificate(time.Now(), c)
require.NoError(t, err)
// Reset the blocklist and block the alternate form fingerprint
caPool.ResetCertBlocklist()
caPool.BlocklistFingerprint(cc.fingerprint2)
err = caPool.VerifyCachedCertificate(time.Now(), cc)
require.EqualError(t, err, "certificate is in the block list")
caPool.ResetCertBlocklist()
err = caPool.VerifyCachedCertificate(time.Now(), cc)
require.NoError(t, err) require.NoError(t, err)
} }
+33
View File
@@ -4,6 +4,8 @@ import (
"fmt" "fmt"
"net/netip" "net/netip"
"time" "time"
"github.com/slackhq/nebula/cert/p256"
) )
type Version uint8 type Version uint8
@@ -110,6 +112,9 @@ type CachedCertificate struct {
InvertedGroups map[string]struct{} InvertedGroups map[string]struct{}
Fingerprint string Fingerprint string
signerFingerprint string signerFingerprint string
// A place to store a 2nd fingerprint if the certificate could have one, such as with P256
fingerprint2 string
} }
func (cc *CachedCertificate) String() string { func (cc *CachedCertificate) String() string {
@@ -152,3 +157,31 @@ func Recombine(v Version, rawCertBytes, publicKey []byte, curve Curve) (Certific
return c, nil return c, nil
} }
// CalculateAlternateFingerprint calculates a 2nd fingerprint representation for P256 certificates
// CAPool blocklist testing through `VerifyCertificate` and `VerifyCachedCertificate` automatically performs this step.
func CalculateAlternateFingerprint(c Certificate) (string, error) {
if c.Curve() != Curve_P256 {
return "", nil
}
nc := c.Copy()
b, err := p256.Swap(nc.Signature())
if err != nil {
return "", err
}
switch v := nc.(type) {
case *certificateV1:
err = v.setSignature(b)
case *certificateV2:
err = v.setSignature(b)
default:
return "", ErrUnknownVersion
}
if err != nil {
return "", err
}
return nc.Fingerprint()
}
+122
View File
@@ -0,0 +1,122 @@
package p256
import (
"crypto/elliptic"
"errors"
"math/big"
"filippo.io/bigmod"
"golang.org/x/crypto/cryptobyte"
"golang.org/x/crypto/cryptobyte/asn1"
)
var halfN = new(big.Int).Rsh(elliptic.P256().Params().N, 1)
var nMod *bigmod.Modulus
func init() {
n, err := bigmod.NewModulus(elliptic.P256().Params().N.Bytes())
if err != nil {
panic(err)
}
nMod = n
}
func IsNormalized(sig []byte) (bool, error) {
r, s, err := parseSignature(sig)
if err != nil {
return false, err
}
return checkLowS(r, s), nil
}
func checkLowS(_, s []byte) bool {
bigS := new(big.Int).SetBytes(s)
// Check if S <= (N/2), because we want to include the midpoint in the set of low-s
return bigS.Cmp(halfN) <= 0
}
func swap(r, s []byte) ([]byte, []byte, error) {
var err error
bigS, err := bigmod.NewNat().SetBytes(s, nMod)
if err != nil {
return nil, nil, err
}
sNormalized := nMod.Nat().Sub(bigS, nMod)
return r, sNormalized.Bytes(nMod), nil
}
func Normalize(sig []byte) ([]byte, error) {
r, s, err := parseSignature(sig)
if err != nil {
return nil, err
}
if checkLowS(r, s) {
return sig, nil
}
newR, newS, err := swap(r, s)
if err != nil {
return nil, err
}
return encodeSignature(newR, newS)
}
// Swap will change sig between its current form to the opposite high or low form.
func Swap(sig []byte) ([]byte, error) {
r, s, err := parseSignature(sig)
if err != nil {
return nil, err
}
newR, newS, err := swap(r, s)
if err != nil {
return nil, err
}
return encodeSignature(newR, newS)
}
// parseSignature taken exactly from crypto/ecdsa/ecdsa.go
func parseSignature(sig []byte) (r, s []byte, err error) {
var inner cryptobyte.String
input := cryptobyte.String(sig)
if !input.ReadASN1(&inner, asn1.SEQUENCE) ||
!input.Empty() ||
!inner.ReadASN1Integer(&r) ||
!inner.ReadASN1Integer(&s) ||
!inner.Empty() {
return nil, nil, errors.New("invalid ASN.1")
}
return r, s, nil
}
func encodeSignature(r, s []byte) ([]byte, error) {
var b cryptobyte.Builder
b.AddASN1(asn1.SEQUENCE, func(b *cryptobyte.Builder) {
addASN1IntBytes(b, r)
addASN1IntBytes(b, s)
})
return b.Bytes()
}
// addASN1IntBytes encodes in ASN.1 a positive integer represented as
// a big-endian byte slice with zero or more leading zeroes.
func addASN1IntBytes(b *cryptobyte.Builder, bytes []byte) {
for len(bytes) > 0 && bytes[0] == 0 {
bytes = bytes[1:]
}
if len(bytes) == 0 {
b.SetError(errors.New("invalid integer"))
return
}
b.AddASN1(asn1.INTEGER, func(c *cryptobyte.Builder) {
if bytes[0]&0x80 != 0 {
c.AddUint8(0)
}
c.AddBytes(bytes)
})
}
+28
View File
@@ -0,0 +1,28 @@
package p256
import (
"crypto/ecdsa"
"crypto/elliptic"
"crypto/rand"
"testing"
"github.com/stretchr/testify/require"
)
func TestFlipping(t *testing.T) {
priv, err1 := ecdsa.GenerateKey(elliptic.P256(), rand.Reader)
require.NoError(t, err1)
out, err := ecdsa.SignASN1(rand.Reader, priv, []byte("big chungus"))
require.NoError(t, err)
r, s, err := parseSignature(out)
require.NoError(t, err)
r, s1, err := swap(r, s)
require.NoError(t, err)
r, s2, err := swap(r, s1)
require.NoError(t, err)
require.Equal(t, s, s2)
require.NotEqual(t, s, s1)
}
+9
View File
@@ -9,6 +9,8 @@ import (
"fmt" "fmt"
"net/netip" "net/netip"
"time" "time"
"github.com/slackhq/nebula/cert/p256"
) )
// TBSCertificate represents a certificate intended to be signed. // TBSCertificate represents a certificate intended to be signed.
@@ -126,6 +128,13 @@ func (t *TBSCertificate) SignWith(signer Certificate, curve Curve, sp SignerLamb
return nil, err return nil, err
} }
if curve == Curve_P256 {
sig, err = p256.Normalize(sig)
if err != nil {
return nil, err
}
}
err = c.setSignature(sig) err = c.setSignature(sig)
if err != nil { if err != nil {
return nil, err return nil, err
+46
View File
@@ -9,6 +9,7 @@ import (
"testing" "testing"
"time" "time"
"github.com/slackhq/nebula/cert/p256"
"github.com/stretchr/testify/assert" "github.com/stretchr/testify/assert"
"github.com/stretchr/testify/require" "github.com/stretchr/testify/require"
) )
@@ -89,3 +90,48 @@ func TestCertificateV1_SignP256(t *testing.T) {
require.NoError(t, err) require.NoError(t, err)
assert.NotNil(t, uc) assert.NotNil(t, uc)
} }
func TestCertificate_SignP256_AlwaysNormalized(t *testing.T) {
before := time.Now().Add(time.Second * -60).Round(time.Second)
after := time.Now().Add(time.Second * 60).Round(time.Second)
pubKey := []byte("01234567890abcedfghij1234567890ab1234567890abcedfghij1234567890ab")
tbs := TBSCertificate{
Version: Version1,
Name: "testing",
Networks: []netip.Prefix{
mustParsePrefixUnmapped("10.1.1.1/24"),
mustParsePrefixUnmapped("10.1.1.2/16"),
},
UnsafeNetworks: []netip.Prefix{
mustParsePrefixUnmapped("9.1.1.2/24"),
mustParsePrefixUnmapped("9.1.1.3/16"),
},
Groups: []string{"test-group1", "test-group2", "test-group3"},
NotBefore: before,
NotAfter: after,
PublicKey: pubKey,
IsCA: true,
Curve: Curve_P256,
}
priv, err := ecdsa.GenerateKey(elliptic.P256(), rand.Reader)
require.NoError(t, err)
pub := elliptic.Marshal(elliptic.P256(), priv.PublicKey.X, priv.PublicKey.Y)
rawPriv := priv.D.FillBytes(make([]byte, 32))
for i := 0; i < 1000; i++ {
if i&1 == 1 {
tbs.Version = Version1
} else {
tbs.Version = Version2
}
c, err := tbs.Sign(nil, Curve_P256, rawPriv)
require.NoError(t, err)
assert.NotNil(t, c)
assert.True(t, c.CheckSignature(pub))
normie, err := p256.IsNormalized(c.Signature())
require.NoError(t, err)
assert.True(t, normie)
}
}
-4
View File
@@ -1,4 +0,0 @@
package main
const NoSuchFileError = "no such file or directory"
const NoSuchDirError = "no such file or directory"
-4
View File
@@ -1,4 +0,0 @@
package main
const NoSuchFileError = "no such file or directory"
const NoSuchDirError = "no such file or directory"
+4 -5
View File
@@ -11,7 +11,6 @@ import (
"github.com/slackhq/nebula/cert" "github.com/slackhq/nebula/cert"
"github.com/slackhq/nebula/config" "github.com/slackhq/nebula/config"
"github.com/slackhq/nebula/test" "github.com/slackhq/nebula/test"
"github.com/slackhq/nebula/test/device"
"github.com/slackhq/nebula/udp" "github.com/slackhq/nebula/udp"
"github.com/stretchr/testify/assert" "github.com/stretchr/testify/assert"
"github.com/stretchr/testify/require" "github.com/stretchr/testify/require"
@@ -53,7 +52,7 @@ func Test_NewConnectionManagerTest(t *testing.T) {
lh := newTestLighthouse() lh := newTestLighthouse()
ifce := &Interface{ ifce := &Interface{
hostMap: hostMap, hostMap: hostMap,
inside: &device.NoopTun{}, inside: &test.NoopTun{},
outside: &udp.NoopConn{}, outside: &udp.NoopConn{},
firewall: &Firewall{}, firewall: &Firewall{},
lightHouse: lh, lightHouse: lh,
@@ -136,7 +135,7 @@ func Test_NewConnectionManagerTest2(t *testing.T) {
lh := newTestLighthouse() lh := newTestLighthouse()
ifce := &Interface{ ifce := &Interface{
hostMap: hostMap, hostMap: hostMap,
inside: &device.NoopTun{}, inside: &test.NoopTun{},
outside: &udp.NoopConn{}, outside: &udp.NoopConn{},
firewall: &Firewall{}, firewall: &Firewall{},
lightHouse: lh, lightHouse: lh,
@@ -221,7 +220,7 @@ func Test_NewConnectionManager_DisconnectInactive(t *testing.T) {
lh := newTestLighthouse() lh := newTestLighthouse()
ifce := &Interface{ ifce := &Interface{
hostMap: hostMap, hostMap: hostMap,
inside: &device.NoopTun{}, inside: &test.NoopTun{},
outside: &udp.NoopConn{}, outside: &udp.NoopConn{},
firewall: &Firewall{}, firewall: &Firewall{},
lightHouse: lh, lightHouse: lh,
@@ -348,7 +347,7 @@ func Test_NewConnectionManagerTest_DisconnectInvalid(t *testing.T) {
lh := newTestLighthouse() lh := newTestLighthouse()
ifce := &Interface{ ifce := &Interface{
hostMap: hostMap, hostMap: hostMap,
inside: &device.NoopTun{}, inside: &test.NoopTun{},
outside: &udp.NoopConn{}, outside: &udp.NoopConn{},
firewall: &Firewall{}, firewall: &Firewall{},
lightHouse: lh, lightHouse: lh,
-1
View File
@@ -1,5 +1,4 @@
//go:build e2e_testing //go:build e2e_testing
// +build e2e_testing
package nebula package nebula
+1 -1
View File
@@ -382,8 +382,8 @@ firewall:
# Rules are comprised of a protocol, port, and one or more of host, group, or CIDR # Rules are comprised of a protocol, port, and one or more of host, group, or CIDR
# Logical evaluation is roughly: port AND proto AND (ca_sha OR ca_name) AND (host OR group OR groups OR cidr) AND (local cidr) # Logical evaluation is roughly: port AND proto AND (ca_sha OR ca_name) AND (host OR group OR groups OR cidr) AND (local cidr)
# - port: Takes `0` or `any` as any, a single number `80`, a range `200-901`, or `fragment` to match second and further fragments of fragmented packets (since there is no port available). # - port: Takes `0` or `any` as any, a single number `80`, a range `200-901`, or `fragment` to match second and further fragments of fragmented packets (since there is no port available).
# code: same as port but makes more sense when talking about ICMP, TODO: this is not currently implemented in a way that works, use `any`
# proto: `any`, `tcp`, `udp`, or `icmp` # proto: `any`, `tcp`, `udp`, or `icmp`
# a port specification is ignored if proto is `icmp`
# host: `any` or a literal hostname, ie `test-host` # host: `any` or a literal hostname, ie `test-host`
# group: `any` or a literal group name, ie `default-group` # group: `any` or a literal group name, ie `default-group`
# groups: Same as group but accepts a list of values. Multiple values are AND'd together and a certificate would have to contain all groups to pass # groups: Same as group but accepts a list of values. Multiple values are AND'd together and a certificate would have to contain all groups to pass
+88 -77
View File
@@ -81,14 +81,8 @@ type FirewallConntrack struct {
Conns map[firewall.Packet]*conn Conns map[firewall.Packet]*conn
TimerWheel *TimerWheel[firewall.Packet] TimerWheel *TimerWheel[firewall.Packet]
// purgeCounter tracks lookups to trigger periodic purge instead of every lookup
purgeCounter uint32
} }
// purgeInterval defines how many lookups between purge attempts
const conntrackPurgeInterval = 1024
// FirewallTable is the entry point for a rule, the evaluation order is: // FirewallTable is the entry point for a rule, the evaluation order is:
// Proto AND port AND (CA SHA or CA name) AND local CIDR AND (group OR groups OR name OR remote CIDR) // Proto AND port AND (CA SHA or CA name) AND local CIDR AND (group OR groups OR name OR remote CIDR)
type FirewallTable struct { type FirewallTable struct {
@@ -255,20 +249,6 @@ func NewFirewallFromConfig(l *logrus.Logger, cs *CertState, c *config.C) (*Firew
// AddRule properly creates the in memory rule structure for a firewall table. // AddRule properly creates the in memory rule structure for a firewall table.
func (f *Firewall) AddRule(incoming bool, proto uint8, startPort int32, endPort int32, groups []string, host string, cidr, localCidr, caName string, caSha string) error { func (f *Firewall) AddRule(incoming bool, proto uint8, startPort int32, endPort int32, groups []string, host string, cidr, localCidr, caName string, caSha string) error {
// We need this rule string because we generate a hash. Removing this will break firewall reload.
ruleString := fmt.Sprintf(
"incoming: %v, proto: %v, startPort: %v, endPort: %v, groups: %v, host: %v, ip: %v, localIp: %v, caName: %v, caSha: %s",
incoming, proto, startPort, endPort, groups, host, cidr, localCidr, caName, caSha,
)
f.rules += ruleString + "\n"
direction := "incoming"
if !incoming {
direction = "outgoing"
}
f.l.WithField("firewallRule", m{"direction": direction, "proto": proto, "startPort": startPort, "endPort": endPort, "groups": groups, "host": host, "cidr": cidr, "localCidr": localCidr, "caName": caName, "caSha": caSha}).
Info("Firewall rule added")
var ( var (
ft *FirewallTable ft *FirewallTable
fp firewallPort fp firewallPort
@@ -286,6 +266,12 @@ func (f *Firewall) AddRule(incoming bool, proto uint8, startPort int32, endPort
case firewall.ProtoUDP: case firewall.ProtoUDP:
fp = ft.UDP fp = ft.UDP
case firewall.ProtoICMP, firewall.ProtoICMPv6: case firewall.ProtoICMP, firewall.ProtoICMPv6:
//ICMP traffic doesn't have ports, so we always coerce to "any", even if a value is provided
if startPort != firewall.PortAny {
f.l.WithField("startPort", startPort).Warn("ignoring port specification for ICMP firewall rule")
}
startPort = firewall.PortAny
endPort = firewall.PortAny
fp = ft.ICMP fp = ft.ICMP
case firewall.ProtoAny: case firewall.ProtoAny:
fp = ft.AnyProto fp = ft.AnyProto
@@ -293,6 +279,20 @@ func (f *Firewall) AddRule(incoming bool, proto uint8, startPort int32, endPort
return fmt.Errorf("unknown protocol %v", proto) return fmt.Errorf("unknown protocol %v", proto)
} }
// We need this rule string because we generate a hash. Removing this will break firewall reload.
ruleString := fmt.Sprintf(
"incoming: %v, proto: %v, startPort: %v, endPort: %v, groups: %v, host: %v, ip: %v, localIp: %v, caName: %v, caSha: %s",
incoming, proto, startPort, endPort, groups, host, cidr, localCidr, caName, caSha,
)
f.rules += ruleString + "\n"
direction := "incoming"
if !incoming {
direction = "outgoing"
}
f.l.WithField("firewallRule", m{"direction": direction, "proto": proto, "startPort": startPort, "endPort": endPort, "groups": groups, "host": host, "cidr": cidr, "localCidr": localCidr, "caName": caName, "caSha": caSha}).
