window

cruft

fix tests

haha yep faster

checksum speed

haha

save pennies

fix

typo!

checkpt

GSO again

.github/workflows/smoke/smoke.sh

@@ -82,7 +82,7 @@ docker exec host4 tcpdump -i eth0 -q -w - -U 2>logs/host4.outside.log >logs/host
 
 docker exec host2 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 host4 ncat -e '/usr/bin/echo helloagainfromhost4' -nkluv 0.0.0.0 4000 &
 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 &
 
@@ -155,11 +155,11 @@ echo " *** Testing conntrack"
 echo
 set -x
 
-# host2 speaking to host4 on UDP 4000 should allow it to reply, when firewall rules would normally not permit this
+# host4's outbound firewall only allows ICMP to the lighthouse, so host4
-docker exec host2 sh -c "/usr/bin/echo host2 | ncat -nuv 192.168.100.4 4000"
+# cannot initiate UDP to host2. Once host2 initiates a flow to host4:4000,
-docker exec host2 ncat -e '/usr/bin/echo helloagainfromhost2' -nkluv 0.0.0.0 4000 &
+# conntrack must let host4's listener reply on that flow. If it doesn't,
-sleep 1
+# the echo back from host4 never reaches host2.
-docker exec host4 sh -c "/usr/bin/echo host4 | ncat -nuv 192.168.100.2 4000"
+docker exec host2 sh -c "(/usr/bin/echo host2; sleep 2) | ncat -nuv 192.168.100.4 4000" | grep -q helloagainfromhost4
 
 docker exec host4 sh -c 'kill 1'
 docker exec host3 sh -c 'kill 1'

.golangci.yaml

@@ -2,7 +2,21 @@ version: "2"
 linters:
   default: none
   enable:
+    - sloglint
     - testifylint
+  settings:
+    sloglint:
+      # Enforce key-value pair form for Info/Debug/Warn/Error/Log/With and
+      # the package-level slog equivalents. Use l.Log(ctx, level, ...) for
+      # custom levels instead of LogAttrs when you can.
+      #
+      # LogAttrs is also flagged by this rule because it takes ...slog.Attr;
+      # the few legitimate sites (where attrs is built up as a []slog.Attr)
+      # carry a //nolint:sloglint with rationale.
+      kv-only: true
+      # no-mixed-args is on by default: forbids mixing kv and attrs in one call.
+      # discard-handler is on by default (since Go 1.24): suggests
+      # slog.DiscardHandler over slog.NewTextHandler(io.Discard, nil).
   exclusions:
     generated: lax
     presets:

batch.go

@@ -1,70 +0,0 @@
-package nebula
-
-import "net/netip"
-
-// sendBatchCap is the maximum number of encrypted packets accumulated before a
-// flush is forced. TSO superpackets segment to at most ~45 packets on
-// reasonable MTUs, so 128 leaves headroom without bloating the backing
-// allocation.
-const sendBatchCap = 128
-
-// sendBatch accumulates encrypted UDP packets for a single sendmmsg flush.
-// One sendBatch is owned by each listenIn goroutine; no locking is needed.
-// The backing storage holds up to batchCap packets of slotCap bytes each;
-// bufs and dsts are parallel slices of committed slots.
-type sendBatch struct {
-	bufs     [][]byte
-	dsts     []netip.AddrPort
-	backing  []byte
-	slotCap  int
-	batchCap int
-	nextSlot int
-}
-
-func newSendBatch(batchCap, slotCap int) *sendBatch {
-	return &sendBatch{
-		bufs:     make([][]byte, 0, batchCap),
-		dsts:     make([]netip.AddrPort, 0, batchCap),
-		backing:  make([]byte, batchCap*slotCap),
-		slotCap:  slotCap,
-		batchCap: batchCap,
-	}
-}
-
-// Next returns a zero-length slice with slotCap capacity over the next unused
-// slot's backing bytes. The caller writes into the returned slice and then
-// calls Commit with the final length and destination. Next returns nil when
-// the batch is full.
-func (b *sendBatch) Next() []byte {
-	if b.nextSlot >= b.batchCap {
-		return nil
-	}
-	start := b.nextSlot * b.slotCap
-	return b.backing[start : start : start+b.slotCap]
-}
-
-// Commit records the slot just returned by Next as a packet of length n
-// destined for dst.
-func (b *sendBatch) Commit(n int, dst netip.AddrPort) {
-	start := b.nextSlot * b.slotCap
-	b.bufs = append(b.bufs, b.backing[start:start+n])
-	b.dsts = append(b.dsts, dst)
-	b.nextSlot++
-}
-
-// Reset clears committed slots; backing storage is retained for reuse.
-func (b *sendBatch) Reset() {
-	b.bufs = b.bufs[:0]
-	b.dsts = b.dsts[:0]
-	b.nextSlot = 0
-}
-
-// Len returns the number of committed packets.
-func (b *sendBatch) Len() int {
-	return len(b.bufs)
-}
-
-// Cap returns the maximum number of slots in the batch.
-func (b *sendBatch) Cap() int {
-	return b.batchCap
-}

bits.go

@@ -1,8 +1,10 @@
 package nebula
 
 import (
+	"context"
+	"log/slog"
+
 	"github.com/rcrowley/go-metrics"
-	"github.com/sirupsen/logrus"
 )
 
 type Bits struct {
@@ -30,7 +32,7 @@ func NewBits(bits uint64) *Bits {
 	return b
 }
 
-func (b *Bits) Check(l *logrus.Logger, i uint64) bool {
+func (b *Bits) Check(l *slog.Logger, i uint64) bool {
 	// If i is the next number, return true.
 	if i > b.current {
 		return true
@@ -42,13 +44,16 @@ func (b *Bits) Check(l *logrus.Logger, i uint64) bool {
 	}
 
 	// Not within the window
-	if l.Level >= logrus.DebugLevel {
+	if l.Enabled(context.Background(), slog.LevelDebug) {
-		l.Debugf("rejected a packet (top) %d %d\n", b.current, i)
+		l.Debug("rejected a packet (top)",
+			"current", b.current,
+			"incoming", i,
+		)
 	}
 	return false
 }
 
-func (b *Bits) Update(l *logrus.Logger, i uint64) bool {
+func (b *Bits) Update(l *slog.Logger, i uint64) bool {
 	// If i is the next number, return true and update current.
 	if i == b.current+1 {
 		// Check if the oldest bit was lost since we are shifting the window by 1 and occupying it with this counter
@@ -87,9 +92,13 @@ func (b *Bits) Update(l *logrus.Logger, i uint64) bool {
 	// Check to see if it's a duplicate
 	if i > b.current-b.length || i < b.length && b.current < b.length {
 		if b.current == i || b.bits[i%b.length] == true {
-			if l.Level >= logrus.DebugLevel {
+			if l.Enabled(context.Background(), slog.LevelDebug) {
-				l.WithField("receiveWindow", m{"accepted": false, "currentCounter": b.current, "incomingCounter": i, "reason": "duplicate"}).
+				l.Debug("Receive window",
-					Debug("Receive window")
+					"accepted", false,
+					"currentCounter", b.current,
+					"incomingCounter", i,
+					"reason", "duplicate",
+				)
 			}
 			b.dupeCounter.Inc(1)
 			return false
@@ -101,12 +110,13 @@ func (b *Bits) Update(l *logrus.Logger, i uint64) bool {
 
 	// In all other cases, fail and don't change current.
 	b.outOfWindowCounter.Inc(1)
-	if l.Level >= logrus.DebugLevel {
+	if l.Enabled(context.Background(), slog.LevelDebug) {
-		l.WithField("accepted", false).
+		l.Debug("Receive window",
-			WithField("currentCounter", b.current).
+			"accepted", false,
-			WithField("incomingCounter", i).
+			"currentCounter", b.current,
-			WithField("reason", "nonsense").
+			"incomingCounter", i,
-			Debug("Receive window")
+			"reason", "nonsense",
+		)
 	}
 	return false
 }

cmd/nebula-service/logs_generic.go

@@ -3,8 +3,15 @@
 
 package main
 
-import "github.com/sirupsen/logrus"
+import (
+	"log/slog"
+	"os"
 
-func HookLogger(l *logrus.Logger) {
+	"github.com/slackhq/nebula/logging"
-	// Do nothing, let the logs flow to stdout/stderr
+)
+
+// newPlatformLogger returns a *slog.Logger that writes to stdout. Non-Windows
+// platforms have no special sink to integrate with.
+func newPlatformLogger() *slog.Logger {
+	return logging.NewLogger(os.Stdout)
 }

cmd/nebula-service/logs_windows.go

@@ -1,54 +1,86 @@
 package main
 
 import (
-	"fmt"
+	"context"
-	"io/ioutil"
+	"log/slog"
-	"os"
+	"strings"
+	"sync"
 
-	"github.com/kardianos/service"
+	"github.com/slackhq/nebula/logging"
-	"github.com/sirupsen/logrus"
 )
 
-// HookLogger routes the logrus logs through the service logger so that they end up in the Windows Event Viewer
+// newPlatformLogger returns a *slog.Logger that routes every log record
-// logrus output will be discarded
+// through the Windows service logger so records end up in the Windows
-func HookLogger(l *logrus.Logger) {
+// Event Log. All the heavy lifting (level management, format swap,
-	l.AddHook(newLogHook(logger))
+// timestamp toggle, WithAttrs/WithGroup) comes from logging.NewHandler;
-	l.SetOutput(ioutil.Discard)
+// this file only contributes:
+//
+//   - an io.Writer that forwards each formatted line to the service
+//     logger at the current record's Event Log severity, and
+//   - a thin severityTag that embeds *logging.Handler and overrides
+//     only Handle / WithAttrs / WithGroup, so Event Viewer's severity
+//     column and severity-based filters keep working the way they did
+//     before the slog migration.
+//
+// Format (text vs json) is carried by the embedded *logging.Handler, so
+// logging.format: json in config still produces JSON lines in Event
+// Viewer, same as the pre-slog logrus setup.
+func newPlatformLogger() *slog.Logger {
+	w := &eventLogWriter{}
+	return slog.New(&severityTag{Handler: logging.NewHandler(w), w: w})
 }
 
-type logHook struct {
+// eventLogWriter forwards slog-formatted lines to the Windows service
-	sl service.Logger
+// logger at the severity most recently stashed by severityTag.Handle.
+// The mutex serializes the stash + inner.Handle + Write cycle per record
+// across all concurrent goroutines; slog's builtin text/json handlers
+// each hold their own mutex around Write, but that only protects the
+// Write call itself, not our stash-then-handle sequence.
+type eventLogWriter struct {
+	mu    sync.Mutex
+	level slog.Level
 }
 
-func newLogHook(sl service.Logger) *logHook {
+func (w *eventLogWriter) Write(p []byte) (int, error) {
-	return &logHook{sl: sl}
+	line := strings.TrimRight(string(p), "\n")
-}
+	switch {
-
+	case w.level >= slog.LevelError:
-func (h *logHook) Fire(entry *logrus.Entry) error {
+		return len(p), logger.Error(line)
-	line, err := entry.String()
+	case w.level >= slog.LevelWarn:
-	if err != nil {
+		return len(p), logger.Warning(line)
-		fmt.Fprintf(os.Stderr, "Unable to read entry, %v", err)
-		return err
-	}
-
-	switch entry.Level {
-	case logrus.PanicLevel:
-		return h.sl.Error(line)
-	case logrus.FatalLevel:
-		return h.sl.Error(line)
-	case logrus.ErrorLevel:
-		return h.sl.Error(line)
-	case logrus.WarnLevel:
-		return h.sl.Warning(line)
-	case logrus.InfoLevel:
-		return h.sl.Info(line)
-	case logrus.DebugLevel:
-		return h.sl.Info(line)
 	default:
-		return nil
+		return len(p), logger.Info(line)
 	}
 }
 
-func (h *logHook) Levels() []logrus.Level {
+// severityTag embeds *logging.Handler to pick up everything it does for
-	return logrus.AllLevels
+// free (Enabled, SetLevel, GetLevel, SetFormat, GetFormat,
+// SetDisableTimestamp) and overrides only Handle / WithAttrs / WithGroup
+// so each record's slog.Level is stashed on the writer before formatting
+// and so derived handlers stay wrapped as severityTag rather than
+// downgrading to bare *logging.Handler.
+type severityTag struct {
+	*logging.Handler
+	w *eventLogWriter
+}
+
+func (s *severityTag) Handle(ctx context.Context, r slog.Record) error {
+	s.w.mu.Lock()
+	defer s.w.mu.Unlock()
+	s.w.level = r.Level
+	return s.Handler.Handle(ctx, r)
+}
+
+func (s *severityTag) WithAttrs(attrs []slog.Attr) slog.Handler {
+	if len(attrs) == 0 {
+		return s
+	}
+	return &severityTag{Handler: s.Handler.WithAttrs(attrs).(*logging.Handler), w: s.w}
+}
+
+func (s *severityTag) WithGroup(name string) slog.Handler {
+	if name == "" {
+		return s
+	}
+	return &severityTag{Handler: s.Handler.WithGroup(name).(*logging.Handler), w: s.w}
 }

cmd/nebula-service/main.go

@@ -7,9 +7,9 @@ import (
 	"runtime/debug"
 	"strings"
 
-	"github.com/sirupsen/logrus"
 	"github.com/slackhq/nebula"
 	"github.com/slackhq/nebula/config"
+	"github.com/slackhq/nebula/logging"
 	"github.com/slackhq/nebula/util"
 )
 
@@ -50,10 +50,15 @@ func main() {
 		os.Exit(0)
 	}
 
+	l := logging.NewLogger(os.Stdout)
+
 	if *serviceFlag != "" {
-		doService(configPath, configTest, Build, serviceFlag)
+		if err := doService(configPath, configTest, Build, serviceFlag); err != nil {
+			l.Error("Service command failed", "error", err)
 			os.Exit(1)
 		}
+		return
+	}
 
 	if *configPath == "" {
 		fmt.Println("-config flag must be set")
@@ -61,9 +66,6 @@ func main() {
 		os.Exit(1)
 	}
 
-	l := logrus.New()
-	l.Out = os.Stdout
-
 	c := config.NewC(l)
 	err := c.Load(*configPath)
 	if err != nil {
@@ -71,6 +73,16 @@ func main() {
 		os.Exit(1)
 	}
 
+	if err := logging.ApplyConfig(l, c); err != nil {
+		fmt.Printf("failed to apply logging config: %s", err)
+		os.Exit(1)
+	}
+	c.RegisterReloadCallback(func(c *config.C) {
+		if err := logging.ApplyConfig(l, c); err != nil {
+			l.Error("Failed to reconfigure logger on reload", "error", err)
+		}
+	})
+
 	ctrl, err := nebula.Main(c, *configTest, Build, l, nil)
 	if err != nil {
 		util.LogWithContextIfNeeded("Failed to start", err, l)
@@ -87,7 +99,7 @@ func main() {
 		go ctrl.ShutdownBlock()
 
 		if err := wait(); err != nil {
-			l.WithError(err).Error("Nebula stopped due to fatal error")
+			l.Error("Nebula stopped due to fatal error", "error", err)
 			os.Exit(2)
 		}
 

cmd/nebula-service/service.go

@@ -7,9 +7,9 @@ import (
 	"path/filepath"
 
 	"github.com/kardianos/service"
-	"github.com/sirupsen/logrus"
 	"github.com/slackhq/nebula"
 	"github.com/slackhq/nebula/config"
+	"github.com/slackhq/nebula/logging"
 )
 
 var logger service.Logger
@@ -25,8 +25,7 @@ func (p *program) Start(s service.Service) error {
 	// Start should not block.
 	logger.Info("Nebula service starting.")
 
-	l := logrus.New()
+	l := newPlatformLogger()
-	HookLogger(l)
 
 	c := config.NewC(l)
 	err := c.Load(*p.configPath)
@@ -34,6 +33,15 @@ func (p *program) Start(s service.Service) error {
 		return fmt.Errorf("failed to load config: %s", err)
 	}
 
+	if err := logging.ApplyConfig(l, c); err != nil {
+		return fmt.Errorf("failed to apply logging config: %s", err)
+	}
+	c.RegisterReloadCallback(func(c *config.C) {
+		if err := logging.ApplyConfig(l, c); err != nil {
+			l.Error("Failed to reconfigure logger on reload", "error", err)
+		}
+	})
+
 	p.control, err = nebula.Main(c, *p.configTest, Build, l, nil)
 	if err != nil {
 		return err
@@ -57,11 +65,11 @@ func fileExists(filename string) bool {
 	return true
 }
 
-func doService(configPath *string, configTest *bool, build string, serviceFlag *string) {
+func doService(configPath *string, configTest *bool, build string, serviceFlag *string) error {
 	if *configPath == "" {
 		ex, err := os.Executable()
 		if err != nil {
-			panic(err)
+			return err
 		}
 		*configPath = filepath.Dir(ex) + "/config.yaml"
 		if !fileExists(*configPath) {
@@ -85,16 +93,16 @@ func doService(configPath *string, configTest *bool, build string, serviceFlag *
 	// Here are what the different loggers are doing:
 	// - `log` is the standard go log utility, meant to be used while the process is still attached to stdout/stderr
 	// - `logger` is the service log utility that may be attached to a special place depending on OS (Windows will have it attached to the event log)
-	// - above, in `Run` we create a `logrus.Logger` which is what nebula expects to use
+	// - in program.Start we build a *slog.Logger via newPlatformLogger; on non-Windows that is a stdout-backed slog logger, on Windows it routes records through the service logger
 	s, err := service.New(prg, svcConfig)
 	if err != nil {
-		log.Fatal(err)
+		return err
 	}
 
 	errs := make(chan error, 5)
 	logger, err = s.Logger(errs)
 	if err != nil {
-		log.Fatal(err)
+		return err
 	}
 
 	go func() {
@@ -109,18 +117,16 @@ func doService(configPath *string, configTest *bool, build string, serviceFlag *
 
 	switch *serviceFlag {
 	case "run":
-		err = s.Run()
+		if err := s.Run(); err != nil {
-		if err != nil {
 			// Route any errors to the system logger
 			logger.Error(err)
 		}
 	default:
-		err := service.Control(s, *serviceFlag)
+		if err := service.Control(s, *serviceFlag); err != nil {
-		if err != nil {
 			log.Printf("Valid actions: %q\n", service.ControlAction)
-			log.Fatal(err)
+			return err
 		}
-		return
 	}
 
+	return nil
 }

cmd/nebula/main.go

@@ -7,9 +7,9 @@ import (
 	"runtime/debug"
 	"strings"
 
-	"github.com/sirupsen/logrus"
 	"github.com/slackhq/nebula"
 	"github.com/slackhq/nebula/config"
+	"github.com/slackhq/nebula/logging"
 	"github.com/slackhq/nebula/util"
 )
 
@@ -55,8 +55,7 @@ func main() {
 		os.Exit(1)
 	}
 
-	l := logrus.New()
+	l := logging.NewLogger(os.Stdout)
-	l.Out = os.Stdout
 
 	c := config.NewC(l)
 	err := c.Load(*configPath)
@@ -65,6 +64,16 @@ func main() {
 		os.Exit(1)
 	}
 
+	if err := logging.ApplyConfig(l, c); err != nil {
+		fmt.Printf("failed to apply logging config: %s", err)
+		os.Exit(1)
+	}
+	c.RegisterReloadCallback(func(c *config.C) {
+		if err := logging.ApplyConfig(l, c); err != nil {
+			l.Error("Failed to reconfigure logger on reload", "error", err)
+		}
+	})
+
 	ctrl, err := nebula.Main(c, *configTest, Build, l, nil)
 	if err != nil {
 		util.LogWithContextIfNeeded("Failed to start", err, l)
@@ -82,7 +91,7 @@ func main() {
 		notifyReady(l)
 
 		if err := wait(); err != nil {
-			l.WithError(err).Error("Nebula stopped due to fatal error")
+			l.Error("Nebula stopped due to fatal error", "error", err)
 			os.Exit(2)
 		}
 

cmd/nebula/notify_linux.go

@@ -1,11 +1,10 @@
 package main
 
 import (
+	"log/slog"
 	"net"
 	"os"
 	"time"
-
-	"github.com/sirupsen/logrus"
 )
 
 // SdNotifyReady tells systemd the service is ready and dependent services can now be started
@@ -13,30 +12,30 @@ import (
 // https://www.freedesktop.org/software/systemd/man/systemd.service.html
 const SdNotifyReady = "READY=1"
 
-func notifyReady(l *logrus.Logger) {
+func notifyReady(l *slog.Logger) {
 	sockName := os.Getenv("NOTIFY_SOCKET")
 	if sockName == "" {
-		l.Debugln("NOTIFY_SOCKET systemd env var not set, not sending ready signal")
+		l.Debug("NOTIFY_SOCKET systemd env var not set, not sending ready signal")
 		return
 	}
 
 	conn, err := net.DialTimeout("unixgram", sockName, time.Second)
 	if err != nil {
-		l.WithError(err).Error("failed to connect to systemd notification socket")
+		l.Error("failed to connect to systemd notification socket", "error", err)
 		return
 	}
 	defer conn.Close()
 
 	err = conn.SetWriteDeadline(time.Now().Add(time.Second))
 	if err != nil {
-		l.WithError(err).Error("failed to set the write deadline for the systemd notification socket")
+		l.Error("failed to set the write deadline for the systemd notification socket", "error", err)
 		return
 	}
 
 	if _, err = conn.Write([]byte(SdNotifyReady)); err != nil {
-		l.WithError(err).Error("failed to signal the systemd notification socket")
+		l.Error("failed to signal the systemd notification socket", "error", err)
 		return
 	}
 
-	l.Debugln("notified systemd the service is ready")
+	l.Debug("notified systemd the service is ready")
 }

cmd/nebula/notify_notlinux.go

@@ -3,8 +3,8 @@
 
 package main
 
-import "github.com/sirupsen/logrus"
+import "log/slog"
 
-func notifyReady(_ *logrus.Logger) {
+func notifyReady(_ *slog.Logger) {
 	// No init service to notify
 }

config/config.go

@@ -4,6 +4,7 @@ import (
 	"context"
 	"errors"
 	"fmt"
+	"log/slog"
 	"math"
 	"os"
 	"os/signal"
@@ -16,7 +17,6 @@ import (
 	"time"
 
 	"dario.cat/mergo"
-	"github.com/sirupsen/logrus"
 	"go.yaml.in/yaml/v3"
 )
 
@@ -26,11 +26,11 @@ type C struct {
 	Settings    map[string]any
 	oldSettings map[string]any
 	callbacks   []func(*C)
-	l           *logrus.Logger
+	l           *slog.Logger
 	reloadLock  sync.Mutex
 }
 
-func NewC(l *logrus.Logger) *C {
+func NewC(l *slog.Logger) *C {
 	return &C{
 		Settings: make(map[string]any),
 		l:        l,
@@ -107,12 +107,18 @@ func (c *C) HasChanged(k string) bool {
 
 	newVals, err := yaml.Marshal(nv)
 	if err != nil {
-		c.l.WithField("config_path", k).WithError(err).Error("Error while marshaling new config")
+		c.l.Error("Error while marshaling new config",
+			"config_path", k,
+			"error", err,
+		)
 	}
 
 	oldVals, err := yaml.Marshal(ov)
 	if err != nil {
-		c.l.WithField("config_path", k).WithError(err).Error("Error while marshaling old config")
+		c.l.Error("Error while marshaling old config",
+			"config_path", k,
+			"error", err,
+		)
 	}
 
 	return string(newVals) != string(oldVals)
@@ -154,7 +160,10 @@ func (c *C) ReloadConfig() {
 
 	err := c.Load(c.path)
 	if err != nil {
-		c.l.WithField("config_path", c.path).WithError(err).Error("Error occurred while reloading config")
+		c.l.Error("Error occurred while reloading config",
+			"config_path", c.path,
+			"error", err,
+		)
 		return
 	}
 

connection_manager.go

@@ -5,13 +5,13 @@ import (
 	"context"
 	"encoding/binary"
 	"fmt"
+	"log/slog"
 	"net/netip"
 	"sync"
 	"sync/atomic"
 	"time"
 
 	"github.com/rcrowley/go-metrics"
-	"github.com/sirupsen/logrus"
 	"github.com/slackhq/nebula/cert"
 	"github.com/slackhq/nebula/config"
 	"github.com/slackhq/nebula/header"
@@ -47,10 +47,10 @@ type connectionManager struct {
 
 	metricsTxPunchy metrics.Counter
 
-	l *logrus.Logger
+	l *slog.Logger
 }
 
-func newConnectionManagerFromConfig(l *logrus.Logger, c *config.C, hm *HostMap, p *Punchy) *connectionManager {
+func newConnectionManagerFromConfig(l *slog.Logger, c *config.C, hm *HostMap, p *Punchy) *connectionManager {
 	cm := &connectionManager{
 		hostMap:         hm,
 		l:               l,
@@ -85,9 +85,10 @@ func (cm *connectionManager) reload(c *config.C, initial bool) {
 		old := cm.getInactivityTimeout()
 		cm.inactivityTimeout.Store((int64)(c.GetDuration("tunnels.inactivity_timeout", 10*time.Minute)))
 		if !initial {
-			cm.l.WithField("oldDuration", old).
+			cm.l.Info("Inactivity timeout has changed",
-				WithField("newDuration", cm.getInactivityTimeout()).
+				"oldDuration", old,
-				Info("Inactivity timeout has changed")
+				"newDuration", cm.getInactivityTimeout(),
+			)
 		}
 	}
 
@@ -95,9 +96,10 @@ func (cm *connectionManager) reload(c *config.C, initial bool) {
 		old := cm.dropInactive.Load()
 		cm.dropInactive.Store(c.GetBool("tunnels.drop_inactive", false))
 		if !initial {
-			cm.l.WithField("oldBool", old).
+			cm.l.Info("Drop inactive setting has changed",
-				WithField("newBool", cm.dropInactive.Load()).
+				"oldBool", old,
-				Info("Drop inactive setting has changed")
+				"newBool", cm.dropInactive.Load(),
+			)
 		}
 	}
 }
@@ -256,7 +258,7 @@ func (cm *connectionManager) migrateRelayUsed(oldhostinfo, newhostinfo *HostInfo
 			var err error
 			index, err = AddRelay(cm.l, newhostinfo, cm.hostMap, r.PeerAddr, nil, r.Type, Requested)
 			if err != nil {
-				cm.l.WithError(err).Error("failed to migrate relay to new hostinfo")
+				cm.l.Error("failed to migrate relay to new hostinfo", "error", err)
 				continue
 			}
 			switch r.Type {
@@ -304,16 +306,16 @@ func (cm *connectionManager) migrateRelayUsed(oldhostinfo, newhostinfo *HostInfo
 
 		msg, err := req.Marshal()
 		if err != nil {
-			cm.l.WithError(err).Error("failed to marshal Control message to migrate relay")
+			cm.l.Error("failed to marshal Control message to migrate relay", "error", err)
 		} else {
 			cm.intf.SendMessageToHostInfo(header.Control, 0, newhostinfo, msg, make([]byte, 12), make([]byte, mtu))
-			cm.l.WithFields(logrus.Fields{
+			cm.l.Info("send CreateRelayRequest",
-				"relayFrom":           req.RelayFromAddr,
+				"relayFrom", req.RelayFromAddr,
-				"relayTo":             req.RelayToAddr,
+				"relayTo", req.RelayToAddr,
-				"initiatorRelayIndex": req.InitiatorRelayIndex,
+				"initiatorRelayIndex", req.InitiatorRelayIndex,
-				"responderRelayIndex": req.ResponderRelayIndex,
+				"responderRelayIndex", req.ResponderRelayIndex,
-				"vpnAddrs":            newhostinfo.vpnAddrs}).
+				"vpnAddrs", newhostinfo.vpnAddrs,
-				Info("send CreateRelayRequest")
+			)
 		}
 	}
 }
@@ -325,7 +327,7 @@ func (cm *connectionManager) makeTrafficDecision(localIndex uint32, now time.Tim
 
 	hostinfo := cm.hostMap.Indexes[localIndex]
 	if hostinfo == nil {
-		cm.l.WithField("localIndex", localIndex).Debugln("Not found in hostmap")
+		cm.l.Debug("Not found in hostmap", "localIndex", localIndex)
 		return doNothing, nil, nil
 	}
 
@@ -345,10 +347,10 @@ func (cm *connectionManager) makeTrafficDecision(localIndex uint32, now time.Tim
 	// A hostinfo is determined alive if there is incoming traffic
 	if inTraffic {
 		decision := doNothing
-		if cm.l.Level >= logrus.DebugLevel {
+		if cm.l.Enabled(context.Background(), slog.LevelDebug) {
-			hostinfo.logger(cm.l).
+			hostinfo.logger(cm.l).Debug("Tunnel status",
-				WithField("tunnelCheck", m{"state": "alive", "method": "passive"}).
+				"tunnelCheck", m{"state": "alive", "method": "passive"},
-				Debug("Tunnel status")
+			)
 		}
 		hostinfo.pendingDeletion.Store(false)
 