Info("Firewall rule added")
return fp.addRule(f, startPort, endPort, groups, host, cidr, localCidr, caName, caSha) return fp.addRule(f, startPort, endPort, groups, host, cidr, localCidr, caName, caSha)
} }
@@ -355,24 +355,31 @@ func AddFirewallRulesFromConfig(l *logrus.Logger, inbound bool, c *config.C, fw
sPort = r.Port sPort = r.Port
} }
startPort, endPort, err := parsePort(sPort)
if err != nil {
return fmt.Errorf("%s rule #%v; %s %s", table, i, errPort, err)
}
var proto uint8 var proto uint8
var startPort, endPort int32
switch r.Proto { switch r.Proto {
case "any": case "any":
proto = firewall.ProtoAny proto = firewall.ProtoAny
startPort, endPort, err = parsePort(sPort)
case "tcp": case "tcp":
proto = firewall.ProtoTCP proto = firewall.ProtoTCP
startPort, endPort, err = parsePort(sPort)
case "udp": case "udp":
proto = firewall.ProtoUDP proto = firewall.ProtoUDP
startPort, endPort, err = parsePort(sPort)
case "icmp": case "icmp":
proto = firewall.ProtoICMP proto = firewall.ProtoICMP
startPort = firewall.PortAny
endPort = firewall.PortAny
if sPort != "" {
l.WithField("port", sPort).Warn("ignoring port specification for ICMP firewall rule")
}
default: default:
return fmt.Errorf("%s rule #%v; proto was not understood; `%s`", table, i, r.Proto) return fmt.Errorf("%s rule #%v; proto was not understood; `%s`", table, i, r.Proto)
} }
if err != nil {
return fmt.Errorf("%s rule #%v; %s %s", table, i, errPort, err)
}
if r.Cidr != "" && r.Cidr != "any" { if r.Cidr != "" && r.Cidr != "any" {
_, err = netip.ParsePrefix(r.Cidr) _, err = netip.ParsePrefix(r.Cidr)
@@ -473,7 +480,7 @@ func (f *Firewall) metrics(incoming bool) firewallMetrics {
} }
} }
// Destroy cleans up any known cyclical references so the object can be free'd my GC. This should be called if a new // Destroy cleans up any known cyclical references so the object can be freed by GC. This should be called if a new
// firewall object is created // firewall object is created
func (f *Firewall) Destroy() { func (f *Firewall) Destroy() {
//TODO: clean references if/when needed //TODO: clean references if/when needed
@@ -498,17 +505,14 @@ func (f *Firewall) inConns(fp firewall.Packet, h *HostInfo, caPool *cert.CAPool,
conntrack := f.Conntrack conntrack := f.Conntrack
conntrack.Lock() conntrack.Lock()
// Periodic purge instead of every lookup (major CPU savings) // Purge every time we test
conntrack.purgeCounter++ ep, has := conntrack.TimerWheel.Purge()
if conntrack.purgeCounter >= conntrackPurgeInterval { if has {
conntrack.purgeCounter = 0 f.evict(ep)
ep, has := conntrack.TimerWheel.Purge()
if has {
f.evict(ep)
}
} }
c, ok := conntrack.Conns[fp] c, ok := conntrack.Conns[fp]
if !ok { if !ok {
conntrack.Unlock() conntrack.Unlock()
return false return false
@@ -669,6 +673,13 @@ func (fp firewallPort) match(p firewall.Packet, incoming bool, c *cert.CachedCer
return false return false
} }
// this branch is here to catch traffic from FirewallTable.Any.match and FirewallTable.ICMP.match
if p.Protocol == firewall.ProtoICMP || p.Protocol == firewall.ProtoICMPv6 {
// port numbers are re-used for connection tracking of ICMP,
// but we don't want to actually filter on them.
return fp[firewall.PortAny].match(p, c, caPool)
}
var port int32 var port int32
if p.Fragment { if p.Fragment {
@@ -813,10 +824,8 @@ func (fr *FirewallRule) isAny(groups []string, host string, cidr string) bool {
return true return true
} }
for _, group := range groups { if slices.Contains(groups, "any") {
if group == "any" { return true
return true
}
} }
if host == "any" { if host == "any" {
@@ -1027,54 +1036,56 @@ func (r *rule) sanity() error {
} }
} }
if r.Code != "" {
return fmt.Errorf("code specified as [%s]. Support for 'code' will be dropped in a future release, as it has never been functional", r.Code)
}
//todo alert on cidr-any //todo alert on cidr-any
return nil return nil
} }
func parsePort(s string) (startPort, endPort int32, err error) { func parsePort(s string) (int32, int32, error) {
var err error
const notAPort int32 = -2
if s == "any" { if s == "any" {
startPort = firewall.PortAny return firewall.PortAny, firewall.PortAny, nil
endPort = firewall.PortAny }
if s == "fragment" {
} else if s == "fragment" { return firewall.PortFragment, firewall.PortFragment, nil
startPort = firewall.PortFragment }
endPort = firewall.PortFragment if !strings.Contains(s, `-`) {
} else if strings.Contains(s, `-`) {
sPorts := strings.SplitN(s, `-`, 2)
sPorts[0] = strings.Trim(sPorts[0], " ")
sPorts[1] = strings.Trim(sPorts[1], " ")
if len(sPorts) != 2 || sPorts[0] == "" || sPorts[1] == "" {
return 0, 0, fmt.Errorf("appears to be a range but could not be parsed; `%s`", s)
}
rStartPort, err := strconv.Atoi(sPorts[0])
if err != nil {
return 0, 0, fmt.Errorf("beginning range was not a number; `%s`", sPorts[0])
}
rEndPort, err := strconv.Atoi(sPorts[1])
if err != nil {
return 0, 0, fmt.Errorf("ending range was not a number; `%s`", sPorts[1])
}
startPort = int32(rStartPort)
endPort = int32(rEndPort)
if startPort == firewall.PortAny {
endPort = firewall.PortAny
}
} else {
rPort, err := strconv.Atoi(s) rPort, err := strconv.Atoi(s)
if err != nil { if err != nil {
return 0, 0, fmt.Errorf("was not a number; `%s`", s) return notAPort, notAPort, fmt.Errorf("was not a number; `%s`", s)
} }
startPort = int32(rPort) return int32(rPort), int32(rPort), nil
endPort = startPort
} }
return sPorts := strings.SplitN(s, `-`, 2)
for i := range sPorts {
sPorts[i] = strings.Trim(sPorts[i], " ")
}
if len(sPorts) != 2 || sPorts[0] == "" || sPorts[1] == "" {
return notAPort, notAPort, fmt.Errorf("appears to be a range but could not be parsed; `%s`", s)
}
rStartPort, err := strconv.Atoi(sPorts[0])
if err != nil {
return notAPort, notAPort, fmt.Errorf("beginning range was not a number; `%s`", sPorts[0])
}
rEndPort, err := strconv.Atoi(sPorts[1])
if err != nil {
return notAPort, notAPort, fmt.Errorf("ending range was not a number; `%s`", sPorts[1])
}
startPort := int32(rStartPort)
endPort := int32(rEndPort)
if startPort == firewall.PortAny {
endPort = firewall.PortAny
}
return startPort, endPort, nil
} }
+6 -1
View File
@@ -22,7 +22,10 @@ const (
type Packet struct { type Packet struct {
LocalAddr netip.Addr LocalAddr netip.Addr
RemoteAddr netip.Addr RemoteAddr netip.Addr
LocalPort uint16 // LocalPort is the destination port for incoming traffic, or the source port for outgoing. Zero for ICMP.
LocalPort uint16
// RemotePort is the source port for incoming traffic, or the destination port for outgoing.
// For ICMP, it's the "identifier". This is only used for connection tracking, actual firewall rules will not filter on ICMP identifier
RemotePort uint16 RemotePort uint16
Protocol uint8 Protocol uint8
Fragment bool Fragment bool
@@ -46,6 +49,8 @@ func (fp Packet) MarshalJSON() ([]byte, error) {
proto = "tcp" proto = "tcp"
case ProtoICMP: case ProtoICMP:
proto = "icmp" proto = "icmp"
case ProtoICMPv6:
proto = "icmpv6"
case ProtoUDP: case ProtoUDP:
proto = "udp" proto = "udp"
default: default:
+158 -6
View File
@@ -87,9 +87,10 @@ func TestFirewall_AddRule(t *testing.T) {
fw = NewFirewall(l, time.Second, time.Minute, time.Hour, c) fw = NewFirewall(l, time.Second, time.Minute, time.Hour, c)
require.NoError(t, fw.AddRule(true, firewall.ProtoICMP, 1, 1, []string{}, "h1", "", "", "", "")) require.NoError(t, fw.AddRule(true, firewall.ProtoICMP, 1, 1, []string{}, "h1", "", "", "", ""))
assert.Nil(t, fw.InRules.ICMP[1].Any.Any) //no matter what port is given for icmp, it should end up as "any"
assert.Empty(t, fw.InRules.ICMP[1].Any.Groups) assert.Nil(t, fw.InRules.ICMP[firewall.PortAny].Any.Any)
assert.Contains(t, fw.InRules.ICMP[1].Any.Hosts, "h1") assert.Empty(t, fw.InRules.ICMP[firewall.PortAny].Any.Groups)
assert.Contains(t, fw.InRules.ICMP[firewall.PortAny].Any.Hosts, "h1")
fw = NewFirewall(l, time.Second, time.Minute, time.Hour, c) fw = NewFirewall(l, time.Second, time.Minute, time.Hour, c)
require.NoError(t, fw.AddRule(false, firewall.ProtoAny, 1, 1, []string{}, "", ti.String(), "", "", "")) require.NoError(t, fw.AddRule(false, firewall.ProtoAny, 1, 1, []string{}, "", ti.String(), "", "", ""))
@@ -734,6 +735,150 @@ func TestFirewall_DropConntrackReload(t *testing.T) {
assert.Equal(t, fw.Drop(p, false, &h, cp, nil), ErrNoMatchingRule) assert.Equal(t, fw.Drop(p, false, &h, cp, nil), ErrNoMatchingRule)
} }
func TestFirewall_ICMPPortBehavior(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"))
network := netip.MustParsePrefix("1.2.3.4/24")
c := cert.CachedCertificate{
Certificate: &dummyCert{
name: "host1",
networks: []netip.Prefix{network},
groups: []string{"default-group"},
issuer: "signer-shasum",
},
InvertedGroups: map[string]struct{}{"default-group": {}},
}
h := HostInfo{
ConnectionState: &ConnectionState{
peerCert: &c,
},
vpnAddrs: []netip.Addr{network.Addr()},
}
h.buildNetworks(myVpnNetworksTable, c.Certificate)
cp := cert.NewCAPool()
templ := firewall.Packet{
LocalAddr: netip.MustParseAddr("1.2.3.4"),
RemoteAddr: netip.MustParseAddr("1.2.3.4"),
Protocol: firewall.ProtoICMP,
Fragment: false,
}
t.Run("ICMP allowed", func(t *testing.T) {
fw := NewFirewall(l, time.Second, time.Minute, time.Hour, c.Certificate)
require.NoError(t, fw.AddRule(true, firewall.ProtoICMP, 0, 0, []string{"any"}, "", "", "", "", ""))
t.Run("zero ports", func(t *testing.T) {
p := templ.Copy()
p.LocalPort = 0
p.RemotePort = 0
// Drop outbound
assert.Equal(t, fw.Drop(*p, false, &h, cp, nil), ErrNoMatchingRule)
// Allow inbound
resetConntrack(fw)
require.NoError(t, fw.Drop(*p, true, &h, cp, nil))
//now also allow outbound
require.NoError(t, fw.Drop(*p, false, &h, cp, nil))
})
t.Run("nonzero ports", func(t *testing.T) {
p := templ.Copy()
p.LocalPort = 0xabcd
p.RemotePort = 0x1234
// Drop outbound
assert.Equal(t, fw.Drop(*p, false, &h, cp, nil), ErrNoMatchingRule)
// Allow inbound
resetConntrack(fw)
require.NoError(t, fw.Drop(*p, true, &h, cp, nil))
//now also allow outbound
require.NoError(t, fw.Drop(*p, false, &h, cp, nil))
})
})
t.Run("Any proto, some ports allowed", func(t *testing.T) {
fw := NewFirewall(l, time.Second, time.Minute, time.Hour, c.Certificate)
require.NoError(t, fw.AddRule(true, firewall.ProtoAny, 80, 444, []string{"any"}, "", "", "", "", ""))
t.Run("zero ports, still blocked", func(t *testing.T) {
p := templ.Copy()
p.LocalPort = 0
p.RemotePort = 0
// Drop outbound
assert.Equal(t, fw.Drop(*p, false, &h, cp, nil), ErrNoMatchingRule)
// Allow inbound
resetConntrack(fw)
assert.Equal(t, fw.Drop(*p, true, &h, cp, nil), ErrNoMatchingRule)
//now also allow outbound
assert.Equal(t, fw.Drop(*p, false, &h, cp, nil), ErrNoMatchingRule)
})
t.Run("nonzero ports, still blocked", func(t *testing.T) {
p := templ.Copy()
p.LocalPort = 0xabcd
p.RemotePort = 0x1234
// Drop outbound
assert.Equal(t, fw.Drop(*p, false, &h, cp, nil), ErrNoMatchingRule)
// Allow inbound
resetConntrack(fw)
assert.Equal(t, fw.Drop(*p, true, &h, cp, nil), ErrNoMatchingRule)
//now also allow outbound
assert.Equal(t, fw.Drop(*p, false, &h, cp, nil), ErrNoMatchingRule)
})
t.Run("nonzero, matching ports, still blocked", func(t *testing.T) {
p := templ.Copy()
p.LocalPort = 80
p.RemotePort = 80
// Drop outbound
assert.Equal(t, fw.Drop(*p, false, &h, cp, nil), ErrNoMatchingRule)
// Allow inbound
resetConntrack(fw)
assert.Equal(t, fw.Drop(*p, true, &h, cp, nil), ErrNoMatchingRule)
//now also allow outbound
assert.Equal(t, fw.Drop(*p, false, &h, cp, nil), ErrNoMatchingRule)
})
})
t.Run("Any proto, any port", func(t *testing.T) {
fw := NewFirewall(l, time.Second, time.Minute, time.Hour, c.Certificate)
require.NoError(t, fw.AddRule(true, firewall.ProtoAny, 0, 0, []string{"any"}, "", "", "", "", ""))
t.Run("zero ports, allowed", func(t *testing.T) {
resetConntrack(fw)
p := templ.Copy()
p.LocalPort = 0
p.RemotePort = 0
// Drop outbound
assert.Equal(t, fw.Drop(*p, false, &h, cp, nil), ErrNoMatchingRule)
// Allow inbound
resetConntrack(fw)
require.NoError(t, fw.Drop(*p, true, &h, cp, nil))
//now also allow outbound
require.NoError(t, fw.Drop(*p, false, &h, cp, nil))
})
t.Run("nonzero ports, allowed", func(t *testing.T) {
resetConntrack(fw)
p := templ.Copy()
p.LocalPort = 0xabcd
p.RemotePort = 0x1234
// Drop outbound
assert.Equal(t, fw.Drop(*p, false, &h, cp, nil), ErrNoMatchingRule)
// Allow inbound
resetConntrack(fw)
require.NoError(t, fw.Drop(*p, true, &h, cp, nil))
//now also allow outbound
require.NoError(t, fw.Drop(*p, false, &h, cp, nil))
//different ID is blocked
p.RemotePort++
require.Equal(t, fw.Drop(*p, false, &h, cp, nil), ErrNoMatchingRule)
})
})
}
func TestFirewall_DropIPSpoofing(t *testing.T) { func TestFirewall_DropIPSpoofing(t *testing.T) {
l := test.NewLogger() l := test.NewLogger()
ob := &bytes.Buffer{} ob := &bytes.Buffer{}
@@ -919,11 +1064,11 @@ func TestNewFirewallFromConfig(t *testing.T) {
// Test code/port error // Test code/port error
conf = config.NewC(l) conf = config.NewC(l)
conf.Settings["firewall"] = map[string]any{"outbound": []any{map[string]any{"code": "a", "host": "testh"}}} conf.Settings["firewall"] = map[string]any{"outbound": []any{map[string]any{"code": "a", "host": "testh", "proto": "any"}}}
_, err = NewFirewallFromConfig(l, cs, conf) _, err = NewFirewallFromConfig(l, cs, conf)
require.EqualError(t, err, "firewall.outbound rule #0; code was not a number; `a`") require.EqualError(t, err, "firewall.outbound rule #0; code was not a number; `a`")
conf.Settings["firewall"] = map[string]any{"outbound": []any{map[string]any{"port": "a", "host": "testh"}}} conf.Settings["firewall"] = map[string]any{"outbound": []any{map[string]any{"port": "a", "host": "testh", "proto": "any"}}}
_, err = NewFirewallFromConfig(l, cs, conf) _, err = NewFirewallFromConfig(l, cs, conf)
require.EqualError(t, err, "firewall.outbound rule #0; port was not a number; `a`") require.EqualError(t, err, "firewall.outbound rule #0; port was not a number; `a`")
@@ -973,7 +1118,14 @@ func TestAddFirewallRulesFromConfig(t *testing.T) {
mf = &mockFirewall{} mf = &mockFirewall{}
conf.Settings["firewall"] = map[string]any{"outbound": []any{map[string]any{"port": "1", "proto": "icmp", "host": "a"}}} conf.Settings["firewall"] = map[string]any{"outbound": []any{map[string]any{"port": "1", "proto": "icmp", "host": "a"}}}
require.NoError(t, AddFirewallRulesFromConfig(l, false, conf, mf)) require.NoError(t, AddFirewallRulesFromConfig(l, false, conf, mf))