@@ -375,9 +377,9 @@ func (cm *connectionManager) makeTrafficDecision(localIndex uint32, now time.Tim
 
 	if hostinfo.pendingDeletion.Load() {
 		// We have already sent a test packet and nothing was returned, this hostinfo is dead
-		hostinfo.logger(cm.l).
+		hostinfo.logger(cm.l).Info("Tunnel status",
-			WithField("tunnelCheck", m{"state": "dead", "method": "active"}).
+			"tunnelCheck", m{"state": "dead", "method": "active"},
-			Info("Tunnel status")
+		)
 
 		return deleteTunnel, hostinfo, nil
 	}
@@ -388,10 +390,10 @@ func (cm *connectionManager) makeTrafficDecision(localIndex uint32, now time.Tim
 			inactiveFor, isInactive := cm.isInactive(hostinfo, now)
 			if isInactive {
 				// Tunnel is inactive, tear it down
-				hostinfo.logger(cm.l).
+				hostinfo.logger(cm.l).Info("Dropping tunnel due to inactivity",
-					WithField("inactiveDuration", inactiveFor).
+					"inactiveDuration", inactiveFor,
-					WithField("primary", mainHostInfo).
+					"primary", mainHostInfo,
-					Info("Dropping tunnel due to inactivity")
+				)
 
 				return closeTunnel, hostinfo, primary
 			}
@@ -410,18 +412,18 @@ func (cm *connectionManager) makeTrafficDecision(localIndex uint32, now time.Tim
 			cm.sendPunch(hostinfo)
 		}
 
-		if cm.l.Level >= logrus.DebugLevel {
+		if cm.l.Enabled(context.Background(), slog.LevelDebug) {
-			hostinfo.logger(cm.l).
+			hostinfo.logger(cm.l).Debug("Tunnel status",
-				WithField("tunnelCheck", m{"state": "testing", "method": "active"}).
+				"tunnelCheck", m{"state": "testing", "method": "active"},
-				Debug("Tunnel status")
+			)
 		}
 
 		// Send a test packet to trigger an authenticated tunnel test, this should suss out any lingering tunnel issues
 		decision = sendTestPacket
 
 	} else {
-		if cm.l.Level >= logrus.DebugLevel {
+		if cm.l.Enabled(context.Background(), slog.LevelDebug) {
-			hostinfo.logger(cm.l).Debugf("Hostinfo sadness")
+			hostinfo.logger(cm.l).Debug("Hostinfo sadness")
 		}
 	}
 
@@ -493,14 +495,16 @@ func (cm *connectionManager) isInvalidCertificate(now time.Time, hostinfo *HostI
 		return false //cert is still valid! yay!
 	} else if err == cert.ErrBlockListed { //avoiding errors.Is for speed
 		// Block listed certificates should always be disconnected
-		hostinfo.logger(cm.l).WithError(err).
+		hostinfo.logger(cm.l).Info("Remote certificate is blocked, tearing down the tunnel",
-			WithField("fingerprint", remoteCert.Fingerprint).
+			"error", err,
-			Info("Remote certificate is blocked, tearing down the tunnel")
+			"fingerprint", remoteCert.Fingerprint,
+		)
 		return true
 	} else if cm.intf.disconnectInvalid.Load() {
-		hostinfo.logger(cm.l).WithError(err).
+		hostinfo.logger(cm.l).Info("Remote certificate is no longer valid, tearing down the tunnel",
-			WithField("fingerprint", remoteCert.Fingerprint).
+			"error", err,
-			Info("Remote certificate is no longer valid, tearing down the tunnel")
+			"fingerprint", remoteCert.Fingerprint,
+		)
 		return true
 	} else {
 		//if we reach here, the cert is no longer valid, but we're configured to keep tunnels from now-invalid certs open
@@ -539,10 +543,11 @@ func (cm *connectionManager) tryRehandshake(hostinfo *HostInfo) {
 	curCrtVersion := curCrt.Version()
 	myCrt := cs.getCertificate(curCrtVersion)
 	if myCrt == nil {
-		cm.l.WithField("vpnAddrs", hostinfo.vpnAddrs).
+		cm.l.Info("Re-handshaking with remote",
-			WithField("version", curCrtVersion).
+			"vpnAddrs", hostinfo.vpnAddrs,
-			WithField("reason", "local certificate removed").
+			"version", curCrtVersion,
-			Info("Re-handshaking with remote")
+			"reason", "local certificate removed",
+		)
 		cm.intf.handshakeManager.StartHandshake(hostinfo.vpnAddrs[0], nil)
 		return
 	}
@@ -550,11 +555,12 @@ func (cm *connectionManager) tryRehandshake(hostinfo *HostInfo) {
 	if peerCrt != nil && curCrtVersion < peerCrt.Certificate.Version() {
 		// if our certificate version is less than theirs, and we have a matching version available, rehandshake?
 		if cs.getCertificate(peerCrt.Certificate.Version()) != nil {
-			cm.l.WithField("vpnAddrs", hostinfo.vpnAddrs).
+			cm.l.Info("Re-handshaking with remote",
-				WithField("version", curCrtVersion).
+				"vpnAddrs", hostinfo.vpnAddrs,
-				WithField("peerVersion", peerCrt.Certificate.Version()).
+				"version", curCrtVersion,
-				WithField("reason", "local certificate version lower than peer, attempting to correct").
+				"peerVersion", peerCrt.Certificate.Version(),
-				Info("Re-handshaking with remote")
+				"reason", "local certificate version lower than peer, attempting to correct",
+			)
 			cm.intf.handshakeManager.StartHandshake(hostinfo.vpnAddrs[0], func(hh *HandshakeHostInfo) {
 				hh.initiatingVersionOverride = peerCrt.Certificate.Version()
 			})
@@ -562,17 +568,19 @@ func (cm *connectionManager) tryRehandshake(hostinfo *HostInfo) {
 		}
 	}
 	if !bytes.Equal(curCrt.Signature(), myCrt.Signature()) {
-		cm.l.WithField("vpnAddrs", hostinfo.vpnAddrs).
+		cm.l.Info("Re-handshaking with remote",
-			WithField("reason", "local certificate is not current").
+			"vpnAddrs", hostinfo.vpnAddrs,
-			Info("Re-handshaking with remote")
+			"reason", "local certificate is not current",
+		)
 
 		cm.intf.handshakeManager.StartHandshake(hostinfo.vpnAddrs[0], nil)
 		return
 	}
 	if curCrtVersion < cs.initiatingVersion {
-		cm.l.WithField("vpnAddrs", hostinfo.vpnAddrs).
+		cm.l.Info("Re-handshaking with remote",
-			WithField("reason", "current cert version < pki.initiatingVersion").
+			"vpnAddrs", hostinfo.vpnAddrs,
-			Info("Re-handshaking with remote")
+			"reason", "current cert version < pki.initiatingVersion",
+		)
 
 		cm.intf.handshakeManager.StartHandshake(hostinfo.vpnAddrs[0], nil)
 		return

connection_manager_test.go

@@ -10,7 +10,7 @@ import (
 	"github.com/flynn/noise"
 	"github.com/slackhq/nebula/cert"
 	"github.com/slackhq/nebula/config"
-	"github.com/slackhq/nebula/overlay"
+	"github.com/slackhq/nebula/overlay/overlaytest"
 	"github.com/slackhq/nebula/test"
 	"github.com/slackhq/nebula/udp"
 	"github.com/stretchr/testify/assert"
@@ -53,7 +53,7 @@ func Test_NewConnectionManagerTest(t *testing.T) {
 	lh := newTestLighthouse()
 	ifce := &Interface{
 		hostMap:          hostMap,
-		inside:           &overlay.NoopTun{},
+		inside:           &overlaytest.NoopTun{},
 		outside:          &udp.NoopConn{},
 		firewall:         &Firewall{},
 		lightHouse:       lh,
@@ -64,9 +64,9 @@ func Test_NewConnectionManagerTest(t *testing.T) {
 	ifce.pki.cs.Store(cs)
 
 	// Create manager
-	conf := config.NewC(l)
+	conf := config.NewC(test.NewLogger())
-	punchy := NewPunchyFromConfig(l, conf)
+	punchy := NewPunchyFromConfig(test.NewLogger(), conf)
-	nc := newConnectionManagerFromConfig(l, conf, hostMap, punchy)
+	nc := newConnectionManagerFromConfig(test.NewLogger(), conf, hostMap, punchy)
 	nc.intf = ifce
 	p := []byte("")
 	nb := make([]byte, 12, 12)
@@ -136,7 +136,7 @@ func Test_NewConnectionManagerTest2(t *testing.T) {
 	lh := newTestLighthouse()
 	ifce := &Interface{
 		hostMap:          hostMap,
-		inside:           &overlay.NoopTun{},
+		inside:           &overlaytest.NoopTun{},
 		outside:          &udp.NoopConn{},
 		firewall:         &Firewall{},
 		lightHouse:       lh,
@@ -147,9 +147,9 @@ func Test_NewConnectionManagerTest2(t *testing.T) {
 	ifce.pki.cs.Store(cs)
 
 	// Create manager
-	conf := config.NewC(l)
+	conf := config.NewC(test.NewLogger())
-	punchy := NewPunchyFromConfig(l, conf)
+	punchy := NewPunchyFromConfig(test.NewLogger(), conf)
-	nc := newConnectionManagerFromConfig(l, conf, hostMap, punchy)
+	nc := newConnectionManagerFromConfig(test.NewLogger(), conf, hostMap, punchy)
 	nc.intf = ifce
 	p := []byte("")
 	nb := make([]byte, 12, 12)
@@ -221,7 +221,7 @@ func Test_NewConnectionManager_DisconnectInactive(t *testing.T) {
 	lh := newTestLighthouse()
 	ifce := &Interface{
 		hostMap:          hostMap,
-		inside:           &overlay.NoopTun{},
+		inside:           &overlaytest.NoopTun{},
 		outside:          &udp.NoopConn{},
 		firewall:         &Firewall{},
 		lightHouse:       lh,
@@ -232,12 +232,12 @@ func Test_NewConnectionManager_DisconnectInactive(t *testing.T) {
 	ifce.pki.cs.Store(cs)
 
 	// Create manager
-	conf := config.NewC(l)
+	conf := config.NewC(test.NewLogger())
 	conf.Settings["tunnels"] = map[string]any{
 		"drop_inactive": true,
 	}
-	punchy := NewPunchyFromConfig(l, conf)
+	punchy := NewPunchyFromConfig(test.NewLogger(), conf)
-	nc := newConnectionManagerFromConfig(l, conf, hostMap, punchy)
+	nc := newConnectionManagerFromConfig(test.NewLogger(), conf, hostMap, punchy)
 	assert.True(t, nc.dropInactive.Load())
 	nc.intf = ifce
 
@@ -348,7 +348,7 @@ func Test_NewConnectionManagerTest_DisconnectInvalid(t *testing.T) {
 	lh := newTestLighthouse()
 	ifce := &Interface{
 		hostMap:          hostMap,
-		inside:           &overlay.NoopTun{},
+		inside:           &overlaytest.NoopTun{},
 		outside:          &udp.NoopConn{},
 		firewall:         &Firewall{},
 		lightHouse:       lh,
@@ -361,9 +361,9 @@ func Test_NewConnectionManagerTest_DisconnectInvalid(t *testing.T) {
 	ifce.disconnectInvalid.Store(true)
 
 	// Create manager
-	conf := config.NewC(l)
+	conf := config.NewC(test.NewLogger())
-	punchy := NewPunchyFromConfig(l, conf)
+	punchy := NewPunchyFromConfig(test.NewLogger(), conf)
-	nc := newConnectionManagerFromConfig(l, conf, hostMap, punchy)
+	nc := newConnectionManagerFromConfig(test.NewLogger(), conf, hostMap, punchy)
 	nc.intf = ifce
 	ifce.connectionManager = nc
 

connection_state.go

@@ -8,12 +8,11 @@ import (
 	"sync/atomic"
 
 	"github.com/flynn/noise"
-	"github.com/sirupsen/logrus"
 	"github.com/slackhq/nebula/cert"
 	"github.com/slackhq/nebula/noiseutil"
 )
 
-const ReplayWindow = 1024
+const ReplayWindow = 1024 //todo I've started seeing out-of-window messages in testing?
 
 type ConnectionState struct {
 	eKey           *NebulaCipherState
@@ -27,7 +26,7 @@ type ConnectionState struct {
 	writeLock      sync.Mutex
 }
 
-func NewConnectionState(l *logrus.Logger, cs *CertState, crt cert.Certificate, initiator bool, pattern noise.HandshakePattern) (*ConnectionState, error) {
+func NewConnectionState(cs *CertState, crt cert.Certificate, initiator bool, pattern noise.HandshakePattern) (*ConnectionState, error) {
 	var dhFunc noise.DHFunc
 	switch crt.Curve() {
 	case cert.Curve_CURVE25519:

control.go

@@ -3,13 +3,13 @@ package nebula
 import (
 	"context"
 	"errors"
+	"log/slog"
 	"net/netip"
 	"os"
 	"os/signal"
 	"sync"
 	"syscall"
 
-	"github.com/sirupsen/logrus"
 	"github.com/slackhq/nebula/cert"
 	"github.com/slackhq/nebula/header"
 	"github.com/slackhq/nebula/overlay"
@@ -46,7 +46,7 @@ type Control struct {
 	state     RunState
 
 	f                      *Interface
-	l                      *logrus.Logger
+	l                      *slog.Logger
 	ctx                    context.Context
 	cancel                 context.CancelFunc
 	sshStart               func()
@@ -151,7 +151,7 @@ func (c *Control) Stop() {
 
 	c.CloseAllTunnels(false)
 	if err := c.f.Close(); err != nil {
-		c.l.WithError(err).Error("Close interface failed")
+		c.l.Error("Close interface failed", "error", err)
 	}
 	c.stateLock.Lock()
 	c.state = StateStopped
@@ -166,7 +166,7 @@ func (c *Control) ShutdownBlock() {
 
 	rawSig := <-sigChan
 	sig := rawSig.String()
-	c.l.WithField("signal", sig).Info("Caught signal, shutting down")
+	c.l.Info("Caught signal, shutting down", "signal", sig)
 	c.Stop()
 }
 
@@ -303,8 +303,10 @@ func (c *Control) CloseAllTunnels(excludeLighthouses bool) (closed int) {
 		c.f.send(header.CloseTunnel, 0, h.ConnectionState, h, []byte{}, make([]byte, 12, 12), make([]byte, mtu))
 		c.f.closeTunnel(h)
 
-		c.l.WithField("vpnAddrs", h.vpnAddrs).WithField("udpAddr", h.remote).
+		c.l.Debug("Sending close tunnel message",
-			Debug("Sending close tunnel message")
+			"vpnAddrs", h.vpnAddrs,
+			"udpAddr", h.remote,
+		)
 		closed++
 	}
 

control_test.go

@@ -6,7 +6,6 @@ import (
 	"reflect"
 	"testing"
 
-	"github.com/sirupsen/logrus"
 	"github.com/slackhq/nebula/cert"
 	"github.com/slackhq/nebula/test"
 	"github.com/stretchr/testify/assert"
@@ -83,7 +82,7 @@ func TestControl_GetHostInfoByVpnIp(t *testing.T) {
 		f: &Interface{
 			hostMap: hm,
 		},
-		l: logrus.New(),
+		l: test.NewLogger(),
 	}
 
 	thi := c.GetHostInfoByVpnAddr(vpnIp, false)

dns_server.go

@@ -1,63 +1,249 @@
 package nebula
 
 import (
+	"context"
 	"fmt"
+	"log/slog"
 	"net"
 	"net/netip"
 	"strconv"
 	"strings"
 	"sync"
+	"sync/atomic"
 
 	"github.com/gaissmai/bart"
 	"github.com/miekg/dns"
-	"github.com/sirupsen/logrus"
 	"github.com/slackhq/nebula/config"
 )
 
-// This whole thing should be rewritten to use context
+type dnsServer struct {
-
-var dnsR *dnsRecords
-var dnsServer *dns.Server
-var dnsAddr string
-
-type dnsRecords struct {
 	sync.RWMutex
-	l               *logrus.Logger
+	l               *slog.Logger
+	ctx             context.Context
 	dnsMap4         map[string]netip.Addr
 	dnsMap6         map[string]netip.Addr
 	hostMap         *HostMap
 	myVpnAddrsTable *bart.Lite
+
+	mux *dns.ServeMux
+
+	// enabled mirrors `lighthouse.serve_dns && lighthouse.am_lighthouse`.
+	// Start, Add, and reload consult it so callers don't need to know the
+	// gating rules. When it toggles off via reload, accumulated records are
+	// cleared so a later re-enable starts with a fresh map populated from
+	// new handshakes.
+	enabled atomic.Bool
+
+	serverMu sync.Mutex
+	server   *dns.Server
+	// started is closed once `server` has finished binding (or after
+	// ListenAndServe returns on a bind failure). Stop waits on it before
+	// calling Shutdown to avoid the miekg/dns "server not started" race
+	// where a Shutdown that arrives before bind completes is silently
+	// ignored, leaving the listener running forever.
+	started chan struct{}
+	addr    string
 }
 
-func newDnsRecords(l *logrus.Logger, cs *CertState, hostMap *HostMap) *dnsRecords {
+// newDnsServerFromConfig builds a dnsServer, applies the initial config, and
-	return &dnsRecords{
+// registers a reload callback. The reload callback is registered before the
+// initial config is applied, so a SIGHUP can later enable, fix, or disable
+// DNS even if the initial application failed.
+//
+// The dnsServer internally gates on `lighthouse.serve_dns &&
+// lighthouse.am_lighthouse`. Start and Add are safe to call unconditionally,
+// they no-op when DNS isn't enabled. Each Start invocation owns a ctx-cancel
+// watcher that tears the listener down on nebula shutdown. The returned
+// pointer is always non-nil, even on error.
+func newDnsServerFromConfig(ctx context.Context, l *slog.Logger, cs *CertState, hostMap *HostMap, c *config.C) (*dnsServer, error) {
+	ds := &dnsServer{
 		l:               l,
+		ctx:             ctx,
 		dnsMap4:         make(map[string]netip.Addr),
 		dnsMap6:         make(map[string]netip.Addr),
 		hostMap:         hostMap,
 		myVpnAddrsTable: cs.myVpnAddrsTable,
 	}
+	ds.mux = dns.NewServeMux()
+	ds.mux.HandleFunc(".", ds.handleDnsRequest)
+
+	c.RegisterReloadCallback(func(c *config.C) {
+		if err := ds.reload(c, false); err != nil {
+			ds.l.Error("Failed to reload DNS responder from config", "error", err)
+		}
+	})
+
+	if err := ds.reload(c, true); err != nil {
+		return ds, err
+	}
+	return ds, nil
 }
 
-func (d *dnsRecords) Query(q uint16, data string) netip.Addr {
+// reload applies the latest config and reconciles the running state with it:
+//   - enabled toggled on  -> spawn a runner
+//   - enabled toggled off -> stop the runner
+//   - listen address changed (while running) -> restart on the new address
+//   - everything else     -> no-op
+//
+// On the initial call it only records configuration; Control.Start is what
+// launches the first runner via dnsStart.
+func (d *dnsServer) reload(c *config.C, initial bool) error {
+	wantsDns := c.GetBool("lighthouse.serve_dns", false)
+	amLighthouse := c.GetBool("lighthouse.am_lighthouse", false)
+	enabled := wantsDns && amLighthouse
+	newAddr := getDnsServerAddr(c)
+
+	d.serverMu.Lock()
+	running := d.server
+	runningStarted := d.started
+	sameAddr := d.addr == newAddr
+	d.addr = newAddr
+	d.enabled.Store(enabled)
+	d.serverMu.Unlock()
+
+	if initial {
+		if wantsDns && !amLighthouse {
+			d.l.Warn("DNS server refusing to run because this host is not a lighthouse.")
+		}
+		return nil
+	}
+
+	if !enabled {
+		if running != nil {
+			d.Stop()
+		}
+		// Drop any records that accumulated while enabled; a later re-enable
+		// will repopulate from fresh handshakes.
+		d.clearRecords()
+		return nil
+	}
+
+	if running == nil {
+		// Was disabled (or never started); bring it up now.
+		go d.Start()
+		return nil
+	}
+
+	if sameAddr {
+		return nil
+	}
+
+	d.shutdownServer(running, runningStarted, "reload")
+	// Old Start goroutine has now exited; bring up a fresh listener on the
+	// new address.
+	go d.Start()
+	return nil
+}
+
+// shutdownServer waits for the server to finish binding (so Shutdown actually
+// stops it rather than no-oping) and then shuts it down.
+func (d *dnsServer) shutdownServer(srv *dns.Server, started chan struct{}, reason string) {
+	if srv == nil {
+		return
+	}
+	if started != nil {
+		<-started
+	}
+	if err := srv.Shutdown(); err != nil {
+		d.l.Warn("Failed to shut down the DNS responder", "reason", reason, "error", err)
+	}
+}
+
+// Start binds and serves the DNS responder. Blocks until Stop is called or
+// the listener errors. Safe to call when DNS is disabled (returns
+// immediately). This is what Control.dnsStart points at.
+//
+// Must be invoked after the tun device is active so that lighthouse.dns.host
+// may bind to a nebula IP.
+func (d *dnsServer) Start() {
+	if !d.enabled.Load() {
+		return
+	}
+
+	started := make(chan struct{})
+	d.serverMu.Lock()
+	if d.ctx.Err() != nil {
+		d.serverMu.Unlock()
+		return
+	}
+	addr := d.addr
+	server := &dns.Server{
+		Addr:              addr,
+		Net:               "udp",
+		Handler:           d.mux,
+		NotifyStartedFunc: func() { close(started) },
+	}
+	d.server = server
+	d.started = started
+	d.serverMu.Unlock()
+
+	// Per-invocation ctx watcher. Exits when Start does, so we don't leak a
+	// watcher per reload-driven restart.
+	done := make(chan struct{})
+	go func() {
+		select {
+		case <-d.ctx.Done():
+			d.shutdownServer(server, started, "shutdown")
+		case <-done:
+		}
+	}()
+
+	d.l.Info("Starting DNS responder", "dnsListener", addr)
+	err := server.ListenAndServe()
+	close(done)
+
+	// If the listener never bound (bind error) NotifyStartedFunc never fires,
+	// so close started here to release any Stop caller waiting on it.
+	select {
+	case <-started:
+	default:
+		close(started)
+	}
+
+	if err != nil {
+		d.l.Warn("Failed to run the DNS responder", "error", err)
+	}
+}
+
+// Stop shuts down the active server, if any. Idempotent.
+func (d *dnsServer) Stop() {
+	d.serverMu.Lock()
+	srv := d.server
+	started := d.started
+	d.server = nil
+	d.started = nil
+	d.serverMu.Unlock()
+	d.shutdownServer(srv, started, "stop")
+}
+
+// Query returns the address for the given name and query type. The second
+// return value reports whether the name is known at all (in either A or AAAA),
+// which lets callers distinguish NODATA from NXDOMAIN.
+func (d *dnsServer) Query(q uint16, data string) (netip.Addr, bool) {
 	data = strings.ToLower(data)
 	d.RLock()
 	defer d.RUnlock()
+	addr4, haveV4 := d.dnsMap4[data]
+	addr6, haveV6 := d.dnsMap6[data]
+	nameExists := haveV4 || haveV6
 	switch q {
 	case dns.TypeA:
-		if r, ok := d.dnsMap4[data]; ok {
+		if haveV4 {
-			return r
+			return addr4, nameExists
 		}
 	case dns.TypeAAAA:
-		if r, ok := d.dnsMap6[data]; ok {
+		if haveV6 {
-			return r
+			return addr6, nameExists
 		}
 	}
 
-	return netip.Addr{}
+	return netip.Addr{}, nameExists
 }
 
-func (d *dnsRecords) QueryCert(data string) string {
+func (d *dnsServer) QueryCert(data string) string {
+	if len(data) < 2 {
+		return ""
+	}
 	ip, err := netip.ParseAddr(data[:len(data)-1])
 	if err != nil {
 		return ""
@@ -80,8 +266,19 @@ func (d *dnsRecords) QueryCert(data string) string {
 	return string(b)
 }
 
+// clearRecords drops all DNS records.
+func (d *dnsServer) clearRecords() {
+	d.Lock()
+	defer d.Unlock()
+	clear(d.dnsMap4)
+	clear(d.dnsMap6)
+}
+
 // Add adds the first IPv4 and IPv6 address that appears in `addresses` as the record for `host`
-func (d *dnsRecords) Add(host string, addresses []netip.Addr) {
+func (d *dnsServer) Add(host string, addresses []netip.Addr) {
+	if !d.enabled.Load() {
+		return
+	}
 	host = strings.ToLower(host)
 	d.Lock()
 	defer d.Unlock()
@@ -101,7 +298,7 @@ func (d *dnsRecords) Add(host string, addresses []netip.Addr) {
 	}
 }
 
-func (d *dnsRecords) isSelfNebulaOrLocalhost(addr string) bool {
+func (d *dnsServer) isSelfNebulaOrLocalhost(addr string) bool {
 	a, _, _ := net.SplitHostPort(addr)
 	b, err := netip.ParseAddr(a)
 	if err != nil {
@@ -116,13 +313,24 @@ func (d *dnsRecords) isSelfNebulaOrLocalhost(addr string) bool {
 	return d.myVpnAddrsTable.Contains(b)
 }
 
-func (d *dnsRecords) parseQuery(m *dns.Msg, w dns.ResponseWriter) {
+func (d *dnsServer) parseQuery(m *dns.Msg, w dns.ResponseWriter) {
+	debugEnabled := d.l.Enabled(context.Background(), slog.LevelDebug)
+	// Per RFC 2308 §2.2, a name that exists but has no record of the requested
+	// type must be answered with NOERROR and an empty answer section (NODATA),
+	// not NXDOMAIN (RFC 2308 §2.1), which is reserved for names that do not
+	// exist at all.
+	anyNameExists := false
 	for _, q := range m.Question {
 		switch q.Qtype {
 		case dns.TypeA, dns.TypeAAAA:
 			qType := dns.TypeToString[q.Qtype]
-			d.l.Debugf("Query for %s %s", qType, q.Name)
+			if debugEnabled {
-			ip := d.Query(q.Qtype, q.Name)
+				d.l.Debug("DNS query", "type", qType, "name", q.Name)
+			}
+			ip, nameExists := d.Query(q.Qtype, q.Name)
+			if nameExists {
+				anyNameExists = true
+			}
 			if ip.IsValid() {
 				rr, err := dns.NewRR(fmt.Sprintf("%s %s %s", q.Name, qType, ip))
 				if err == nil {
@@ -134,7 +342,9 @@ func (d *dnsRecords) parseQuery(m *dns.Msg, w dns.ResponseWriter) {
 			if !d.isSelfNebulaOrLocalhost(w.RemoteAddr().String()) {
 				return
 			}
-			d.l.Debugf("Query for TXT %s", q.Name)
+			if debugEnabled {
+				d.l.Debug("DNS query", "type", "TXT", "name", q.Name)
+			}
 			ip := d.QueryCert(q.Name)
 			if ip != "" {
 				rr, err := dns.NewRR(fmt.Sprintf("%s TXT %s", q.Name, ip))
@@ -145,12 +355,12 @@ func (d *dnsRecords) parseQuery(m *dns.Msg, w dns.ResponseWriter) {
 		}
 	}
 
-	if len(m.Answer) == 0 {
+	if len(m.Answer) == 0 && !anyNameExists {
 		m.Rcode = dns.RcodeNameError
 	}
 }
 
-func (d *dnsRecords) handleDnsRequest(w dns.ResponseWriter, r *dns.Msg) {
+func (d *dnsServer) handleDnsRequest(w dns.ResponseWriter, r *dns.Msg) {
 	m := new(dns.Msg)
 	m.SetReply(r)
 	m.Compress = false
@@ -163,21 +373,6 @@ func (d *dnsRecords) handleDnsRequest(w dns.ResponseWriter, r *dns.Msg) {
 	w.WriteMsg(m)
 }
 
-func dnsMain(l *logrus.Logger, cs *CertState, hostMap *HostMap, c *config.C) func() {
-	dnsR = newDnsRecords(l, cs, hostMap)
-
-	// attach request handler func
-	dns.HandleFunc(".", dnsR.handleDnsRequest)
-
-	c.RegisterReloadCallback(func(c *config.C) {
-		reloadDns(l, c)
-	})
-
-	return func() {
-		startDns(l, c)
-	}
-}
-
 func getDnsServerAddr(c *config.C) string {
 	dnsHost := strings.TrimSpace(c.GetString("lighthouse.dns.host", ""))
 	// Old guidance was to provide the literal `[::]` in `lighthouse.dns.host` but that won't resolve.
@@ -186,25 +381,3 @@ func getDnsServerAddr(c *config.C) string {
 	}
 	return net.JoinHostPort(dnsHost, strconv.Itoa(c.GetInt("lighthouse.dns.port", 53)))
 }
-
-func startDns(l *logrus.Logger, c *config.C) {
-	dnsAddr = getDnsServerAddr(c)
-	dnsServer = &dns.Server{Addr: dnsAddr, Net: "udp"}
-	l.WithField("dnsListener", dnsAddr).Info("Starting DNS responder")
-	err := dnsServer.ListenAndServe()
-	defer dnsServer.Shutdown()
-	if err != nil {
-		l.Errorf("Failed to start server: %s\n ", err.Error())
-	}
-}
-
-func reloadDns(l *logrus.Logger, c *config.C) {
-	if dnsAddr == getDnsServerAddr(c) {
-		l.Debug("No DNS server config change detected")
-		return
-	}
-
-	l.Debug("Restarting DNS server")
-	dnsServer.Shutdown()
-	go startDns(l, c)
-}

dns_server_test.go

@@ -1,19 +1,43 @@
 package nebula
 
 import (
+	"context"
+	"log/slog"
+	"net"
 	"net/netip"
+	"strconv"
 	"testing"
+	"time"
 
 	"github.com/miekg/dns"
-	"github.com/sirupsen/logrus"
 	"github.com/slackhq/nebula/config"
 	"github.com/stretchr/testify/assert"
+	"github.com/stretchr/testify/require"
 )
 
+type stubDNSWriter struct{}
+
+func (stubDNSWriter) LocalAddr() net.Addr { return &net.UDPAddr{} }
+func (stubDNSWriter) RemoteAddr() net.Addr {
+	return &net.UDPAddr{IP: net.ParseIP("127.0.0.1"), Port: 5353}
+}
+func (stubDNSWriter) Write([]byte) (int, error) { return 0, nil }
+func (stubDNSWriter) WriteMsg(*dns.Msg) error   { return nil }
+func (stubDNSWriter) Close() error              { return nil }
+func (stubDNSWriter) TsigStatus() error         { return nil }
+func (stubDNSWriter) TsigTimersOnly(bool)       {}
+func (stubDNSWriter) Hijack()                   {}
+
 func TestParsequery(t *testing.T) {
-	l := logrus.New()
+	l := slog.New(slog.DiscardHandler)
 	hostMap := &HostMap{}
-	ds := newDnsRecords(l, &CertState{}, hostMap)
+	ds := &dnsServer{
+		l:       l,
+		dnsMap4: make(map[string]netip.Addr),
+		dnsMap6: make(map[string]netip.Addr),
+		hostMap: hostMap,
+	}
+	ds.enabled.Store(true)
 	addrs := []netip.Addr{
 		netip.MustParseAddr("1.2.3.4"),
 		netip.MustParseAddr("1.2.3.5"),
@@ -21,18 +45,56 @@ func TestParsequery(t *testing.T) {
 		netip.MustParseAddr("fd01::25"),
 	}
 	ds.Add("test.com.com", addrs)
+	ds.Add("v4only.com.com", []netip.Addr{netip.MustParseAddr("1.2.3.6")})
+	ds.Add("v6only.com.com", []netip.Addr{netip.MustParseAddr("fd01::26")})
 
 	m := &dns.Msg{}
 	m.SetQuestion("test.com.com", dns.TypeA)
 	ds.parseQuery(m, nil)
 	assert.NotNil(t, m.Answer)
 	assert.Equal(t, "1.2.3.4", m.Answer[0].(*dns.A).A.String())
+	assert.Equal(t, dns.RcodeSuccess, m.Rcode)
 
 	m = &dns.Msg{}
 	m.SetQuestion("test.com.com", dns.TypeAAAA)
 	ds.parseQuery(m, nil)
 	assert.NotNil(t, m.Answer)
 	assert.Equal(t, "fd01::24", m.Answer[0].(*dns.AAAA).AAAA.String())
+	assert.Equal(t, dns.RcodeSuccess, m.Rcode)
+
+	// A known name with no record of the requested type should return NODATA
+	// (NOERROR with empty answer), not NXDOMAIN.
+	m = &dns.Msg{}
+	m.SetQuestion("v4only.com.com", dns.TypeAAAA)
+	ds.parseQuery(m, nil)
+	assert.Empty(t, m.Answer)
+	assert.Equal(t, dns.RcodeSuccess, m.Rcode)
+
+	m = &dns.Msg{}
+	m.SetQuestion("v6only.com.com", dns.TypeA)
+	ds.parseQuery(m, nil)
+	assert.Empty(t, m.Answer)
+	assert.Equal(t, dns.RcodeSuccess, m.Rcode)
+
+	// An unknown name should still return NXDOMAIN.
+	m = &dns.Msg{}
+	m.SetQuestion("unknown.com.com", dns.TypeA)
+	ds.parseQuery(m, nil)
+	assert.Empty(t, m.Answer)
+	assert.Equal(t, dns.RcodeNameError, m.Rcode)
+
+	// short lookups should not fail
+	m = &dns.Msg{}
+	m.Question = []dns.Question{{Name: "", Qtype: dns.TypeTXT, Qclass: dns.ClassINET}}
+	ds.parseQuery(m, stubDNSWriter{})
+	assert.Empty(t, m.Answer)
+	assert.Equal(t, dns.RcodeNameError, m.Rcode)
+
+	m = &dns.Msg{}
+	m.Question = []dns.Question{{Name: ".", Qtype: dns.TypeTXT, Qclass: dns.ClassINET}}
+	ds.parseQuery(m, stubDNSWriter{})
+	assert.Empty(t, m.Answer)
+	assert.Equal(t, dns.RcodeNameError, m.Rcode)
 }
 
 func Test_getDnsServerAddr(t *testing.T) {
@@ -71,3 +133,208 @@ func Test_getDnsServerAddr(t *testing.T) {
 	}
 	assert.Equal(t, "[::]:1", getDnsServerAddr(c))
 }
+
+func newTestDnsServer(t *testing.T) (*dnsServer, *config.C) {
+	t.Helper()
+	sl := slog.New(slog.DiscardHandler)
+	ds := &dnsServer{
+		l:       sl,
+		ctx:     context.Background(),
+		dnsMap4: make(map[string]netip.Addr),
+		dnsMap6: make(map[string]netip.Addr),
+		hostMap: &HostMap{},
+	}
+	ds.mux = dns.NewServeMux()
+	ds.mux.HandleFunc(".", ds.handleDnsRequest)
+	return ds, config.NewC(nil)
+}
+
+func setDnsConfig(c *config.C, host string, port string, amLighthouse, serveDns bool) {
+	c.Settings["lighthouse"] = map[string]any{
+		"am_lighthouse": amLighthouse,
+		"serve_dns":     serveDns,
+		"dns": map[string]any{
+			"host": host,
+			"port": port,
+		},
+	}
+}
+
+func TestDnsServer_reload_initial_disabled(t *testing.T) {
+	ds, c := newTestDnsServer(t)
+	setDnsConfig(c, "127.0.0.1", "0", true, false)
+
+	require.NoError(t, ds.reload(c, true))
+	assert.False(t, ds.enabled.Load())
+	assert.Equal(t, "127.0.0.1:0", ds.addr)
+	assert.Nil(t, ds.server)
+}
+
+func TestDnsServer_reload_initial_enabled(t *testing.T) {
+	ds, c := newTestDnsServer(t)
+	setDnsConfig(c, "127.0.0.1", "0", true, true)
+
+	require.NoError(t, ds.reload(c, true))
+	assert.True(t, ds.enabled.Load())
+	assert.Equal(t, "127.0.0.1:0", ds.addr)
+	// initial never starts a runner; that's Control.Start's job
+	assert.Nil(t, ds.server)
+}
+
+func TestDnsServer_reload_initial_serveDnsWithoutLighthouse(t *testing.T) {
+	ds, c := newTestDnsServer(t)
+	setDnsConfig(c, "127.0.0.1", "0", false, true)
+
+	require.NoError(t, ds.reload(c, true))
+	// Wants DNS but isn't a lighthouse: gated off, no runner.
+	assert.False(t, ds.enabled.Load())
+}
+
+func TestDnsServer_reload_sameAddr_noOp(t *testing.T) {
+	ds, c := newTestDnsServer(t)
+	setDnsConfig(c, "127.0.0.1", "0", true, true)
+
+	require.NoError(t, ds.reload(c, true))
+	// No server running yet, no addr change. Reload should not spawn anything.
+	require.NoError(t, ds.reload(c, false))
+	assert.True(t, ds.enabled.Load())
+	assert.Nil(t, ds.server)
+}
+
+func TestDnsServer_StartStop_lifecycle(t *testing.T) {
+	// Bind to a real (random) UDP port so we exercise the actual
+	// ListenAndServe + Shutdown plumbing including the started-chan race fix.
+	port := freeUDPPort(t)
+
+	ds, c := newTestDnsServer(t)
+	setDnsConfig(c, "127.0.0.1", port, true, true)
+	require.NoError(t, ds.reload(c, true))
+
+	done := make(chan struct{})
+	go func() {
+		ds.Start()
+		close(done)
+	}()
+
+	waitFor(t, func() bool {
+		ds.serverMu.Lock()
+		started := ds.started
+		ds.serverMu.Unlock()
+		if started == nil {
+			return false
+		}
+		select {
+		case <-started:
+			return true
+		default:
+			return false
+		}
+	})
+
+	ds.Stop()
+	select {
+	case <-done:
+	case <-time.After(5 * time.Second):
+		t.Fatal("Start did not return after Stop")
+	}
+}
+
+func TestDnsServer_Stop_beforeBind_doesNotHang(t *testing.T) {
+	// Stop called immediately after Start should not deadlock even if bind
+	// hasn't completed yet. This exercises the started-chan close-on-bind-fail
+	// path: by binding to an obviously bad port (privileged) we get a fast
+	// bind error before NotifyStartedFunc fires.
+	ds, c := newTestDnsServer(t)
+	// Use a port that should fail to bind (negative would be invalid, use a
+	// host that won't resolve to ensure listenUDP fails quickly).
+	setDnsConfig(c, "256.256.256.256", "53", true, true)
+	require.NoError(t, ds.reload(c, true))
+
+	done := make(chan struct{})
+	go func() {
+		ds.Start()
+		close(done)
+	}()
+
+	// Give Start a moment to attempt the bind and fail.
+	select {
+	case <-done:
+		// Bind failed and Start returned; Stop should be a no-op.
+	case <-time.After(time.Second):
+		t.Fatal("Start did not return after a bad bind")
+	}
+
+	stopped := make(chan struct{})
+	go func() {
+		ds.Stop()
+		close(stopped)
+	}()
+	select {
+	case <-stopped:
+	case <-time.After(time.Second):
+		t.Fatal("Stop hung after a failed bind")
+	}
+}
+
+func TestDnsServer_reload_disable_stopsRunningServer(t *testing.T) {
+	port := freeUDPPort(t)
+	ds, c := newTestDnsServer(t)
+	setDnsConfig(c, "127.0.0.1", port, true, true)
+	require.NoError(t, ds.reload(c, true))
+
+	startReturned := make(chan struct{})
+	go func() {
+		ds.Start()
+		close(startReturned)
+	}()
+	waitForBind(t, ds)
+
+	// Toggle serve_dns off; reload should shut the running server down.
+	setDnsConfig(c, "127.0.0.1", port, true, false)
+	require.NoError(t, ds.reload(c, false))
+	select {
+	case <-startReturned:
+	case <-time.After(5 * time.Second):
+		t.Fatal("Start did not return after reload disabled DNS")
+	}
+	assert.False(t, ds.enabled.Load())
+}
+
+func freeUDPPort(t *testing.T) string {
+	t.Helper()
+	conn, err := net.ListenPacket("udp", "127.0.0.1:0")
+	require.NoError(t, err)
+	port := conn.LocalAddr().(*net.UDPAddr).Port
+	require.NoError(t, conn.Close())
+	return strconv.Itoa(port)
+}
+
+func waitForBind(t *testing.T, ds *dnsServer) {
+	t.Helper()
+	waitFor(t, func() bool {
+		ds.serverMu.Lock()
+		started := ds.started
+		ds.serverMu.Unlock()
+		if started == nil {
+			return false
+		}
+		select {
+		case <-started:
+			return true
+		default:
+			return false
+		}
+	})
+}
+
+func waitFor(t *testing.T, cond func() bool) {
+	t.Helper()
+	deadline := time.Now().Add(5 * time.Second)
+	for time.Now().Before(deadline) {
+		if cond() {
+			return
+		}
+		time.Sleep(5 * time.Millisecond)
+	}
+	t.Fatal("timed out waiting for condition")
+}

e2e/handshakes_test.go

@@ -11,7 +11,6 @@ import (
 
 	"github.com/google/gopacket"
 	"github.com/google/gopacket/layers"
-	"github.com/sirupsen/logrus"
 	"github.com/slackhq/nebula"
 	"github.com/slackhq/nebula/cert"
 	"github.com/slackhq/nebula/cert_test"
@@ -749,7 +748,6 @@ func TestStage1RaceRelays2(t *testing.T) {
 	myControl, myVpnIpNet, myUdpAddr, _ := newSimpleServer(cert.Version1, ca, caKey, "me     ", "10.128.0.1/24", m{"relay": m{"use_relays": true}})
 	relayControl, relayVpnIpNet, relayUdpAddr, _ := newSimpleServer(cert.Version1, ca, caKey, "relay  ", "10.128.0.128/24", m{"relay": m{"am_relay": true}})
 	theirControl, theirVpnIpNet, theirUdpAddr, _ := newSimpleServer(cert.Version1, ca, caKey, "them   ", "10.128.0.2/24", m{"relay": m{"use_relays": true}})
-	l := NewTestLogger()
 
 	// Teach my how to get to the relay and that their can be reached via the relay
 	myControl.InjectLightHouseAddr(relayVpnIpNet[0].Addr(), relayUdpAddr)
@@ -771,49 +769,41 @@ func TestStage1RaceRelays2(t *testing.T) {
 	theirControl.Start()
 
 	r.Log("Get a tunnel between me and relay")
-	l.Info("Get a tunnel between me and relay")
 	assertTunnel(t, myVpnIpNet[0].Addr(), relayVpnIpNet[0].Addr(), myControl, relayControl, r)
 
 	r.Log("Get a tunnel between them and relay")
-	l.Info("Get a tunnel between them and relay")
 	assertTunnel(t, theirVpnIpNet[0].Addr(), relayVpnIpNet[0].Addr(), theirControl, relayControl, r)
 
 	r.Log("Trigger a handshake from both them and me via relay to them and me")
-	l.Info("Trigger a handshake from both them and me via relay to them and me")
 	myControl.InjectTunUDPPacket(theirVpnIpNet[0].Addr(), 80, myVpnIpNet[0].Addr(), 80, []byte("Hi from me"))
 	theirControl.InjectTunUDPPacket(myVpnIpNet[0].Addr(), 80, theirVpnIpNet[0].Addr(), 80, []byte("Hi from them"))
 
 	//r.RouteUntilAfterMsgType(myControl, header.Control, header.MessageNone)
 	//r.RouteUntilAfterMsgType(theirControl, header.Control, header.MessageNone)
 
-	r.Log("Wait for a packet from them to me")
+	r.Log("Wait for a packet from them to me; myControl")
-	l.Info("Wait for a packet from them to me; myControl")
 	r.RouteForAllUntilTxTun(myControl)
-	l.Info("Wait for a packet from them to me; theirControl")
+	r.Log("Wait for a packet from them to me; theirControl")
 	r.RouteForAllUntilTxTun(theirControl)
 
 	r.Log("Assert the tunnel works")
-	l.Info("Assert the tunnel works")
 	assertTunnel(t, theirVpnIpNet[0].Addr(), myVpnIpNet[0].Addr(), theirControl, myControl, r)
 
 	t.Log("Wait until we remove extra tunnels")
-	l.Info("Wait until we remove extra tunnels")
+	t.Logf("Waiting for hostinfos to be removed... myControl=%d theirControl=%d relayControl=%d",
-	l.WithFields(
+		len(myControl.GetHostmap().Indexes),
-		logrus.Fields{
+		len(theirControl.GetHostmap().Indexes),
-			"myControl":    len(myControl.GetHostmap().Indexes),
+		len(relayControl.GetHostmap().Indexes),
-			"theirControl": len(theirControl.GetHostmap().Indexes),
+	)
-			"relayControl": len(relayControl.GetHostmap().Indexes),
-		}).Info("Waiting for hostinfos to be removed...")
 	hostInfos := len(myControl.GetHostmap().Indexes) + len(theirControl.GetHostmap().Indexes) + len(relayControl.GetHostmap().Indexes)
 	retries := 60
 	for hostInfos > 6 && retries > 0 {
 		hostInfos = len(myControl.GetHostmap().Indexes) + len(theirControl.GetHostmap().Indexes) + len(relayControl.GetHostmap().Indexes)
-		l.WithFields(
+		t.Logf("Waiting for hostinfos to be removed... myControl=%d theirControl=%d relayControl=%d",
-			logrus.Fields{
+			len(myControl.GetHostmap().Indexes),
-				"myControl":    len(myControl.GetHostmap().Indexes),
+			len(theirControl.GetHostmap().Indexes),
-				"theirControl": len(theirControl.GetHostmap().Indexes),
+			len(relayControl.GetHostmap().Indexes),
-				"relayControl": len(relayControl.GetHostmap().Indexes),
+		)
-			}).Info("Waiting for hostinfos to be removed...")
 		assertTunnel(t, myVpnIpNet[0].Addr(), theirVpnIpNet[0].Addr(), myControl, theirControl, r)
 		t.Log("Connection manager hasn't ticked yet")
 		time.Sleep(time.Second)
@@ -821,7 +811,6 @@ func TestStage1RaceRelays2(t *testing.T) {
 	}
 
 	r.Log("Assert the tunnel works")
-	l.Info("Assert the tunnel works")
 	assertTunnel(t, theirVpnIpNet[0].Addr(), myVpnIpNet[0].Addr(), theirControl, myControl, r)
 
 	myControl.Stop()
@@ -1369,6 +1358,81 @@ func TestV2NonPrimaryWithOffNetLighthouse(t *testing.T) {
 	theirControl.Stop()
 }
 
+func TestLighthouseUpdateOnReload(t *testing.T) {
+	ca, _, caKey, _ := cert_test.NewTestCaCert(cert.Version2, cert.Curve_CURVE25519, time.Now(), time.Now().Add(10*time.Minute), nil, nil, []string{})
+
+	// Create the lighthouse
+	lhControl, lhVpnIpNet, lhUdpAddr, _ := newSimpleServer(cert.Version2, ca, caKey, "lh", "10.128.0.1/24", m{"lighthouse": m{"am_lighthouse": true}})
+
+	// Create a client with NO lighthouse configured and a long update interval.
+	// The initial SendUpdate at startup will be a no-op since no lighthouses are known.
+	myControl, myVpnIpNet, _, myConfig := newSimpleServer(cert.Version2, ca, caKey, "me", "10.128.0.2/24", m{
+		"lighthouse": m{
+			"interval": 600,
+			"local_allow_list": m{
+				"10.0.0.0/24": true,
+				"::/0":        false,
+			},
+		},
+	})
+
+	r := router.NewR(t, lhControl, myControl)
+	defer r.RenderFlow()
+
+	lhControl.Start()
+	myControl.Start()
+
+	// Drain any startup packets (there should be none meaningful)
+	r.FlushAll()
+
+	// Verify lighthouse has no knowledge of the client
+	assert.Nil(t, lhControl.QueryLighthouse(myVpnIpNet[0].Addr()))
+
+	// Build a new config that adds the lighthouse
+	newSettings := make(m)
+	for k, v := range myConfig.Settings {
+		newSettings[k] = v
+	}
+	newSettings["static_host_map"] = m{
+		lhVpnIpNet[0].Addr().String(): []any{lhUdpAddr.String()},
+	}
+	newSettings["lighthouse"] = m{
+		"hosts":    []any{lhVpnIpNet[0].Addr().String()},
+		"interval": 600,
+		"local_allow_list": m{
+			"10.0.0.0/24": true,
+			"::/0":        false,
+		},
+	}
+	newCfg, err := yaml.Marshal(newSettings)
+	require.NoError(t, err)
+
+	// Reload the config. The lighthouse.hosts change triggers TriggerUpdate,
+	// which wakes the update worker. It calls SendUpdate, initiating a
+	// handshake to the new lighthouse and caching the HostUpdateNotification.
+	require.NoError(t, myConfig.ReloadConfigString(string(newCfg)))
+
+	// Route until the lighthouse receives the HostUpdateNotification.
+	// This covers: handshake stage 1, stage 2, then the cached update.
+	done := make(chan struct{})
+	go func() {
+		r.RouteForAllUntilAfterMsgTypeTo(lhControl, header.LightHouse, 0)
+		close(done)
+	}()
+	select {
+	case <-done:
+	case <-time.After(5 * time.Second):
+		t.Fatal("timed out waiting for lighthouse update after config reload")
+	}
+
+	// Verify lighthouse now has the client's addresses
+	assert.NotNil(t, lhControl.QueryLighthouse(myVpnIpNet[0].Addr()))
+
+	r.RenderHostmaps("Final hostmaps", lhControl, myControl)
+	lhControl.Stop()
+	myControl.Stop()
+}
+
 func TestGoodHandshakeUnsafeDest(t *testing.T) {
 	unsafePrefix := "192.168.6.0/24"
 	ca, _, caKey, _ := cert_test.NewTestCaCert(cert.Version2, cert.Curve_CURVE25519, time.Now(), time.Now().Add(10*time.Minute), nil, nil, []string{})

e2e/helpers_test.go

@@ -4,7 +4,6 @@
 package e2e
 
 import (
-	"fmt"
 	"io"
 	"net/netip"
 	"os"
@@ -12,15 +11,18 @@ import (
 	"testing"
 	"time"
 
+	"log/slog"
+
 	"dario.cat/mergo"
 	"github.com/google/gopacket"
 	"github.com/google/gopacket/layers"
-	"github.com/sirupsen/logrus"
+
 	"github.com/slackhq/nebula"
 	"github.com/slackhq/nebula/cert"
 	"github.com/slackhq/nebula/cert_test"
 	"github.com/slackhq/nebula/config"
 	"github.com/slackhq/nebula/e2e/router"
+	"github.com/slackhq/nebula/logging"
 	"github.com/stretchr/testify/assert"
 	"github.com/stretchr/testify/require"
 	"go.yaml.in/yaml/v3"
@@ -132,8 +134,7 @@ func newSimpleServerWithUdpAndUnsafeNetworks(v cert.Version, caCrt cert.Certific
 			"port": udpAddr.Port(),
 		},
 		"logging": m{
-			"timestamp_format": fmt.Sprintf("%v 15:04:05.000000", name),
+			"level": testLogLevelName(),
-			"level":            l.Level.String(),
 		},
 		"timers": m{
 			"pending_deletion_interval": 2,
@@ -234,8 +235,7 @@ func newServer(caCrt []cert.Certificate, certs []cert.Certificate, key []byte, o
 			"port": udpAddr.Port(),
 		},
 		"logging": m{
-			"timestamp_format": fmt.Sprintf("%v 15:04:05.000000", certs[0].Name()),
+			"level": testLogLevelName(),
-			"level":            l.Level.String(),
 		},
 		"timers": m{
 			"pending_deletion_interval": 2,
@@ -379,24 +379,32 @@ func getAddrs(ns []netip.Prefix) []netip.Addr {
 	return a
 }
 
-func NewTestLogger() *logrus.Logger {
+func NewTestLogger() *slog.Logger {
-	l := logrus.New()
-
 	v := os.Getenv("TEST_LOGS")
 	if v == "" {
-		l.SetOutput(io.Discard)
+		return slog.New(slog.NewTextHandler(io.Discard, nil))
-		l.SetLevel(logrus.PanicLevel)
-		return l
 	}
 
+	level := slog.LevelInfo
 	switch v {
 	case "2":
-		l.SetLevel(logrus.DebugLevel)
+		level = slog.LevelDebug
 	case "3":
-		l.SetLevel(logrus.TraceLevel)
+		level = logging.LevelTrace
-	default:
+	}
-		l.SetLevel(logrus.InfoLevel)
+	return slog.New(slog.NewTextHandler(os.Stderr, &slog.HandlerOptions{Level: level}))
 }
 
-	return l
+// testLogLevelName returns the level name string accepted by logging.ApplyConfig
+// for the current TEST_LOGS setting. Kept in sync with NewTestLogger.
+func testLogLevelName() string {
+	switch os.Getenv("TEST_LOGS") {
+	case "2":
+		return "debug"
+	case "3":
+		return "trace"
+	case "":
+		return "info"
+	}
+	return "info"
 }

examples/config.yml

@@ -292,24 +292,21 @@ tun:
 
 # Configure logging level
 logging:
-  # panic, fatal, error, warning, info, or debug. Default is info and is reloadable.
+  # trace, debug, info, warn, or error. Default is info and is reloadable.
-  #NOTE: Debug mode can log remotely controlled/untrusted data which can quickly fill a disk in some
+  # fatal and panic are accepted for backwards compatibility and map to error.
-  # scenarios. Debug logging is also CPU intensive and will decrease performance overall.
+  #NOTE: Debug and trace modes can log remotely controlled/untrusted data which can quickly fill a disk in some
-  # Only enable debug logging while actively investigating an issue.
+  # scenarios. Debug and trace logging are also CPU intensive and will decrease performance overall.
+  # Only enable debug or trace logging while actively investigating an issue.
   level: info
-  # json or text formats currently available. Default is text
+  # json or text formats currently available. Default is text.
   format: text
-  # Disable timestamp logging. useful when output is redirected to logging system that already adds timestamps. Default is false
+  # Disable timestamp logging. Useful when output is redirected to a logging system that already adds timestamps. Default is false.
   #disable_timestamp: true
-  # timestamp format is specified in Go time format, see:
+  # Timestamps use RFC3339Nano ("2006-01-02T15:04:05.999999999Z07:00") and are not configurable.
-  #     https://golang.org/pkg/time/#pkg-constants
-  # default when `format: json`: "2006-01-02T15:04:05Z07:00" (RFC3339)
-  # default when `format: text`:
-  #     when TTY attached: seconds since beginning of execution
-  #     otherwise: "2006-01-02T15:04:05Z07:00" (RFC3339)
-  # As an example, to log as RFC3339 with millisecond precision, set to:
-  #timestamp_format: "2006-01-02T15:04:05.000Z07:00"
 
+# The stats section is reloadable. A HUP may change the backend, toggle stats
+# on or off, switch the listen/host address, or pick up new DNS for the
+# configured graphite host.
 #stats:
   #type: graphite
   #prefix: nebula
@@ -327,10 +324,12 @@ logging:
   # enables counter metrics for meta packets
   #   e.g.: `messages.tx.handshake`
   # NOTE: `message.{tx,rx}.recv_error` is always emitted
+  # Not reloadable.
   #message_metrics: false
 
   # enables detailed counter metrics for lighthouse packets
   #   e.g.: `lighthouse.rx.HostQuery`
+  # Not reloadable.
   #lighthouse_metrics: false
 
 # Handshake Manager Settings

examples/go_service/main.go

@@ -7,9 +7,9 @@ import (
 	"net"
 	"os"
 
-	"github.com/sirupsen/logrus"
 	"github.com/slackhq/nebula"
 	"github.com/slackhq/nebula/config"
+	"github.com/slackhq/nebula/logging"
 	"github.com/slackhq/nebula/overlay"
 	"github.com/slackhq/nebula/service"
 )
@@ -64,8 +64,7 @@ pki:
 		return err
 	}
 
-	logger := logrus.New()
+	logger := logging.NewLogger(os.Stdout)
-	logger.Out = os.Stdout
 
 	ctrl, err := nebula.Main(&cfg, false, "custom-app", logger, overlay.NewUserDeviceFromConfig)
 	if err != nil {

firewall.go

@@ -1,11 +1,13 @@
 package nebula
 
 import (
+	"context"
 	"crypto/sha256"
 	"encoding/hex"
 	"errors"
 	"fmt"
 	"hash/fnv"
+	"log/slog"
 	"net/netip"
 	"reflect"
 	"slices"
@@ -16,7 +18,6 @@ import (
 
 	"github.com/gaissmai/bart"
 	"github.com/rcrowley/go-metrics"
-	"github.com/sirupsen/logrus"
 	"github.com/slackhq/nebula/cert"
 	"github.com/slackhq/nebula/config"
 	"github.com/slackhq/nebula/firewall"
@@ -67,7 +68,7 @@ type Firewall struct {
 	incomingMetrics     firewallMetrics
 	outgoingMetrics     firewallMetrics
 
-	l *logrus.Logger
+	l *slog.Logger
 }
 
 type firewallMetrics struct {
@@ -131,7 +132,7 @@ type firewallLocalCIDR struct {
 