assert.Equal(t, addRuleCall{incoming: false, proto: firewall.ProtoICMP, startPort: 1, endPort: 1, groups: nil, host: "a", ip: "", localIp: ""}, mf.lastCall) assert.Equal(t, addRuleCall{incoming: false, proto: firewall.ProtoICMP, startPort: firewall.PortAny, endPort: firewall.PortAny, groups: nil, host: "a", ip: "", localIp: ""}, mf.lastCall)
// Test adding icmp rule no port
conf = config.NewC(l)
mf = &mockFirewall{}
conf.Settings["firewall"] = map[string]any{"outbound": []any{map[string]any{"proto": "icmp", "host": "a"}}}
require.NoError(t, AddFirewallRulesFromConfig(l, false, conf, mf))
assert.Equal(t, addRuleCall{incoming: false, proto: firewall.ProtoICMP, startPort: firewall.PortAny, endPort: firewall.PortAny, groups: nil, host: "a", ip: "", localIp: ""}, mf.lastCall)
// Test adding any rule // Test adding any rule
conf = config.NewC(l) conf = config.NewC(l)
+4 -3
View File
@@ -4,6 +4,7 @@ go 1.25
require ( require (
dario.cat/mergo v1.0.2 dario.cat/mergo v1.0.2
filippo.io/bigmod v0.1.0
github.com/anmitsu/go-shlex v0.0.0-20200514113438-38f4b401e2be github.com/anmitsu/go-shlex v0.0.0-20200514113438-38f4b401e2be
github.com/armon/go-radix v1.0.0 github.com/armon/go-radix v1.0.0
github.com/cyberdelia/go-metrics-graphite v0.0.0-20161219230853-39f87cc3b432 github.com/cyberdelia/go-metrics-graphite v0.0.0-20161219230853-39f87cc3b432
@@ -30,11 +31,11 @@ require (
golang.org/x/sys v0.40.0 golang.org/x/sys v0.40.0
golang.org/x/term v0.39.0 golang.org/x/term v0.39.0
golang.zx2c4.com/wintun v0.0.0-20230126152724-0fa3db229ce2 golang.zx2c4.com/wintun v0.0.0-20230126152724-0fa3db229ce2
golang.zx2c4.com/wireguard v0.0.0-20250521234502-f333402bd9cb golang.zx2c4.com/wireguard v0.0.0-20230325221338-052af4a8072b
golang.zx2c4.com/wireguard/windows v0.5.3 golang.zx2c4.com/wireguard/windows v0.5.3
google.golang.org/protobuf v1.36.11 google.golang.org/protobuf v1.36.11
gopkg.in/yaml.v3 v3.0.1 gopkg.in/yaml.v3 v3.0.1
gvisor.dev/gvisor v0.0.0-20250503011706-39ed1f5ac29c gvisor.dev/gvisor v0.0.0-20240423190808-9d7a357edefe
) )
require ( require (
@@ -50,6 +51,6 @@ require (
github.com/vishvananda/netns v0.0.5 // indirect github.com/vishvananda/netns v0.0.5 // indirect
go.yaml.in/yaml/v2 v2.4.2 // indirect go.yaml.in/yaml/v2 v2.4.2 // indirect
golang.org/x/mod v0.31.0 // indirect golang.org/x/mod v0.31.0 // indirect
golang.org/x/time v0.7.0 // indirect golang.org/x/time v0.5.0 // indirect
golang.org/x/tools v0.40.0 // indirect golang.org/x/tools v0.40.0 // indirect
) )
+8 -6
View File
@@ -1,6 +1,8 @@
cloud.google.com/go v0.34.0/go.mod h1:aQUYkXzVsufM+DwF1aE+0xfcU+56JwCaLick0ClmMTw= cloud.google.com/go v0.34.0/go.mod h1:aQUYkXzVsufM+DwF1aE+0xfcU+56JwCaLick0ClmMTw=
dario.cat/mergo v1.0.2 h1:85+piFYR1tMbRrLcDwR18y4UKJ3aH1Tbzi24VRW1TK8= dario.cat/mergo v1.0.2 h1:85+piFYR1tMbRrLcDwR18y4UKJ3aH1Tbzi24VRW1TK8=
dario.cat/mergo v1.0.2/go.mod h1:E/hbnu0NxMFBjpMIE34DRGLWqDy0g5FuKDhCb31ngxA= dario.cat/mergo v1.0.2/go.mod h1:E/hbnu0NxMFBjpMIE34DRGLWqDy0g5FuKDhCb31ngxA=
filippo.io/bigmod v0.1.0 h1:UNzDk7y9ADKST+axd9skUpBQeW7fG2KrTZyOE4uGQy8=
filippo.io/bigmod v0.1.0/go.mod h1:OjOXDNlClLblvXdwgFFOQFJEocLhhtai8vGLy0JCZlI=
github.com/alecthomas/template v0.0.0-20160405071501-a0175ee3bccc/go.mod h1:LOuyumcjzFXgccqObfd/Ljyb9UuFJ6TxHnclSeseNhc= github.com/alecthomas/template v0.0.0-20160405071501-a0175ee3bccc/go.mod h1:LOuyumcjzFXgccqObfd/Ljyb9UuFJ6TxHnclSeseNhc=
github.com/alecthomas/template v0.0.0-20190718012654-fb15b899a751/go.mod h1:LOuyumcjzFXgccqObfd/Ljyb9UuFJ6TxHnclSeseNhc= github.com/alecthomas/template v0.0.0-20190718012654-fb15b899a751/go.mod h1:LOuyumcjzFXgccqObfd/Ljyb9UuFJ6TxHnclSeseNhc=
github.com/alecthomas/units v0.0.0-20151022065526-2efee857e7cf/go.mod h1:ybxpYRFXyAe+OPACYpWeL0wqObRcbAqCMya13uyzqw0= github.com/alecthomas/units v0.0.0-20151022065526-2efee857e7cf/go.mod h1:ybxpYRFXyAe+OPACYpWeL0wqObRcbAqCMya13uyzqw0=
@@ -216,8 +218,8 @@ golang.org/x/term v0.39.0/go.mod h1:yxzUCTP/U+FzoxfdKmLaA0RV1WgE0VY7hXBwKtY/4ww=
golang.org/x/text v0.3.0/go.mod h1:NqM8EUOU14njkJ3fqMW+pc6Ldnwhi/IjpwHt7yyuwOQ= 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 v0.3.2/go.mod h1:bEr9sfX3Q8Zfm5fL9x+3itogRgK3+ptLWKqgva+5dAk=
golang.org/x/text v0.3.3/go.mod h1:5Zoc/QRtKVWzQhOtBMvqHzDpF6irO9z98xDceosuGiQ= golang.org/x/text v0.3.3/go.mod h1:5Zoc/QRtKVWzQhOtBMvqHzDpF6irO9z98xDceosuGiQ=
golang.org/x/time v0.7.0 h1:ntUhktv3OPE6TgYxXWv9vKvUSJyIFJlyohwbkEwPrKQ= golang.org/x/time v0.5.0 h1:o7cqy6amK/52YcAKIPlM3a+Fpj35zvRj2TP+e1xFSfk=
golang.org/x/time v0.7.0/go.mod h1:3BpzKBy/shNhVucY/MWOyx10tF3SFh9QdLuxbVysPQM= golang.org/x/time v0.5.0/go.mod h1:3BpzKBy/shNhVucY/MWOyx10tF3SFh9QdLuxbVysPQM=
golang.org/x/tools v0.0.0-20180917221912-90fa682c2a6e/go.mod h1:n7NCudcB/nEzxVGmLbDWY5pfWTLqBcC2KZ6jyYvM4mQ= golang.org/x/tools v0.0.0-20180917221912-90fa682c2a6e/go.mod h1:n7NCudcB/nEzxVGmLbDWY5pfWTLqBcC2KZ6jyYvM4mQ=
golang.org/x/tools v0.0.0-20191119224855-298f0cb1881e/go.mod h1:b+2E5dAYhXwXZwtnZ6UAqBI28+e2cm9otk0dWdXHAEo= golang.org/x/tools v0.0.0-20191119224855-298f0cb1881e/go.mod h1:b+2E5dAYhXwXZwtnZ6UAqBI28+e2cm9otk0dWdXHAEo=
golang.org/x/tools v0.0.0-20200130002326-2f3ba24bd6e7/go.mod h1:TB2adYChydJhpapKDTa4BR/hXlZSLoq2Wpct/0txZ28= golang.org/x/tools v0.0.0-20200130002326-2f3ba24bd6e7/go.mod h1:TB2adYChydJhpapKDTa4BR/hXlZSLoq2Wpct/0txZ28=
@@ -231,8 +233,8 @@ golang.org/x/xerrors v0.0.0-20191204190536-9bdfabe68543/go.mod h1:I/5z698sn9Ka8T
golang.org/x/xerrors v0.0.0-20200804184101-5ec99f83aff1/go.mod h1:I/5z698sn9Ka8TeJc9MKroUUfqBBauWjQqLJ2OPfmY0= golang.org/x/xerrors v0.0.0-20200804184101-5ec99f83aff1/go.mod h1:I/5z698sn9Ka8TeJc9MKroUUfqBBauWjQqLJ2OPfmY0=
golang.zx2c4.com/wintun v0.0.0-20230126152724-0fa3db229ce2 h1:B82qJJgjvYKsXS9jeunTOisW56dUokqW/FOteYJJ/yg= golang.zx2c4.com/wintun v0.0.0-20230126152724-0fa3db229ce2 h1:B82qJJgjvYKsXS9jeunTOisW56dUokqW/FOteYJJ/yg=
golang.zx2c4.com/wintun v0.0.0-20230126152724-0fa3db229ce2/go.mod h1:deeaetjYA+DHMHg+sMSMI58GrEteJUUzzw7en6TJQcI= golang.zx2c4.com/wintun v0.0.0-20230126152724-0fa3db229ce2/go.mod h1:deeaetjYA+DHMHg+sMSMI58GrEteJUUzzw7en6TJQcI=
golang.zx2c4.com/wireguard v0.0.0-20250521234502-f333402bd9cb h1:whnFRlWMcXI9d+ZbWg+4sHnLp52d5yiIPUxMBSt4X9A= golang.zx2c4.com/wireguard v0.0.0-20230325221338-052af4a8072b h1:J1CaxgLerRR5lgx3wnr6L04cJFbWoceSK9JWBdglINo=
golang.zx2c4.com/wireguard v0.0.0-20250521234502-f333402bd9cb/go.mod h1:rpwXGsirqLqN2L0JDJQlwOboGHmptD5ZD6T2VmcqhTw= golang.zx2c4.com/wireguard v0.0.0-20230325221338-052af4a8072b/go.mod h1:tqur9LnfstdR9ep2LaJT4lFUl0EjlHtge+gAjmsHUG4=
golang.zx2c4.com/wireguard/windows v0.5.3 h1:On6j2Rpn3OEMXqBq00QEDC7bWSZrPIHKIus8eIuExIE= golang.zx2c4.com/wireguard/windows v0.5.3 h1:On6j2Rpn3OEMXqBq00QEDC7bWSZrPIHKIus8eIuExIE=
golang.zx2c4.com/wireguard/windows v0.5.3/go.mod h1:9TEe8TJmtwyQebdFwAkEWOPr3prrtqm+REGFifP60hI= golang.zx2c4.com/wireguard/windows v0.5.3/go.mod h1:9TEe8TJmtwyQebdFwAkEWOPr3prrtqm+REGFifP60hI=
google.golang.org/appengine v1.4.0/go.mod h1:xpcJRLb0r/rnEns0DIKYYv+WjYCduHsrkT7/EB5XEv4= google.golang.org/appengine v1.4.0/go.mod h1:xpcJRLb0r/rnEns0DIKYYv+WjYCduHsrkT7/EB5XEv4=
@@ -258,5 +260,5 @@ gopkg.in/yaml.v2 v2.3.0/go.mod h1:hI93XBmqTisBFMUTm0b8Fm+jr3Dg1NNxqwp+5A1VGuI=
gopkg.in/yaml.v3 v3.0.0-20200313102051-9f266ea9e77c/go.mod h1:K4uyk7z7BCEPqu6E+C64Yfv1cQ7kz7rIZviUmN+EgEM= gopkg.in/yaml.v3 v3.0.0-20200313102051-9f266ea9e77c/go.mod h1:K4uyk7z7BCEPqu6E+C64Yfv1cQ7kz7rIZviUmN+EgEM=
gopkg.in/yaml.v3 v3.0.1 h1:fxVm/GzAzEWqLHuvctI91KS9hhNmmWOoWu0XTYJS7CA= gopkg.in/yaml.v3 v3.0.1 h1:fxVm/GzAzEWqLHuvctI91KS9hhNmmWOoWu0XTYJS7CA=
gopkg.in/yaml.v3 v3.0.1/go.mod h1:K4uyk7z7BCEPqu6E+C64Yfv1cQ7kz7rIZviUmN+EgEM= gopkg.in/yaml.v3 v3.0.1/go.mod h1:K4uyk7z7BCEPqu6E+C64Yfv1cQ7kz7rIZviUmN+EgEM=
gvisor.dev/gvisor v0.0.0-20250503011706-39ed1f5ac29c h1:m/r7OM+Y2Ty1sgBQ7Qb27VgIMBW8ZZhT4gLnUyDIhzI= gvisor.dev/gvisor v0.0.0-20240423190808-9d7a357edefe h1:fre4i6mv4iBuz5lCMOzHD1rH1ljqHWSICFmZRbbgp3g=
gvisor.dev/gvisor v0.0.0-20250503011706-39ed1f5ac29c/go.mod h1:3r5CMtNQMKIvBlrmM9xWUNamjKBYPOWyXOjmg5Kts3g= gvisor.dev/gvisor v0.0.0-20240423190808-9d7a357edefe/go.mod h1:sxc3Uvk/vHcd3tj7/DHVBoR5wvWT/MmRq2pj7HRJnwU=
+1 -1
View File
@@ -590,7 +590,7 @@ func (hm *HandshakeManager) allocateIndex(hh *HandshakeHostInfo) error {
hm.Lock() hm.Lock()
defer hm.Unlock() defer hm.Unlock()
for i := 0; i < 32; i++ { for range 32 {
index, err := generateIndex(hm.l) index, err := generateIndex(hm.l)
if err != nil { if err != nil {
return err return err
-1
View File
@@ -1,5 +1,4 @@
//go:build e2e_testing //go:build e2e_testing
// +build e2e_testing
package nebula package nebula
-254
View File
@@ -2,8 +2,6 @@ package nebula
import ( import (
"net/netip" "net/netip"
"sync"
"time"
"github.com/sirupsen/logrus" "github.com/sirupsen/logrus"
"github.com/slackhq/nebula/firewall" "github.com/slackhq/nebula/firewall"
@@ -13,258 +11,6 @@ import (
"github.com/slackhq/nebula/routing" "github.com/slackhq/nebula/routing"
) )
// preEncryptionPacket holds packet data before batch encryption
type preEncryptionPacket struct {
hostinfo *HostInfo
ci *ConnectionState
packet []byte
out []byte
}
// Pool for preEncryptionBatch slices to reduce allocations
var preEncryptionBatchPool = sync.Pool{
New: func() any {
// Pre-allocate with reasonable capacity
batch := make([]preEncryptionPacket, 0, 128)
return &batch
},
}
// consumeInsidePackets processes multiple packets in a batch for improved performance
// packets: slice of packet buffers to process
// sizes: slice of packet sizes
// count: number of packets to process
// outs: slice of output buffers (one per packet) with virtio headroom
// q: queue index
// localCache: firewall conntrack cache
// batchPackets: pre-allocated slice for accumulating encrypted packets
// batchAddrs: pre-allocated slice for accumulating destination addresses
func (f *Interface) consumeInsidePackets(packets [][]byte, sizes []int, count int, outs [][]byte, nb []byte, q int, localCache firewall.ConntrackCache, batchPackets *[][]byte, batchAddrs *[]netip.AddrPort) {
// Reusable per-packet state
fwPacket := &firewall.Packet{}
// Reset batch accumulation slices (reuse capacity)
*batchPackets = (*batchPackets)[:0]
*batchAddrs = (*batchAddrs)[:0]
// Collect packets for batched encryption
preEncryptionBatch := make([]preEncryptionPacket, 0, count)
// Process each packet in the batch
for i := 0; i < count; i++ {
packet := packets[i][:sizes[i]]
out := outs[i]
// Inline the consumeInsidePacket logic for better performance
err := newPacket(packet, false, fwPacket)
if err != nil {
if f.l.Level >= logrus.DebugLevel {
f.l.WithField("packet", packet).Debugf("Error while validating outbound packet: %s", err)
}
continue
}
// Ignore local broadcast packets
if f.dropLocalBroadcast {
if f.myBroadcastAddrsTable.Contains(fwPacket.RemoteAddr) {
continue
}
}
if f.myVpnAddrsTable.Contains(fwPacket.RemoteAddr) {
// Immediately forward packets from self to self.
if immediatelyForwardToSelf {
_, err := f.readers[q].Write(packet)
if err != nil {
f.l.WithError(err).Error("Failed to forward to tun")
}
}
continue
}
// Ignore multicast packets
if f.dropMulticast && fwPacket.RemoteAddr.IsMulticast() {
continue
}
hostinfo, ready := f.getOrHandshakeConsiderRouting(fwPacket, func(hh *HandshakeHostInfo) {
hh.cachePacket(f.l, header.Message, 0, packet, f.sendMessageNow, f.cachedPacketMetrics)
})
if hostinfo == nil {
f.rejectInside(packet, out, q)
if f.l.Level >= logrus.DebugLevel {
f.l.WithField("vpnAddr", fwPacket.RemoteAddr).
WithField("fwPacket", fwPacket).
Debugln("dropping outbound packet, vpnAddr not in our vpn networks or in unsafe networks")
}
continue
}
if !ready {
continue
}
dropReason := f.firewall.Drop(*fwPacket, false, hostinfo, f.pki.GetCAPool(), localCache)
if dropReason != nil {
f.rejectInside(packet, out, q)
if f.l.Level >= logrus.DebugLevel {
hostinfo.logger(f.l).
WithField("fwPacket", fwPacket).
WithField("reason", dropReason).
Debugln("dropping outbound packet")
}
continue
}
// Prepare packet for batch encryption
ci := hostinfo.ConnectionState
if ci.eKey == nil {
continue
}
// Check if this needs relay - if so, send immediately and skip batching
useRelay := !hostinfo.remote.IsValid()
if useRelay {
// Handle relay sends individually (less common path)
f.sendNoMetrics(header.Message, 0, ci, hostinfo, netip.AddrPort{}, packet, nb, out, q)
continue
}
// Collect for batched encryption
preEncryptionBatch = append(preEncryptionBatch, preEncryptionPacket{
hostinfo: hostinfo,
ci: ci,
packet: packet,
out: out,
})
}
// BATCH ENCRYPTION: Process all collected packets
if len(preEncryptionBatch) > 0 {
f.encryptBatch(preEncryptionBatch, nb, batchPackets, batchAddrs)
}
// Send all accumulated packets in one batch
if len(*batchPackets) > 0 {
batchSize := len(*batchPackets)
n, err := f.writers[q].WriteMulti(*batchPackets, *batchAddrs)
if err != nil {
f.l.WithError(err).WithField("sent", n).WithField("total", batchSize).Error("Failed to send batch")
}
}
}
// encryptBatch processes multiple packets, grouping by ConnectionState to reduce lock acquisitions
func (f *Interface) encryptBatch(batch []preEncryptionPacket, nb []byte, batchPackets *[][]byte, batchAddrs *[]netip.AddrPort) {
lockStart := time.Now()
lockAcquisitions := int64(0)
if noiseutil.EncryptLockNeeded {
// Group packets by ConnectionState to minimize lock acquisitions
// Process packets in order but batch lock acquisitions per CI
var currentCI *ConnectionState
var lockHeld bool
for i := range batch {
ci := batch[i].ci
hostinfo := batch[i].hostinfo
// Validate packet data to prevent nil pointer dereference
if ci == nil || hostinfo == nil {
continue
}
// Switch locks if we're moving to a different ConnectionState
if ci != currentCI {
if lockHeld {
currentCI.writeLock.Unlock()
}
ci.writeLock.Lock()
lockAcquisitions++
currentCI = ci
lockHeld = true
}
c := ci.messageCounter.Add(1)
out := header.Encode(batch[i].out, header.Version, header.Message, 0, hostinfo.remoteIndexId, c)
f.connectionManager.Out(hostinfo)
// Query lighthouse if needed
if hostinfo.lastRebindCount != f.rebindCount {
f.lightHouse.QueryServer(hostinfo.vpnAddrs[0])
hostinfo.lastRebindCount = f.rebindCount
if f.l.Level >= logrus.DebugLevel {
f.l.WithField("vpnAddrs", hostinfo.vpnAddrs).Debug("Lighthouse update triggered for punch due to rebind counter")
}
}
var err error
out, err = ci.eKey.EncryptDanger(out, out, batch[i].packet, c, nb)
if err != nil {
hostinfo.logger(f.l).WithError(err).