 // NewFirewall creates a new Firewall object. A TimerWheel is created for you from the provided timeouts.
 // The certificate provided should be the highest version loaded in memory.
-func NewFirewall(l *logrus.Logger, tcpTimeout, UDPTimeout, defaultTimeout time.Duration, c cert.Certificate) *Firewall {
+func NewFirewall(l *slog.Logger, tcpTimeout, UDPTimeout, defaultTimeout time.Duration, c cert.Certificate) *Firewall {
 	//TODO: error on 0 duration
 	var tmin, tmax time.Duration
 
@@ -191,7 +192,7 @@ func NewFirewall(l *logrus.Logger, tcpTimeout, UDPTimeout, defaultTimeout time.D
 	}
 }
 
-func NewFirewallFromConfig(l *logrus.Logger, cs *CertState, c *config.C) (*Firewall, error) {
+func NewFirewallFromConfig(l *slog.Logger, cs *CertState, c *config.C) (*Firewall, error) {
 	certificate := cs.getCertificate(cert.Version2)
 	if certificate == nil {
 		certificate = cs.getCertificate(cert.Version1)
@@ -219,7 +220,7 @@ func NewFirewallFromConfig(l *logrus.Logger, cs *CertState, c *config.C) (*Firew
 	case "drop":
 		fw.InSendReject = false
 	default:
-		l.WithField("action", inboundAction).Warn("invalid firewall.inbound_action, defaulting to `drop`")
+		l.Warn("invalid firewall.inbound_action, defaulting to `drop`", "action", inboundAction)
 		fw.InSendReject = false
 	}
 
@@ -230,7 +231,7 @@ func NewFirewallFromConfig(l *logrus.Logger, cs *CertState, c *config.C) (*Firew
 	case "drop":
 		fw.OutSendReject = false
 	default:
-		l.WithField("action", outboundAction).Warn("invalid firewall.outbound_action, defaulting to `drop`")
+		l.Warn("invalid firewall.outbound_action, defaulting to `drop`", "action", outboundAction)
 		fw.OutSendReject = false
 	}
 
@@ -268,7 +269,7 @@ func (f *Firewall) AddRule(incoming bool, proto uint8, startPort int32, endPort
 	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")
+			f.l.Warn("ignoring port specification for ICMP firewall rule", "startPort", startPort)
 		}
 		startPort = firewall.PortAny
 		endPort = firewall.PortAny
@@ -290,8 +291,9 @@ func (f *Firewall) AddRule(incoming bool, proto uint8, startPort int32, endPort
 	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}).
+	f.l.Info("Firewall rule added",
-		Info("Firewall rule added")
+		"firewallRule", m{"direction": direction, "proto": proto, "startPort": startPort, "endPort": endPort, "groups": groups, "host": host, "cidr": cidr, "localCidr": localCidr, "caName": caName, "caSha": caSha},
+	)
 
 	return fp.addRule(f, startPort, endPort, groups, host, cidr, localCidr, caName, caSha)
 }
@@ -314,7 +316,7 @@ func (f *Firewall) GetRuleHashes() string {
 	return "SHA:" + f.GetRuleHash() + ",FNV:" + strconv.FormatUint(uint64(f.GetRuleHashFNV()), 10)
 }
 
-func AddFirewallRulesFromConfig(l *logrus.Logger, inbound bool, c *config.C, fw FirewallInterface) error {
+func AddFirewallRulesFromConfig(l *slog.Logger, inbound bool, c *config.C, fw FirewallInterface) error {
 	var table string
 	if inbound {
 		table = "firewall.inbound"
@@ -372,7 +374,7 @@ func AddFirewallRulesFromConfig(l *logrus.Logger, inbound bool, c *config.C, fw
 			startPort = firewall.PortAny
 			endPort = firewall.PortAny
 			if sPort != "" {
-				l.WithField("port", sPort).Warn("ignoring port specification for ICMP firewall rule")
+				l.Warn("ignoring port specification for ICMP firewall rule", "port", sPort)
 			}
 		default:
 			return fmt.Errorf("%s rule #%v; proto was not understood; `%s`", table, i, r.Proto)
@@ -396,7 +398,11 @@ func AddFirewallRulesFromConfig(l *logrus.Logger, inbound bool, c *config.C, fw
 		}
 
 		if warning := r.sanity(); warning != nil {
-			l.Warnf("%s rule #%v; %s", table, i, warning)
+			l.Warn("firewall rule sanity check",
+				"table", table,
+				"rule", i,
+				"warning", warning,
+			)
 		}
 
 		err = fw.AddRule(inbound, proto, startPort, endPort, r.Groups, r.Host, r.Cidr, r.LocalCidr, r.CAName, r.CASha)
@@ -528,26 +534,26 @@ func (f *Firewall) inConns(fp firewall.Packet, h *HostInfo, caPool *cert.CAPool,
 
 		// We now know which firewall table to check against
 		if !table.match(fp, c.incoming, h.ConnectionState.peerCert, caPool) {
-			if f.l.Level >= logrus.DebugLevel {
+			if f.l.Enabled(context.Background(), slog.LevelDebug) {
-				h.logger(f.l).
+				h.logger(f.l).Debug("dropping old conntrack entry, does not match new ruleset",
-					WithField("fwPacket", fp).
+					"fwPacket", fp,
-					WithField("incoming", c.incoming).
+					"incoming", c.incoming,
-					WithField("rulesVersion", f.rulesVersion).
+					"rulesVersion", f.rulesVersion,
-					WithField("oldRulesVersion", c.rulesVersion).
+					"oldRulesVersion", c.rulesVersion,
-					Debugln("dropping old conntrack entry, does not match new ruleset")
+				)
 			}
 			delete(conntrack.Conns, fp)
 			conntrack.Unlock()
 			return false
 		}
 
-		if f.l.Level >= logrus.DebugLevel {
+		if f.l.Enabled(context.Background(), slog.LevelDebug) {
-			h.logger(f.l).
+			h.logger(f.l).Debug("keeping old conntrack entry, does match new ruleset",
-				WithField("fwPacket", fp).
+				"fwPacket", fp,
-				WithField("incoming", c.incoming).
+				"incoming", c.incoming,
-				WithField("rulesVersion", f.rulesVersion).
+				"rulesVersion", f.rulesVersion,
-				WithField("oldRulesVersion", c.rulesVersion).
+				"oldRulesVersion", c.rulesVersion,
-				Debugln("keeping old conntrack entry, does match new ruleset")
+			)
 		}
 
 		c.rulesVersion = f.rulesVersion
@@ -935,7 +941,7 @@ type rule struct {
 	CASha     string
 }
 
-func convertRule(l *logrus.Logger, p any, table string, i int) (rule, error) {
+func convertRule(l *slog.Logger, p any, table string, i int) (rule, error) {
 	r := rule{}
 
 	m, ok := p.(map[string]any)
@@ -966,7 +972,10 @@ func convertRule(l *logrus.Logger, p any, table string, i int) (rule, error) {
 			return r, errors.New("group should contain a single value, an array with more than one entry was provided")
 		}
 
-		l.Warnf("%s rule #%v; group was an array with a single value, converting to simple value", table, i)
+		l.Warn("group was an array with a single value, converting to simple value",
+			"table", table,
+			"rule", i,
+		)
 		m["group"] = v[0]
 	}
 

firewall/cache.go

@@ -1,10 +1,10 @@
 package firewall
 
 import (
+	"context"
+	"log/slog"
 	"sync/atomic"
 	"time"
-
-	"github.com/sirupsen/logrus"
 )
 
 // ConntrackCache is used as a local routine cache to know if a given flow
@@ -15,41 +15,49 @@ type ConntrackCacheTicker struct {
 	cacheV    uint64
 	cacheTick atomic.Uint64
 
+	l     *slog.Logger
 	cache ConntrackCache
 }
 
-func NewConntrackCacheTicker(d time.Duration) *ConntrackCacheTicker {
+func NewConntrackCacheTicker(ctx context.Context, l *slog.Logger, d time.Duration) *ConntrackCacheTicker {
 	if d == 0 {
 		return nil
 	}
 
 	c := &ConntrackCacheTicker{
+		l:     l,
 		cache: ConntrackCache{},
 	}
 
-	go c.tick(d)
+	go c.tick(ctx, d)
 
 	return c
 }
 
-func (c *ConntrackCacheTicker) tick(d time.Duration) {
+func (c *ConntrackCacheTicker) tick(ctx context.Context, d time.Duration) {
+	t := time.NewTicker(d)
+	defer t.Stop()
 	for {
-		time.Sleep(d)
+		select {
+		case <-ctx.Done():
+			return
+		case <-t.C:
 			c.cacheTick.Add(1)
 		}
 	}
+}
 
 // Get checks if the cache ticker has moved to the next version before returning
 // the map. If it has moved, we reset the map.
-func (c *ConntrackCacheTicker) Get(l *logrus.Logger) ConntrackCache {
+func (c *ConntrackCacheTicker) Get() ConntrackCache {
 	if c == nil {
 		return nil
 	}
 	if tick := c.cacheTick.Load(); tick != c.cacheV {
 		c.cacheV = tick
 		if ll := len(c.cache); ll > 0 {
-			if l.Level == logrus.DebugLevel {
+			if c.l.Enabled(context.Background(), slog.LevelDebug) {
-				l.WithField("len", ll).Debug("resetting conntrack cache")
+				c.l.Debug("resetting conntrack cache", "len", ll)
 			}
 			c.cache = make(ConntrackCache, ll)
 		}

firewall/cache_test.go

@@ -0,0 +1,69 @@
+package firewall
+
+import (
+	"bytes"
+	"log/slog"
+	"strings"
+	"testing"
+
+	"github.com/slackhq/nebula/test"
+	"github.com/stretchr/testify/assert"
+)
+
+// The tests below pin the log format produced by ConntrackCacheTicker.Get
+// so changes cannot silently break what operators are grepping for. The
+// ticker's internal state (cache + cacheTick) is poked directly to avoid
+// racing a goroutine-driven tick in tests.
+
+func newFixedTicker(t *testing.T, l *slog.Logger, cacheLen int) *ConntrackCacheTicker {
+	t.Helper()
+	c := &ConntrackCacheTicker{
+		l:     l,
+		cache: make(ConntrackCache, cacheLen),
+	}
+	for i := 0; i < cacheLen; i++ {
+		c.cache[Packet{LocalPort: uint16(i) + 1}] = struct{}{}
+	}
+	c.cacheTick.Store(1) // cacheV starts at 0, so Get() takes the reset path
+	return c
+}
+
+func TestConntrackCacheTicker_Get_TextFormat(t *testing.T) {
+	buf := &bytes.Buffer{}
+	l := test.NewLoggerWithOutputAndLevel(buf, slog.LevelDebug)
+
+	c := newFixedTicker(t, l, 3)
+	c.Get()
+
+	assert.Equal(t, "level=DEBUG msg=\"resetting conntrack cache\" len=3\n", buf.String())
+}
+
+func TestConntrackCacheTicker_Get_JSONFormat(t *testing.T) {
+	buf := &bytes.Buffer{}
+	l := test.NewJSONLoggerWithOutput(buf, slog.LevelDebug)
+
+	c := newFixedTicker(t, l, 2)
+	c.Get()
+
+	assert.JSONEq(t, `{"level":"DEBUG","msg":"resetting conntrack cache","len":2}`, strings.TrimSpace(buf.String()))
+}
+
+func TestConntrackCacheTicker_Get_QuietBelowDebug(t *testing.T) {
+	buf := &bytes.Buffer{}
+	l := test.NewLoggerWithOutputAndLevel(buf, slog.LevelInfo)
+
+	c := newFixedTicker(t, l, 5)
+	c.Get()
+
+	assert.Empty(t, buf.String())
+}
+
+func TestConntrackCacheTicker_Get_QuietWhenCacheEmpty(t *testing.T) {
+	buf := &bytes.Buffer{}
+	l := test.NewLoggerWithOutputAndLevel(buf, slog.LevelDebug)
+
+	c := newFixedTicker(t, l, 0)
+	c.Get()
+
+	assert.Empty(t, buf.String())
+}

firewall_test.go

@@ -3,13 +3,13 @@ package nebula
 import (
 	"bytes"
 	"errors"
+	"log/slog"
 	"math"
 	"net/netip"
 	"testing"
 	"time"
 
 	"github.com/gaissmai/bart"
-	"github.com/sirupsen/logrus"
 	"github.com/slackhq/nebula/cert"
 	"github.com/slackhq/nebula/config"
 	"github.com/slackhq/nebula/firewall"
@@ -58,9 +58,8 @@ func TestNewFirewall(t *testing.T) {
 }
 
 func TestFirewall_AddRule(t *testing.T) {
-	l := test.NewLogger()
 	ob := &bytes.Buffer{}
-	l.SetOutput(ob)
+	l := test.NewLoggerWithOutput(ob)
 
 	c := &dummyCert{}
 	fw := NewFirewall(l, time.Second, time.Minute, time.Hour, c)
@@ -177,9 +176,8 @@ func TestFirewall_AddRule(t *testing.T) {
 }
 
 func TestFirewall_Drop(t *testing.T) {
-	l := test.NewLogger()
 	ob := &bytes.Buffer{}
-	l.SetOutput(ob)
+	l := test.NewLoggerWithOutput(ob)
 	myVpnNetworksTable := new(bart.Lite)
 	myVpnNetworksTable.Insert(netip.MustParsePrefix("1.1.1.1/8"))
 	p := firewall.Packet{
@@ -254,9 +252,8 @@ func TestFirewall_Drop(t *testing.T) {
 }
 
 func TestFirewall_DropV6(t *testing.T) {
-	l := test.NewLogger()
 	ob := &bytes.Buffer{}
-	l.SetOutput(ob)
+	l := test.NewLoggerWithOutput(ob)
 
 	myVpnNetworksTable := new(bart.Lite)
 	myVpnNetworksTable.Insert(netip.MustParsePrefix("fd00::/7"))
@@ -485,9 +482,8 @@ func BenchmarkFirewallTable_match(b *testing.B) {
 }
 
 func TestFirewall_Drop2(t *testing.T) {
-	l := test.NewLogger()
 	ob := &bytes.Buffer{}
-	l.SetOutput(ob)
+	l := test.NewLoggerWithOutput(ob)
 	myVpnNetworksTable := new(bart.Lite)
 	myVpnNetworksTable.Insert(netip.MustParsePrefix("1.1.1.1/8"))
 
@@ -544,9 +540,8 @@ func TestFirewall_Drop2(t *testing.T) {
 }
 
 func TestFirewall_Drop3(t *testing.T) {
-	l := test.NewLogger()
 	ob := &bytes.Buffer{}
-	l.SetOutput(ob)
+	l := test.NewLoggerWithOutput(ob)
 	myVpnNetworksTable := new(bart.Lite)
 	myVpnNetworksTable.Insert(netip.MustParsePrefix("1.1.1.1/8"))
 
@@ -633,9 +628,8 @@ func TestFirewall_Drop3(t *testing.T) {
 }
 
 func TestFirewall_Drop3V6(t *testing.T) {
-	l := test.NewLogger()
 	ob := &bytes.Buffer{}
-	l.SetOutput(ob)
+	l := test.NewLoggerWithOutput(ob)
 	myVpnNetworksTable := new(bart.Lite)
 	myVpnNetworksTable.Insert(netip.MustParsePrefix("fd00::/7"))
 
@@ -671,9 +665,8 @@ func TestFirewall_Drop3V6(t *testing.T) {
 }
 
 func TestFirewall_DropConntrackReload(t *testing.T) {
-	l := test.NewLogger()
 	ob := &bytes.Buffer{}
-	l.SetOutput(ob)
+	l := test.NewLoggerWithOutput(ob)
 	myVpnNetworksTable := new(bart.Lite)
 	myVpnNetworksTable.Insert(netip.MustParsePrefix("1.1.1.1/8"))
 
@@ -736,9 +729,8 @@ func TestFirewall_DropConntrackReload(t *testing.T) {
 }
 
 func TestFirewall_ICMPPortBehavior(t *testing.T) {
-	l := test.NewLogger()
 	ob := &bytes.Buffer{}
-	l.SetOutput(ob)
+	l := test.NewLoggerWithOutput(ob)
 	myVpnNetworksTable := new(bart.Lite)
 	myVpnNetworksTable.Insert(netip.MustParsePrefix("1.1.1.1/8"))
 
@@ -880,9 +872,8 @@ func TestFirewall_ICMPPortBehavior(t *testing.T) {
 }
 
 func TestFirewall_DropIPSpoofing(t *testing.T) {
-	l := test.NewLogger()
 	ob := &bytes.Buffer{}
-	l.SetOutput(ob)
+	l := test.NewLoggerWithOutput(ob)
 	myVpnNetworksTable := new(bart.Lite)
 	myVpnNetworksTable.Insert(netip.MustParsePrefix("192.0.2.1/24"))
 
@@ -1045,25 +1036,25 @@ func TestNewFirewallFromConfig(t *testing.T) {
 	cs, err := newCertState(cert.Version2, nil, c, false, cert.Curve_CURVE25519, nil)
 	require.NoError(t, err)
 
-	conf := config.NewC(l)
+	conf := config.NewC(test.NewLogger())
 	conf.Settings["firewall"] = map[string]any{"outbound": "asdf"}
 	_, err = NewFirewallFromConfig(l, cs, conf)
 	require.EqualError(t, err, "firewall.outbound failed to parse, should be an array of rules")
 
 	// Test both port and code
-	conf = config.NewC(l)
+	conf = config.NewC(test.NewLogger())
 	conf.Settings["firewall"] = map[string]any{"outbound": []any{map[string]any{"port": "1", "code": "2"}}}
 	_, err = NewFirewallFromConfig(l, cs, conf)
 	require.EqualError(t, err, "firewall.outbound rule #0; only one of port or code should be provided")
 
 	// Test missing host, group, cidr, ca_name and ca_sha
-	conf = config.NewC(l)
+	conf = config.NewC(test.NewLogger())
 	conf.Settings["firewall"] = map[string]any{"outbound": []any{map[string]any{}}}
 	_, err = NewFirewallFromConfig(l, cs, conf)
 	require.EqualError(t, err, "firewall.outbound rule #0; at least one of host, group, cidr, local_cidr, ca_name, or ca_sha must be provided")
 
 	// Test code/port error
-	conf = config.NewC(l)
+	conf = config.NewC(test.NewLogger())
 	conf.Settings["firewall"] = map[string]any{"outbound": []any{map[string]any{"code": "a", "host": "testh", "proto": "any"}}}
 	_, err = NewFirewallFromConfig(l, cs, conf)
 	require.EqualError(t, err, "firewall.outbound rule #0; code was not a number; `a`")
@@ -1073,25 +1064,25 @@ func TestNewFirewallFromConfig(t *testing.T) {
 	require.EqualError(t, err, "firewall.outbound rule #0; port was not a number; `a`")
 
 	// Test proto error
-	conf = config.NewC(l)
+	conf = config.NewC(test.NewLogger())
 	conf.Settings["firewall"] = map[string]any{"outbound": []any{map[string]any{"code": "1", "host": "testh"}}}
 	_, err = NewFirewallFromConfig(l, cs, conf)
 	require.EqualError(t, err, "firewall.outbound rule #0; proto was not understood; ``")
 
 	// Test cidr parse error
-	conf = config.NewC(l)
+	conf = config.NewC(test.NewLogger())
 	conf.Settings["firewall"] = map[string]any{"outbound": []any{map[string]any{"code": "1", "cidr": "testh", "proto": "any"}}}
 	_, err = NewFirewallFromConfig(l, cs, conf)
 	require.EqualError(t, err, "firewall.outbound rule #0; cidr did not parse; netip.ParsePrefix(\"testh\"): no '/'")
 
 	// Test local_cidr parse error
-	conf = config.NewC(l)
+	conf = config.NewC(test.NewLogger())
 	conf.Settings["firewall"] = map[string]any{"outbound": []any{map[string]any{"code": "1", "local_cidr": "testh", "proto": "any"}}}
 	_, err = NewFirewallFromConfig(l, cs, conf)
 	require.EqualError(t, err, "firewall.outbound rule #0; local_cidr did not parse; netip.ParsePrefix(\"testh\"): no '/'")
 
 	// Test both group and groups
-	conf = config.NewC(l)
+	conf = config.NewC(test.NewLogger())
 	conf.Settings["firewall"] = map[string]any{"inbound": []any{map[string]any{"port": "1", "proto": "any", "group": "a", "groups": []string{"b", "c"}}}}
 	_, err = NewFirewallFromConfig(l, cs, conf)
 	require.EqualError(t, err, "firewall.inbound rule #0; only one of group or groups should be defined, both provided")
@@ -1100,35 +1091,35 @@ func TestNewFirewallFromConfig(t *testing.T) {
 func TestAddFirewallRulesFromConfig(t *testing.T) {
 	l := test.NewLogger()
 	// Test adding tcp rule
-	conf := config.NewC(l)
+	conf := config.NewC(test.NewLogger())
 	mf := &mockFirewall{}
 	conf.Settings["firewall"] = map[string]any{"outbound": []any{map[string]any{"port": "1", "proto": "tcp", "host": "a"}}}
 	require.NoError(t, AddFirewallRulesFromConfig(l, false, conf, mf))
 	assert.Equal(t, addRuleCall{incoming: false, proto: firewall.ProtoTCP, startPort: 1, endPort: 1, groups: nil, host: "a", ip: "", localIp: ""}, mf.lastCall)
 
 	// Test adding udp rule
-	conf = config.NewC(l)
+	conf = config.NewC(test.NewLogger())
 	mf = &mockFirewall{}
 	conf.Settings["firewall"] = map[string]any{"outbound": []any{map[string]any{"port": "1", "proto": "udp", "host": "a"}}}
 	require.NoError(t, AddFirewallRulesFromConfig(l, false, conf, mf))
 	assert.Equal(t, addRuleCall{incoming: false, proto: firewall.ProtoUDP, startPort: 1, endPort: 1, groups: nil, host: "a", ip: "", localIp: ""}, mf.lastCall)
 
 	// Test adding icmp rule
-	conf = config.NewC(l)
+	conf = config.NewC(test.NewLogger())
 	mf = &mockFirewall{}
 	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))
 	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)
+	conf = config.NewC(test.NewLogger())
 	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
-	conf = config.NewC(l)
+	conf = config.NewC(test.NewLogger())
 	mf = &mockFirewall{}
 	conf.Settings["firewall"] = map[string]any{"inbound": []any{map[string]any{"port": "1", "proto": "any", "host": "a"}}}
 	require.NoError(t, AddFirewallRulesFromConfig(l, true, conf, mf))
@@ -1136,14 +1127,14 @@ func TestAddFirewallRulesFromConfig(t *testing.T) {
 
 	// Test adding rule with cidr
 	cidr := netip.MustParsePrefix("10.0.0.0/8")
-	conf = config.NewC(l)
+	conf = config.NewC(test.NewLogger())
 	mf = &mockFirewall{}
 	conf.Settings["firewall"] = map[string]any{"inbound": []any{map[string]any{"port": "1", "proto": "any", "cidr": cidr.String()}}}
 	require.NoError(t, AddFirewallRulesFromConfig(l, true, conf, mf))
 	assert.Equal(t, addRuleCall{incoming: true, proto: firewall.ProtoAny, startPort: 1, endPort: 1, groups: nil, ip: cidr.String(), localIp: ""}, mf.lastCall)
 
 	// Test adding rule with local_cidr
-	conf = config.NewC(l)
+	conf = config.NewC(test.NewLogger())
 	mf = &mockFirewall{}
 	conf.Settings["firewall"] = map[string]any{"inbound": []any{map[string]any{"port": "1", "proto": "any", "local_cidr": cidr.String()}}}
 	require.NoError(t, AddFirewallRulesFromConfig(l, true, conf, mf))
@@ -1151,82 +1142,82 @@ func TestAddFirewallRulesFromConfig(t *testing.T) {
 
 	// Test adding rule with cidr ipv6
 	cidr6 := netip.MustParsePrefix("fd00::/8")
-	conf = config.NewC(l)
+	conf = config.NewC(test.NewLogger())
 	mf = &mockFirewall{}
 	conf.Settings["firewall"] = map[string]any{"inbound": []any{map[string]any{"port": "1", "proto": "any", "cidr": cidr6.String()}}}
 	require.NoError(t, AddFirewallRulesFromConfig(l, true, conf, mf))
 	assert.Equal(t, addRuleCall{incoming: true, proto: firewall.ProtoAny, startPort: 1, endPort: 1, groups: nil, ip: cidr6.String(), localIp: ""}, mf.lastCall)
 
 	// Test adding rule with any cidr
-	conf = config.NewC(l)
+	conf = config.NewC(test.NewLogger())
 	mf = &mockFirewall{}
 	conf.Settings["firewall"] = map[string]any{"inbound": []any{map[string]any{"port": "1", "proto": "any", "cidr": "any"}}}
 	require.NoError(t, AddFirewallRulesFromConfig(l, true, conf, mf))
 	assert.Equal(t, addRuleCall{incoming: true, proto: firewall.ProtoAny, startPort: 1, endPort: 1, groups: nil, ip: "any", localIp: ""}, mf.lastCall)
 
 	// Test adding rule with junk cidr
-	conf = config.NewC(l)
+	conf = config.NewC(test.NewLogger())
 	mf = &mockFirewall{}
 	conf.Settings["firewall"] = map[string]any{"inbound": []any{map[string]any{"port": "1", "proto": "any", "cidr": "junk/junk"}}}
 	require.EqualError(t, AddFirewallRulesFromConfig(l, true, conf, mf), "firewall.inbound rule #0; cidr did not parse; netip.ParsePrefix(\"junk/junk\"): ParseAddr(\"junk\"): unable to parse IP")
 
 	// Test adding rule with local_cidr ipv6
-	conf = config.NewC(l)
+	conf = config.NewC(test.NewLogger())
 	mf = &mockFirewall{}
 	conf.Settings["firewall"] = map[string]any{"inbound": []any{map[string]any{"port": "1", "proto": "any", "local_cidr": cidr6.String()}}}
 	require.NoError(t, AddFirewallRulesFromConfig(l, true, conf, mf))
 	assert.Equal(t, addRuleCall{incoming: true, proto: firewall.ProtoAny, startPort: 1, endPort: 1, groups: nil, ip: "", localIp: cidr6.String()}, mf.lastCall)
 
 	// Test adding rule with any local_cidr
-	conf = config.NewC(l)
+	conf = config.NewC(test.NewLogger())
 	mf = &mockFirewall{}
 	conf.Settings["firewall"] = map[string]any{"inbound": []any{map[string]any{"port": "1", "proto": "any", "local_cidr": "any"}}}
 	require.NoError(t, AddFirewallRulesFromConfig(l, true, conf, mf))
 	assert.Equal(t, addRuleCall{incoming: true, proto: firewall.ProtoAny, startPort: 1, endPort: 1, groups: nil, localIp: "any"}, mf.lastCall)
 
 	// Test adding rule with junk local_cidr
-	conf = config.NewC(l)
+	conf = config.NewC(test.NewLogger())
 	mf = &mockFirewall{}
 	conf.Settings["firewall"] = map[string]any{"inbound": []any{map[string]any{"port": "1", "proto": "any", "local_cidr": "junk/junk"}}}
 	require.EqualError(t, AddFirewallRulesFromConfig(l, true, conf, mf), "firewall.inbound rule #0; local_cidr did not parse; netip.ParsePrefix(\"junk/junk\"): ParseAddr(\"junk\"): unable to parse IP")
 
 	// Test adding rule with ca_sha
-	conf = config.NewC(l)
+	conf = config.NewC(test.NewLogger())
 	mf = &mockFirewall{}
 	conf.Settings["firewall"] = map[string]any{"inbound": []any{map[string]any{"port": "1", "proto": "any", "ca_sha": "12312313123"}}}
 	require.NoError(t, AddFirewallRulesFromConfig(l, true, conf, mf))
 	assert.Equal(t, addRuleCall{incoming: true, proto: firewall.ProtoAny, startPort: 1, endPort: 1, groups: nil, ip: "", localIp: "", caSha: "12312313123"}, mf.lastCall)
 
 	// Test adding rule with ca_name
-	conf = config.NewC(l)
+	conf = config.NewC(test.NewLogger())
 	mf = &mockFirewall{}
 	conf.Settings["firewall"] = map[string]any{"inbound": []any{map[string]any{"port": "1", "proto": "any", "ca_name": "root01"}}}
 	require.NoError(t, AddFirewallRulesFromConfig(l, true, conf, mf))
 	assert.Equal(t, addRuleCall{incoming: true, proto: firewall.ProtoAny, startPort: 1, endPort: 1, groups: nil, ip: "", localIp: "", caName: "root01"}, mf.lastCall)
 
 	// Test single group
-	conf = config.NewC(l)
+	conf = config.NewC(test.NewLogger())
 	mf = &mockFirewall{}
 	conf.Settings["firewall"] = map[string]any{"inbound": []any{map[string]any{"port": "1", "proto": "any", "group": "a"}}}
 	require.NoError(t, AddFirewallRulesFromConfig(l, true, conf, mf))
 	assert.Equal(t, addRuleCall{incoming: true, proto: firewall.ProtoAny, startPort: 1, endPort: 1, groups: []string{"a"}, ip: "", localIp: ""}, mf.lastCall)
 