WithField("counter", c).
Error("Failed to encrypt outgoing packet")
continue
}
// Add to output batches
*batchPackets = append(*batchPackets, out)
*batchAddrs = append(*batchAddrs, hostinfo.remote)
}
// Release final lock
if lockHeld {
currentCI.writeLock.Unlock()
}
} else {
// No locks needed - process directly
for i := range batch {
ci := batch[i].ci
hostinfo := batch[i].hostinfo
// Validate packet data to prevent nil pointer dereference
if ci == nil || hostinfo == nil {
continue
}
c := ci.messageCounter.Add(1)
out := header.Encode(batch[i].out, header.Version, header.Message, 0, hostinfo.remoteIndexId, c)
f.connectionManager.Out(hostinfo)
// Query lighthouse if needed
if hostinfo.lastRebindCount != f.rebindCount {
f.lightHouse.QueryServer(hostinfo.vpnAddrs[0])
hostinfo.lastRebindCount = f.rebindCount
if f.l.Level >= logrus.DebugLevel {
f.l.WithField("vpnAddrs", hostinfo.vpnAddrs).Debug("Lighthouse update triggered for punch due to rebind counter")
}
}
var err error
out, err = ci.eKey.EncryptDanger(out, out, batch[i].packet, c, nb)
if err != nil {
hostinfo.logger(f.l).WithError(err).
WithField("counter", c).
Error("Failed to encrypt outgoing packet")
continue
}
// Add to output batches
*batchPackets = append(*batchPackets, out)
*batchAddrs = append(*batchAddrs, hostinfo.remote)
}
}
// Record metrics
encryptionTime := time.Since(lockStart)
if noiseutil.EncryptLockNeeded {
f.batchMetrics.lockAcquisitions.Inc(lockAcquisitions)
}
f.batchMetrics.encryptionTime.Update(encryptionTime.Nanoseconds())
f.batchMetrics.batchSize.Update(int64(len(batch)))
}
func (f *Interface) consumeInsidePacket(packet []byte, fwPacket *firewall.Packet, nb, out []byte, q int, localCache firewall.ConntrackCache) { func (f *Interface) consumeInsidePacket(packet []byte, fwPacket *firewall.Packet, nb, out []byte, q int, localCache firewall.ConntrackCache) {
err := newPacket(packet, false, fwPacket) err := newPacket(packet, false, fwPacket)
if err != nil { if err != nil {
-1
View File
@@ -1,5 +1,4 @@
//go:build darwin || dragonfly || freebsd || netbsd || openbsd //go:build darwin || dragonfly || freebsd || netbsd || openbsd
// +build darwin dragonfly freebsd netbsd openbsd
package nebula package nebula
-1
View File
@@ -1,5 +1,4 @@
//go:build !darwin && !dragonfly && !freebsd && !netbsd && !openbsd //go:build !darwin && !dragonfly && !freebsd && !netbsd && !openbsd
// +build !darwin,!dragonfly,!freebsd,!netbsd,!openbsd
package nebula package nebula
+16 -61
View File
@@ -4,6 +4,7 @@ import (
"context" "context"
"errors" "errors"
"fmt" "fmt"
"io"
"net/netip" "net/netip"
"os" "os"
"runtime" "runtime"
@@ -21,7 +22,6 @@ import (
) )
const mtu = 9001 const mtu = 9001
const virtioNetHdrLen = overlay.VirtioNetHdrLen
type InterfaceConfig struct { type InterfaceConfig struct {
HostMap *HostMap HostMap *HostMap
@@ -50,13 +50,6 @@ type InterfaceConfig struct {
l *logrus.Logger l *logrus.Logger
} }
type batchMetrics struct {
udpReadSize metrics.Histogram
encryptionTime metrics.Histogram // Time spent in encryption (including lock waits)
batchSize metrics.Histogram // Dynamic batch sizes being used
lockAcquisitions metrics.Counter // Number of lock acquisitions (should be minimal)
}
type Interface struct { type Interface struct {
hostMap *HostMap hostMap *HostMap
outside udp.Conn outside udp.Conn
@@ -94,12 +87,11 @@ type Interface struct {
conntrackCacheTimeout time.Duration conntrackCacheTimeout time.Duration
writers []udp.Conn writers []udp.Conn
readers []overlay.BatchReadWriter readers []io.ReadWriteCloser
metricHandshakes metrics.Histogram metricHandshakes metrics.Histogram
messageMetrics *MessageMetrics messageMetrics *MessageMetrics
cachedPacketMetrics *cachedPacketMetrics cachedPacketMetrics *cachedPacketMetrics
batchMetrics *batchMetrics
l *logrus.Logger l *logrus.Logger
} }
@@ -186,7 +178,7 @@ func NewInterface(ctx context.Context, c *InterfaceConfig) (*Interface, error) {
routines: c.routines, routines: c.routines,
version: c.version, version: c.version,
writers: make([]udp.Conn, c.routines), writers: make([]udp.Conn, c.routines),
readers: make([]overlay.BatchReadWriter, c.routines), readers: make([]io.ReadWriteCloser, c.routines),
myVpnNetworks: cs.myVpnNetworks, myVpnNetworks: cs.myVpnNetworks,
myVpnNetworksTable: cs.myVpnNetworksTable, myVpnNetworksTable: cs.myVpnNetworksTable,
myVpnAddrs: cs.myVpnAddrs, myVpnAddrs: cs.myVpnAddrs,
@@ -202,12 +194,6 @@ func NewInterface(ctx context.Context, c *InterfaceConfig) (*Interface, error) {
sent: metrics.GetOrRegisterCounter("hostinfo.cached_packets.sent", nil), sent: metrics.GetOrRegisterCounter("hostinfo.cached_packets.sent", nil),
dropped: metrics.GetOrRegisterCounter("hostinfo.cached_packets.dropped", nil), dropped: metrics.GetOrRegisterCounter("hostinfo.cached_packets.dropped", nil),
}, },
batchMetrics: &batchMetrics{
udpReadSize: metrics.GetOrRegisterHistogram("batch.udp_read_size", nil, metrics.NewUniformSample(1024)),
encryptionTime: metrics.GetOrRegisterHistogram("batch.encryption_time_ns", nil, metrics.NewUniformSample(1024)),
batchSize: metrics.GetOrRegisterHistogram("batch.size", nil, metrics.NewUniformSample(1024)),
lockAcquisitions: metrics.GetOrRegisterCounter("batch.lock_acquisitions", nil),
},
l: c.l, l: c.l,
} }
@@ -247,7 +233,7 @@ func (f *Interface) activate() {
metrics.GetOrRegisterGauge("routines", nil).Update(int64(f.routines)) metrics.GetOrRegisterGauge("routines", nil).Update(int64(f.routines))
// Prepare n tun queues // Prepare n tun queues
var reader overlay.BatchReadWriter = f.inside var reader io.ReadWriteCloser = f.inside
for i := 0; i < f.routines; i++ { for i := 0; i < f.routines; i++ {
if i > 0 { if i > 0 {
reader, err = f.inside.NewMultiQueueReader() reader, err = f.inside.NewMultiQueueReader()
@@ -277,8 +263,6 @@ func (f *Interface) run() {
} }
func (f *Interface) listenOut(i int) { func (f *Interface) listenOut(i int) {
runtime.LockOSThread()
var li udp.Conn var li udp.Conn
if i > 0 { if i > 0 {
li = f.writers[i] li = f.writers[i]
@@ -288,68 +272,39 @@ func (f *Interface) listenOut(i int) {
ctCache := firewall.NewConntrackCacheTicker(f.conntrackCacheTimeout) ctCache := firewall.NewConntrackCacheTicker(f.conntrackCacheTimeout)
lhh := f.lightHouse.NewRequestHandler() lhh := f.lightHouse.NewRequestHandler()
plaintext := make([]byte, udp.MTU)
// Pre-allocate output buffers for batch processing
batchSize := li.BatchSize()
outs := make([][]byte, batchSize)
for idx := range outs {
// Allocate full buffer with virtio header space
outs[idx] = make([]byte, virtioNetHdrLen, virtioNetHdrLen+udp.MTU)
}
h := &header.H{} h := &header.H{}
fwPacket := &firewall.Packet{} fwPacket := &firewall.Packet{}
nb := make([]byte, 12) nb := make([]byte, 12, 12)
li.ListenOutBatch(func(addrs []netip.AddrPort, payloads [][]byte, count int) {
f.readOutsidePacketsBatch(addrs, payloads, count, outs[:count], nb, i, h, fwPacket, lhh, ctCache.Get(f.l))
li.ListenOut(func(fromUdpAddr netip.AddrPort, payload []byte) {
f.readOutsidePackets(ViaSender{UdpAddr: fromUdpAddr}, plaintext[:0], payload, h, fwPacket, lhh, nb, i, ctCache.Get(f.l))
}) })
} }
func (f *Interface) listenIn(reader overlay.BatchReadWriter, i int) { func (f *Interface) listenIn(reader io.ReadWriteCloser, i int) {
runtime.LockOSThread() runtime.LockOSThread()
batchSize := reader.BatchSize() packet := make([]byte, mtu)
out := make([]byte, mtu)
// Allocate buffers for batch reading fwPacket := &firewall.Packet{}
bufs := make([][]byte, batchSize) nb := make([]byte, 12, 12)
for idx := range bufs {
bufs[idx] = make([]byte, mtu)
}
sizes := make([]int, batchSize)
// Allocate output buffers for batch processing (one per packet)
// Each has virtio header headroom to avoid copies on write
outs := make([][]byte, batchSize)
for idx := range outs {
outBuf := make([]byte, virtioNetHdrLen+mtu)
outs[idx] = outBuf[virtioNetHdrLen:] // Slice starting after headroom
}
// Pre-allocate batch accumulation buffers for sending
batchPackets := make([][]byte, 0, batchSize)
batchAddrs := make([]netip.AddrPort, 0, batchSize)
// Pre-allocate nonce buffer (reused for all encryptions)
nb := make([]byte, 12)
conntrackCache := firewall.NewConntrackCacheTicker(f.conntrackCacheTimeout) conntrackCache := firewall.NewConntrackCacheTicker(f.conntrackCacheTimeout)
for { for {
n, err := reader.BatchRead(bufs, sizes) n, err := reader.Read(packet)
if err != nil { if err != nil {
if errors.Is(err, os.ErrClosed) && f.closed.Load() { if errors.Is(err, os.ErrClosed) && f.closed.Load() {
return return
} }
f.l.WithError(err).Error("Error while batch reading outbound packets") f.l.WithError(err).Error("Error while reading outbound packet")
// This only seems to happen when something fatal happens to the fd, so exit. // This only seems to happen when something fatal happens to the fd, so exit.
os.Exit(2) os.Exit(2)
} }
// Process all packets in the batch at once f.consumeInsidePacket(packet[:n], fwPacket, nb, out, i, conntrackCache.Get(f.l))
f.consumeInsidePackets(bufs, sizes, n, outs, nb, i, conntrackCache.Get(f.l), &batchPackets, &batchAddrs)
} }
} }
+3 -10
View File
@@ -713,21 +713,14 @@ func (lh *LightHouse) unlockedShouldAddV6(vpnAddr netip.Addr, to *V6AddrPort) bo
func (lh *LightHouse) IsLighthouseAddr(vpnAddr netip.Addr) bool { func (lh *LightHouse) IsLighthouseAddr(vpnAddr netip.Addr) bool {
l := lh.GetLighthouses() l := lh.GetLighthouses()
for i := range l { return slices.Contains(l, vpnAddr)
if l[i] == vpnAddr {
return true
}
}
return false
} }
func (lh *LightHouse) IsAnyLighthouseAddr(vpnAddrs []netip.Addr) bool { func (lh *LightHouse) IsAnyLighthouseAddr(vpnAddrs []netip.Addr) bool {
l := lh.GetLighthouses() l := lh.GetLighthouses()
for i := range vpnAddrs { for i := range vpnAddrs {
for j := range l { if slices.Contains(l, vpnAddrs[i]) {
if l[j] == vpnAddrs[i] { return true
return true
}
} }
} }
return false return false
+2 -6
View File
@@ -105,11 +105,7 @@ func Main(c *config.C, configTest bool, buildVersion string, logger *logrus.Logg
// deprecated and undocumented // deprecated and undocumented
tunQueues := c.GetInt("tun.routines", 1) tunQueues := c.GetInt("tun.routines", 1)
udpQueues := c.GetInt("listen.routines", 1) udpQueues := c.GetInt("listen.routines", 1)
if tunQueues > udpQueues { routines = max(tunQueues, udpQueues)
routines = tunQueues
} else {
routines = udpQueues
}
if routines != 1 { if routines != 1 {
l.WithField("routines", routines).Warn("Setting tun.routines and listen.routines is deprecated. Use `routines` instead") l.WithField("routines", routines).Warn("Setting tun.routines and listen.routines is deprecated. Use `routines` instead")
} }
@@ -171,7 +167,7 @@ func Main(c *config.C, configTest bool, buildVersion string, logger *logrus.Logg
for i := 0; i < routines; i++ { for i := 0; i < routines; i++ {
l.Infof("listening on %v", netip.AddrPortFrom(listenHost, uint16(port))) l.Infof("listening on %v", netip.AddrPortFrom(listenHost, uint16(port)))
udpServer, err := udp.NewListener(l, listenHost, port, routines > 1, c.GetInt("listen.batch", 128)) udpServer, err := udp.NewListener(l, listenHost, port, routines > 1, c.GetInt("listen.batch", 64))
if err != nil { if err != nil {
return nil, util.NewContextualError("Failed to open udp listener", m{"queue": i}, err) return nil, util.NewContextualError("Failed to open udp listener", m{"queue": i}, err)
} }
-1
View File
@@ -1,5 +1,4 @@
//go:build !boringcrypto //go:build !boringcrypto
// +build !boringcrypto
package nebula package nebula
+46 -133
View File
@@ -102,7 +102,7 @@ func (f *Interface) readOutsidePackets(via ViaSender, out []byte, packet []byte,
relay: relay, relay: relay,
IsRelayed: true, IsRelayed: true,
} }
f.readOutsidePackets(via, out[:virtioNetHdrLen], signedPayload, h, fwPacket, lhf, nb, q, localCache) f.readOutsidePackets(via, out[:0], signedPayload, h, fwPacket, lhf, nb, q, localCache)
return return
case ForwardingType: case ForwardingType:
// Find the target HostInfo relay object // Find the target HostInfo relay object
@@ -145,7 +145,7 @@ func (f *Interface) readOutsidePackets(via ViaSender, out []byte, packet []byte,
} }
//TODO: assert via is not relayed //TODO: assert via is not relayed
lhf.HandleRequest(via.UdpAddr, hostinfo.vpnAddrs, d[virtioNetHdrLen:], f) lhf.HandleRequest(via.UdpAddr, hostinfo.vpnAddrs, d, f)
// Fallthrough to the bottom to record incoming traffic // Fallthrough to the bottom to record incoming traffic
@@ -167,7 +167,7 @@ func (f *Interface) readOutsidePackets(via ViaSender, out []byte, packet []byte,
// This testRequest might be from TryPromoteBest, so we should roam // This testRequest might be from TryPromoteBest, so we should roam
// to the new IP address before responding // to the new IP address before responding
f.handleHostRoaming(hostinfo, via) f.handleHostRoaming(hostinfo, via)
f.send(header.Test, header.TestReply, ci, hostinfo, d[virtioNetHdrLen:], nb, out) f.send(header.Test, header.TestReply, ci, hostinfo, d, nb, out)
} }
// Fallthrough to the bottom to record incoming traffic // Fallthrough to the bottom to record incoming traffic
@@ -210,7 +210,7 @@ func (f *Interface) readOutsidePackets(via ViaSender, out []byte, packet []byte,
return return
} }
f.relayManager.HandleControlMsg(hostinfo, d[virtioNetHdrLen:], f) f.relayManager.HandleControlMsg(hostinfo, d, f)
default: default:
f.messageMetrics.Rx(h.Type, h.Subtype, 1) f.messageMetrics.Rx(h.Type, h.Subtype, 1)
@@ -327,13 +327,29 @@ func parseV6(data []byte, incoming bool, fp *firewall.Packet) error {
proto := layers.IPProtocol(data[protoAt]) proto := layers.IPProtocol(data[protoAt])
switch proto { switch proto {
case layers.IPProtocolICMPv6, layers.IPProtocolESP, layers.IPProtocolNoNextHeader: case layers.IPProtocolESP, layers.IPProtocolNoNextHeader:
fp.Protocol = uint8(proto) fp.Protocol = uint8(proto)
fp.RemotePort = 0 fp.RemotePort = 0
fp.LocalPort = 0 fp.LocalPort = 0
fp.Fragment = false fp.Fragment = false
return nil return nil
case layers.IPProtocolICMPv6:
if dataLen < offset+6 {
return ErrIPv6PacketTooShort
}
fp.Protocol = uint8(proto)
fp.LocalPort = 0 //incoming vs outgoing doesn't matter for icmpv6
icmptype := data[offset+1]
switch icmptype {
case layers.ICMPv6TypeEchoRequest, layers.ICMPv6TypeEchoReply:
fp.RemotePort = binary.BigEndian.Uint16(data[offset+4 : offset+6]) //identifier
default:
fp.RemotePort = 0
}
fp.Fragment = false
return nil
case layers.IPProtocolTCP, layers.IPProtocolUDP: case layers.IPProtocolTCP, layers.IPProtocolUDP:
if dataLen < offset+4 { if dataLen < offset+4 {
return ErrIPv6PacketTooShort return ErrIPv6PacketTooShort
@@ -423,34 +439,38 @@ func parseV4(data []byte, incoming bool, fp *firewall.Packet) error {
// Accounting for a variable header length, do we have enough data for our src/dst tuples? // Accounting for a variable header length, do we have enough data for our src/dst tuples?
minLen := ihl minLen := ihl
if !fp.Fragment && fp.Protocol != firewall.ProtoICMP { if !fp.Fragment {
minLen += minFwPacketLen if fp.Protocol == firewall.ProtoICMP {
minLen += minFwPacketLen + 2
} else {
minLen += minFwPacketLen
}
} }
if len(data) < minLen { if len(data) < minLen {
return ErrIPv4InvalidHeaderLength return ErrIPv4InvalidHeaderLength
} }
// Firewall packets are locally oriented if incoming { // Firewall packets are locally oriented
if incoming {
fp.RemoteAddr, _ = netip.AddrFromSlice(data[12:16]) fp.RemoteAddr, _ = netip.AddrFromSlice(data[12:16])
fp.LocalAddr, _ = netip.AddrFromSlice(data[16:20]) fp.LocalAddr, _ = netip.AddrFromSlice(data[16:20])
if fp.Fragment || fp.Protocol == firewall.ProtoICMP {
fp.RemotePort = 0
fp.LocalPort = 0
} else {
fp.RemotePort = binary.BigEndian.Uint16(data[ihl : ihl+2])
fp.LocalPort = binary.BigEndian.Uint16(data[ihl+2 : ihl+4])
}
} else { } else {
fp.LocalAddr, _ = netip.AddrFromSlice(data[12:16]) fp.LocalAddr, _ = netip.AddrFromSlice(data[12:16])
fp.RemoteAddr, _ = netip.AddrFromSlice(data[16:20]) fp.RemoteAddr, _ = netip.AddrFromSlice(data[16:20])
if fp.Fragment || fp.Protocol == firewall.ProtoICMP { }
fp.RemotePort = 0
fp.LocalPort = 0 if fp.Fragment {
} else { fp.RemotePort = 0
fp.LocalPort = binary.BigEndian.Uint16(data[ihl : ihl+2]) fp.LocalPort = 0
fp.RemotePort = binary.BigEndian.Uint16(data[ihl+2 : ihl+4]) } else if fp.Protocol == firewall.ProtoICMP { //note that orientation doesn't matter on ICMP
} fp.RemotePort = binary.BigEndian.Uint16(data[ihl+4 : ihl+6]) //identifier
fp.LocalPort = 0 //code would be uint16(data[ihl+1])
} else if incoming {
fp.RemotePort = binary.BigEndian.Uint16(data[ihl : ihl+2]) //src port
fp.LocalPort = binary.BigEndian.Uint16(data[ihl+2 : ihl+4]) //dst port
} else {
fp.LocalPort = binary.BigEndian.Uint16(data[ihl : ihl+2]) //src port
fp.RemotePort = binary.BigEndian.Uint16(data[ihl+2 : ihl+4]) //dst port
} }
return nil return nil
@@ -481,11 +501,9 @@ func (f *Interface) decryptToTun(hostinfo *HostInfo, messageCounter uint64, out
return false return false
} }
packetData := out[virtioNetHdrLen:] err = newPacket(out, true, fwPacket)
err = newPacket(packetData, true, fwPacket)
if err != nil { if err != nil {
hostinfo.logger(f.l).WithError(err).WithField("packet", packetData). hostinfo.logger(f.l).WithError(err).WithField("packet", out).
Warnf("Error while validating inbound packet") Warnf("Error while validating inbound packet")
return false return false
} }
@@ -500,7 +518,7 @@ func (f *Interface) decryptToTun(hostinfo *HostInfo, messageCounter uint64, out
if dropReason != nil { if dropReason != nil {
// NOTE: We give `packet` as the `out` here since we already decrypted from it and we don't need it anymore // NOTE: We give `packet` as the `out` here since we already decrypted from it and we don't need it anymore
// This gives us a buffer to build the reject packet in // This gives us a buffer to build the reject packet in
f.rejectOutside(packetData, hostinfo.ConnectionState, hostinfo, nb, packet, q) f.rejectOutside(out, hostinfo.ConnectionState, hostinfo, nb, packet, q)
if f.l.Level >= logrus.DebugLevel { if f.l.Level >= logrus.DebugLevel {
hostinfo.logger(f.l).WithField("fwPacket", fwPacket). hostinfo.logger(f.l).WithField("fwPacket", fwPacket).
WithField("reason", dropReason). WithField("reason", dropReason).
@@ -564,108 +582,3 @@ func (f *Interface) handleRecvError(addr netip.AddrPort, h *header.H) {
// We also delete it from pending hostmap to allow for fast reconnect. // We also delete it from pending hostmap to allow for fast reconnect.
f.handshakeManager.DeleteHostInfo(hostinfo) f.handshakeManager.DeleteHostInfo(hostinfo)
} }
// readOutsidePacketsBatch processes multiple packets received from UDP in a batch
// and writes all successfully decrypted packets to TUN in a single operation
func (f *Interface) readOutsidePacketsBatch(addrs []netip.AddrPort, payloads [][]byte, count int, outs [][]byte, nb []byte, q int, h *header.H, fwPacket *firewall.Packet, lhf *LightHouseHandler, localCache firewall.ConntrackCache) {
// Pre-allocate slice for accumulating successful decryptions
tunPackets := make([][]byte, 0, count)
for i := 0; i < count; i++ {
payload := payloads[i]
addr := addrs[i]
out := outs[i]
// Parse header
err := h.Parse(payload)
if err != nil {
if len(payload) > 1 {
f.l.WithField("packet", payload).Infof("Error while parsing inbound packet from %s: %s", addr, err)
}
continue
}
if addr.IsValid() {
if f.myVpnNetworksTable.Contains(addr.Addr()) {
if f.l.Level >= logrus.DebugLevel {
f.l.WithField("udpAddr", addr).Debug("Refusing to process double encrypted packet")
}
continue
}
}
var hostinfo *HostInfo
if h.Type == header.Message && h.Subtype == header.MessageRelay {
hostinfo = f.hostMap.QueryRelayIndex(h.RemoteIndex)
} else {
hostinfo = f.hostMap.QueryIndex(h.RemoteIndex)
}
var ci *ConnectionState
if hostinfo != nil {
ci = hostinfo.ConnectionState
}
switch h.Type {
case header.Message:
if !f.handleEncrypted(ci, ViaSender{UdpAddr: addr}, h) {
continue
}
switch h.Subtype {
case header.MessageNone:
// Decrypt packet
out, err = hostinfo.ConnectionState.dKey.DecryptDanger(out, payload[:header.Len], payload[header.Len:], h.MessageCounter, nb)
if err != nil {
hostinfo.logger(f.l).WithError(err).Error("Failed to decrypt packet")
continue
}
packetData := out[virtioNetHdrLen:]
err = newPacket(packetData, true, fwPacket)
if err != nil {
hostinfo.logger(f.l).WithError(err).WithField("packet", packetData).Warnf("Error while validating inbound packet")
continue
}
if !hostinfo.ConnectionState.window.Update(f.l, h.MessageCounter) {
hostinfo.logger(f.l).WithField("fwPacket", fwPacket).Debugln("dropping out of window packet")
continue
}
dropReason := f.firewall.Drop(*fwPacket, true, hostinfo, f.pki.GetCAPool(), localCache)
if dropReason != nil {
f.rejectOutside(packetData, hostinfo.ConnectionState, hostinfo, nb, payload, q)
if f.l.Level >= logrus.DebugLevel {
hostinfo.logger(f.l).WithField("fwPacket", fwPacket).WithField("reason", dropReason).Debugln("dropping inbound packet")
}
continue
}
f.connectionManager.In(hostinfo)
// Add to batch for TUN write
tunPackets = append(tunPackets, out)
case header.MessageRelay:
// Skip relay packets in batch mode for now (less common path)
f.readOutsidePackets(ViaSender{UdpAddr: addr}, out[:virtioNetHdrLen], payload, h, fwPacket, lhf, nb, q, localCache)
default:
hostinfo.logger(f.l).Debugf("unexpected message subtype %d", h.Subtype)
}
default:
// Handle non-Message types using single-packet path
f.readOutsidePackets(ViaSender{UdpAddr: addr}, out[:virtioNetHdrLen], payload, h, fwPacket, lhf, nb, q, localCache)
}
}
if len(tunPackets) > 0 {
n, err := f.readers[q].WriteBatch(tunPackets, virtioNetHdrLen)
if err != nil {
f.l.WithError(err).WithField("sent", n).WithField("total", len(tunPackets)).Error("Failed to batch write to tun")
}
}
}
+17 -10
View File
@@ -155,6 +155,7 @@ func Test_newPacket_v6(t *testing.T) {
// next layer, missing length byte // next layer, missing length byte
err = newPacket(buffer.Bytes()[:49], true, p) err = newPacket(buffer.Bytes()[:49], true, p)
require.ErrorIs(t, err, ErrIPv6CouldNotFindPayload) require.ErrorIs(t, err, ErrIPv6CouldNotFindPayload)
err = nil
// A good ICMP packet // A good ICMP packet
ip = layers.IPv6{ ip = layers.IPv6{
@@ -165,20 +166,26 @@ func Test_newPacket_v6(t *testing.T) {
DstIP: net.IPv6linklocalallnodes, DstIP: net.IPv6linklocalallnodes,
} }
icmp := layers.ICMPv6{} icmp := layers.ICMPv6{
TypeCode: layers.ICMPv6TypeEchoRequest,
buffer.Clear() Checksum: 0x1234,
err = gopacket.SerializeLayers(buffer, opt, &ip, &icmp)
if err != nil {
panic(err)
} }
err = newPacket(buffer.Bytes(), true, p) buffer.Clear()
require.NoError(t, err) require.NoError(t, gopacket.SerializeLayers(buffer, opt, &ip, &icmp))
require.Error(t, newPacket(buffer.Bytes(), true, p))
buffer.Clear()
echo := layers.ICMPv6Echo{
Identifier: 0xabcd,
SeqNumber: 1234,
}
require.NoError(t, gopacket.SerializeLayers(buffer, opt, &ip, &icmp, &echo))
require.NoError(t, newPacket(buffer.Bytes(), true, p))
assert.Equal(t, uint8(layers.IPProtocolICMPv6), p.Protocol) assert.Equal(t, uint8(layers.IPProtocolICMPv6), p.Protocol)
assert.Equal(t, netip.MustParseAddr("ff02::2"), p.RemoteAddr) assert.Equal(t, netip.MustParseAddr("ff02::2"), p.RemoteAddr)
assert.Equal(t, netip.MustParseAddr("ff02::1"), p.LocalAddr) assert.Equal(t, netip.MustParseAddr("ff02::1"), p.LocalAddr)
assert.Equal(t, uint16(0), p.RemotePort) assert.Equal(t, uint16(0xabcd), p.RemotePort)
assert.Equal(t, uint16(0), p.LocalPort) assert.Equal(t, uint16(0), p.LocalPort)
assert.False(t, p.Fragment) assert.False(t, p.Fragment)
@@ -574,7 +581,7 @@ func BenchmarkParseV6(b *testing.B) {
} }
evilBytes := buffer.Bytes() evilBytes := buffer.Bytes()
for i := 0; i < 200; i++ { for range 200 {
evilBytes = append(evilBytes, hopHeader...) evilBytes = append(evilBytes, hopHeader...)
} }
evilBytes = append(evilBytes, lastHopHeader...) evilBytes = append(evilBytes, lastHopHeader...)
+2 -16
View File
@@ -7,26 +7,12 @@ import (
"github.com/slackhq/nebula/routing" "github.com/slackhq/nebula/routing"
) )
// BatchReadWriter extends io.ReadWriteCloser with batch I/O operations
type BatchReadWriter interface {
io.ReadWriteCloser
// BatchRead reads multiple packets at once
BatchRead(bufs [][]byte, sizes []int) (int, error)
// WriteBatch writes multiple packets at once
WriteBatch(bufs [][]byte, offset int) (int, error)
// BatchSize returns the optimal batch size for this device
BatchSize() int
}
type Device interface { type Device interface {
BatchReadWriter io.ReadWriteCloser
Activate() error Activate() error
Networks() []netip.Prefix Networks() []netip.Prefix
Name() string Name() string
RoutesFor(netip.Addr) routing.Gateways RoutesFor(netip.Addr) routing.Gateways
SupportsMultiqueue() bool SupportsMultiqueue() bool
NewMultiQueueReader() (BatchReadWriter, error) NewMultiQueueReader() (io.ReadWriteCloser, error)
} }
-1
View File
@@ -11,7 +11,6 @@ import (
) )
const DefaultMTU = 1300 const DefaultMTU = 1300
const VirtioNetHdrLen = 10 // Size of virtio_net_hdr structure
type NameError struct { type NameError struct {
Name string Name string
+1 -24
View File
@@ -99,29 +99,6 @@ func (t *tun) SupportsMultiqueue() bool {
return false return false
} }
func (t *tun) NewMultiQueueReader() (BatchReadWriter, error) { func (t *tun) NewMultiQueueReader() (io.ReadWriteCloser, error) {
return nil, fmt.Errorf("TODO: multiqueue not implemented for android") return nil, fmt.Errorf("TODO: multiqueue not implemented for android")
} }
func (t *tun) BatchRead(bufs [][]byte, sizes []int) (int, error) {
n, err := t.Read(bufs[0])
if err != nil {
return 0, err
}
sizes[0] = n
return 1, nil
}
func (t *tun) WriteBatch(bufs [][]byte, offset int) (int, error) {
for i, buf := range bufs {
_, err := t.Write(buf[offset:])
if err != nil {
return i, err
}
}
return len(bufs), nil
}
func (t *tun) BatchSize() int {
return 1
}
+1 -27
View File
@@ -553,32 +553,6 @@ func (t *tun) SupportsMultiqueue() bool {
return false return false
} }
func (t *tun) NewMultiQueueReader() (BatchReadWriter, error) { func (t *tun) NewMultiQueueReader() (io.ReadWriteCloser, error) {
return nil, fmt.Errorf("TODO: multiqueue not implemented for darwin") return nil, fmt.Errorf("TODO: multiqueue not implemented for darwin")
} }
// BatchRead reads a single packet (batch size 1 for non-Linux platforms)
func (t *tun) BatchRead(bufs [][]byte, sizes []int) (int, error) {
n, err := t.Read(bufs[0])
if err != nil {
return 0, err
}
sizes[0] = n
return 1, nil
}
// WriteBatch writes packets individually (no batching for non-Linux platforms)
func (t *tun) WriteBatch(bufs [][]byte, offset int) (int, error) {
for i, buf := range bufs {
_, err := t.Write(buf[offset:])
if err != nil {
return i, err
}
}
return len(bufs), nil
}
// BatchSize returns 1 for non-Linux platforms (no batching)
func (t *tun) BatchSize() int {
return 1
}
+1 -27
View File
@@ -109,36 +109,10 @@ func (t *disabledTun) SupportsMultiqueue() bool {
return true return true
} }
func (t *disabledTun) NewMultiQueueReader() (BatchReadWriter, error) { func (t *disabledTun) NewMultiQueueReader() (io.ReadWriteCloser, error) {
return t, nil return t, nil
} }
// BatchRead reads a single packet (batch size 1 for disabled tun)
func (t *disabledTun) BatchRead(bufs [][]byte, sizes []int) (int, error) {
n, err := t.Read(bufs[0])
if err != nil {
return 0, err
}
sizes[0] = n
return 1, nil
}
// WriteBatch writes packets individually (no batching for disabled tun)
func (t *disabledTun) WriteBatch(bufs [][]byte, offset int) (int, error) {
for i, buf := range bufs {
_, err := t.Write(buf[offset:])
if err != nil {
return i, err
}
}
return len(bufs), nil
}
// BatchSize returns 1 for disabled tun (no batching)
func (t *disabledTun) BatchSize() int {
return 1
}
func (t *disabledTun) Close() error { func (t *disabledTun) Close() error {
if t.read != nil { if t.read != nil {
close(t.read) close(t.read)
+2 -27
View File
@@ -7,6 +7,7 @@ import (
"bytes" "bytes"
"errors" "errors"
"fmt" "fmt"
"io"
"io/fs" "io/fs"
"net/netip" "net/netip"
"sync/atomic" "sync/atomic"
@@ -453,36 +454,10 @@ func (t *tun) SupportsMultiqueue() bool {
return false return false
} }
func (t *tun) NewMultiQueueReader() (BatchReadWriter, error) { func (t *tun) NewMultiQueueReader() (io.ReadWriteCloser, error) {
return nil, fmt.Errorf("TODO: multiqueue not implemented for freebsd") return nil, fmt.Errorf("TODO: multiqueue not implemented for freebsd")
} }
// BatchRead reads a single packet (batch size 1 for FreeBSD)
func (t *tun) BatchRead(bufs [][]byte, sizes []int) (int, error) {
n, err := t.Read(bufs[0])
if err != nil {
return 0, err
}
sizes[0] = n
return 1, nil
}
// WriteBatch writes packets individually (no batching for FreeBSD)
func (t *tun) WriteBatch(bufs [][]byte, offset int) (int, error) {
for i, buf := range bufs {
_, err := t.Write(buf[offset:])
if err != nil {
return i, err
}
}
return len(bufs), nil
}
// BatchSize returns 1 for FreeBSD (no batching)
func (t *tun) BatchSize() int {
return 1
}
func (t *tun) addRoutes(logErrors bool) error { func (t *tun) addRoutes(logErrors bool) error {
routes := *t.Routes.Load() routes := *t.Routes.Load()
for _, r := range routes { for _, r := range routes {
+1 -24
View File
@@ -155,29 +155,6 @@ func (t *tun) SupportsMultiqueue() bool {
return false return false
} }
func (t *tun) NewMultiQueueReader() (BatchReadWriter, error) { func (t *tun) NewMultiQueueReader() (io.ReadWriteCloser, error) {
return nil, fmt.Errorf("TODO: multiqueue not implemented for ios") return nil, fmt.Errorf("TODO: multiqueue not implemented for ios")
} }
func (t *tun) BatchRead(bufs [][]byte, sizes []int) (int, error) {
n, err := t.Read(bufs[0])
if err != nil {
return 0, err
}
sizes[0] = n
return 1, nil
}
func (t *tun) WriteBatch(bufs [][]byte, offset int) (int, error) {
for i, buf := range bufs {
_, err := t.Write(buf[offset:])
if err != nil {
return i, err
}
}
return len(bufs), nil
}
func (t *tun) BatchSize() int {
return 1
}
+32 -227
View File
@@ -9,6 +9,7 @@ import (
"net" "net"
"net/netip" "net/netip"
"os" "os"
"strings"
"sync" "sync"
"sync/atomic" "sync/atomic"
"time" "time"
@@ -21,12 +22,10 @@ import (
"github.com/slackhq/nebula/util" "github.com/slackhq/nebula/util"
"github.com/vishvananda/netlink" "github.com/vishvananda/netlink"
"golang.org/x/sys/unix" "golang.org/x/sys/unix"
wgtun "golang.zx2c4.com/wireguard/tun"
) )
type tun struct { type tun struct {
io.ReadWriteCloser io.ReadWriteCloser
wgDevice wgtun.Device
fd int fd int
Device string Device string
vpnNetworks []netip.Prefix vpnNetworks []netip.Prefix
@@ -72,154 +71,63 @@ type ifreqQLEN struct {
pad [8]byte pad [8]byte
} }
// wgDeviceWrapper wraps a wireguard Device to implement io.ReadWriteCloser
// This allows multiqueue readers to use the same wireguard Device batching as the main device
type wgDeviceWrapper struct {
dev wgtun.Device
buf []byte // Reusable buffer for single packet reads
}
func (w *wgDeviceWrapper) Read(b []byte) (int, error) {
// Use wireguard Device's batch API for single packet
bufs := [][]byte{b}
sizes := make([]int, 1)
n, err := w.dev.Read(bufs, sizes, 0)
if err != nil {
return 0, err
}
if n == 0 {
return 0, io.EOF
}
return sizes[0], nil
}
func (w *wgDeviceWrapper) Write(b []byte) (int, error) {
// Buffer b should have virtio header space (10 bytes) at the beginning
// The decrypted packet data starts at offset 10
// Pass the full buffer to WireGuard with offset=virtioNetHdrLen
bufs := [][]byte{b}
n, err := w.dev.Write(bufs, VirtioNetHdrLen)
if err != nil {
return 0, err
}
if n == 0 {
return 0, io.ErrShortWrite
}
return len(b), nil
}
func (w *wgDeviceWrapper) WriteBatch(bufs [][]byte, offset int) (int, error) {
// Pass all buffers to WireGuard's batch write
return w.dev.Write(bufs, offset)
}
func (w *wgDeviceWrapper) Close() error {
return w.dev.Close()
}
// BatchRead implements batching for multiqueue readers
func (w *wgDeviceWrapper) BatchRead(bufs [][]byte, sizes []int) (int, error) {
// The zero here is offset.