 	// Test single groups
-	conf = config.NewC(l)
+	conf = config.NewC(test.NewLogger())
 	mf = &mockFirewall{}
 	conf.Settings["firewall"] = map[string]any{"inbound": []any{map[string]any{"port": "1", "proto": "any", "groups": "a"}}}
 	require.NoError(t, AddFirewallRulesFromConfig(l, true, conf, mf))
 	assert.Equal(t, addRuleCall{incoming: true, proto: firewall.ProtoAny, startPort: 1, endPort: 1, groups: []string{"a"}, ip: "", localIp: ""}, mf.lastCall)
 
 	// Test multiple AND groups
-	conf = config.NewC(l)
+	conf = config.NewC(test.NewLogger())
 	mf = &mockFirewall{}
 	conf.Settings["firewall"] = map[string]any{"inbound": []any{map[string]any{"port": "1", "proto": "any", "groups": []string{"a", "b"}}}}
 	require.NoError(t, AddFirewallRulesFromConfig(l, true, conf, mf))
 	assert.Equal(t, addRuleCall{incoming: true, proto: firewall.ProtoAny, startPort: 1, endPort: 1, groups: []string{"a", "b"}, ip: "", localIp: ""}, mf.lastCall)
 
 	// Test Add error
-	conf = config.NewC(l)
+	conf = config.NewC(test.NewLogger())
 	mf = &mockFirewall{}
 	mf.nextCallReturn = errors.New("test error")
 	conf.Settings["firewall"] = map[string]any{"inbound": []any{map[string]any{"port": "1", "proto": "any", "host": "a"}}}
@@ -1234,9 +1225,8 @@ func TestAddFirewallRulesFromConfig(t *testing.T) {
 }
 
 func TestFirewall_convertRule(t *testing.T) {
-	l := test.NewLogger()
 	ob := &bytes.Buffer{}
-	l.SetOutput(ob)
+	l := test.NewLoggerWithOutput(ob)
 
 	// Ensure group array of 1 is converted and a warning is printed
 	c := map[string]any{
@@ -1244,7 +1234,9 @@ func TestFirewall_convertRule(t *testing.T) {
 	}
 
 	r, err := convertRule(l, c, "test", 1)
-	assert.Contains(t, ob.String(), "test rule #1; group was an array with a single value, converting to simple value")
+	assert.Contains(t, ob.String(), "group was an array with a single value, converting to simple value")
+	assert.Contains(t, ob.String(), "table=test")
+	assert.Contains(t, ob.String(), "rule=1")
 	require.NoError(t, err)
 	assert.Equal(t, []string{"group1"}, r.Groups)
 
@@ -1270,9 +1262,8 @@ func TestFirewall_convertRule(t *testing.T) {
 }
 
 func TestFirewall_convertRuleSanity(t *testing.T) {
-	l := test.NewLogger()
 	ob := &bytes.Buffer{}
-	l.SetOutput(ob)
+	l := test.NewLoggerWithOutput(ob)
 
 	noWarningPlease := []map[string]any{
 		{"group": "group1"},
@@ -1386,7 +1377,7 @@ type testsetup struct {
 	fw                 *Firewall
 }
 
-func newSetup(t *testing.T, l *logrus.Logger, myPrefixes ...netip.Prefix) testsetup {
+func newSetup(t *testing.T, l *slog.Logger, myPrefixes ...netip.Prefix) testsetup {
 	c := dummyCert{
 		name:     "me",
 		networks: myPrefixes,
@@ -1397,7 +1388,7 @@ func newSetup(t *testing.T, l *logrus.Logger, myPrefixes ...netip.Prefix) testse
 	return newSetupFromCert(t, l, c)
 }
 
-func newSetupFromCert(t *testing.T, l *logrus.Logger, c dummyCert) testsetup {
+func newSetupFromCert(t *testing.T, l *slog.Logger, c dummyCert) testsetup {
 	myVpnNetworksTable := new(bart.Lite)
 	for _, prefix := range c.Networks() {
 		myVpnNetworksTable.Insert(prefix)
@@ -1414,9 +1405,8 @@ func newSetupFromCert(t *testing.T, l *logrus.Logger, c dummyCert) testsetup {
 
 func TestFirewall_Drop_EnforceIPMatch(t *testing.T) {
 	t.Parallel()
-	l := test.NewLogger()
 	ob := &bytes.Buffer{}
-	l.SetOutput(ob)
+	l := test.NewLoggerWithOutput(ob)
 
 	myPrefix := netip.MustParsePrefix("1.1.1.1/8")
 	// for now, it's okay that these are all "incoming", the logic this test tries to check doesn't care about in/out

go.mod

@@ -18,7 +18,6 @@ require (
 	github.com/nbrownus/go-metrics-prometheus v0.0.0-20210712211119-974a6260965f
 	github.com/prometheus/client_golang v1.23.2
 	github.com/rcrowley/go-metrics v0.0.0-20201227073835-cf1acfcdf475
-	github.com/sirupsen/logrus v1.9.4
 	github.com/skip2/go-qrcode v0.0.0-20200617195104-da1b6568686e
 	github.com/stefanberger/go-pkcs11uri v0.0.0-20230803200340-78284954bff6
 	github.com/stretchr/testify v1.11.1

go.sum

@@ -133,8 +133,6 @@ github.com/rogpeppe/go-internal v1.10.0/go.mod h1:UQnix2H7Ngw/k4C5ijL5+65zddjncj
 github.com/sirupsen/logrus v1.2.0/go.mod h1:LxeOpSwHxABJmUn/MG1IvRgCAasNZTLOkJPxbbu5VWo=
 github.com/sirupsen/logrus v1.4.2/go.mod h1:tLMulIdttU9McNUspp0xgXVQah82FyeX6MwdIuYE2rE=
 github.com/sirupsen/logrus v1.6.0/go.mod h1:7uNnSEd1DgxDLC74fIahvMZmmYsHGZGEOFrfsX/uA88=
-github.com/sirupsen/logrus v1.9.4 h1:TsZE7l11zFCLZnZ+teH4Umoq5BhEIfIzfRDZ1Uzql2w=
-github.com/sirupsen/logrus v1.9.4/go.mod h1:ftWc9WdOfJ0a92nsE2jF5u5ZwH8Bv2zdeOC42RjbV2g=
 github.com/skip2/go-qrcode v0.0.0-20200617195104-da1b6568686e h1:MRM5ITcdelLK2j1vwZ3Je0FKVCfqOLp5zO6trqMLYs0=
 github.com/skip2/go-qrcode v0.0.0-20200617195104-da1b6568686e/go.mod h1:XV66xRDqSt+GTGFMVlhk3ULuV0y9ZmzeVGR4mloJI3M=
 github.com/stefanberger/go-pkcs11uri v0.0.0-20230803200340-78284954bff6 h1:pnnLyeX7o/5aX8qUQ69P/mLojDqwda8hFOCBTmP/6hw=

handshake_ix.go

@@ -2,11 +2,12 @@ package nebula
 
 import (
 	"bytes"
+	"context"
+	"log/slog"
 	"net/netip"
 	"time"
 
 	"github.com/flynn/noise"
-	"github.com/sirupsen/logrus"
 	"github.com/slackhq/nebula/cert"
 	"github.com/slackhq/nebula/header"
 )
@@ -18,8 +19,11 @@ import (
 func ixHandshakeStage0(f *Interface, hh *HandshakeHostInfo) bool {
 	err := f.handshakeManager.allocateIndex(hh)
 	if err != nil {
-		f.l.WithError(err).WithField("vpnAddrs", hh.hostinfo.vpnAddrs).
+		f.l.Error("Failed to generate index",
-			WithField("handshake", m{"stage": 0, "style": "ix_psk0"}).Error("Failed to generate index")
+			"error", err,
+			"vpnAddrs", hh.hostinfo.vpnAddrs,
+			"handshake", m{"stage": 0, "style": "ix_psk0"},
+		)
 		return false
 	}
 
@@ -39,28 +43,32 @@ func ixHandshakeStage0(f *Interface, hh *HandshakeHostInfo) bool {
 
 	crt := cs.getCertificate(v)
 	if crt == nil {
-		f.l.WithField("vpnAddrs", hh.hostinfo.vpnAddrs).
+		f.l.Error("Unable to handshake with host because no certificate is available",
-			WithField("handshake", m{"stage": 0, "style": "ix_psk0"}).
+			"vpnAddrs", hh.hostinfo.vpnAddrs,
-			WithField("certVersion", v).
+			"handshake", m{"stage": 0, "style": "ix_psk0"},
-			Error("Unable to handshake with host because no certificate is available")
+			"certVersion", v,
+		)
 		return false
 	}
 
 	crtHs := cs.getHandshakeBytes(v)
 	if crtHs == nil {
-		f.l.WithField("vpnAddrs", hh.hostinfo.vpnAddrs).
+		f.l.Error("Unable to handshake with host because no certificate handshake bytes is available",
-			WithField("handshake", m{"stage": 0, "style": "ix_psk0"}).
+			"vpnAddrs", hh.hostinfo.vpnAddrs,
-			WithField("certVersion", v).
+			"handshake", m{"stage": 0, "style": "ix_psk0"},
-			Error("Unable to handshake with host because no certificate handshake bytes is available")
+			"certVersion", v,
+		)
 		return false
 	}
 
-	ci, err := NewConnectionState(f.l, cs, crt, true, noise.HandshakeIX)
+	ci, err := NewConnectionState(cs, crt, true, noise.HandshakeIX)
 	if err != nil {
-		f.l.WithError(err).WithField("vpnAddrs", hh.hostinfo.vpnAddrs).
+		f.l.Error("Failed to create connection state",
-			WithField("handshake", m{"stage": 0, "style": "ix_psk0"}).
+			"error", err,
-			WithField("certVersion", v).
+			"vpnAddrs", hh.hostinfo.vpnAddrs,
-			Error("Failed to create connection state")
+			"handshake", m{"stage": 0, "style": "ix_psk0"},
+			"certVersion", v,
+		)
 		return false
 	}
 	hh.hostinfo.ConnectionState = ci
@@ -76,9 +84,12 @@ func ixHandshakeStage0(f *Interface, hh *HandshakeHostInfo) bool {
 
 	hsBytes, err := hs.Marshal()
 	if err != nil {
-		f.l.WithError(err).WithField("vpnAddrs", hh.hostinfo.vpnAddrs).
+		f.l.Error("Failed to marshal handshake message",
-			WithField("certVersion", v).
+			"error", err,
-			WithField("handshake", m{"stage": 0, "style": "ix_psk0"}).Error("Failed to marshal handshake message")
+			"vpnAddrs", hh.hostinfo.vpnAddrs,
+			"certVersion", v,
+			"handshake", m{"stage": 0, "style": "ix_psk0"},
+		)
 		return false
 	}
 
@@ -86,8 +97,11 @@ func ixHandshakeStage0(f *Interface, hh *HandshakeHostInfo) bool {
 
 	msg, _, _, err := ci.H.WriteMessage(h, hsBytes)
 	if err != nil {
-		f.l.WithError(err).WithField("vpnAddrs", hh.hostinfo.vpnAddrs).
+		f.l.Error("Failed to call noise.WriteMessage",
-			WithField("handshake", m{"stage": 0, "style": "ix_psk0"}).Error("Failed to call noise.WriteMessage")
+			"error", err,
+			"vpnAddrs", hh.hostinfo.vpnAddrs,
+			"handshake", m{"stage": 0, "style": "ix_psk0"},
+		)
 		return false
 	}
 
@@ -104,18 +118,21 @@ func ixHandshakeStage1(f *Interface, via ViaSender, packet []byte, h *header.H)
 	cs := f.pki.getCertState()
 	crt := cs.GetDefaultCertificate()
 	if crt == nil {
-		f.l.WithField("from", via).
+		f.l.Error("Unable to handshake with host because no certificate is available",
-			WithField("handshake", m{"stage": 0, "style": "ix_psk0"}).
+			"from", via,
-			WithField("certVersion", cs.initiatingVersion).
+			"handshake", m{"stage": 0, "style": "ix_psk0"},
-			Error("Unable to handshake with host because no certificate is available")
+			"certVersion", cs.initiatingVersion,
+		)
 		return
 	}
 
-	ci, err := NewConnectionState(f.l, cs, crt, false, noise.HandshakeIX)
+	ci, err := NewConnectionState(cs, crt, false, noise.HandshakeIX)
 	if err != nil {
-		f.l.WithError(err).WithField("from", via).
+		f.l.Error("Failed to create connection state",
-			WithField("handshake", m{"stage": 1, "style": "ix_psk0"}).
+			"error", err,
-			Error("Failed to create connection state")
+			"from", via,
+			"handshake", m{"stage": 1, "style": "ix_psk0"},
+		)
 		return
 	}
 
@@ -124,26 +141,32 @@ func ixHandshakeStage1(f *Interface, via ViaSender, packet []byte, h *header.H)
 
 	msg, _, _, err := ci.H.ReadMessage(nil, packet[header.Len:])
 	if err != nil {
-		f.l.WithError(err).WithField("from", via).
+		f.l.Error("Failed to call noise.ReadMessage",
-			WithField("handshake", m{"stage": 1, "style": "ix_psk0"}).
+			"error", err,
-			Error("Failed to call noise.ReadMessage")
+			"from", via,
+			"handshake", m{"stage": 1, "style": "ix_psk0"},
+		)
 		return
 	}
 
 	hs := &NebulaHandshake{}
 	err = hs.Unmarshal(msg)
 	if err != nil || hs.Details == nil {
-		f.l.WithError(err).WithField("from", via).
+		f.l.Error("Failed unmarshal handshake message",
-			WithField("handshake", m{"stage": 1, "style": "ix_psk0"}).
+			"error", err,
-			Error("Failed unmarshal handshake message")
+			"from", via,
+			"handshake", m{"stage": 1, "style": "ix_psk0"},
+		)
 		return
 	}
 
 	rc, err := cert.Recombine(cert.Version(hs.Details.CertVersion), hs.Details.Cert, ci.H.PeerStatic(), ci.Curve())
 	if err != nil {
-		f.l.WithError(err).WithField("from", via).
+		f.l.Info("Handshake did not contain a certificate",
-			WithField("handshake", m{"stage": 1, "style": "ix_psk0"}).
+			"error", err,
-			Info("Handshake did not contain a certificate")
+			"from", via,
+			"handshake", m{"stage": 1, "style": "ix_psk0"},
+		)
 		return
 	}
 
@@ -154,23 +177,30 @@ func ixHandshakeStage1(f *Interface, via ViaSender, packet []byte, h *header.H)
 			fp = "<error generating certificate fingerprint>"
 		}
 
-		e := f.l.WithError(err).WithField("from", via).
+		attrs := []slog.Attr{
-			WithField("handshake", m{"stage": 1, "style": "ix_psk0"}).
+			slog.Any("error", err),
-			WithField("certVpnNetworks", rc.Networks()).
+			slog.Any("from", via),
-			WithField("certFingerprint", fp)
+			slog.Any("handshake", m{"stage": 1, "style": "ix_psk0"}),
-
+			slog.Any("certVpnNetworks", rc.Networks()),
-		if f.l.Level >= logrus.DebugLevel {
+			slog.String("certFingerprint", fp),
-			e = e.WithField("cert", rc)
+		}
+		if f.l.Enabled(context.Background(), slog.LevelDebug) {
+			attrs = append(attrs, slog.Any("cert", rc))
 		}
 
-		e.Info("Invalid certificate from host")
+		// LogAttrs is intentional: attrs is a pre-built []slog.Attr slice that
+		// callers grow conditionally, which has no pair-form equivalent.
+		//nolint:sloglint
+		f.l.LogAttrs(context.Background(), slog.LevelInfo, "Invalid certificate from host", attrs...)
 		return
 	}
 
 	if !bytes.Equal(remoteCert.Certificate.PublicKey(), ci.H.PeerStatic()) {
-		f.l.WithField("from", via).
+		f.l.Info("public key mismatch between certificate and handshake",
-			WithField("handshake", m{"stage": 1, "style": "ix_psk0"}).
+			"from", via,
-			WithField("cert", remoteCert).Info("public key mismatch between certificate and handshake")
+			"handshake", m{"stage": 1, "style": "ix_psk0"},
+			"cert", remoteCert,
+		)
 		return
 	}
 
@@ -178,12 +208,13 @@ func ixHandshakeStage1(f *Interface, via ViaSender, packet []byte, h *header.H)
 		// We started off using the wrong certificate version, lets see if we can match the version that was sent to us
 		myCertOtherVersion := cs.getCertificate(remoteCert.Certificate.Version())
 		if myCertOtherVersion == nil {
-			if f.l.Level >= logrus.DebugLevel {
+			if f.l.Enabled(context.Background(), slog.LevelDebug) {
-				f.l.WithError(err).WithFields(m{
+				f.l.Debug("Might be unable to handshake with host due to missing certificate version",
-					"from":      via,
+					"error", err,
-					"handshake": m{"stage": 1, "style": "ix_psk0"},
+					"from", via,
-					"cert":      remoteCert,
+					"handshake", m{"stage": 1, "style": "ix_psk0"},
-				}).Debug("Might be unable to handshake with host due to missing certificate version")
+					"cert", remoteCert,
+				)
 			}
 		} else {
 			// Record the certificate we are actually using
@@ -192,10 +223,12 @@ func ixHandshakeStage1(f *Interface, via ViaSender, packet []byte, h *header.H)
 	}
 
 	if len(remoteCert.Certificate.Networks()) == 0 {
-		f.l.WithError(err).WithField("from", via).
+		f.l.Info("No networks in certificate",
-			WithField("cert", remoteCert).
+			"error", err,
-			WithField("handshake", m{"stage": 1, "style": "ix_psk0"}).
+			"from", via,
-			Info("No networks in certificate")
+			"cert", remoteCert,
+			"handshake", m{"stage": 1, "style": "ix_psk0"},
+		)
 		return
 	}
 
@@ -209,12 +242,15 @@ func ixHandshakeStage1(f *Interface, via ViaSender, packet []byte, h *header.H)
 	vpnAddrs := make([]netip.Addr, len(vpnNetworks))
 	for i, network := range vpnNetworks {
 		if f.myVpnAddrsTable.Contains(network.Addr()) {
-			f.l.WithField("vpnNetworks", vpnNetworks).WithField("from", via).
+			f.l.Error("Refusing to handshake with myself",
-				WithField("certName", certName).
+				"vpnNetworks", vpnNetworks,
-				WithField("certVersion", certVersion).
+				"from", via,
-				WithField("fingerprint", fingerprint).
+				"certName", certName,
-				WithField("issuer", issuer).
+				"certVersion", certVersion,
-				WithField("handshake", m{"stage": 1, "style": "ix_psk0"}).Error("Refusing to handshake with myself")
+				"fingerprint", fingerprint,
+				"issuer", issuer,
+				"handshake", m{"stage": 1, "style": "ix_psk0"},
+			)
 			return
 		}
 		vpnAddrs[i] = network.Addr()
@@ -226,20 +262,28 @@ func ixHandshakeStage1(f *Interface, via ViaSender, packet []byte, h *header.H)
 	if !via.IsRelayed {
 		// We only want to apply the remote allow list for direct tunnels here
 		if !f.lightHouse.GetRemoteAllowList().AllowAll(vpnAddrs, via.UdpAddr.Addr()) {
-			f.l.WithField("vpnAddrs", vpnAddrs).WithField("from", via).
+			if f.l.Enabled(context.Background(), slog.LevelDebug) {
-				Debug("lighthouse.remote_allow_list denied incoming handshake")
+				f.l.Debug("lighthouse.remote_allow_list denied incoming handshake",
+					"vpnAddrs", vpnAddrs,
+					"from", via,
+				)
+			}
 			return
 		}
 	}
 
 	myIndex, err := generateIndex(f.l)
 	if err != nil {
-		f.l.WithError(err).WithField("vpnAddrs", vpnAddrs).WithField("from", via).
+		f.l.Error("Failed to generate index",
-			WithField("certName", certName).
+			"error", err,
-			WithField("certVersion", certVersion).
+			"vpnAddrs", vpnAddrs,
-			WithField("fingerprint", fingerprint).
+			"from", via,
-			WithField("issuer", issuer).
+			"certName", certName,
-			WithField("handshake", m{"stage": 1, "style": "ix_psk0"}).Error("Failed to generate index")
+			"certVersion", certVersion,
+			"fingerprint", fingerprint,
+			"issuer", issuer,
+			"handshake", m{"stage": 1, "style": "ix_psk0"},
+		)
 		return
 	}
 
@@ -257,18 +301,18 @@ func ixHandshakeStage1(f *Interface, via ViaSender, packet []byte, h *header.H)
 		},
 	}
 
-	msgRxL := f.l.WithFields(m{
+	msgRxL := f.l.With(
-		"vpnAddrs":       vpnAddrs,
+		"vpnAddrs", vpnAddrs,
-		"from":           via,
+		"from", via,
-		"certName":       certName,
+		"certName", certName,
-		"certVersion":    certVersion,
+		"certVersion", certVersion,
-		"fingerprint":    fingerprint,
+		"fingerprint", fingerprint,
-		"issuer":         issuer,
+		"issuer", issuer,
-		"initiatorIndex": hs.Details.InitiatorIndex,
+		"initiatorIndex", hs.Details.InitiatorIndex,
-		"responderIndex": hs.Details.ResponderIndex,
+		"responderIndex", hs.Details.ResponderIndex,
-		"remoteIndex":    h.RemoteIndex,
+		"remoteIndex", h.RemoteIndex,
-		"handshake":      m{"stage": 1, "style": "ix_psk0"},
+		"handshake", m{"stage": 1, "style": "ix_psk0"},
-	})
+	)
 
 	if anyVpnAddrsInCommon {
 		msgRxL.Info("Handshake message received")
@@ -280,8 +324,9 @@ func ixHandshakeStage1(f *Interface, via ViaSender, packet []byte, h *header.H)
 	hs.Details.ResponderIndex = myIndex
 	hs.Details.Cert = cs.getHandshakeBytes(ci.myCert.Version())
 	if hs.Details.Cert == nil {
-		msgRxL.WithField("myCertVersion", ci.myCert.Version()).
+		msgRxL.Error("Unable to handshake with host because no certificate handshake bytes is available",
-			Error("Unable to handshake with host because no certificate handshake bytes is available")
+			"myCertVersion", ci.myCert.Version(),
+		)
 		return
 	}
 
@@ -291,32 +336,43 @@ func ixHandshakeStage1(f *Interface, via ViaSender, packet []byte, h *header.H)
 
 	hsBytes, err := hs.Marshal()
 	if err != nil {
-		f.l.WithError(err).WithField("vpnAddrs", hostinfo.vpnAddrs).WithField("from", via).
+		f.l.Error("Failed to marshal handshake message",
-			WithField("certName", certName).
+			"error", err,
-			WithField("certVersion", certVersion).
+			"vpnAddrs", hostinfo.vpnAddrs,
-			WithField("fingerprint", fingerprint).
+			"from", via,
-			WithField("issuer", issuer).
+			"certName", certName,
-			WithField("handshake", m{"stage": 1, "style": "ix_psk0"}).Error("Failed to marshal handshake message")
+			"certVersion", certVersion,
+			"fingerprint", fingerprint,
+			"issuer", issuer,
+			"handshake", m{"stage": 1, "style": "ix_psk0"},
+		)
 		return
 	}
 
 	nh := header.Encode(make([]byte, header.Len), header.Version, header.Handshake, header.HandshakeIXPSK0, hs.Details.InitiatorIndex, 2)
 	msg, dKey, eKey, err := ci.H.WriteMessage(nh, hsBytes)
 	if err != nil {
-		f.l.WithError(err).WithField("vpnAddrs", hostinfo.vpnAddrs).WithField("from", via).
+		f.l.Error("Failed to call noise.WriteMessage",
-			WithField("certName", certName).
+			"error", err,
-			WithField("certVersion", certVersion).
+			"vpnAddrs", hostinfo.vpnAddrs,
-			WithField("fingerprint", fingerprint).
+			"from", via,
-			WithField("issuer", issuer).
+			"certName", certName,
-			WithField("handshake", m{"stage": 1, "style": "ix_psk0"}).Error("Failed to call noise.WriteMessage")
+			"certVersion", certVersion,
+			"fingerprint", fingerprint,
+			"issuer", issuer,
+			"handshake", m{"stage": 1, "style": "ix_psk0"},
+		)
 		return
 	} else if dKey == nil || eKey == nil {
-		f.l.WithField("vpnAddrs", hostinfo.vpnAddrs).WithField("from", via).
+		f.l.Error("Noise did not arrive at a key",
-			WithField("certName", certName).
+			"vpnAddrs", hostinfo.vpnAddrs,
-			WithField("certVersion", certVersion).
+			"from", via,
-			WithField("fingerprint", fingerprint).
+			"certName", certName,
-			WithField("issuer", issuer).
+			"certVersion", certVersion,
-			WithField("handshake", m{"stage": 1, "style": "ix_psk0"}).Error("Noise did not arrive at a key")
+			"fingerprint", fingerprint,
+			"issuer", issuer,
+			"handshake", m{"stage": 1, "style": "ix_psk0"},
+		)
 		return
 	}
 
@@ -358,13 +414,20 @@ func ixHandshakeStage1(f *Interface, via ViaSender, packet []byte, h *header.H)
 			if !via.IsRelayed {
 				err := f.outside.WriteTo(msg, via.UdpAddr)
 				if err != nil {
-					f.l.WithField("vpnAddrs", existing.vpnAddrs).WithField("from", via).
+					f.l.Error("Failed to send handshake message",
-						WithField("handshake", m{"stage": 2, "style": "ix_psk0"}).WithField("cached", true).
+						"vpnAddrs", existing.vpnAddrs,
-						WithError(err).Error("Failed to send handshake message")
+						"from", via,
+						"handshake", m{"stage": 2, "style": "ix_psk0"},
+						"cached", true,
+						"error", err,
+					)
 				} else {
-					f.l.WithField("vpnAddrs", existing.vpnAddrs).WithField("from", via).
+					f.l.Info("Handshake message sent",
-						WithField("handshake", m{"stage": 2, "style": "ix_psk0"}).WithField("cached", true).
+						"vpnAddrs", existing.vpnAddrs,
-						Info("Handshake message sent")
+						"from", via,
+						"handshake", m{"stage": 2, "style": "ix_psk0"},
+						"cached", true,
+					)
 				}
 				return
 			} else {
@@ -374,50 +437,67 @@ func ixHandshakeStage1(f *Interface, via ViaSender, packet []byte, h *header.H)
 				}
 				hostinfo.relayState.InsertRelayTo(via.relayHI.vpnAddrs[0])
 				f.SendVia(via.relayHI, via.relay, msg, make([]byte, 12), make([]byte, mtu), false)
-				f.l.WithField("vpnAddrs", existing.vpnAddrs).WithField("relay", via.relayHI.vpnAddrs[0]).
+				f.l.Info("Handshake message sent",
-					WithField("handshake", m{"stage": 2, "style": "ix_psk0"}).WithField("cached", true).
+					"vpnAddrs", existing.vpnAddrs,
-					Info("Handshake message sent")
+					"relay", via.relayHI.vpnAddrs[0],
+					"handshake", m{"stage": 2, "style": "ix_psk0"},
+					"cached", true,
+				)
 				return
 			}
 		case ErrExistingHostInfo:
 			// This means there was an existing tunnel and this handshake was older than the one we are currently based on
-			f.l.WithField("vpnAddrs", vpnAddrs).WithField("from", via).
+			f.l.Info("Handshake too old",
-				WithField("certName", certName).
+				"vpnAddrs", vpnAddrs,
-				WithField("certVersion", certVersion).
+				"from", via,
-				WithField("oldHandshakeTime", existing.lastHandshakeTime).
+				"certName", certName,
-				WithField("newHandshakeTime", hostinfo.lastHandshakeTime).
+				"certVersion", certVersion,
-				WithField("fingerprint", fingerprint).
+				"oldHandshakeTime", existing.lastHandshakeTime,
-				WithField("issuer", issuer).
+				"newHandshakeTime", hostinfo.lastHandshakeTime,
-				WithField("initiatorIndex", hs.Details.InitiatorIndex).WithField("responderIndex", hs.Details.ResponderIndex).
+				"fingerprint", fingerprint,
-				WithField("remoteIndex", h.RemoteIndex).WithField("handshake", m{"stage": 1, "style": "ix_psk0"}).
+				"issuer", issuer,
-				Info("Handshake too old")
+				"initiatorIndex", hs.Details.InitiatorIndex,
+				"responderIndex", hs.Details.ResponderIndex,
+				"remoteIndex", h.RemoteIndex,
+				"handshake", m{"stage": 1, "style": "ix_psk0"},
+			)
 