return w.dev.Read(bufs, sizes, 0)
}
// BatchSize returns the optimal batch size
func (w *wgDeviceWrapper) BatchSize() int {
return w.dev.BatchSize()
}
func newTunFromFd(c *config.C, l *logrus.Logger, deviceFd int, vpnNetworks []netip.Prefix) (*tun, error) { func newTunFromFd(c *config.C, l *logrus.Logger, deviceFd int, vpnNetworks []netip.Prefix) (*tun, error) {
wgDev, name, err := wgtun.CreateUnmonitoredTUNFromFD(deviceFd) file := os.NewFile(uintptr(deviceFd), "/dev/net/tun")
if err != nil {
return nil, fmt.Errorf("failed to create TUN from FD: %w", err)
}
file := wgDev.File()
t, err := newTunGeneric(c, l, file, vpnNetworks) t, err := newTunGeneric(c, l, file, vpnNetworks)
if err != nil { if err != nil {
_ = wgDev.Close()
return nil, err return nil, err
} }
t.wgDevice = wgDev t.Device = "tun0"
t.Device = name
return t, nil return t, nil
} }
func newTun(c *config.C, l *logrus.Logger, vpnNetworks []netip.Prefix, multiqueue bool) (*tun, error) { func newTun(c *config.C, l *logrus.Logger, vpnNetworks []netip.Prefix, multiqueue bool) (*tun, error) {
// Check if /dev/net/tun exists, create if needed (for docker containers)
if _, err := os.Stat("/dev/net/tun"); os.IsNotExist(err) {
if err := os.MkdirAll("/dev/net", 0755); err != nil {
return nil, fmt.Errorf("/dev/net/tun doesn't exist, failed to mkdir -p /dev/net: %w", err)
}
if err := unix.Mknod("/dev/net/tun", unix.S_IFCHR|0600, int(unix.Mkdev(10, 200))); err != nil {
return nil, fmt.Errorf("failed to create /dev/net/tun: %w", err)
}
}
devName := c.GetString("tun.dev", "")
mtu := c.GetInt("tun.mtu", DefaultMTU)
// Create TUN device manually to support multiqueue
fd, err := unix.Open("/dev/net/tun", os.O_RDWR, 0) fd, err := unix.Open("/dev/net/tun", os.O_RDWR, 0)
if err != nil { if err != nil {
return nil, err // If /dev/net/tun doesn't exist, try to create it (will happen in docker)
if os.IsNotExist(err) {
err = os.MkdirAll("/dev/net", 0755)
if err != nil {
return nil, fmt.Errorf("/dev/net/tun doesn't exist, failed to mkdir -p /dev/net: %w", err)
}
err = unix.Mknod("/dev/net/tun", unix.S_IFCHR|0600, int(unix.Mkdev(10, 200)))
if err != nil {
return nil, fmt.Errorf("failed to create /dev/net/tun: %w", err)
}
fd, err = unix.Open("/dev/net/tun", os.O_RDWR, 0)
if err != nil {
return nil, fmt.Errorf("created /dev/net/tun, but still failed: %w", err)
}
} else {
return nil, err
}
} }
var req ifReq var req ifReq
req.Flags = uint16(unix.IFF_TUN | unix.IFF_NO_PI | unix.IFF_VNET_HDR) req.Flags = uint16(unix.IFF_TUN | unix.IFF_NO_PI)
if multiqueue { if multiqueue {
req.Flags |= unix.IFF_MULTI_QUEUE req.Flags |= unix.IFF_MULTI_QUEUE
} }
copy(req.Name[:], devName) nameStr := c.GetString("tun.dev", "")
copy(req.Name[:], nameStr)
if err = ioctl(uintptr(fd), uintptr(unix.TUNSETIFF), uintptr(unsafe.Pointer(&req))); err != nil { if err = ioctl(uintptr(fd), uintptr(unix.TUNSETIFF), uintptr(unsafe.Pointer(&req))); err != nil {
unix.Close(fd) return nil, &NameError{
return nil, err Name: nameStr,
} Underlying: err,
}
// Set nonblocking
if err = unix.SetNonblock(fd, true); err != nil {
unix.Close(fd)
return nil, err
}
// Enable TCP and UDP offload (TSO/GRO) for performance
// This allows the kernel to handle segmentation/coalescing
const (
tunTCPOffloads = unix.TUN_F_CSUM | unix.TUN_F_TSO4 | unix.TUN_F_TSO6
tunUDPOffloads = unix.TUN_F_USO4 | unix.TUN_F_USO6
)
offloads := tunTCPOffloads | tunUDPOffloads
if err = unix.IoctlSetInt(fd, unix.TUNSETOFFLOAD, offloads); err != nil {
// Log warning but don't fail - offload is optional
l.WithError(err).Warn("Failed to enable TUN offload (TSO/GRO), performance may be reduced")
} }
name := strings.Trim(string(req.Name[:]), "\x00")
file := os.NewFile(uintptr(fd), "/dev/net/tun") file := os.NewFile(uintptr(fd), "/dev/net/tun")
// Create wireguard device from file descriptor
wgDev, err := wgtun.CreateTUNFromFile(file, mtu)
if err != nil {
file.Close()
return nil, fmt.Errorf("failed to create TUN from file: %w", err)
}
name, err := wgDev.Name()
if err != nil {
_ = wgDev.Close()
return nil, fmt.Errorf("failed to get TUN device name: %w", err)
}
// file is now owned by wgDev, get a new reference
file = wgDev.File()
t, err := newTunGeneric(c, l, file, vpnNetworks) t, err := newTunGeneric(c, l, file, vpnNetworks)
if err != nil { if err != nil {
_ = wgDev.Close()
return nil, err return nil, err
} }
t.wgDevice = wgDev
t.Device = name t.Device = name
return t, nil return t, nil
@@ -330,44 +238,22 @@ func (t *tun) SupportsMultiqueue() bool {
return true return true
} }
func (t *tun) NewMultiQueueReader() (BatchReadWriter, error) { func (t *tun) NewMultiQueueReader() (io.ReadWriteCloser, error) {
fd, err := unix.Open("/dev/net/tun", os.O_RDWR, 0) fd, err := unix.Open("/dev/net/tun", os.O_RDWR, 0)
if err != nil { if err != nil {
return nil, err return nil, err
} }
var req ifReq var req ifReq
// MUST match the flags used in newTun - includes IFF_VNET_HDR req.Flags = uint16(unix.IFF_TUN | unix.IFF_NO_PI | unix.IFF_MULTI_QUEUE)
req.Flags = uint16(unix.IFF_TUN | unix.IFF_NO_PI | unix.IFF_VNET_HDR | unix.IFF_MULTI_QUEUE)
copy(req.Name[:], t.Device) copy(req.Name[:], t.Device)
if err = ioctl(uintptr(fd), uintptr(unix.TUNSETIFF), uintptr(unsafe.Pointer(&req))); err != nil { if err = ioctl(uintptr(fd), uintptr(unix.TUNSETIFF), uintptr(unsafe.Pointer(&req))); err != nil {
unix.Close(fd)
return nil, err return nil, err
} }
// Set nonblocking mode - CRITICAL for proper netpoller integration
if err = unix.SetNonblock(fd, true); err != nil {
unix.Close(fd)
return nil, err
}
// Get MTU from main device
mtu := t.MaxMTU
if mtu == 0 {
mtu = DefaultMTU
}
file := os.NewFile(uintptr(fd), "/dev/net/tun") file := os.NewFile(uintptr(fd), "/dev/net/tun")
// Create wireguard Device from the file descriptor (just like the main device) return file, nil
wgDev, err := wgtun.CreateTUNFromFile(file, mtu)
if err != nil {
file.Close()
return nil, fmt.Errorf("failed to create multiqueue TUN device: %w", err)
}
// Return a wrapper that uses the wireguard Device for all I/O
return &wgDeviceWrapper{dev: wgDev}, nil
} }
func (t *tun) RoutesFor(ip netip.Addr) routing.Gateways { func (t *tun) RoutesFor(ip netip.Addr) routing.Gateways {
@@ -375,68 +261,7 @@ func (t *tun) RoutesFor(ip netip.Addr) routing.Gateways {
return r return r
} }
func (t *tun) Read(b []byte) (int, error) {
if t.wgDevice != nil {
// Use wireguard device which handles virtio headers internally
bufs := [][]byte{b}
sizes := make([]int, 1)
n, err := t.wgDevice.Read(bufs, sizes, 0)
if err != nil {
return 0, err
}
if n == 0 {
return 0, io.EOF
}
return sizes[0], nil
}
// Fallback: direct read from file (shouldn't happen in normal operation)
return t.ReadWriteCloser.Read(b)
}
// BatchRead reads multiple packets at once for improved performance
// bufs: slice of buffers to read into
// sizes: slice that will be filled with packet sizes
// Returns number of packets read
func (t *tun) BatchRead(bufs [][]byte, sizes []int) (int, error) {
if t.wgDevice != nil {
return t.wgDevice.Read(bufs, sizes, 0)
}
// Fallback: single packet read
n, err := t.ReadWriteCloser.Read(bufs[0])
if err != nil {
return 0, err
}
sizes[0] = n
return 1, nil
}
// BatchSize returns the optimal number of packets to read/write in a batch
func (t *tun) BatchSize() int {
if t.wgDevice != nil {
return t.wgDevice.BatchSize()
}
return 1
}
func (t *tun) Write(b []byte) (int, error) { func (t *tun) Write(b []byte) (int, error) {
if t.wgDevice != nil {
// Buffer b should have virtio header space (10 bytes) at the beginning
// The decrypted packet data starts at offset 10
// Pass the full buffer to WireGuard with offset=virtioNetHdrLen
bufs := [][]byte{b}
n, err := t.wgDevice.Write(bufs, VirtioNetHdrLen)
if err != nil {
return 0, err
}
if n == 0 {
return 0, io.ErrShortWrite
}
return len(b), nil
}
// Fallback: direct write (shouldn't happen in normal operation)
var nn int var nn int
maximum := len(b) maximum := len(b)
@@ -459,22 +284,6 @@ func (t *tun) Write(b []byte) (int, error) {
} }
} }
// WriteBatch writes multiple packets to the TUN device in a single syscall
func (t *tun) WriteBatch(bufs [][]byte, offset int) (int, error) {
if t.wgDevice != nil {
return t.wgDevice.Write(bufs, offset)
}
// Fallback: write individually (shouldn't happen in normal operation)
for i, buf := range bufs {
_, err := t.Write(buf)
if err != nil {
return i, err
}
}
return len(bufs), nil
}
func (t *tun) deviceBytes() (o [16]byte) { func (t *tun) deviceBytes() (o [16]byte) {
for i, c := range t.Device { for i, c := range t.Device {
o[i] = byte(c) o[i] = byte(c)
@@ -902,10 +711,6 @@ func (t *tun) Close() error {
close(t.routeChan) close(t.routeChan)
} }
if t.wgDevice != nil {
_ = t.wgDevice.Close()
}
if t.ReadWriteCloser != nil { if t.ReadWriteCloser != nil {
_ = t.ReadWriteCloser.Close() _ = t.ReadWriteCloser.Close()
} }
+2 -24
View File
@@ -6,6 +6,7 @@ package overlay
import ( import (
"errors" "errors"
"fmt" "fmt"
"io"
"net/netip" "net/netip"
"os" "os"
"regexp" "regexp"
@@ -393,33 +394,10 @@ func (t *tun) SupportsMultiqueue() bool {
return false return false
} }
func (t *tun) NewMultiQueueReader() (BatchReadWriter, error) { func (t *tun) NewMultiQueueReader() (io.ReadWriteCloser, error) {
return nil, fmt.Errorf("TODO: multiqueue not implemented for netbsd") return nil, fmt.Errorf("TODO: multiqueue not implemented for netbsd")
} }
func (t *tun) BatchRead(bufs [][]byte, sizes []int) (int, error) {
n, err := t.Read(bufs[0])
if err != nil {
return 0, err
}
sizes[0] = n
return 1, nil
}
func (t *tun) WriteBatch(bufs [][]byte, offset int) (int, error) {
for i, buf := range bufs {
_, err := t.Write(buf[offset:])
if err != nil {
return i, err
}
}
return len(bufs), nil
}
func (t *tun) BatchSize() int {
return 1
}
func (t *tun) addRoutes(logErrors bool) error { func (t *tun) addRoutes(logErrors bool) error {
routes := *t.Routes.Load() routes := *t.Routes.Load()
+2 -24
View File
@@ -6,6 +6,7 @@ package overlay
import ( import (
"errors" "errors"
"fmt" "fmt"
"io"
"net/netip" "net/netip"
"os" "os"
"regexp" "regexp"
@@ -313,33 +314,10 @@ func (t *tun) SupportsMultiqueue() bool {
return false return false
} }
func (t *tun) NewMultiQueueReader() (BatchReadWriter, error) { func (t *tun) NewMultiQueueReader() (io.ReadWriteCloser, error) {
return nil, fmt.Errorf("TODO: multiqueue not implemented for openbsd") return nil, fmt.Errorf("TODO: multiqueue not implemented for openbsd")
} }
func (t *tun) BatchRead(bufs [][]byte, sizes []int) (int, error) {
n, err := t.Read(bufs[0])
if err != nil {
return 0, err
}
sizes[0] = n
return 1, nil
}
func (t *tun) WriteBatch(bufs [][]byte, offset int) (int, error) {
for i, buf := range bufs {
_, err := t.Write(buf[offset:])
if err != nil {
return i, err
}
}
return len(bufs), nil
}
func (t *tun) BatchSize() int {
return 1
}
func (t *tun) addRoutes(logErrors bool) error { func (t *tun) addRoutes(logErrors bool) error {
routes := *t.Routes.Load() routes := *t.Routes.Load()
+3 -35
View File
@@ -109,12 +109,8 @@ func (t *TestTun) Write(b []byte) (n int, err error) {
return 0, io.ErrClosedPipe return 0, io.ErrClosedPipe
} }
// Skip virtio header (consistent with production Linux tun) packet := make([]byte, len(b), len(b))
// The buffer b has VirtioNetHdrLen bytes of header followed by the actual packet copy(packet, b)
data := b[VirtioNetHdrLen:]
packet := make([]byte, len(data))
copy(packet, data)
t.TxPackets <- packet t.TxPackets <- packet
return len(b), nil return len(b), nil
} }
@@ -140,34 +136,6 @@ func (t *TestTun) SupportsMultiqueue() bool {
return false return false
} }
func (t *TestTun) NewMultiQueueReader() (BatchReadWriter, error) { func (t *TestTun) NewMultiQueueReader() (io.ReadWriteCloser, error) {
return nil, fmt.Errorf("TODO: multiqueue not implemented") return nil, fmt.Errorf("TODO: multiqueue not implemented")
} }
func (t *TestTun) BatchRead(bufs [][]byte, sizes []int) (int, error) {
n, err := t.Read(bufs[0])
if err != nil {
return 0, err
}
sizes[0] = n
return 1, nil
}
func (t *TestTun) WriteBatch(bufs [][]byte, offset int) (int, error) {
if t.closed.Load() {
return 0, io.ErrClosedPipe
}
for _, buf := range bufs {
// Strip the header at offset and send directly to channel
data := buf[offset:]
packet := make([]byte, len(data))
copy(packet, data)
t.TxPackets <- packet
}
return len(bufs), nil
}
func (t *TestTun) BatchSize() int {
return 1
}
+2 -27
View File
@@ -6,6 +6,7 @@ package overlay
import ( import (
"crypto" "crypto"
"fmt" "fmt"
"io"
"net/netip" "net/netip"
"os" "os"
"path/filepath" "path/filepath"
@@ -240,36 +241,10 @@ func (t *winTun) SupportsMultiqueue() bool {
return false return false
} }
func (t *winTun) NewMultiQueueReader() (BatchReadWriter, error) { func (t *winTun) NewMultiQueueReader() (io.ReadWriteCloser, error) {
return nil, fmt.Errorf("TODO: multiqueue not implemented for windows") return nil, fmt.Errorf("TODO: multiqueue not implemented for windows")
} }
// BatchRead reads a single packet (batch size 1 for Windows)
func (t *winTun) BatchRead(bufs [][]byte, sizes []int) (int, error) {
n, err := t.Read(bufs[0])
if err != nil {
return 0, err
}
sizes[0] = n
return 1, nil
}
// WriteBatch writes packets individually (no batching for Windows)
func (t *winTun) WriteBatch(bufs [][]byte, offset int) (int, error) {
for i, buf := range bufs {
_, err := t.Write(buf[offset:])
if err != nil {
return i, err
}
}
return len(bufs), nil
}
// BatchSize returns 1 for Windows (no batching)
func (t *winTun) BatchSize() int {
return 1
}
func (t *winTun) Close() error { func (t *winTun) Close() error {
// It seems that the Windows networking stack doesn't like it when we destroy interfaces that have active routes, // It seems that the Windows networking stack doesn't like it when we destroy interfaces that have active routes,
// so to be certain, just remove everything before destroying. // so to be certain, just remove everything before destroying.