 			// Send a test packet to trigger an authenticated tunnel test, this should suss out any lingering tunnel issues
 			f.SendMessageToVpnAddr(header.Test, header.TestRequest, vpnAddrs[0], []byte(""), make([]byte, 12, 12), make([]byte, mtu))
 			return
 		case ErrLocalIndexCollision:
 			// This means we failed to insert because of collision on localIndexId. Just let the next handshake packet retry
-			f.l.WithField("vpnAddrs", vpnAddrs).WithField("from", via).
+			f.l.Error("Failed to add HostInfo due to localIndex collision",
-				WithField("certName", certName).
+				"vpnAddrs", vpnAddrs,
-				WithField("certVersion", certVersion).
+				"from", via,
-				WithField("fingerprint", fingerprint).
+				"certName", certName,
-				WithField("issuer", issuer).
+				"certVersion", certVersion,
-				WithField("initiatorIndex", hs.Details.InitiatorIndex).WithField("responderIndex", hs.Details.ResponderIndex).
+				"fingerprint", fingerprint,
-				WithField("remoteIndex", h.RemoteIndex).WithField("handshake", m{"stage": 1, "style": "ix_psk0"}).
+				"issuer", issuer,
-				WithField("localIndex", hostinfo.localIndexId).WithField("collision", existing.vpnAddrs).
+				"initiatorIndex", hs.Details.InitiatorIndex,
-				Error("Failed to add HostInfo due to localIndex collision")
+				"responderIndex", hs.Details.ResponderIndex,
+				"remoteIndex", h.RemoteIndex,
+				"handshake", m{"stage": 1, "style": "ix_psk0"},
+				"localIndex", hostinfo.localIndexId,
+				"collision", existing.vpnAddrs,
+			)
 			return
 		default:
 			// Shouldn't happen, but just in case someone adds a new error type to CheckAndComplete
 			// And we forget to update it here
-			f.l.WithError(err).WithField("vpnAddrs", vpnAddrs).WithField("from", via).
+			f.l.Error("Failed to add HostInfo to HostMap",
-				WithField("certName", certName).
+				"error", err,
-				WithField("certVersion", certVersion).
+				"vpnAddrs", vpnAddrs,
-				WithField("fingerprint", fingerprint).
+				"from", via,
-				WithField("issuer", issuer).
+				"certName", certName,
-				WithField("initiatorIndex", hs.Details.InitiatorIndex).WithField("responderIndex", hs.Details.ResponderIndex).
+				"certVersion", certVersion,
-				WithField("remoteIndex", h.RemoteIndex).WithField("handshake", m{"stage": 1, "style": "ix_psk0"}).
+				"fingerprint", fingerprint,
-				Error("Failed to add HostInfo to HostMap")
+				"issuer", issuer,
+				"initiatorIndex", hs.Details.InitiatorIndex,
+				"responderIndex", hs.Details.ResponderIndex,
+				"remoteIndex", h.RemoteIndex,
+				"handshake", m{"stage": 1, "style": "ix_psk0"},
+			)
 			return
 		}
 	}
@@ -426,15 +506,20 @@ func ixHandshakeStage1(f *Interface, via ViaSender, packet []byte, h *header.H)
 	f.messageMetrics.Tx(header.Handshake, header.MessageSubType(msg[1]), 1)
 	if !via.IsRelayed {
 		err = f.outside.WriteTo(msg, via.UdpAddr)
-		log := f.l.WithField("vpnAddrs", vpnAddrs).WithField("from", via).
+		log := f.l.With(
-			WithField("certName", certName).
+			"vpnAddrs", vpnAddrs,
-			WithField("certVersion", certVersion).
+			"from", via,
-			WithField("fingerprint", fingerprint).
+			"certName", certName,
-			WithField("issuer", issuer).
+			"certVersion", certVersion,
-			WithField("initiatorIndex", hs.Details.InitiatorIndex).WithField("responderIndex", hs.Details.ResponderIndex).
+			"fingerprint", fingerprint,
-			WithField("remoteIndex", h.RemoteIndex).WithField("handshake", m{"stage": 2, "style": "ix_psk0"})
+			"issuer", issuer,
+			"initiatorIndex", hs.Details.InitiatorIndex,
+			"responderIndex", hs.Details.ResponderIndex,
+			"remoteIndex", h.RemoteIndex,
+			"handshake", m{"stage": 2, "style": "ix_psk0"},
+		)
 		if err != nil {
-			log.WithError(err).Error("Failed to send handshake")
+			log.Error("Failed to send handshake", "error", err)
 		} else {
 			log.Info("Handshake message sent")
 		}
@@ -448,20 +533,29 @@ func ixHandshakeStage1(f *Interface, via ViaSender, packet []byte, h *header.H)
 		// it's correctly marked as working.
 		via.relayHI.relayState.UpdateRelayForByIdxState(via.remoteIdx, Established)
 		f.SendVia(via.relayHI, via.relay, msg, make([]byte, 12), make([]byte, mtu), false)
-		f.l.WithField("vpnAddrs", vpnAddrs).WithField("relay", via.relayHI.vpnAddrs[0]).
+		f.l.Info("Handshake message sent",
-			WithField("certName", certName).
+			"vpnAddrs", vpnAddrs,
-			WithField("certVersion", certVersion).
+			"relay", via.relayHI.vpnAddrs[0],
-			WithField("fingerprint", fingerprint).
+			"certName", certName,
-			WithField("issuer", issuer).
+			"certVersion", certVersion,
-			WithField("initiatorIndex", hs.Details.InitiatorIndex).WithField("responderIndex", hs.Details.ResponderIndex).
+			"fingerprint", fingerprint,
-			WithField("remoteIndex", h.RemoteIndex).WithField("handshake", m{"stage": 2, "style": "ix_psk0"}).
+			"issuer", issuer,
-			Info("Handshake message sent")
+			"initiatorIndex", hs.Details.InitiatorIndex,
+			"responderIndex", hs.Details.ResponderIndex,
+			"remoteIndex", h.RemoteIndex,
+			"handshake", m{"stage": 2, "style": "ix_psk0"},
+		)
 	}
 
 	f.connectionManager.AddTrafficWatch(hostinfo)
 
 	hostinfo.remotes.RefreshFromHandshake(vpnAddrs)
 
+	// Don't wait for UpdateWorker
+	if f.lightHouse.IsAnyLighthouseAddr(vpnAddrs) {
+		f.lightHouse.TriggerUpdate()
+	}
+
 	return
 }
 
@@ -478,7 +572,12 @@ func ixHandshakeStage2(f *Interface, via ViaSender, hh *HandshakeHostInfo, packe
 	if !via.IsRelayed {
 		// The vpnAddr we know about is the one we tried to handshake with, use it to apply the remote allow list.
 		if !f.lightHouse.GetRemoteAllowList().AllowAll(hostinfo.vpnAddrs, via.UdpAddr.Addr()) {
-			f.l.WithField("vpnAddrs", hostinfo.vpnAddrs).WithField("from", via).Debug("lighthouse.remote_allow_list denied incoming handshake")
+			if f.l.Enabled(context.Background(), slog.LevelDebug) {
+				f.l.Debug("lighthouse.remote_allow_list denied incoming handshake",
+					"vpnAddrs", hostinfo.vpnAddrs,
+					"from", via,
+				)
+			}
 			return false
 		}
 	}
@@ -486,18 +585,24 @@ func ixHandshakeStage2(f *Interface, via ViaSender, hh *HandshakeHostInfo, packe
 	ci := hostinfo.ConnectionState
 	msg, eKey, dKey, err := ci.H.ReadMessage(nil, packet[header.Len:])
 	if err != nil {
-		f.l.WithError(err).WithField("vpnAddrs", hostinfo.vpnAddrs).WithField("from", via).
+		f.l.Error("Failed to call noise.ReadMessage",
-			WithField("handshake", m{"stage": 2, "style": "ix_psk0"}).WithField("header", h).
+			"error", err,
-			Error("Failed to call noise.ReadMessage")
+			"vpnAddrs", hostinfo.vpnAddrs,
+			"from", via,
+			"handshake", m{"stage": 2, "style": "ix_psk0"},
+			"header", h,
+		)
 
 		// We don't want to tear down the connection on a bad ReadMessage because it could be an attacker trying
 		// to DOS us. Every other error condition after should to allow a possible good handshake to complete in the
 		// near future
 		return false
 	} else if dKey == nil || eKey == nil {
-		f.l.WithField("vpnAddrs", hostinfo.vpnAddrs).WithField("from", via).
+		f.l.Error("Noise did not arrive at a key",
-			WithField("handshake", m{"stage": 2, "style": "ix_psk0"}).
+			"vpnAddrs", hostinfo.vpnAddrs,
-			Error("Noise did not arrive at a key")
+			"from", via,
+			"handshake", m{"stage": 2, "style": "ix_psk0"},
+		)
 
 		// This should be impossible in IX but just in case, if we get here then there is no chance to recover
 		// the handshake state machine. Tear it down
@@ -507,8 +612,12 @@ func ixHandshakeStage2(f *Interface, via ViaSender, hh *HandshakeHostInfo, packe
 	hs := &NebulaHandshake{}
 	err = hs.Unmarshal(msg)
 	if err != nil || hs.Details == nil {
-		f.l.WithError(err).WithField("vpnAddrs", hostinfo.vpnAddrs).WithField("from", via).
+		f.l.Error("Failed unmarshal handshake message",
-			WithField("handshake", m{"stage": 2, "style": "ix_psk0"}).Error("Failed unmarshal handshake message")
+			"error", err,
+			"vpnAddrs", hostinfo.vpnAddrs,
+			"from", via,
+			"handshake", m{"stage": 2, "style": "ix_psk0"},
+		)
 
 		// The handshake state machine is complete, if things break now there is no chance to recover. Tear down and start again
 		return true
@@ -516,10 +625,12 @@ func ixHandshakeStage2(f *Interface, via ViaSender, hh *HandshakeHostInfo, packe
 
 	rc, err := cert.Recombine(cert.Version(hs.Details.CertVersion), hs.Details.Cert, ci.H.PeerStatic(), ci.Curve())
 	if err != nil {
-		f.l.WithError(err).WithField("from", via).
+		f.l.Info("Handshake did not contain a certificate",
-			WithField("vpnAddrs", hostinfo.vpnAddrs).
+			"error", err,
-			WithField("handshake", m{"stage": 2, "style": "ix_psk0"}).
+			"from", via,
-			Info("Handshake did not contain a certificate")
+			"vpnAddrs", hostinfo.vpnAddrs,
+			"handshake", m{"stage": 2, "style": "ix_psk0"},
+		)
 		return true
 	}
 
@@ -530,32 +641,41 @@ func ixHandshakeStage2(f *Interface, via ViaSender, hh *HandshakeHostInfo, packe
 			fp = "<error generating certificate fingerprint>"
 		}
 
-		e := f.l.WithError(err).WithField("from", via).
+		attrs := []slog.Attr{
-			WithField("vpnAddrs", hostinfo.vpnAddrs).
+			slog.Any("error", err),
-			WithField("handshake", m{"stage": 2, "style": "ix_psk0"}).
+			slog.Any("from", via),
-			WithField("certFingerprint", fp).
+			slog.Any("vpnAddrs", hostinfo.vpnAddrs),
-			WithField("certVpnNetworks", rc.Networks())
+			slog.Any("handshake", m{"stage": 2, "style": "ix_psk0"}),
-
+			slog.String("certFingerprint", fp),
-		if f.l.Level >= logrus.DebugLevel {
+			slog.Any("certVpnNetworks", rc.Networks()),
-			e = e.WithField("cert", rc)
+		}
+		if f.l.Enabled(context.Background(), slog.LevelDebug) {
+			attrs = append(attrs, slog.Any("cert", rc))
 		}
 
-		e.Info("Invalid certificate from host")
+		// LogAttrs is intentional: attrs is a pre-built []slog.Attr slice that
+		// callers grow conditionally, which has no pair-form equivalent.
+		//nolint:sloglint
+		f.l.LogAttrs(context.Background(), slog.LevelInfo, "Invalid certificate from host", attrs...)
 		return true
 	}
 	if !bytes.Equal(remoteCert.Certificate.PublicKey(), ci.H.PeerStatic()) {
-		f.l.WithField("from", via).
+		f.l.Info("public key mismatch between certificate and handshake",
-			WithField("handshake", m{"stage": 2, "style": "ix_psk0"}).
+			"from", via,
-			WithField("cert", remoteCert).Info("public key mismatch between certificate and handshake")
+			"handshake", m{"stage": 2, "style": "ix_psk0"},
+			"cert", remoteCert,
+		)
 		return true
 	}
 
 	if len(remoteCert.Certificate.Networks()) == 0 {
-		f.l.WithError(err).WithField("from", via).
+		f.l.Info("No networks in certificate",
-			WithField("vpnAddrs", hostinfo.vpnAddrs).
+			"error", err,
-			WithField("cert", remoteCert).
+			"from", via,
-			WithField("handshake", m{"stage": 2, "style": "ix_psk0"}).
+			"vpnAddrs", hostinfo.vpnAddrs,
-			Info("No networks in certificate")
+			"cert", remoteCert,
+			"handshake", m{"stage": 2, "style": "ix_psk0"},
+		)
 		return true
 	}
 
@@ -596,12 +716,14 @@ func ixHandshakeStage2(f *Interface, via ViaSender, hh *HandshakeHostInfo, packe
 
 	// Ensure the right host responded
 	if !correctHostResponded {
-		f.l.WithField("intendedVpnAddrs", hostinfo.vpnAddrs).WithField("haveVpnNetworks", vpnNetworks).
+		f.l.Info("Incorrect host responded to handshake",
-			WithField("from", via).
+			"intendedVpnAddrs", hostinfo.vpnAddrs,
-			WithField("certName", certName).
+			"haveVpnNetworks", vpnNetworks,
-			WithField("certVersion", certVersion).
+			"from", via,
-			WithField("handshake", m{"stage": 2, "style": "ix_psk0"}).
+			"certName", certName,
-			Info("Incorrect host responded to handshake")
+			"certVersion", certVersion,
+			"handshake", m{"stage": 2, "style": "ix_psk0"},
+		)
 
 		// Release our old handshake from pending, it should not continue
 		f.handshakeManager.DeleteHostInfo(hostinfo)
@@ -613,10 +735,11 @@ func ixHandshakeStage2(f *Interface, via ViaSender, hh *HandshakeHostInfo, packe
 			newHH.hostinfo.remotes = hostinfo.remotes
 			newHH.hostinfo.remotes.BlockRemote(via)
 
-			f.l.WithField("blockedUdpAddrs", newHH.hostinfo.remotes.CopyBlockedRemotes()).
+			f.l.Info("Blocked addresses for handshakes",
-				WithField("vpnNetworks", vpnNetworks).
+				"blockedUdpAddrs", newHH.hostinfo.remotes.CopyBlockedRemotes(),
-				WithField("remotes", newHH.hostinfo.remotes.CopyAddrs(f.hostMap.GetPreferredRanges())).
+				"vpnNetworks", vpnNetworks,
-				Info("Blocked addresses for handshakes")
+				"remotes", newHH.hostinfo.remotes.CopyAddrs(f.hostMap.GetPreferredRanges()),
+			)
 
 			// Swap the packet store to benefit the original intended recipient
 			newHH.packetStore = hh.packetStore
@@ -634,15 +757,20 @@ func ixHandshakeStage2(f *Interface, via ViaSender, hh *HandshakeHostInfo, packe
 	ci.window.Update(f.l, 2)
 
 	duration := time.Since(hh.startTime).Nanoseconds()
-	msgRxL := f.l.WithField("vpnAddrs", vpnAddrs).WithField("from", via).
+	msgRxL := f.l.With(
-		WithField("certName", certName).
+		"vpnAddrs", vpnAddrs,
-		WithField("certVersion", certVersion).
+		"from", via,
-		WithField("fingerprint", fingerprint).
+		"certName", certName,
-		WithField("issuer", issuer).
+		"certVersion", certVersion,
-		WithField("initiatorIndex", hs.Details.InitiatorIndex).WithField("responderIndex", hs.Details.ResponderIndex).
+		"fingerprint", fingerprint,
-		WithField("remoteIndex", h.RemoteIndex).WithField("handshake", m{"stage": 2, "style": "ix_psk0"}).
+		"issuer", issuer,
-		WithField("durationNs", duration).
+		"initiatorIndex", hs.Details.InitiatorIndex,
-		WithField("sentCachedPackets", len(hh.packetStore))
+		"responderIndex", hs.Details.ResponderIndex,
+		"remoteIndex", h.RemoteIndex,
+		"handshake", m{"stage": 2, "style": "ix_psk0"},
+		"durationNs", duration,
+		"sentCachedPackets", len(hh.packetStore),
+	)
 	if anyVpnAddrsInCommon {
 		msgRxL.Info("Handshake message received")
 	} else {
@@ -658,8 +786,10 @@ func ixHandshakeStage2(f *Interface, via ViaSender, hh *HandshakeHostInfo, packe
 	f.handshakeManager.Complete(hostinfo, f)
 	f.connectionManager.AddTrafficWatch(hostinfo)
 
-	if f.l.Level >= logrus.DebugLevel {
+	if f.l.Enabled(context.Background(), slog.LevelDebug) {
-		hostinfo.logger(f.l).Debugf("Sending %d stored packets", len(hh.packetStore))
+		hostinfo.logger(f.l).Debug("Sending stored packets",
+			"count", len(hh.packetStore),
+		)
 	}
 
 	if len(hh.packetStore) > 0 {
@@ -674,5 +804,10 @@ func ixHandshakeStage2(f *Interface, via ViaSender, hh *HandshakeHostInfo, packe
 	hostinfo.remotes.RefreshFromHandshake(vpnAddrs)
 	f.metricHandshakes.Update(duration)
 
+	// Don't wait for UpdateWorker
+	if f.lightHouse.IsAnyLighthouseAddr(vpnAddrs) {
+		f.lightHouse.TriggerUpdate()
+	}
+
 	return false
 }

handshake_manager.go

@@ -6,13 +6,13 @@ import (
 	"crypto/rand"
 	"encoding/binary"
 	"errors"
+	"log/slog"
 	"net/netip"
 	"slices"
 	"sync"
 	"time"
 
 	"github.com/rcrowley/go-metrics"
-	"github.com/sirupsen/logrus"
 	"github.com/slackhq/nebula/cert"
 	"github.com/slackhq/nebula/header"
 	"github.com/slackhq/nebula/udp"
@@ -59,7 +59,7 @@ type HandshakeManager struct {
 	metricInitiated        metrics.Counter
 	metricTimedOut         metrics.Counter
 	f                      *Interface
-	l                      *logrus.Logger
+	l                      *slog.Logger
 
 	// can be used to trigger outbound handshake for the given vpnIp
 	trigger chan netip.Addr
@@ -78,32 +78,32 @@ type HandshakeHostInfo struct {
 	hostinfo *HostInfo
 }
 
-func (hh *HandshakeHostInfo) cachePacket(l *logrus.Logger, t header.MessageType, st header.MessageSubType, packet []byte, f packetCallback, m *cachedPacketMetrics) {
+func (hh *HandshakeHostInfo) cachePacket(l *slog.Logger, t header.MessageType, st header.MessageSubType, packet []byte, f packetCallback, m *cachedPacketMetrics) {
 	if len(hh.packetStore) < 100 {
 		tempPacket := make([]byte, len(packet))
 		copy(tempPacket, packet)
 
 		hh.packetStore = append(hh.packetStore, &cachedPacket{t, st, f, tempPacket})
-		if l.Level >= logrus.DebugLevel {
+		if l.Enabled(context.Background(), slog.LevelDebug) {
-			hh.hostinfo.logger(l).
+			hh.hostinfo.logger(l).Debug("Packet store",
-				WithField("length", len(hh.packetStore)).
+				"length", len(hh.packetStore),
-				WithField("stored", true).
+				"stored", true,
-				Debugf("Packet store")
+			)
 		}
 
 	} else {
 		m.dropped.Inc(1)
 
-		if l.Level >= logrus.DebugLevel {
+		if l.Enabled(context.Background(), slog.LevelDebug) {
-			hh.hostinfo.logger(l).
+			hh.hostinfo.logger(l).Debug("Packet store",
-				WithField("length", len(hh.packetStore)).
+				"length", len(hh.packetStore),
-				WithField("stored", false).
+				"stored", false,
-				Debugf("Packet store")
+			)
 		}
 	}
 }
 
-func NewHandshakeManager(l *logrus.Logger, mainHostMap *HostMap, lightHouse *LightHouse, outside udp.Conn, config HandshakeConfig) *HandshakeManager {
+func NewHandshakeManager(l *slog.Logger, mainHostMap *HostMap, lightHouse *LightHouse, outside udp.Conn, config HandshakeConfig) *HandshakeManager {
 	return &HandshakeManager{
 		vpnIps:                 map[netip.Addr]*HandshakeHostInfo{},
 		indexes:                map[uint32]*HandshakeHostInfo{},
@@ -140,7 +140,7 @@ func (hm *HandshakeManager) HandleIncoming(via ViaSender, packet []byte, h *head
 	// First remote allow list check before we know the vpnIp
 	if !via.IsRelayed {
 		if !hm.lightHouse.GetRemoteAllowList().AllowUnknownVpnAddr(via.UdpAddr.Addr()) {
-			hm.l.WithField("from", via).Debug("lighthouse.remote_allow_list denied incoming handshake")
+			hm.l.Debug("lighthouse.remote_allow_list denied incoming handshake", "from", via)
 			return
 		}
 	}
@@ -183,12 +183,13 @@ func (hm *HandshakeManager) handleOutbound(vpnIp netip.Addr, lighthouseTriggered
 	hostinfo := hh.hostinfo
 	// If we are out of time, clean up
 	if hh.counter >= hm.config.retries {
-		hh.hostinfo.logger(hm.l).WithField("udpAddrs", hh.hostinfo.remotes.CopyAddrs(hm.mainHostMap.GetPreferredRanges())).
+		hh.hostinfo.logger(hm.l).Info("Handshake timed out",
-			WithField("initiatorIndex", hh.hostinfo.localIndexId).
+			"udpAddrs", hh.hostinfo.remotes.CopyAddrs(hm.mainHostMap.GetPreferredRanges()),
-			WithField("remoteIndex", hh.hostinfo.remoteIndexId).
+			"initiatorIndex", hh.hostinfo.localIndexId,
-			WithField("handshake", m{"stage": 1, "style": "ix_psk0"}).
+			"remoteIndex", hh.hostinfo.remoteIndexId,
-			WithField("durationNs", time.Since(hh.startTime).Nanoseconds()).
+			"handshake", m{"stage": 1, "style": "ix_psk0"},
-			Info("Handshake timed out")
+			"durationNs", time.Since(hh.startTime).Nanoseconds(),
+		)
 		hm.metricTimedOut.Inc(1)
 		hm.DeleteHostInfo(hostinfo)
 		return
@@ -241,10 +242,12 @@ func (hm *HandshakeManager) handleOutbound(vpnIp netip.Addr, lighthouseTriggered
 		hm.messageMetrics.Tx(header.Handshake, header.MessageSubType(hostinfo.HandshakePacket[0][1]), 1)
 		err := hm.outside.WriteTo(hostinfo.HandshakePacket[0], addr)
 		if err != nil {
-			hostinfo.logger(hm.l).WithField("udpAddr", addr).
+			hostinfo.logger(hm.l).Error("Failed to send handshake message",
-				WithField("initiatorIndex", hostinfo.localIndexId).
+				"udpAddr", addr,
-				WithField("handshake", m{"stage": 1, "style": "ix_psk0"}).
+				"initiatorIndex", hostinfo.localIndexId,
-				WithError(err).Error("Failed to send handshake message")
+				"handshake", m{"stage": 1, "style": "ix_psk0"},
+				"error", err,
+			)
 
 		} else {
 			sentTo = append(sentTo, addr)
@@ -254,19 +257,21 @@ func (hm *HandshakeManager) handleOutbound(vpnIp netip.Addr, lighthouseTriggered
 	// Don't be too noisy or confusing if we fail to send a handshake - if we don't get through we'll eventually log a timeout,
 	// so only log when the list of remotes has changed
 	if remotesHaveChanged {
-		hostinfo.logger(hm.l).WithField("udpAddrs", sentTo).
+		hostinfo.logger(hm.l).Info("Handshake message sent",
-			WithField("initiatorIndex", hostinfo.localIndexId).
+			"udpAddrs", sentTo,
-			WithField("handshake", m{"stage": 1, "style": "ix_psk0"}).
+			"initiatorIndex", hostinfo.localIndexId,
-			Info("Handshake message sent")
+			"handshake", m{"stage": 1, "style": "ix_psk0"},
-	} else if hm.l.Level >= logrus.DebugLevel {
+		)
-		hostinfo.logger(hm.l).WithField("udpAddrs", sentTo).
+	} else if hm.l.Enabled(context.Background(), slog.LevelDebug) {
-			WithField("initiatorIndex", hostinfo.localIndexId).
+		hostinfo.logger(hm.l).Debug("Handshake message sent",
-			WithField("handshake", m{"stage": 1, "style": "ix_psk0"}).
+			"udpAddrs", sentTo,
-			Debug("Handshake message sent")
+			"initiatorIndex", hostinfo.localIndexId,
+			"handshake", m{"stage": 1, "style": "ix_psk0"},
+		)
 	}
 
 	if hm.config.useRelays && len(hostinfo.remotes.relays) > 0 {
-		hostinfo.logger(hm.l).WithField("relays", hostinfo.remotes.relays).Info("Attempt to relay through hosts")
+		hostinfo.logger(hm.l).Info("Attempt to relay through hosts", "relays", hostinfo.remotes.relays)
 		// Send a RelayRequest to all known Relay IP's
 		for _, relay := range hostinfo.remotes.relays {
 			// Don't relay through the host I'm trying to connect to
@@ -281,7 +286,7 @@ func (hm *HandshakeManager) handleOutbound(vpnIp netip.Addr, lighthouseTriggered
 
 			relayHostInfo := hm.mainHostMap.QueryVpnAddr(relay)
 			if relayHostInfo == nil || !relayHostInfo.remote.IsValid() {
-				hostinfo.logger(hm.l).WithField("relay", relay.String()).Info("Establish tunnel to relay target")
+				hostinfo.logger(hm.l).Info("Establish tunnel to relay target", "relay", relay.String())
 				hm.f.Handshake(relay)
 				continue
 			}
@@ -292,7 +297,7 @@ func (hm *HandshakeManager) handleOutbound(vpnIp netip.Addr, lighthouseTriggered
 				if relayHostInfo.remote.IsValid() {
 					idx, err := AddRelay(hm.l, relayHostInfo, hm.mainHostMap, vpnIp, nil, TerminalType, Requested)
 					if err != nil {
-						hostinfo.logger(hm.l).WithField("relay", relay.String()).WithError(err).Info("Failed to add relay to hostmap")
+						hostinfo.logger(hm.l).Info("Failed to add relay to hostmap", "relay", relay.String(), "error", err)
 					}
 
 					m := NebulaControl{
@@ -326,17 +331,15 @@ func (hm *HandshakeManager) handleOutbound(vpnIp netip.Addr, lighthouseTriggered
 
 					msg, err := m.Marshal()
 					if err != nil {
-						hostinfo.logger(hm.l).
+						hostinfo.logger(hm.l).Error("Failed to marshal Control message to create relay", "error", err)
-							WithError(err).
-							Error("Failed to marshal Control message to create relay")
 					} else {
 						hm.f.SendMessageToHostInfo(header.Control, 0, relayHostInfo, msg, make([]byte, 12), make([]byte, mtu))
-						hm.l.WithFields(logrus.Fields{
+						hm.l.Info("send CreateRelayRequest",
-							"relayFrom":           hm.f.myVpnAddrs[0],
+							"relayFrom", hm.f.myVpnAddrs[0],
-							"relayTo":             vpnIp,
+							"relayTo", vpnIp,
-							"initiatorRelayIndex": idx,
+							"initiatorRelayIndex", idx,
-							"relay":               relay}).
+							"relay", relay,
-							Info("send CreateRelayRequest")
+						)
 					}
 				}
 				continue
@@ -344,14 +347,14 @@ func (hm *HandshakeManager) handleOutbound(vpnIp netip.Addr, lighthouseTriggered
 
 			switch existingRelay.State {
 			case Established:
-				hostinfo.logger(hm.l).WithField("relay", relay.String()).Info("Send handshake via relay")
+				hostinfo.logger(hm.l).Info("Send handshake via relay", "relay", relay.String())
 				hm.f.SendVia(relayHostInfo, existingRelay, hostinfo.HandshakePacket[0], make([]byte, 12), make([]byte, mtu), false)
 			case Disestablished:
 				// Mark this relay as 'requested'
 				relayHostInfo.relayState.UpdateRelayForByIpState(vpnIp, Requested)
 				fallthrough
 			case Requested:
-				hostinfo.logger(hm.l).WithField("relay", relay.String()).Info("Re-send CreateRelay request")
+				hostinfo.logger(hm.l).Info("Re-send CreateRelay request", "relay", relay.String())
 				// Re-send the CreateRelay request, in case the previous one was lost.
 				m := NebulaControl{
 					Type:                NebulaControl_CreateRelayRequest,
@@ -383,28 +386,26 @@ func (hm *HandshakeManager) handleOutbound(vpnIp netip.Addr, lighthouseTriggered
 				}
 				msg, err := m.Marshal()
 				if err != nil {
-					hostinfo.logger(hm.l).
+					hostinfo.logger(hm.l).Error("Failed to marshal Control message to create relay", "error", err)
-						WithError(err).
-						Error("Failed to marshal Control message to create relay")
 				} else {
 					// This must send over the hostinfo, not over hm.Hosts[ip]
 					hm.f.SendMessageToHostInfo(header.Control, 0, relayHostInfo, msg, make([]byte, 12), make([]byte, mtu))
-					hm.l.WithFields(logrus.Fields{
+					hm.l.Info("send CreateRelayRequest",
-						"relayFrom":           hm.f.myVpnAddrs[0],
+						"relayFrom", hm.f.myVpnAddrs[0],
-						"relayTo":             vpnIp,
+						"relayTo", vpnIp,
-						"initiatorRelayIndex": existingRelay.LocalIndex,
+						"initiatorRelayIndex", existingRelay.LocalIndex,
-						"relay":               relay}).
+						"relay", relay,
-						Info("send CreateRelayRequest")
+					)
 				}
 			case PeerRequested:
 				// PeerRequested only occurs in Forwarding relays, not Terminal relays, and this is a Terminal relay case.
 				fallthrough
 			default:
-				hostinfo.logger(hm.l).
+				hostinfo.logger(hm.l).Error("Relay unexpected state",
-					WithField("vpnIp", vpnIp).
+					"vpnIp", vpnIp,
-					WithField("state", existingRelay.State).
+					"state", existingRelay.State,
-					WithField("relay", relay).
+					"relay", relay,
-					Errorf("Relay unexpected state")
+				)
 
 			}
 		}
@@ -549,9 +550,10 @@ func (hm *HandshakeManager) CheckAndComplete(hostinfo *HostInfo, handshakePacket
 	if found && existingRemoteIndex != nil && existingRemoteIndex.vpnAddrs[0] != hostinfo.vpnAddrs[0] {
 		// We have a collision, but this can happen since we can't control
 		// the remote ID. Just log about the situation as a note.
-		hostinfo.logger(hm.l).
+		hostinfo.logger(hm.l).Info("New host shadows existing host remoteIndex",
-			WithField("remoteIndex", hostinfo.remoteIndexId).WithField("collision", existingRemoteIndex.vpnAddrs).
+			"remoteIndex", hostinfo.remoteIndexId,
-			Info("New host shadows existing host remoteIndex")
+			"collision", existingRemoteIndex.vpnAddrs,
+		)
 	}
 
 	hm.mainHostMap.unlockedAddHostInfo(hostinfo, f)
@@ -571,9 +573,10 @@ func (hm *HandshakeManager) Complete(hostinfo *HostInfo, f *Interface) {
 	if found && existingRemoteIndex != nil {
 		// We have a collision, but this can happen since we can't control
 		// the remote ID. Just log about the situation as a note.
-		hostinfo.logger(hm.l).
+		hostinfo.logger(hm.l).Info("New host shadows existing host remoteIndex",
-			WithField("remoteIndex", hostinfo.remoteIndexId).WithField("collision", existingRemoteIndex.vpnAddrs).
+			"remoteIndex", hostinfo.remoteIndexId,
-			Info("New host shadows existing host remoteIndex")
+			"collision", existingRemoteIndex.vpnAddrs,
+		)
 	}
 
 	// We need to remove from the pending hostmap first to avoid undoing work when after to the main hostmap.
@@ -629,10 +632,11 @@ func (hm *HandshakeManager) unlockedDeleteHostInfo(hostinfo *HostInfo) {
 		hm.indexes = map[uint32]*HandshakeHostInfo{}
 	}
 
-	if hm.l.Level >= logrus.DebugLevel {
+	if hm.l.Enabled(context.Background(), slog.LevelDebug) {
-		hm.l.WithField("hostMap", m{"mapTotalSize": len(hm.vpnIps),
+		hm.l.Debug("Pending hostmap hostInfo deleted",
-			"vpnAddrs": hostinfo.vpnAddrs, "indexNumber": hostinfo.localIndexId, "remoteIndexNumber": hostinfo.remoteIndexId}).
+			"hostMap", m{"mapTotalSize": len(hm.vpnIps),
-			Debug("Pending hostmap hostInfo deleted")
+				"vpnAddrs": hostinfo.vpnAddrs, "indexNumber": hostinfo.localIndexId, "remoteIndexNumber": hostinfo.remoteIndexId},
+		)
 	}
 }
 
@@ -700,7 +704,7 @@ func (hm *HandshakeManager) EmitStats() {
 
 // Utility functions below
 
-func generateIndex(l *logrus.Logger) (uint32, error) {
+func generateIndex(l *slog.Logger) (uint32, error) {
 	b := make([]byte, 4)
 
 	// Let zero mean we don't know the ID, so don't generate zero
@@ -708,16 +712,15 @@ func generateIndex(l *logrus.Logger) (uint32, error) {
 	for index == 0 {
 		_, err := rand.Read(b)
 		if err != nil {
-			l.Errorln(err)
+			l.Error("Failed to generate index", "error", err)
 			return 0, err
 		}
 
 		index = binary.BigEndian.Uint32(b)
 	}
 
-	if l.Level >= logrus.DebugLevel {
+	if l.Enabled(context.Background(), slog.LevelDebug) {
-		l.WithField("index", index).
+		l.Debug("Generated index", "index", index)
-			Debug("Generated index")
 	}
 	return index, nil
 }

hostmap.go

@@ -1,9 +1,11 @@
 package nebula
 
 import (
+	"context"
 	"encoding/json"
 	"errors"
 	"fmt"
+	"log/slog"
 	"net"
 	"net/netip"
 	"slices"
@@ -13,10 +15,10 @@ import (
 
 	"github.com/gaissmai/bart"
 	"github.com/rcrowley/go-metrics"
-	"github.com/sirupsen/logrus"
 	"github.com/slackhq/nebula/cert"
 	"github.com/slackhq/nebula/config"
 	"github.com/slackhq/nebula/header"
+	"github.com/slackhq/nebula/logging"
 )
 
 const defaultPromoteEvery = 1000       // Count of packets sent before we try moving a tunnel to a preferred underlay ip address
@@ -60,7 +62,7 @@ type HostMap struct {
 	RemoteIndexes   map[uint32]*HostInfo
 	Hosts           map[netip.Addr]*HostInfo
 	preferredRanges atomic.Pointer[[]netip.Prefix]
-	l               *logrus.Logger
+	l               *slog.Logger
 }
 
 // For synchronization, treat the pointed-to Relay struct as immutable. To edit the Relay
@@ -313,7 +315,7 @@ type cachedPacketMetrics struct {
 	dropped metrics.Counter
 }
 
-func NewHostMapFromConfig(l *logrus.Logger, c *config.C) *HostMap {
+func NewHostMapFromConfig(l *slog.Logger, c *config.C) *HostMap {
 	hm := newHostMap(l)
 
 	hm.reload(c, true)
@@ -321,13 +323,12 @@ func NewHostMapFromConfig(l *logrus.Logger, c *config.C) *HostMap {
 		hm.reload(c, false)
 	})
 
-	l.WithField("preferredRanges", hm.GetPreferredRanges()).
+	l.Info("Main HostMap created", "preferredRanges", hm.GetPreferredRanges())
-		Info("Main HostMap created")
 
 	return hm
 }
 
-func newHostMap(l *logrus.Logger) *HostMap {
+func newHostMap(l *slog.Logger) *HostMap {
 	return &HostMap{
 		Indexes:       map[uint32]*HostInfo{},
 		Relays:        map[uint32]*HostInfo{},
@@ -346,7 +347,10 @@ func (hm *HostMap) reload(c *config.C, initial bool) {
 			preferredRange, err := netip.ParsePrefix(rawPreferredRange)
 
 			if err != nil {
-				hm.l.WithError(err).WithField("range", rawPreferredRanges).Warn("Failed to parse preferred ranges, ignoring")
+				hm.l.Warn("Failed to parse preferred ranges, ignoring",
+					"error", err,
+					"range", rawPreferredRanges,
+				)
 				continue
 			}
 
@@ -355,7 +359,10 @@ func (hm *HostMap) reload(c *config.C, initial bool) {
 
 		oldRanges := hm.preferredRanges.Swap(&preferredRanges)
 		if !initial {
-			hm.l.WithField("oldPreferredRanges", *oldRanges).WithField("newPreferredRanges", preferredRanges).Info("preferred_ranges changed")
+			hm.l.Info("preferred_ranges changed",
+				"oldPreferredRanges", *oldRanges,
+				"newPreferredRanges", preferredRanges,
+			)
 		}
 	}
 }
@@ -488,10 +495,11 @@ func (hm *HostMap) unlockedInnerDeleteHostInfo(hostinfo *HostInfo, addr netip.Ad
 		hm.Indexes = map[uint32]*HostInfo{}
 	}
 
-	if hm.l.Level >= logrus.DebugLevel {
+	if hm.l.Enabled(context.Background(), slog.LevelDebug) {
-		hm.l.WithField("hostMap", m{"mapTotalSize": len(hm.Hosts),
+		hm.l.Debug("Hostmap hostInfo deleted",
-			"vpnAddrs": hostinfo.vpnAddrs, "indexNumber": hostinfo.localIndexId, "remoteIndexNumber": hostinfo.remoteIndexId}).
+			"hostMap", m{"mapTotalSize": len(hm.Hosts),
-			Debug("Hostmap hostInfo deleted")
+				"vpnAddrs": hostinfo.vpnAddrs, "indexNumber": hostinfo.localIndexId, "remoteIndexNumber": hostinfo.remoteIndexId},
+		)
 	}
 
 	if isLastHostinfo {
@@ -604,9 +612,9 @@ func (hm *HostMap) queryVpnAddr(vpnIp netip.Addr, promoteIfce *Interface) *HostI
 // unlockedAddHostInfo assumes you have a write-lock and will add a hostinfo object to the hostmap Indexes and RemoteIndexes maps.
 // If an entry exists for the Hosts table (vpnIp -> hostinfo) then the provided hostinfo will be made primary
 func (hm *HostMap) unlockedAddHostInfo(hostinfo *HostInfo, f *Interface) {
-	if f.serveDns {
+	if f.dnsServer != nil {
 		remoteCert := hostinfo.ConnectionState.peerCert
-		dnsR.Add(remoteCert.Certificate.Name()+".", hostinfo.vpnAddrs)
+		f.dnsServer.Add(remoteCert.Certificate.Name()+".", hostinfo.vpnAddrs)
 	}
 	for _, addr := range hostinfo.vpnAddrs {
 		hm.unlockedInnerAddHostInfo(addr, hostinfo, f)
@@ -615,10 +623,11 @@ func (hm *HostMap) unlockedAddHostInfo(hostinfo *HostInfo, f *Interface) {
 	hm.Indexes[hostinfo.localIndexId] = hostinfo
 	hm.RemoteIndexes[hostinfo.remoteIndexId] = hostinfo
 
-	if hm.l.Level >= logrus.DebugLevel {
+	if hm.l.Enabled(context.Background(), slog.LevelDebug) {
-		hm.l.WithField("hostMap", m{"vpnAddrs": hostinfo.vpnAddrs, "mapTotalSize": len(hm.Hosts),
+		hm.l.Debug("Hostmap vpnIp added",
-			"hostinfo": m{"existing": true, "localIndexId": hostinfo.localIndexId, "vpnAddrs": hostinfo.vpnAddrs}}).
+			"hostMap", m{"vpnAddrs": hostinfo.vpnAddrs, "mapTotalSize": len(hm.Hosts),
-			Debug("Hostmap vpnIp added")
+				"hostinfo": m{"existing": true, "localIndexId": hostinfo.localIndexId, "vpnAddrs": hostinfo.vpnAddrs}},
+		)
 	}
 }
 
@@ -784,18 +793,21 @@ func (i *HostInfo) buildNetworks(myVpnNetworksTable *bart.Lite, c cert.Certifica
 	}
 }
 
-func (i *HostInfo) logger(l *logrus.Logger) *logrus.Entry {
+// logger returns a derived slog.Logger with per-hostinfo fields pre-bound.
+func (i *HostInfo) logger(l *slog.Logger) *slog.Logger {
 	if i == nil {
-		return logrus.NewEntry(l)
+		return l
 	}
 
-	li := l.WithField("vpnAddrs", i.vpnAddrs).
+	li := l.With(
-		WithField("localIndex", i.localIndexId).
+		"vpnAddrs", i.vpnAddrs,
-		WithField("remoteIndex", i.remoteIndexId)
+		"localIndex", i.localIndexId,
+		"remoteIndex", i.remoteIndexId,
+	)
 
 	if connState := i.ConnectionState; connState != nil {
 		if peerCert := connState.peerCert; peerCert != nil {
-			li = li.WithField("certName", peerCert.Certificate.Name())
+			li = li.With("certName", peerCert.Certificate.Name())
 		}
 	}
 
@@ -804,14 +816,17 @@ func (i *HostInfo) logger(l *logrus.Logger) *logrus.Entry {
 
 // Utility functions
 
-func localAddrs(l *logrus.Logger, allowList *LocalAllowList) []netip.Addr {
+func localAddrs(l *slog.Logger, allowList *LocalAllowList) []netip.Addr {
 	//FIXME: This function is pretty garbage
 	var finalAddrs []netip.Addr
 	ifaces, _ := net.Interfaces()
 	for _, i := range ifaces {
 		allow := allowList.AllowName(i.Name)
-		if l.Level >= logrus.TraceLevel {
+		if l.Enabled(context.Background(), logging.LevelTrace) {
-			l.WithField("interfaceName", i.Name).WithField("allow", allow).Trace("localAllowList.AllowName")
+			l.Log(context.Background(), logging.LevelTrace, "localAllowList.AllowName",
+				"interfaceName", i.Name,
+				"allow", allow,
+			)
 		}
 
 		if !allow {
@@ -829,8 +844,8 @@ func localAddrs(l *logrus.Logger, allowList *LocalAllowList) []netip.Addr {
 			}
 
 			if !addr.IsValid() {
-				if l.Level >= logrus.DebugLevel {
+				if l.Enabled(context.Background(), slog.LevelDebug) {
-					l.WithField("localAddr", rawAddr).Debug("addr was invalid")
+					l.Debug("addr was invalid", "localAddr", rawAddr)
 				}
 				continue
 			}
@@ -838,8 +853,11 @@ func localAddrs(l *logrus.Logger, allowList *LocalAllowList) []netip.Addr {
 
 			if addr.IsLoopback() == false && addr.IsLinkLocalUnicast() == false {
 				isAllowed := allowList.Allow(addr)
-				if l.Level >= logrus.TraceLevel {
+				if l.Enabled(context.Background(), logging.LevelTrace) {
-					l.WithField("localAddr", addr).WithField("allowed", isAllowed).Trace("localAllowList.Allow")
+					l.Log(context.Background(), logging.LevelTrace, "localAllowList.Allow",
+						"localAddr", addr,
+						"allowed", isAllowed,
+					)
 				}
 				if !isAllowed {
 					continue

hostmap_test.go

@@ -196,7 +196,7 @@ func TestHostMap_DeleteHostInfo(t *testing.T) {
 
 func TestHostMap_reload(t *testing.T) {
 	l := test.NewLogger()
-	c := config.NewC(l)
+	c := config.NewC(test.NewLogger())
 
 	hm := NewHostMapFromConfig(l, c)
 

inside.go

@@ -1,21 +1,26 @@
 package nebula
 
 import (
+	"context"
+	"log/slog"
 	"net/netip"
 
-	"github.com/sirupsen/logrus"
 	"github.com/slackhq/nebula/firewall"
 	"github.com/slackhq/nebula/header"
 	"github.com/slackhq/nebula/iputil"
 	"github.com/slackhq/nebula/noiseutil"
+	"github.com/slackhq/nebula/overlay/batch"
 	"github.com/slackhq/nebula/routing"
 )
 
-func (f *Interface) consumeInsidePacket(packet []byte, fwPacket *firewall.Packet, nb []byte, batch *sendBatch, rejectBuf []byte, q int, localCache firewall.ConntrackCache) {
+func (f *Interface) consumeInsidePacket(packet []byte, fwPacket *firewall.Packet, nb []byte, sendBatch batch.TxBatcher, rejectBuf []byte, q int, localCache firewall.ConntrackCache) {
 	err := newPacket(packet, false, fwPacket)
 	if err != nil {
-		if f.l.Level >= logrus.DebugLevel {
+		if f.l.Enabled(context.Background(), slog.LevelDebug) {
-			f.l.WithField("packet", packet).Debugf("Error while validating outbound packet: %s", err)
+			f.l.Debug("Error while validating outbound packet",
+				"packet", packet,
+				"error", err,
+			)
 		}
 		return
 	}
@@ -33,9 +38,9 @@ func (f *Interface) consumeInsidePacket(packet []byte, fwPacket *firewall.Packet
 		// routes packets from the Nebula addr to the Nebula addr through the Nebula
 		// TUN device.
 		if immediatelyForwardToSelf {
-			_, err := f.readers[q].WriteReject(packet)
+			_, err := f.readers[q].Write(packet)
 			if err != nil {
-				f.l.WithError(err).Error("Failed to forward to tun")
+				f.l.Error("Failed to forward to tun", "error", err)
 			}
 		}
 		// Otherwise, drop. On linux, we should never see these packets - Linux
@@ -54,10 +59,11 @@ func (f *Interface) consumeInsidePacket(packet []byte, fwPacket *firewall.Packet
 
 	if hostinfo == nil {
 		f.rejectInside(packet, rejectBuf, q)
-		if f.l.Level >= logrus.DebugLevel {
+		if f.l.Enabled(context.Background(), slog.LevelDebug) {
-			f.l.WithField("vpnAddr", fwPacket.RemoteAddr).
+			f.l.Debug("dropping outbound packet, vpnAddr not in our vpn networks or in unsafe networks",
-				WithField("fwPacket", fwPacket).
+				"vpnAddr", fwPacket.RemoteAddr,
-				Debugln("dropping outbound packet, vpnAddr not in our vpn networks or in unsafe networks")
+				"fwPacket", fwPacket,
+			)
 		}
 		return
 	}
@@ -68,15 +74,14 @@ func (f *Interface) consumeInsidePacket(packet []byte, fwPacket *firewall.Packet
 
 	dropReason := f.firewall.Drop(*fwPacket, false, hostinfo, f.pki.GetCAPool(), localCache)
 	if dropReason == nil {
-		f.sendInsideMessage(hostinfo, packet, nb, batch, rejectBuf, q)
+		f.sendInsideMessage(hostinfo, packet, nb, sendBatch, rejectBuf, q)
-
 	} else {
 		f.rejectInside(packet, rejectBuf, q)
-		if f.l.Level >= logrus.DebugLevel {
+		if f.l.Enabled(context.Background(), slog.LevelDebug) {
-			hostinfo.logger(f.l).
+			hostinfo.logger(f.l).Debug("dropping outbound packet",
-				WithField("fwPacket", fwPacket).
+				"fwPacket", fwPacket,
-				WithField("reason", dropReason).
+				"reason", dropReason,
-				Debugln("dropping outbound packet")
+			)
 		}
 	}
 }
@@ -85,7 +90,7 @@ func (f *Interface) consumeInsidePacket(packet []byte, fwPacket *firewall.Packet
 // caller's batch slot for later sendmmsg flush. When hostinfo.remote is not
 // valid we fall through to the relay slow path via the unbatched sendNoMetrics
 // so relay behavior is unchanged.
-func (f *Interface) sendInsideMessage(hostinfo *HostInfo, p, nb []byte, batch *sendBatch, rejectBuf []byte, q int) {
+func (f *Interface) sendInsideMessage(hostinfo *HostInfo, p, nb []byte, sendBatch batch.TxBatcher, rejectBuf []byte, q int) {
 	ci := hostinfo.ConnectionState
 	if ci.eKey == nil {
 		return
@@ -98,7 +103,7 @@ func (f *Interface) sendInsideMessage(hostinfo *HostInfo, p, nb []byte, batch *s
 		return
 	}
 
-	scratch := batch.Next()
+	scratch := sendBatch.Next()
 	if scratch == nil {
 		// Batch full: bypass batching and send this packet directly so we
 		// never drop traffic on over-subscribed iterations.
@@ -119,8 +124,10 @@ func (f *Interface) sendInsideMessage(hostinfo *HostInfo, p, nb []byte, batch *s
 		// finally used again. This tunnel would eventually be torn down and recreated if this action didn't help.
 		f.lightHouse.QueryServer(hostinfo.vpnAddrs[0])
 		hostinfo.lastRebindCount = f.rebindCount
-		if f.l.Level >= logrus.DebugLevel {
+		if f.l.Enabled(context.Background(), slog.LevelDebug) {
-			f.l.WithField("vpnAddrs", hostinfo.vpnAddrs).Debug("Lighthouse update triggered for punch due to rebind counter")
+			hostinfo.logger(f.l).Debug("Lighthouse update triggered for punch due to rebind counter",
+				"vpnAddrs", hostinfo.vpnAddrs,
+			)
 		}
 	}
 
@@ -129,13 +136,15 @@ func (f *Interface) sendInsideMessage(hostinfo *HostInfo, p, nb []byte, batch *s
 		ci.writeLock.Unlock()
 	}
 	if err != nil {
-		hostinfo.logger(f.l).WithError(err).
+		hostinfo.logger(f.l).Error("Failed to encrypt outgoing packet",
-			WithField("udpAddr", hostinfo.remote).WithField("counter", c).
+			"error", err,
-			Error("Failed to encrypt outgoing packet")
+			"udpAddr", hostinfo.remote,
+			"counter", c,
+		)
 		return
 	}
 
-	batch.Commit(len(out), hostinfo.remote)
+	sendBatch.Commit(len(out), hostinfo.remote)
 }
 
 func (f *Interface) rejectInside(packet []byte, out []byte, q int) {
@@ -148,9 +157,9 @@ func (f *Interface) rejectInside(packet []byte, out []byte, q int) {
 		return
 	}
 
-	_, err := f.readers[q].WriteReject(out)
+	_, err := f.readers[q].Write(out)
 	if err != nil {
-		f.l.WithError(err).Error("Failed to write to tun")
+		f.l.Error("Failed to write to tun", "error", err)
 	}
 }
 
@@ -165,11 +174,11 @@ func (f *Interface) rejectOutside(packet []byte, ci *ConnectionState, hostinfo *
 	}
 
 	if len(out) > iputil.MaxRejectPacketSize {
-		if f.l.GetLevel() >= logrus.InfoLevel {
+		if f.l.Enabled(context.Background(), slog.LevelInfo) {
-			f.l.
+			f.l.Info("rejectOutside: packet too big, not sending",
-				WithField("packet", packet).
+				"packet", packet,
-				WithField("outPacket", out).
+				"outPacket", out,
-				Info("rejectOutside: packet too big, not sending")
+			)
 		}
 		return
 	}
@@ -241,10 +250,11 @@ func (f *Interface) getOrHandshakeConsiderRouting(fwPacket *firewall.Packet, cac
 		// This would also need to interact with unsafe_route updates through reloading the config or
 		// use of the use_system_route_table option
 
-		if f.l.Level >= logrus.DebugLevel {
+		if f.l.Enabled(context.Background(), slog.LevelDebug) {
-			f.l.WithField("destination", destinationAddr).
+			f.l.Debug("Calculated gateway for ECMP not available, attempting other gateways",
-				WithField("originalGateway", gatewayAddr).
+				"destination", destinationAddr,
-				Debugln("Calculated gateway for ECMP not available, attempting other gateways")
+				"originalGateway", gatewayAddr,
+			)
 		}
 
 		for i := range gateways {
@@ -270,17 +280,18 @@ func (f *Interface) sendMessageNow(t header.MessageType, st header.MessageSubTyp
 	fp := &firewall.Packet{}
 	err := newPacket(p, false, fp)
 	if err != nil {
-		f.l.Warnf("error while parsing outgoing packet for firewall check; %v", err)
+		f.l.Warn("error while parsing outgoing packet for firewall check", "error", err)
 		return
 	}
 
 	// check if packet is in outbound fw rules
 	dropReason := f.firewall.Drop(*fp, false, hostinfo, f.pki.GetCAPool(), nil)
 	if dropReason != nil {
-		if f.l.Level >= logrus.DebugLevel {
+		if f.l.Enabled(context.Background(), slog.LevelDebug) {
-			f.l.WithField("fwPacket", fp).
+			f.l.Debug("dropping cached packet",
-				WithField("reason", dropReason).
+				"fwPacket", fp,
-				Debugln("dropping cached packet")
+				"reason", dropReason,
+			)
 		}
 		return
 	}
@@ -296,9 +307,10 @@ func (f *Interface) SendMessageToVpnAddr(t header.MessageType, st header.Message
 	})
 
 	if hostInfo == nil {
-		if f.l.Level >= logrus.DebugLevel {
+		if f.l.Enabled(context.Background(), slog.LevelDebug) {
-			f.l.WithField("vpnAddr", vpnAddr).
+			f.l.Debug("dropping SendMessageToVpnAddr, vpnAddr not in our vpn networks or in unsafe routes",
-				Debugln("dropping SendMessageToVpnAddr, vpnAddr not in our vpn networks or in unsafe routes")
+				"vpnAddr", vpnAddr,
+			)
 		}
 		return
 	}
@@ -354,12 +366,12 @@ func (f *Interface) SendVia(via *HostInfo,
 		if noiseutil.EncryptLockNeeded {
 			via.ConnectionState.writeLock.Unlock()
 		}
-		via.logger(f.l).
+		via.logger(f.l).Error("SendVia out buffer not large enough for relay",
-			WithField("outCap", cap(out)).
+			"outCap", cap(out),
-			WithField("payloadLen", len(ad)).
+			"payloadLen", len(ad),
-			WithField("headerLen", len(out)).
+			"headerLen", len(out),
-			WithField("cipherOverhead", via.ConnectionState.eKey.Overhead()).
+			"cipherOverhead", via.ConnectionState.eKey.Overhead(),
-			Error("SendVia out buffer not large enough for relay")
+		)
 		return
 	}
 
@@ -379,12 +391,12 @@ func (f *Interface) SendVia(via *HostInfo,
 		via.ConnectionState.writeLock.Unlock()
 	}
 	if err != nil {
-		via.logger(f.l).WithError(err).Info("Failed to EncryptDanger in sendVia")
+		via.logger(f.l).Info("Failed to EncryptDanger in sendVia", "error", err)
 		return
 	}
 	err = f.writers[0].WriteTo(out, via.remote)
 	if err != nil {
-		via.logger(f.l).WithError(err).Info("Failed to WriteTo in sendVia")
+		via.logger(f.l).Info("Failed to WriteTo in sendVia", "error", err)
 	}
 	f.connectionManager.RelayUsed(relay.LocalIndex)
 }
@@ -423,8 +435,10 @@ func (f *Interface) sendNoMetrics(t header.MessageType, st header.MessageSubType
 		// finally used again. This tunnel would eventually be torn down and recreated if this action didn't help.
 		f.lightHouse.QueryServer(hostinfo.vpnAddrs[0])
 		hostinfo.lastRebindCount = f.rebindCount
-		if f.l.Level >= logrus.DebugLevel {
+		if f.l.Enabled(context.Background(), slog.LevelDebug) {
-			f.l.WithField("vpnAddrs", hostinfo.vpnAddrs).Debug("Lighthouse update triggered for punch due to rebind counter")
+			f.l.Debug("Lighthouse update triggered for punch due to rebind counter",
+				"vpnAddrs", hostinfo.vpnAddrs,
+			)
 		}
 	}
 