+1 -27
View File
@@ -50,36 +50,10 @@ func (d *UserDevice) SupportsMultiqueue() bool {
return true return true
} }
func (d *UserDevice) NewMultiQueueReader() (BatchReadWriter, error) { func (d *UserDevice) NewMultiQueueReader() (io.ReadWriteCloser, error) {
return d, nil return d, nil
} }
// BatchRead reads a single packet (batch size 1 for UserDevice)
func (d *UserDevice) BatchRead(bufs [][]byte, sizes []int) (int, error) {
n, err := d.Read(bufs[0])
if err != nil {
return 0, err
}
sizes[0] = n
return 1, nil
}
// WriteBatch writes packets individually (no batching for UserDevice)
func (d *UserDevice) WriteBatch(bufs [][]byte, offset int) (int, error) {
for i, buf := range bufs {
_, err := d.Write(buf[offset:])
if err != nil {
return i, err
}
}
return len(bufs), nil
}
// BatchSize returns 1 for UserDevice (no batching)
func (d *UserDevice) BatchSize() int {
return 1
}
func (d *UserDevice) Pipe() (*io.PipeReader, *io.PipeWriter) { func (d *UserDevice) Pipe() (*io.PipeReader, *io.PipeWriter) {
return d.inboundReader, d.outboundWriter return d.inboundReader, d.outboundWriter
} }
+1 -1
View File
@@ -55,7 +55,7 @@ func (rm *relayManager) setAmRelay(v bool) {
func AddRelay(l *logrus.Logger, relayHostInfo *HostInfo, hm *HostMap, vpnIp netip.Addr, remoteIdx *uint32, relayType int, state int) (uint32, error) { func AddRelay(l *logrus.Logger, relayHostInfo *HostInfo, hm *HostMap, vpnIp netip.Addr, remoteIdx *uint32, relayType int, state int) (uint32, error) {
hm.Lock() hm.Lock()
defer hm.Unlock() defer hm.Unlock()
for i := 0; i < 32; i++ { for range 32 {
index, err := generateIndex(l) index, err := generateIndex(l)
if err != nil { if err != nil {
return 0, err return 0, err
+1 -6
View File
@@ -404,12 +404,7 @@ func (r *RemoteList) Rebuild(preferredRanges []netip.Prefix) {
// unlockedIsBad assumes you have the write lock and checks if the remote matches any entry in the blocked address list // unlockedIsBad assumes you have the write lock and checks if the remote matches any entry in the blocked address list
func (r *RemoteList) unlockedIsBad(remote netip.AddrPort) bool { func (r *RemoteList) unlockedIsBad(remote netip.AddrPort) bool {
for _, v := range r.badRemotes { return slices.Contains(r.badRemotes, remote)
if v == remote {
return true
}
}
return false
} }
// unlockedSetLearnedV4 assumes you have the write lock and sets the current learned address for this owner and marks the // unlockedSetLearnedV4 assumes you have the write lock and sets the current learned address for this owner and marks the
+2 -3
View File
@@ -6,6 +6,7 @@ import (
"errors" "errors"
"flag" "flag"
"fmt" "fmt"
"maps"
"net" "net"
"net/netip" "net/netip"
"os" "os"
@@ -831,9 +832,7 @@ func sshPrintRelays(ifce *Interface, fs any, a []string, w sshd.StringWriter) er
relays := map[uint32]*HostInfo{} relays := map[uint32]*HostInfo{}
ifce.hostMap.Lock() ifce.hostMap.Lock()
for k, v := range ifce.hostMap.Relays { maps.Copy(relays, ifce.hostMap.Relays)
relays[k] = v
}
ifce.hostMap.Unlock() ifce.hostMap.Unlock()
type RelayFor struct { type RelayFor struct {
-1
View File
@@ -6,7 +6,6 @@ import (
"log" "log"
"net" "net"
"net/http" "net/http"
_ "net/http/pprof"
"runtime" "runtime"
"strconv" "strconv"
"time" "time"
+2 -18
View File
@@ -1,11 +1,10 @@
package device package test
import ( import (
"errors" "errors"
"io" "io"
"net/netip" "net/netip"
"github.com/slackhq/nebula/overlay"
"github.com/slackhq/nebula/routing" "github.com/slackhq/nebula/routing"
) )
@@ -39,25 +38,10 @@ func (NoopTun) SupportsMultiqueue() bool {
return false return false
} }
func (NoopTun) NewMultiQueueReader() (overlay.BatchReadWriter, error) { func (NoopTun) NewMultiQueueReader() (io.ReadWriteCloser, error) {
return nil, errors.New("unsupported") return nil, errors.New("unsupported")
} }
func (NoopTun) Close() error { func (NoopTun) Close() error {
return nil return nil
} }
// BatchRead implements BatchReadWriter interface
func (NoopTun) BatchRead(bufs [][]byte, sizes []int) (int, error) {
return 0, io.EOF
}
// WriteBatch implements BatchReadWriter interface
func (NoopTun) WriteBatch(bufs [][]byte, offset int) (int, error) {
return len(bufs), nil
}
// BatchSize implements BatchReadWriter interface
func (NoopTun) BatchSize() int {
return 1
}
+2 -2
View File
@@ -134,7 +134,7 @@ func TestTimerWheel_Purge(t *testing.T) {
assert.True(t, tw.lastTick.After(lastTick)) assert.True(t, tw.lastTick.After(lastTick))
// Make sure we get all 4 packets back // Make sure we get all 4 packets back
for i := 0; i < 4; i++ { for i := range 4 {
p, has := tw.Purge() p, has := tw.Purge()
assert.True(t, has) assert.True(t, has)
assert.Equal(t, fps[i], p) assert.Equal(t, fps[i], p)
@@ -149,7 +149,7 @@ func TestTimerWheel_Purge(t *testing.T) {
// Make sure we cached the free'd items // Make sure we cached the free'd items
assert.Equal(t, 4, tw.itemsCached) assert.Equal(t, 4, tw.itemsCached)
ci := tw.itemCache ci := tw.itemCache
for i := 0; i < 4; i++ { for range 4 {
assert.NotNil(t, ci) assert.NotNil(t, ci)
ci = ci.Next ci = ci.Next
} }
+6 -22
View File
@@ -13,21 +13,13 @@ type EncReader func(
payload []byte, payload []byte,
) )
type EncBatchReader func(
addrs []netip.AddrPort,
payloads [][]byte,
count int,
)
type Conn interface { type Conn interface {
Rebind() error Rebind() error
LocalAddr() (netip.AddrPort, error) LocalAddr() (netip.AddrPort, error)
ListenOutBatch(r EncBatchReader) ListenOut(r EncReader)
WriteTo(b []byte, addr netip.AddrPort) error WriteTo(b []byte, addr netip.AddrPort) error
WriteMulti(packets [][]byte, addrs []netip.AddrPort) (int, error)
ReloadConfig(c *config.C) ReloadConfig(c *config.C)
SupportsMultipleReaders() bool SupportsMultipleReaders() bool
BatchSize() int
Close() error Close() error
} }
@@ -39,25 +31,17 @@ func (NoopConn) Rebind() error {
func (NoopConn) LocalAddr() (netip.AddrPort, error) { func (NoopConn) LocalAddr() (netip.AddrPort, error) {
return netip.AddrPort{}, nil return netip.AddrPort{}, nil
} }
func (NoopConn) ListenOut(_ EncReader) {
func (NoopConn) ListenOut(_ EncReader) {} return
}
func (NoopConn) SupportsMultipleReaders() bool { func (NoopConn) SupportsMultipleReaders() bool {
return false return false
} }
func (NoopConn) ListenOutBatch(_ EncBatchReader) {}
func (NoopConn) WriteTo(_ []byte, _ netip.AddrPort) error { func (NoopConn) WriteTo(_ []byte, _ netip.AddrPort) error {
return nil return nil
} }
func (NoopConn) WriteMulti(_ [][]byte, _ []netip.AddrPort) (int, error) { func (NoopConn) ReloadConfig(_ *config.C) {
return 0, nil return
}
func (NoopConn) ReloadConfig(_ *config.C) {}
func (NoopConn) BatchSize() int {
return 1
} }
func (NoopConn) Close() error { func (NoopConn) Close() error {
return nil return nil
-39
View File
@@ -140,17 +140,6 @@ func (u *StdConn) WriteTo(b []byte, ap netip.AddrPort) error {
} }
} }
// WriteMulti sends multiple packets - fallback implementation without sendmmsg
func (u *StdConn) WriteMulti(packets [][]byte, addrs []netip.AddrPort) (int, error) {
for i := range packets {
err := u.WriteTo(packets[i], addrs[i])
if err != nil {
return i, err
}
}
return len(packets), nil
}
func (u *StdConn) LocalAddr() (netip.AddrPort, error) { func (u *StdConn) LocalAddr() (netip.AddrPort, error) {
a := u.UDPConn.LocalAddr() a := u.UDPConn.LocalAddr()
@@ -199,34 +188,6 @@ func (u *StdConn) SupportsMultipleReaders() bool {
return false return false
} }
// ListenOutBatch - fallback to single-packet reads for Darwin
func (u *StdConn) ListenOutBatch(r EncBatchReader) {
buffer := make([]byte, MTU)
addrs := make([]netip.AddrPort, 1)
payloads := make([][]byte, 1)
for {
// Just read one packet at a time and call batch callback with count=1
n, rua, err := u.ReadFromUDPAddrPort(buffer)
if err != nil {
if errors.Is(err, net.ErrClosed) {
u.l.WithError(err).Debug("udp socket is closed, exiting read loop")
return
}
u.l.WithError(err).Error("unexpected udp socket receive error")
}
addrs[0] = netip.AddrPortFrom(rua.Addr().Unmap(), rua.Port())
payloads[0] = buffer[:n]
r(addrs, payloads, 1)
}
}
func (u *StdConn) BatchSize() int {
return 1
}
func (u *StdConn) Rebind() error { func (u *StdConn) Rebind() error {
var err error var err error
if u.isV4 { if u.isV4 {
-35
View File
@@ -101,38 +101,3 @@ func (u *GenericConn) ListenOut(r EncReader) {
func (u *GenericConn) SupportsMultipleReaders() bool { func (u *GenericConn) SupportsMultipleReaders() bool {
return false return false
} }
// ListenOutBatch - fallback to single-packet reads for generic platforms
func (u *GenericConn) ListenOutBatch(r EncBatchReader) {
buffer := make([]byte, MTU)
addrs := make([]netip.AddrPort, 1)
payloads := make([][]byte, 1)
for {
// Just read one packet at a time and call batch callback with count=1
n, rua, err := u.ReadFromUDPAddrPort(buffer)
if err != nil {
u.l.WithError(err).Debug("udp socket is closed, exiting read loop")
return
}
addrs[0] = netip.AddrPortFrom(rua.Addr().Unmap(), rua.Port())
payloads[0] = buffer[:n]
r(addrs, payloads, 1)
}
}
// WriteMulti sends multiple packets - fallback implementation
func (u *GenericConn) WriteMulti(packets [][]byte, addrs []netip.AddrPort) (int, error) {
for i := range packets {
err := u.WriteTo(packets[i], addrs[i])
if err != nil {
return i, err
}
}
return len(packets), nil
}
func (u *GenericConn) BatchSize() int {
return 1
}
+165 -280
View File
@@ -4,6 +4,7 @@
package udp package udp
import ( import (
"context"
"encoding/binary" "encoding/binary"
"fmt" "fmt"
"net" "net"
@@ -18,82 +19,66 @@ import (
) )
type StdConn struct { type StdConn struct {
sysFd int udpConn *net.UDPConn
isV4 bool rawConn syscall.RawConn
l *logrus.Logger sockFd int
batch int isV4 bool
l *logrus.Logger
// Pre-allocated buffers for batch writes (sized for IPv6, works for both) batch int
writeMsgs []rawMessage
writeIovecs []iovec
writeNames [][]byte
} }
func maybeIPV4(ip net.IP) (net.IP, bool) { func setReusePort(network, address string, c syscall.RawConn) error {
ip4 := ip.To4() var opErr error
if ip4 != nil { err := c.Control(func(fd uintptr) {
return ip4, true opErr = unix.SetsockoptInt(int(fd), unix.SOL_SOCKET, unix.SO_REUSEPORT, 1)
//CloseOnExec already set by the runtime
})
if err != nil {
return err
} }
return ip, false return opErr
} }
func NewListener(l *logrus.Logger, ip netip.Addr, port int, multi bool, batch int) (Conn, error) { func NewListener(l *logrus.Logger, ip netip.Addr, port int, multi bool, batch int) (Conn, error) {
af := unix.AF_INET6 listen := netip.AddrPortFrom(ip, uint16(port))
if ip.Is4() { lc := net.ListenConfig{}
af = unix.AF_INET if multi {
lc.Control = setReusePort
} }
syscall.ForkLock.RLock() network := "udp"
fd, err := unix.Socket(af, unix.SOCK_DGRAM, unix.IPPROTO_UDP) if ip.Is4() { //todo I am not a fan of this, would prefer a default where we always bind dual-stack unless given a non-wildcard v4 addr
if err == nil { network = "udp4"
unix.CloseOnExec(fd)
} }
syscall.ForkLock.RUnlock() //this context is only used during the bind operation, you can't cancel it to kill the socket
pc, err := lc.ListenPacket(context.Background(), network, listen.String())
if err != nil { if err != nil {
unix.Close(fd)
return nil, fmt.Errorf("unable to open socket: %s", err) return nil, fmt.Errorf("unable to open socket: %s", err)
} }
udpConn := pc.(*net.UDPConn)
if multi { rawConn, err := udpConn.SyscallConn()
if err = unix.SetsockoptInt(fd, unix.SOL_SOCKET, unix.SO_REUSEPORT, 1); err != nil { if err != nil {
return nil, fmt.Errorf("unable to set SO_REUSEPORT: %s", err) _ = udpConn.Close()
} return nil, err
} }
var sa unix.Sockaddr //steal the socket's fd for sending
if ip.Is4() { var sockFd uintptr
sa4 := &unix.SockaddrInet4{Port: port} err = rawConn.Control(func(fd uintptr) {
sa4.Addr = ip.As4() sockFd = fd
sa = sa4 })
} else { if err != nil {
sa6 := &unix.SockaddrInet6{Port: port} _ = udpConn.Close()
sa6.Addr = ip.As16() return nil, err
sa = sa6
}
if err = unix.Bind(fd, sa); err != nil {
return nil, fmt.Errorf("unable to bind to socket: %s", err)
} }
c := &StdConn{sysFd: fd, isV4: ip.Is4(), l: l, batch: batch} return &StdConn{
udpConn: udpConn,
// Pre-allocate write message structures for batching (sized for IPv6, works for both) rawConn: rawConn,
c.writeMsgs = make([]rawMessage, batch) sockFd: int(sockFd),
c.writeIovecs = make([]iovec, batch) isV4: ip.Is4(),
c.writeNames = make([][]byte, batch) l: l,
batch: batch,
for i := range c.writeMsgs { }, err
// Allocate for IPv6 size (larger than IPv4, works for both)
c.writeNames[i] = make([]byte, unix.SizeofSockaddrInet6)
// Point to the iovec in the slice
c.writeMsgs[i].Hdr.Iov = &c.writeIovecs[i]
c.writeMsgs[i].Hdr.Iovlen = 1
c.writeMsgs[i].Hdr.Name = &c.writeNames[i][0]
// Namelen will be set appropriately in writeMulti4/writeMulti6
}
return c, err
} }
func (u *StdConn) SupportsMultipleReaders() bool { func (u *StdConn) SupportsMultipleReaders() bool {
@@ -104,126 +89,147 @@ func (u *StdConn) Rebind() error {
return nil return nil
} }
func (u *StdConn) getSockOptInt(opt int) (int, error) {
if u.rawConn == nil {
return 0, fmt.Errorf("no UDP connection")
}
var out int
var opErr error
err := u.rawConn.Control(func(fd uintptr) {
out, opErr = unix.GetsockoptInt(int(fd), unix.SOL_SOCKET, opt)
})
if err != nil {
return 0, err
}
return out, opErr
}
func (u *StdConn) setSockOptInt(opt int, n int) error {
if u.rawConn == nil {
return fmt.Errorf("no UDP connection")
}
var opErr error
err := u.rawConn.Control(func(fd uintptr) {
opErr = unix.SetsockoptInt(int(fd), unix.SOL_SOCKET, opt, n)
})
if err != nil {
return err
}
return opErr
}
func (u *StdConn) SetRecvBuffer(n int) error { func (u *StdConn) SetRecvBuffer(n int) error {
return unix.SetsockoptInt(u.sysFd, unix.SOL_SOCKET, unix.SO_RCVBUFFORCE, n) return u.setSockOptInt(unix.SO_RCVBUFFORCE, n)
} }
func (u *StdConn) SetSendBuffer(n int) error { func (u *StdConn) SetSendBuffer(n int) error {
return unix.SetsockoptInt(u.sysFd, unix.SOL_SOCKET, unix.SO_SNDBUFFORCE, n) return u.setSockOptInt(unix.SO_SNDBUFFORCE, n)
} }
func (u *StdConn) SetSoMark(mark int) error { func (u *StdConn) SetSoMark(mark int) error {
return unix.SetsockoptInt(u.sysFd, unix.SOL_SOCKET, unix.SO_MARK, mark) return u.setSockOptInt(unix.SO_MARK, mark)
} }
func (u *StdConn) GetRecvBuffer() (int, error) { func (u *StdConn) GetRecvBuffer() (int, error) {
return unix.GetsockoptInt(int(u.sysFd), unix.SOL_SOCKET, unix.SO_RCVBUF) return u.getSockOptInt(unix.SO_RCVBUF)
} }
func (u *StdConn) GetSendBuffer() (int, error) { func (u *StdConn) GetSendBuffer() (int, error) {
return unix.GetsockoptInt(int(u.sysFd), unix.SOL_SOCKET, unix.SO_SNDBUF) return u.getSockOptInt(unix.SO_SNDBUF)
} }
func (u *StdConn) GetSoMark() (int, error) { func (u *StdConn) GetSoMark() (int, error) {
return unix.GetsockoptInt(int(u.sysFd), unix.SOL_SOCKET, unix.SO_MARK) return u.getSockOptInt(unix.SO_MARK)
} }
func (u *StdConn) LocalAddr() (netip.AddrPort, error) { func (u *StdConn) LocalAddr() (netip.AddrPort, error) {
sa, err := unix.Getsockname(u.sysFd) addr := u.udpConn.LocalAddr()
if err != nil { return netip.ParseAddrPort(addr.String())
return netip.AddrPort{}, err
}
switch sa := sa.(type) {
case *unix.SockaddrInet4:
return netip.AddrPortFrom(netip.AddrFrom4(sa.Addr), uint16(sa.Port)), nil
case *unix.SockaddrInet6:
return netip.AddrPortFrom(netip.AddrFrom16(sa.Addr), uint16(sa.Port)), nil
default:
return netip.AddrPort{}, fmt.Errorf("unsupported sock type: %T", sa)
}
} }
func (u *StdConn) ListenOutBatch(r EncBatchReader) { func recvmmsg(fd uintptr, msgs []rawMessage) (int, bool, error) {
var ip netip.Addr var errno syscall.Errno
n, _, errno := unix.Syscall6(
msgs, buffers, names := u.PrepareRawMessages(u.batch) unix.SYS_RECVMMSG,
read := u.ReadMulti fd,
if u.batch == 1 { uintptr(unsafe.Pointer(&msgs[0])),
read = u.ReadSingle uintptr(len(msgs)),
unix.MSG_WAITFORONE,
0,
0,
)
if errno == syscall.EAGAIN || errno == syscall.EWOULDBLOCK {
// No data available, block for I/O and try again.
return int(n), false, nil
} }
if errno != 0 {
return int(n), true, &net.OpError{Op: "recvmmsg", Err: errno}
}
return int(n), true, nil
}
udpBatchHist := metrics.GetOrRegisterHistogram("batch.udp_read_size", nil, metrics.NewUniformSample(1024)) func (u *StdConn) listenOutSingle(r EncReader) {
var err error
// Pre-allocate slices for batch callback var n int
addrs := make([]netip.AddrPort, u.batch) var from netip.AddrPort
payloads := make([][]byte, u.batch) buffer := make([]byte, MTU)
for { for {
n, err := read(msgs) n, from, err = u.udpConn.ReadFromUDPAddrPort(buffer)
if err != nil { if err != nil {
u.l.WithError(err).Debug("udp socket is closed, exiting read loop") u.l.WithError(err).Debug("udp socket is closed, exiting read loop")
return return
} }
from = netip.AddrPortFrom(from.Addr().Unmap(), from.Port())
r(from, buffer[:n])
}
}
udpBatchHist.Update(int64(n)) func (u *StdConn) listenOutBatch(r EncReader) {
var ip netip.Addr
var n int
var operr error
msgs, buffers, names := u.PrepareRawMessages(u.batch)
//reader needs to capture variables from this function, since it's used as a lambda with rawConn.Read
//defining it outside the loop so it gets re-used
reader := func(fd uintptr) (done bool) {
n, done, operr = recvmmsg(fd, msgs)
return done
}
for {
err := u.rawConn.Read(reader)
if err != nil {
u.l.WithError(err).Debug("udp socket is closed, exiting read loop")
return
}
if operr != nil {
u.l.WithError(operr).Debug("operr: udp socket is closed, exiting read loop")
return
}
// Prepare batch data
for i := 0; i < n; i++ { for i := 0; i < n; i++ {
// Its ok to skip the ok check here, the slicing is the only error that can occur and it will panic
if u.isV4 { if u.isV4 {
ip, _ = netip.AddrFromSlice(names[i][4:8]) ip, _ = netip.AddrFromSlice(names[i][4:8])
} else { } else {
ip, _ = netip.AddrFromSlice(names[i][8:24]) ip, _ = netip.AddrFromSlice(names[i][8:24])
} }
addrs[i] = netip.AddrPortFrom(ip.Unmap(), binary.BigEndian.Uint16(names[i][2:4])) r(netip.AddrPortFrom(ip.Unmap(), binary.BigEndian.Uint16(names[i][2:4])), buffers[i][:msgs[i].Len])
payloads[i] = buffers[i][:msgs[i].Len]
} }
// Call batch callback with all packets
r(addrs, payloads, n)
} }
} }
func (u *StdConn) ReadSingle(msgs []rawMessage) (int, error) { func (u *StdConn) ListenOut(r EncReader) {
for { if u.batch == 1 {
n, _, err := unix.Syscall6( //save some ram by not calling PrepareRawMessages for fields we won't use
unix.SYS_RECVMSG, //we could also make this path more common by calling recvmmsg with msgs[:1],
uintptr(u.sysFd), //but that's still the recvmmsg syscall, which would be a change
uintptr(unsafe.Pointer(&(msgs[0].Hdr))), u.listenOutSingle(r)
0, } else {
0, u.listenOutBatch(r)
0,
0,
)
if err != 0 {
return 0, &net.OpError{Op: "recvmsg", Err: err}
}
msgs[0].Len = uint32(n)
return 1, nil
}
}
func (u *StdConn) ReadMulti(msgs []rawMessage) (int, error) {
for {
n, _, err := unix.Syscall6(
unix.SYS_RECVMMSG,
uintptr(u.sysFd),
uintptr(unsafe.Pointer(&msgs[0])),
uintptr(len(msgs)),
unix.MSG_WAITFORONE,
0,
0,
)
if err != 0 {
return 0, &net.OpError{Op: "recvmmsg", Err: err}
}
return int(n), nil
} }
} }
@@ -234,19 +240,6 @@ func (u *StdConn) WriteTo(b []byte, ip netip.AddrPort) error {
return u.writeTo6(b, ip) return u.writeTo6(b, ip)
} }
func (u *StdConn) WriteMulti(packets [][]byte, addrs []netip.AddrPort) (int, error) {
if len(packets) != len(addrs) {
return 0, fmt.Errorf("packets and addrs length mismatch")
}
if len(packets) == 0 {
return 0, nil
}
if u.isV4 {
return u.writeMulti4(packets, addrs)
}
return u.writeMulti6(packets, addrs)
}
func (u *StdConn) writeTo6(b []byte, ip netip.AddrPort) error { func (u *StdConn) writeTo6(b []byte, ip netip.AddrPort) error {
var rsa unix.RawSockaddrInet6 var rsa unix.RawSockaddrInet6
rsa.Family = unix.AF_INET6 rsa.Family = unix.AF_INET6
@@ -256,7 +249,7 @@ func (u *StdConn) writeTo6(b []byte, ip netip.AddrPort) error {
for { for {
_, _, err := unix.Syscall6( _, _, err := unix.Syscall6(
unix.SYS_SENDTO, unix.SYS_SENDTO,
uintptr(u.sysFd), uintptr(u.sockFd),
uintptr(unsafe.Pointer(&b[0])), uintptr(unsafe.Pointer(&b[0])),
uintptr(len(b)), uintptr(len(b)),
uintptr(0), uintptr(0),
@@ -285,7 +278,7 @@ func (u *StdConn) writeTo4(b []byte, ip netip.AddrPort) error {
for { for {
_, _, err := unix.Syscall6( _, _, err := unix.Syscall6(
unix.SYS_SENDTO, unix.SYS_SENDTO,
uintptr(u.sysFd), uintptr(u.sockFd),
uintptr(unsafe.Pointer(&b[0])), uintptr(unsafe.Pointer(&b[0])),
uintptr(len(b)), uintptr(len(b)),
uintptr(0), uintptr(0),
@@ -301,123 +294,6 @@ func (u *StdConn) writeTo4(b []byte, ip netip.AddrPort) error {
} }
} }
func (u *StdConn) writeMulti4(packets [][]byte, addrs []netip.AddrPort) (int, error) {
sent := 0
for sent < len(packets) {
// Determine batch size based on remaining packets and buffer capacity
batchSize := len(packets) - sent
if batchSize > len(u.writeMsgs) {
batchSize = len(u.writeMsgs)
}
// Use pre-allocated buffers
msgs := u.writeMsgs[:batchSize]
iovecs := u.writeIovecs[:batchSize]
names := u.writeNames[:batchSize]
// Setup message structures for this batch
for i := 0; i < batchSize; i++ {
pktIdx := sent + i
if !addrs[pktIdx].Addr().Is4() {
return sent + i, ErrInvalidIPv6RemoteForSocket
}
// Setup the packet buffer
iovecs[i].Base = &packets[pktIdx][0]
iovecs[i].Len = uint(len(packets[pktIdx]))
// Setup the destination address
rsa := (*unix.RawSockaddrInet4)(unsafe.Pointer(&names[i][0]))
rsa.Family = unix.AF_INET
rsa.Addr = addrs[pktIdx].Addr().As4()
binary.BigEndian.PutUint16((*[2]byte)(unsafe.Pointer(&rsa.Port))[:], addrs[pktIdx].Port())
// Set the appropriate address length for IPv4
msgs[i].Hdr.Namelen = unix.SizeofSockaddrInet4
}
// Send this batch
nsent, _, err := unix.Syscall6(
unix.SYS_SENDMMSG,
uintptr(u.sysFd),
uintptr(unsafe.Pointer(&msgs[0])),
uintptr(batchSize),
0,
0,
0,
)
if err != 0 {
return sent + int(nsent), &net.OpError{Op: "sendmmsg", Err: err}
}
sent += int(nsent)
if int(nsent) < batchSize {
// Couldn't send all packets in batch, return what we sent
return sent, nil
}
}
return sent, nil
}
func (u *StdConn) writeMulti6(packets [][]byte, addrs []netip.AddrPort) (int, error) {
sent := 0
for sent < len(packets) {
// Determine batch size based on remaining packets and buffer capacity
batchSize := len(packets) - sent
if batchSize > len(u.writeMsgs) {
batchSize = len(u.writeMsgs)
}
// Use pre-allocated buffers
msgs := u.writeMsgs[:batchSize]
iovecs := u.writeIovecs[:batchSize]
names := u.writeNames[:batchSize]
// Setup message structures for this batch
for i := 0; i < batchSize; i++ {
pktIdx := sent + i
// Setup the packet buffer
iovecs[i].Base = &packets[pktIdx][0]
iovecs[i].Len = uint(len(packets[pktIdx]))
// Setup the destination address
rsa := (*unix.RawSockaddrInet6)(unsafe.Pointer(&names[i][0]))
rsa.Family = unix.AF_INET6
rsa.Addr = addrs[pktIdx].Addr().As16()
binary.BigEndian.PutUint16((*[2]byte)(unsafe.Pointer(&rsa.Port))[:], addrs[pktIdx].Port())
// Set the appropriate address length for IPv6
msgs[i].Hdr.Namelen = unix.SizeofSockaddrInet6
}
// Send this batch
nsent, _, err := unix.Syscall6(
unix.SYS_SENDMMSG,
uintptr(u.sysFd),
uintptr(unsafe.Pointer(&msgs[0])),
uintptr(batchSize),
0,
0,
0,
)
if err != 0 {
return sent + int(nsent), &net.OpError{Op: "sendmmsg", Err: err}
}
sent += int(nsent)
if int(nsent) < batchSize {
// Couldn't send all packets in batch, return what we sent
return sent, nil
}
}
return sent, nil
}
func (u *StdConn) ReloadConfig(c *config.C) { func (u *StdConn) ReloadConfig(c *config.C) {
b := c.GetInt("listen.read_buffer", 0) b := c.GetInt("listen.read_buffer", 0)
if b > 0 { if b > 0 {
@@ -468,19 +344,28 @@ func (u *StdConn) ReloadConfig(c *config.C) {
func (u *StdConn) getMemInfo(meminfo *[unix.SK_MEMINFO_VARS]uint32) error { func (u *StdConn) getMemInfo(meminfo *[unix.SK_MEMINFO_VARS]uint32) error {
var vallen uint32 = 4 * unix.SK_MEMINFO_VARS var vallen uint32 = 4 * unix.SK_MEMINFO_VARS
_, _, err := unix.Syscall6(unix.SYS_GETSOCKOPT, uintptr(u.sysFd), uintptr(unix.SOL_SOCKET), uintptr(unix.SO_MEMINFO), uintptr(unsafe.Pointer(meminfo)), uintptr(unsafe.Pointer(&vallen)), 0)
if err != 0 { if u.rawConn == nil {
return fmt.Errorf("no UDP connection")
}
var opErr error
err := u.rawConn.Control(func(fd uintptr) {
_, _, syserr := unix.Syscall6(unix.SYS_GETSOCKOPT, fd, uintptr(unix.SOL_SOCKET), uintptr(unix.SO_MEMINFO), uintptr(unsafe.Pointer(meminfo)), uintptr(unsafe.Pointer(&vallen)), 0)
if syserr != 0 {
opErr = syserr
}
})
if err != nil {
return err return err
} }
return nil return opErr
}
func (u *StdConn) BatchSize() int {
return u.batch
} }
func (u *StdConn) Close() error { func (u *StdConn) Close() error {
return syscall.Close(u.sysFd) if u.udpConn != nil {
return u.udpConn.Close()
}
return nil
} }
func NewUDPStatsEmitter(udpConns []Conn) func() { func NewUDPStatsEmitter(udpConns []Conn) func() {
+2 -2
View File
@@ -12,7 +12,7 @@ import (
type iovec struct { type iovec struct {
Base *byte Base *byte
Len uint Len uint32
} }
type msghdr struct { type msghdr struct {
@@ -40,7 +40,7 @@ func (u *StdConn) PrepareRawMessages(n int) ([]rawMessage, [][]byte, [][]byte) {
names[i] = make([]byte, unix.SizeofSockaddrInet6) names[i] = make([]byte, unix.SizeofSockaddrInet6)
vs := []iovec{ vs := []iovec{
{Base: &buffers[i][0], Len: uint(len(buffers[i]))}, {Base: &buffers[i][0], Len: uint32(len(buffers[i]))},
} }
msgs[i].Hdr.Iov = &vs[0] msgs[i].Hdr.Iov = &vs[0]
+2 -2
View File
@@ -12,7 +12,7 @@ import (
type iovec struct { type iovec struct {
Base *byte Base *byte
Len uint Len uint64
} }
type msghdr struct { type msghdr struct {
@@ -43,7 +43,7 @@ func (u *StdConn) PrepareRawMessages(n int) ([]rawMessage, [][]byte, [][]byte) {
names[i] = make([]byte, unix.SizeofSockaddrInet6) names[i] = make([]byte, unix.SizeofSockaddrInet6)
vs := []iovec{ vs := []iovec{
{Base: &buffers[i][0], Len: uint(len(buffers[i]))}, {Base: &buffers[i][0], Len: uint64(len(buffers[i]))},
} }
msgs[i].Hdr.Iov = &vs[0] msgs[i].Hdr.Iov = &vs[0]
-44
View File
@@ -338,50 +338,6 @@ func (u *RIOConn) Rebind() error {
func (u *RIOConn) ReloadConfig(*config.C) {} func (u *RIOConn) ReloadConfig(*config.C) {}
// BatchSize returns 1 since RIO reads packets one at a time
func (u *RIOConn) BatchSize() int {
return 1
}
// ListenOutBatch - fallback to single-packet reads for RIO
func (u *RIOConn) ListenOutBatch(r EncBatchReader) {
buffer := make([]byte, MTU)
addrs := make([]netip.AddrPort, 1)
payloads := make([][]byte, 1)
var lastRecvErr time.Time
for {
n, rua, err := u.receive(buffer)
if err != nil {
if errors.Is(err, net.ErrClosed) {
u.l.WithError(err).Debug("udp socket is closed, exiting read loop")
return
}
if lastRecvErr.IsZero() || time.Since(lastRecvErr) > time.Minute {
lastRecvErr = time.Now()
u.l.WithError(err).Warn("unexpected udp socket receive error")
}
continue
}
addrs[0] = netip.AddrPortFrom(netip.AddrFrom16(rua.Addr).Unmap(), (rua.Port>>8)|((rua.Port&0xff)<<8))
payloads[0] = buffer[:n]
r(addrs, payloads, 1)
}
}
// WriteMulti sends multiple packets - fallback implementation
func (u *RIOConn) WriteMulti(packets [][]byte, addrs []netip.AddrPort) (int, error) {
for i := range packets {
err := u.WriteTo(packets[i], addrs[i])
if err != nil {
return i, err
}
}
return len(packets), nil
}
func (u *RIOConn) Close() error { func (u *RIOConn) Close() error {
if !u.isOpen.CompareAndSwap(true, false) { if !u.isOpen.CompareAndSwap(true, false) {
return nil return nil
-29
View File
@@ -116,31 +116,6 @@ func (u *TesterConn) ListenOut(r EncReader) {
} }
} }
func (u *TesterConn) ListenOutBatch(r EncBatchReader) {
addrs := make([]netip.AddrPort, 1)
payloads := make([][]byte, 1)
for {
p, ok := <-u.RxPackets
if !ok {
return
}
addrs[0] = p.From
payloads[0] = p.Data
r(addrs, payloads, 1)
}
}
func (u *TesterConn) WriteMulti(packets [][]byte, addrs []netip.AddrPort) (int, error) {
for i := range packets {
err := u.WriteTo(packets[i], addrs[i])
if err != nil {
return i, err
}
}
return len(packets), nil
}
func (u *TesterConn) ReloadConfig(*config.C) {} func (u *TesterConn) ReloadConfig(*config.C) {}
func NewUDPStatsEmitter(_ []Conn) func() { func NewUDPStatsEmitter(_ []Conn) func() {
@@ -156,10 +131,6 @@ func (u *TesterConn) SupportsMultipleReaders() bool {
return false return false
} }
func (u *TesterConn) BatchSize() int {
return 1
}
func (u *TesterConn) Rebind() error { func (u *TesterConn) Rebind() error {
return nil return nil
} }