@@ -434,24 +448,30 @@ func (f *Interface) sendNoMetrics(t header.MessageType, st header.MessageSubType
 		ci.writeLock.Unlock()
 	}
 	if err != nil {
-		hostinfo.logger(f.l).WithError(err).
+		hostinfo.logger(f.l).Error("Failed to encrypt outgoing packet",
-			WithField("udpAddr", remote).WithField("counter", c).
+			"error", err,
-			WithField("attemptedCounter", c).
+			"udpAddr", remote,
-			Error("Failed to encrypt outgoing packet")
+			"counter", c,
+			"attemptedCounter", c,
+		)
 		return
 	}
 
 	if remote.IsValid() {
 		err = f.writers[q].WriteTo(out, remote)
 		if err != nil {
-			hostinfo.logger(f.l).WithError(err).
+			hostinfo.logger(f.l).Error("Failed to write outgoing packet",
-				WithField("udpAddr", remote).Error("Failed to write outgoing packet")
+				"error", err,
+				"udpAddr", remote,
+			)
 		}
 	} else if hostinfo.remote.IsValid() {
 		err = f.writers[q].WriteTo(out, hostinfo.remote)
 		if err != nil {
-			hostinfo.logger(f.l).WithError(err).
+			hostinfo.logger(f.l).Error("Failed to write outgoing packet",
-				WithField("udpAddr", remote).Error("Failed to write outgoing packet")
+				"error", err,
+				"udpAddr", remote,
+			)
 		}
 	} else {
 		// Try to send via a relay
@@ -459,7 +479,10 @@ func (f *Interface) sendNoMetrics(t header.MessageType, st header.MessageSubType
 			relayHostInfo, relay, err := f.hostMap.QueryVpnAddrsRelayFor(hostinfo.vpnAddrs, relayIP)
 			if err != nil {
 				hostinfo.relayState.DeleteRelay(relayIP)
-				hostinfo.logger(f.l).WithField("relay", relayIP).WithError(err).Info("sendNoMetrics failed to find HostInfo")
+				hostinfo.logger(f.l).Info("sendNoMetrics failed to find HostInfo",
+					"relay", relayIP,
+					"error", err,
+				)
 				continue
 			}
 			f.SendVia(relayHostInfo, relay, out, nb, fullOut[:header.Len+len(out)], true)

interface.go

@@ -4,6 +4,7 @@ import (
 	"context"
 	"errors"
 	"fmt"
+	"log/slog"
 	"net/netip"
 	"sync"
 	"sync/atomic"
@@ -11,12 +12,12 @@ import (
 
 	"github.com/gaissmai/bart"
 	"github.com/rcrowley/go-metrics"
-	"github.com/sirupsen/logrus"
+
 	"github.com/slackhq/nebula/config"
 	"github.com/slackhq/nebula/firewall"
 	"github.com/slackhq/nebula/header"
 	"github.com/slackhq/nebula/overlay"
-	"github.com/slackhq/nebula/overlay/coalesce"
+	"github.com/slackhq/nebula/overlay/batch"
 	"github.com/slackhq/nebula/overlay/tio"
 	"github.com/slackhq/nebula/udp"
 )
@@ -30,7 +31,7 @@ type InterfaceConfig struct {
 	pki                *PKI
 	Cipher             string
 	Firewall           *Firewall
-	ServeDns           bool
+	DnsServer          *dnsServer
 	HandshakeManager   *HandshakeManager
 	lightHouse         *LightHouse
 	connectionManager  *connectionManager
@@ -47,7 +48,7 @@ type InterfaceConfig struct {
 	reQueryWait     time.Duration
 
 	ConntrackCacheTimeout time.Duration
-	l                     *logrus.Logger
+	l                     *slog.Logger
 }
 
 type Interface struct {
@@ -58,7 +59,7 @@ type Interface struct {
 	firewall              *Firewall
 	connectionManager     *connectionManager
 	handshakeManager      *HandshakeManager
-	serveDns              bool
+	dnsServer             *dnsServer
 	createTime            time.Time
 	lightHouse            *LightHouse
 	myBroadcastAddrsTable *bart.Lite
@@ -86,12 +87,13 @@ type Interface struct {
 
 	conntrackCacheTimeout time.Duration
 
+	ctx     context.Context
 	writers []udp.Conn
 	readers []tio.Queue
-	// tunCoalescers is one tcpCoalescer per tun queue, wrapping readers[i].
+	// batchers is one per tun queue, wrapping readers[i].
-	// decryptToTun sends plaintext into the coalescer; listenOut calls its
+	// decryptToTun sends plaintext into the batch.RxBatcher;
-	// Flush at the end of each UDP recvmmsg batch.
+	// listenOut calls its Flush at the end of each UDP recvmmsg batch.
-	tunCoalescers []*coalesce.TCPCoalescer
+	batchers []batch.RxBatcher
 	wg       sync.WaitGroup
 
 	// fatalErr holds the first unexpected reader error that caused shutdown.
@@ -104,7 +106,7 @@ type Interface struct {
 	messageMetrics      *MessageMetrics
 	cachedPacketMetrics *cachedPacketMetrics
 
-	l *logrus.Logger
+	l *slog.Logger
 }
 
 type EncWriter interface {
@@ -175,12 +177,13 @@ func NewInterface(ctx context.Context, c *InterfaceConfig) (*Interface, error) {
 
 	cs := c.pki.getCertState()
 	ifce := &Interface{
+		ctx:                   ctx,
 		pki:                   c.pki,
 		hostMap:               c.HostMap,
 		outside:               c.Outside,
 		inside:                c.Inside,
 		firewall:              c.Firewall,
-		serveDns:              c.ServeDns,
+		dnsServer:             c.DnsServer,
 		handshakeManager:      c.HandshakeManager,
 		createTime:            time.Now(),
 		lightHouse:            c.lightHouse,
@@ -190,7 +193,7 @@ func NewInterface(ctx context.Context, c *InterfaceConfig) (*Interface, error) {
 		version:               c.version,
 		writers:               make([]udp.Conn, c.routines),
 		readers:               make([]tio.Queue, c.routines),
-		tunCoalescers:         make([]*coalesce.TCPCoalescer, c.routines),
+		batchers:              make([]batch.RxBatcher, c.routines),
 		myVpnNetworks:         cs.myVpnNetworks,
 		myVpnNetworksTable:    cs.myVpnNetworksTable,
 		myVpnAddrs:            cs.myVpnAddrs,
@@ -227,13 +230,16 @@ func (f *Interface) activate() error {
 
 	addr, err := f.outside.LocalAddr()
 	if err != nil {
-		f.l.WithError(err).Error("Failed to get udp listen address")
+		f.l.Error("Failed to get udp listen address", "error", err)
 	}
 
-	f.l.WithField("interface", f.inside.Name()).WithField("networks", f.myVpnNetworks).
+	f.l.Info("Nebula interface is active",
-		WithField("build", f.version).WithField("udpAddr", addr).
+		"interface", f.inside.Name(),
-		WithField("boringcrypto", boringEnabled()).
+		"networks", f.myVpnNetworks,
-		Info("Nebula interface is active")
+		"build", f.version,
+		"udpAddr", addr,
+		"boringcrypto", boringEnabled(),
+	)
 
 	if f.routines > 1 {
 		if !f.inside.SupportsMultiqueue() || !f.outside.SupportsMultipleReaders() {
@@ -247,15 +253,14 @@ func (f *Interface) activate() error {
 	// Prepare n tun queues
 	for i := 0; i < f.routines; i++ {
 		if i > 0 {
-			err = f.inside.NewMultiQueueReader()
+			if err = f.inside.NewMultiQueueReader(); err != nil {
-			if err != nil {
 				return err
 			}
 		}
 	}
 	f.readers = f.inside.Readers()
 	for i := range f.readers {
-		f.tunCoalescers[i] = coalesce.NewTCPCoalescer(f.readers[i]) //todo don't always do this
+		f.batchers[i] = batch.NewTCPCoalescer(f.readers[i])
 	}
 
 	f.wg.Add(1) // for us to wait on Close() to return
@@ -311,77 +316,73 @@ func (f *Interface) listenOut(i int) {
 		li = f.outside
 	}
 
-	ctCache := firewall.NewConntrackCacheTicker(f.conntrackCacheTimeout)
+	ctCache := firewall.NewConntrackCacheTicker(f.ctx, f.l, f.conntrackCacheTimeout)
 	lhh := f.lightHouse.NewRequestHandler()
 	h := &header.H{}
 	fwPacket := &firewall.Packet{}
 	nb := make([]byte, 12, 12)
 
-	// plaintexts is a ring of decrypt scratches, one per packet in a UDP
+	coalescer := f.batchers[i]
-	// recvmmsg batch. The coalescer borrows payload slices from here and
+
-	// requires they stay valid until Flush — so we rotate each packet and
+	listener := func(fromUdpAddr netip.AddrPort, payload []byte) {
-	// reset only in the batch-end flush callback.
+		plaintext := f.batchers[i].Reserve(len(payload))
-	var plaintexts [][]byte
+		f.readOutsidePackets(ViaSender{UdpAddr: fromUdpAddr}, plaintext[:0], payload, h, fwPacket, lhh, nb, i, ctCache.Get())
-	idx := 0
-	coalescer := f.tunCoalescers[i]
-	err := li.ListenOut(func(fromUdpAddr netip.AddrPort, payload []byte) {
-		if idx >= len(plaintexts) {
-			plaintexts = append(plaintexts, make([]byte, udp.MTU))
 	}
-		f.readOutsidePackets(ViaSender{UdpAddr: fromUdpAddr}, plaintexts[idx][:0], payload, h, fwPacket, lhh, nb, i, ctCache.Get(f.l))
+
-		idx++
+	flusher := func() {
-	}, func() {
 		if err := coalescer.Flush(); err != nil {
-			f.l.WithError(err).Error("Failed to flush tun coalescer")
+			f.l.Error("Failed to flush tun coalescer", "error", err)
 		}
-		idx = 0
+	}
-	})
+
+	err := li.ListenOut(listener, flusher)
 
 	if err != nil && !f.closed.Load() {
-		f.l.WithError(err).Error("Error while reading inbound packet, closing")
+		f.l.Error("Error while reading inbound packet, closing", "error", err)
 		f.onFatal(err)
 	}
 
-	f.l.Debugf("underlay reader %v is done", i)
+	f.l.Debug("underlay reader is done", "reader", i)
 }
 
 func (f *Interface) listenIn(reader tio.Queue, i int) {
 	rejectBuf := make([]byte, mtu)
-	batch := newSendBatch(sendBatchCap, udp.MTU+32)
+	sb := batch.NewSendBatch(batch.SendBatchCap, udp.MTU+32)
 	fwPacket := &firewall.Packet{}
 	nb := make([]byte, 12, 12)
 
-	conntrackCache := firewall.NewConntrackCacheTicker(f.conntrackCacheTimeout)
+	conntrackCache := firewall.NewConntrackCacheTicker(f.ctx, f.l, f.conntrackCacheTimeout)
 
 	for {
 		pkts, err := reader.Read()
 		if err != nil {
 			if !f.closed.Load() {
-				f.l.WithError(err).WithField("reader", i).Error("Error while reading outbound packet, closing")
+				f.l.Error("Error while reading outbound packet, closing", "error", err, "reader", i)
 				f.onFatal(err)
 			}
 			break
 		}
 
-		batch.Reset()
+		sb.Reset()
 		for _, pkt := range pkts {
-			if batch.Len() >= batch.Cap() {
+			if sb.Len() >= sb.Cap() {
-				f.flushBatch(batch, i)
+				f.flushBatch(sb, i)
-				batch.Reset()
+				sb.Reset()
 			}
-			f.consumeInsidePacket(pkt, fwPacket, nb, batch, rejectBuf, i, conntrackCache.Get(f.l))
+			f.consumeInsidePacket(pkt, fwPacket, nb, sb, rejectBuf, i, conntrackCache.Get())
 		}
-		if batch.Len() > 0 {
+		if sb.Len() > 0 {
-			f.flushBatch(batch, i)
+			f.flushBatch(sb, i)
 		}
 	}
 
-	f.l.Debugf("overlay reader %v is done", i)
+	f.l.Debug("overlay reader is done", "reader", i)
 }
 
-func (f *Interface) flushBatch(batch *sendBatch, q int) {
+func (f *Interface) flushBatch(sb batch.TxBatcher, q int) {
-	if err := f.writers[q].WriteBatch(batch.bufs, batch.dsts); err != nil {
+	bufs, dsts := sb.Get()
-		f.l.WithError(err).WithField("writer", q).Error("Failed to write outgoing batch")
+	if err := f.writers[q].WriteBatch(bufs, dsts); err != nil {
+		f.l.Error("Failed to write outgoing batch", "error", err, "writer", q)
 	}
 }
 
@@ -402,7 +403,7 @@ func (f *Interface) reloadDisconnectInvalid(c *config.C) {
 	if initial || c.HasChanged("pki.disconnect_invalid") {
 		f.disconnectInvalid.Store(c.GetBool("pki.disconnect_invalid", true))
 		if !initial {
-			f.l.Infof("pki.disconnect_invalid changed to %v", f.disconnectInvalid.Load())
+			f.l.Info("pki.disconnect_invalid changed", "value", f.disconnectInvalid.Load())
 		}
 	}
 }
@@ -416,7 +417,7 @@ func (f *Interface) reloadFirewall(c *config.C) {
 
 	fw, err := NewFirewallFromConfig(f.l, f.pki.getCertState(), c)
 	if err != nil {
-		f.l.WithError(err).Error("Error while creating firewall during reload")
+		f.l.Error("Error while creating firewall during reload", "error", err)
 		return
 	}
 
@@ -429,10 +430,11 @@ func (f *Interface) reloadFirewall(c *config.C) {
 	// If rulesVersion is back to zero, we have wrapped all the way around. Be
 	// safe and just reset conntrack in this case.
 	if fw.rulesVersion == 0 {
-		f.l.WithField("firewallHashes", fw.GetRuleHashes()).
+		f.l.Warn("firewall rulesVersion has overflowed, resetting conntrack",
-			WithField("oldFirewallHashes", oldFw.GetRuleHashes()).
+			"firewallHashes", fw.GetRuleHashes(),
-			WithField("rulesVersion", fw.rulesVersion).
+			"oldFirewallHashes", oldFw.GetRuleHashes(),
-			Warn("firewall rulesVersion has overflowed, resetting conntrack")
+			"rulesVersion", fw.rulesVersion,
+		)
 	} else {
 		fw.Conntrack = conntrack
 	}
@@ -440,10 +442,11 @@ func (f *Interface) reloadFirewall(c *config.C) {
 	f.firewall = fw
 
 	oldFw.Destroy()
-	f.l.WithField("firewallHashes", fw.GetRuleHashes()).
+	f.l.Info("New firewall has been installed",
-		WithField("oldFirewallHashes", oldFw.GetRuleHashes()).
+		"firewallHashes", fw.GetRuleHashes(),
-		WithField("rulesVersion", fw.rulesVersion).
+		"oldFirewallHashes", oldFw.GetRuleHashes(),
-		Info("New firewall has been installed")
+		"rulesVersion", fw.rulesVersion,
+	)
 }
 
 func (f *Interface) reloadSendRecvError(c *config.C) {
@@ -465,8 +468,7 @@ func (f *Interface) reloadSendRecvError(c *config.C) {
 			}
 		}
 
-		f.l.WithField("sendRecvError", f.sendRecvErrorConfig.String()).
+		f.l.Info("Loaded send_recv_error config", "sendRecvError", f.sendRecvErrorConfig.String())
-			Info("Loaded send_recv_error config")
 	}
 }
 
@@ -489,8 +491,7 @@ func (f *Interface) reloadAcceptRecvError(c *config.C) {
 			}
 		}
 
-		f.l.WithField("acceptRecvError", f.acceptRecvErrorConfig.String()).
+		f.l.Info("Loaded accept_recv_error config", "acceptRecvError", f.acceptRecvErrorConfig.String())
-			Info("Loaded accept_recv_error config")
 	}
 }
 
@@ -564,7 +565,7 @@ func (f *Interface) Close() error {
 	for i, u := range f.writers {
 		err := u.Close()
 		if err != nil {
-			f.l.WithError(err).WithField("writer", i).Error("Error while closing udp socket")
+			f.l.Error("Error while closing udp socket", "error", err, "writer", i)
 			errs = append(errs, err)
 		}
 	}

lighthouse.go

@@ -5,6 +5,7 @@ import (
 	"encoding/binary"
 	"errors"
 	"fmt"
+	"log/slog"
 	"net"
 	"net/netip"
 	"slices"
@@ -15,10 +16,10 @@ import (
 
 	"github.com/gaissmai/bart"
 	"github.com/rcrowley/go-metrics"
-	"github.com/sirupsen/logrus"
 	"github.com/slackhq/nebula/cert"
 	"github.com/slackhq/nebula/config"
 	"github.com/slackhq/nebula/header"
+	"github.com/slackhq/nebula/logging"
 	"github.com/slackhq/nebula/udp"
 	"github.com/slackhq/nebula/util"
 )
@@ -69,18 +70,19 @@ type LightHouse struct {
 	// Addr's of relays that can be used by peers to access me
 	relaysForMe atomic.Pointer[[]netip.Addr]
 
+	updateTrigger chan struct{}
 	queryChan     chan netip.Addr
 
 	calculatedRemotes atomic.Pointer[bart.Table[[]*calculatedRemote]] // Maps VpnAddr to []*calculatedRemote
 
 	metrics           *MessageMetrics
 	metricHolepunchTx metrics.Counter
-	l                 *logrus.Logger
+	l                 *slog.Logger
 }
 
 // NewLightHouseFromConfig will build a Lighthouse struct from the values provided in the config object
 // addrMap should be nil unless this is during a config reload
-func NewLightHouseFromConfig(ctx context.Context, l *logrus.Logger, c *config.C, cs *CertState, pc udp.Conn, p *Punchy) (*LightHouse, error) {
+func NewLightHouseFromConfig(ctx context.Context, l *slog.Logger, c *config.C, cs *CertState, pc udp.Conn, p *Punchy) (*LightHouse, error) {
 	amLighthouse := c.GetBool("lighthouse.am_lighthouse", false)
 	nebulaPort := uint32(c.GetInt("listen.port", 0))
 	if amLighthouse && nebulaPort == 0 {
@@ -105,6 +107,7 @@ func NewLightHouseFromConfig(ctx context.Context, l *logrus.Logger, c *config.C,
 		nebulaPort:         nebulaPort,
 		punchConn:          pc,
 		punchy:             p,
+		updateTrigger:      make(chan struct{}, 1),
 		queryChan:          make(chan netip.Addr, c.GetUint32("handshakes.query_buffer", 64)),
 		l:                  l,
 	}
@@ -131,7 +134,7 @@ func NewLightHouseFromConfig(ctx context.Context, l *logrus.Logger, c *config.C,
 		case *util.ContextualError:
 			v.Log(l)
 		case error:
-			l.WithError(err).Error("failed to reload lighthouse")
+			l.Error("failed to reload lighthouse", "error", err)
 		}
 	})
 
@@ -203,8 +206,10 @@ func (lh *LightHouse) reload(c *config.C, initial bool) error {
 			//TODO: we could technically insert all returned addrs instead of just the first one if a dns lookup was used
 			addr := addrs[0].Unmap()
 			if lh.myVpnNetworksTable.Contains(addr) {
-				lh.l.WithField("addr", rawAddr).WithField("entry", i+1).
+				lh.l.Warn("Ignoring lighthouse.advertise_addrs report because it is within the nebula network range",
-					Warn("Ignoring lighthouse.advertise_addrs report because it is within the nebula network range")
+					"addr", rawAddr,
+					"entry", i+1,
+				)
 				continue
 			}
 
@@ -222,7 +227,9 @@ func (lh *LightHouse) reload(c *config.C, initial bool) error {
 		lh.interval.Store(int64(c.GetInt("lighthouse.interval", 10)))
 
 		if !initial {
-			lh.l.Infof("lighthouse.interval changed to %v", lh.interval.Load())
+			lh.l.Info("lighthouse.interval changed",
+				"interval", lh.interval.Load(),
+			)
 
 			if lh.updateCancel != nil {
 				// May not always have a running routine
@@ -316,6 +323,7 @@ func (lh *LightHouse) reload(c *config.C, initial bool) error {
 		if !initial {
 			//NOTE: we are not tearing down existing lighthouse connections because they might be used for non lighthouse traffic
 			lh.l.Info("lighthouse.hosts has changed")
+			lh.TriggerUpdate()
 		}
 	}
 
@@ -333,9 +341,12 @@ func (lh *LightHouse) reload(c *config.C, initial bool) error {
 			for _, v := range c.GetStringSlice("relay.relays", nil) {
 				configRIP, err := netip.ParseAddr(v)
 				if err != nil {
-					lh.l.WithField("relay", v).WithError(err).Warn("Parse relay from config failed")
+					lh.l.Warn("Parse relay from config failed",
+						"relay", v,
+						"error", err,
+					)
 				} else {
-					lh.l.WithField("relay", v).Info("Read relay from config")
+					lh.l.Info("Read relay from config", "relay", v)
 					relaysForMe = append(relaysForMe, configRIP)
 				}
 			}
@@ -360,8 +371,10 @@ func (lh *LightHouse) parseLighthouses(c *config.C) ([]netip.Addr, error) {
 		}
 
 		if !lh.myVpnNetworksTable.Contains(addr) {
-			lh.l.WithFields(m{"vpnAddr": addr, "networks": lh.myVpnNetworks}).
+			lh.l.Warn("lighthouse host is not within our networks, lighthouse functionality will work but layer 3 network traffic to the lighthouse will not",
-				Warn("lighthouse host is not within our networks, lighthouse functionality will work but layer 3 network traffic to the lighthouse will not")
+				"vpnAddr", addr,
+				"networks", lh.myVpnNetworks,
+			)
 		}
 		out[i] = addr
 	}
@@ -432,8 +445,11 @@ func (lh *LightHouse) loadStaticMap(c *config.C, staticList map[netip.Addr]struc
 		}
 
 		if !lh.myVpnNetworksTable.Contains(vpnAddr) {
-			lh.l.WithFields(m{"vpnAddr": vpnAddr, "networks": lh.myVpnNetworks, "entry": i + 1}).
+			lh.l.Warn("static_host_map key is not within our networks, layer 3 network traffic to this host will not work",
-				Warn("static_host_map key is not within our networks, layer 3 network traffic to this host will not work")
+				"vpnAddr", vpnAddr,
+				"networks", lh.myVpnNetworks,
+				"entry", i+1,
+			)
 		}
 
 		vals, ok := v.([]any)
@@ -534,12 +550,13 @@ func (lh *LightHouse) DeleteVpnAddrs(allVpnAddrs []netip.Addr) {
 	lh.Lock()
 	rm, ok := lh.addrMap[allVpnAddrs[0]]
 	if ok {
+		debugEnabled := lh.l.Enabled(context.Background(), slog.LevelDebug)
 		for _, addr := range allVpnAddrs {
 			srm := lh.addrMap[addr]
 			if srm == rm {
 				delete(lh.addrMap, addr)
-				if lh.l.Level >= logrus.DebugLevel {
+				if debugEnabled {
-					lh.l.Debugf("deleting %s from lighthouse.", addr)
+					lh.l.Debug("deleting from lighthouse", "vpnAddr", addr)
 				}
 			}
 		}
@@ -656,9 +673,12 @@ func (lh *LightHouse) unlockedGetRemoteList(allAddrs []netip.Addr) *RemoteList {
 
 func (lh *LightHouse) shouldAdd(vpnAddrs []netip.Addr, to netip.Addr) bool {
 	allow := lh.GetRemoteAllowList().AllowAll(vpnAddrs, to)
-	if lh.l.Level >= logrus.TraceLevel {
+	if lh.l.Enabled(context.Background(), logging.LevelTrace) {
-		lh.l.WithField("vpnAddrs", vpnAddrs).WithField("udpAddr", to).WithField("allow", allow).
+		lh.l.Log(context.Background(), logging.LevelTrace, "remoteAllowList.Allow",
-			Trace("remoteAllowList.Allow")
+			"vpnAddrs", vpnAddrs,
+			"udpAddr", to,
+			"allow", allow,
+		)
 	}
 	if !allow {
 		return false
@@ -675,9 +695,12 @@ func (lh *LightHouse) shouldAdd(vpnAddrs []netip.Addr, to netip.Addr) bool {
 func (lh *LightHouse) unlockedShouldAddV4(vpnAddr netip.Addr, to *V4AddrPort) bool {
 	udpAddr := protoV4AddrPortToNetAddrPort(to)
 	allow := lh.GetRemoteAllowList().Allow(vpnAddr, udpAddr.Addr())
-	if lh.l.Level >= logrus.TraceLevel {
+	if lh.l.Enabled(context.Background(), logging.LevelTrace) {
-		lh.l.WithField("vpnAddr", vpnAddr).WithField("udpAddr", udpAddr).WithField("allow", allow).
+		lh.l.Log(context.Background(), logging.LevelTrace, "remoteAllowList.Allow",
-			Trace("remoteAllowList.Allow")
+			"vpnAddr", vpnAddr,
+			"udpAddr", udpAddr,
+			"allow", allow,
+		)
 	}
 
 	if !allow {
@@ -695,9 +718,12 @@ func (lh *LightHouse) unlockedShouldAddV4(vpnAddr netip.Addr, to *V4AddrPort) bo
 func (lh *LightHouse) unlockedShouldAddV6(vpnAddr netip.Addr, to *V6AddrPort) bool {
 	udpAddr := protoV6AddrPortToNetAddrPort(to)
 	allow := lh.GetRemoteAllowList().Allow(vpnAddr, udpAddr.Addr())
-	if lh.l.Level >= logrus.TraceLevel {
+	if lh.l.Enabled(context.Background(), logging.LevelTrace) {
-		lh.l.WithField("vpnAddr", vpnAddr).WithField("udpAddr", udpAddr).WithField("allow", allow).
+		lh.l.Log(context.Background(), logging.LevelTrace, "remoteAllowList.Allow",
-			Trace("remoteAllowList.Allow")
+			"vpnAddr", vpnAddr,
+			"udpAddr", udpAddr,
+			"allow", allow,
+		)
 	}
 
 	if !allow {
@@ -772,8 +798,10 @@ func (lh *LightHouse) innerQueryServer(addr netip.Addr, nb, out []byte) {
 
 		if v == cert.Version1 {
 			if !addr.Is4() {
-				lh.l.WithField("queryVpnAddr", addr).WithField("lighthouseAddr", lhVpnAddr).
+				lh.l.Error("Can't query lighthouse for v6 address using a v1 protocol",
-					Error("Can't query lighthouse for v6 address using a v1 protocol")
+					"queryVpnAddr", addr,
+					"lighthouseAddr", lhVpnAddr,
+				)
 				continue
 			}
 
@@ -784,9 +812,11 @@ func (lh *LightHouse) innerQueryServer(addr netip.Addr, nb, out []byte) {
 
 				v1Query, err = msg.Marshal()
 				if err != nil {
-					lh.l.WithError(err).WithField("queryVpnAddr", addr).
+					lh.l.Error("Failed to marshal lighthouse v1 query payload",
-						WithField("lighthouseAddr", lhVpnAddr).
+						"error", err,
-						Error("Failed to marshal lighthouse v1 query payload")
+						"queryVpnAddr", addr,
+						"lighthouseAddr", lhVpnAddr,
+					)
 					continue
 				}
 			}
@@ -801,9 +831,11 @@ func (lh *LightHouse) innerQueryServer(addr netip.Addr, nb, out []byte) {
 
 				v2Query, err = msg.Marshal()
 				if err != nil {
-					lh.l.WithError(err).WithField("queryVpnAddr", addr).
+					lh.l.Error("Failed to marshal lighthouse v2 query payload",
-						WithField("lighthouseAddr", lhVpnAddr).
+						"error", err,
-						Error("Failed to marshal lighthouse v2 query payload")
+						"queryVpnAddr", addr,
+						"lighthouseAddr", lhVpnAddr,
+					)
 					continue
 				}
 			}
@@ -812,7 +844,11 @@ func (lh *LightHouse) innerQueryServer(addr netip.Addr, nb, out []byte) {
 			queried++
 
 		} else {
-			lh.l.Debugf("Can not query lighthouse for %v using unknown protocol version: %v", addr, v)
+			lh.l.Debug("unsupported protocol version",
+				"op", "query",
+				"queryVpnAddr", addr,
+				"version", v,
+			)
 			continue
 		}
 	}
@@ -841,11 +877,24 @@ func (lh *LightHouse) StartUpdateWorker() {
 				return
 			case <-clockSource.C:
 				continue
+			case <-lh.updateTrigger:
+				continue
 			}
 		}
 	}()
 }
 
+// TriggerUpdate requests an immediate lighthouse update. This is a non-blocking
+// operation intended to be called after a handshake completes with a lighthouse,
+// so the lighthouse has our current addresses without waiting for the next
+// periodic update.
+func (lh *LightHouse) TriggerUpdate() {
+	select {
+	case lh.updateTrigger <- struct{}{}:
+	default:
+	}
+}
+
 func (lh *LightHouse) SendUpdate() {
 	var v4 []*V4AddrPort
 	var v6 []*V6AddrPort
@@ -891,8 +940,9 @@ func (lh *LightHouse) SendUpdate() {
 		if v == cert.Version1 {
 			if v1Update == nil {
 				if !lh.myVpnNetworks[0].Addr().Is4() {
-					lh.l.WithField("lighthouseAddr", lhVpnAddr).
+					lh.l.Warn("cannot update lighthouse using v1 protocol without an IPv4 address",
-						Warn("cannot update lighthouse using v1 protocol without an IPv4 address")
+						"lighthouseAddr", lhVpnAddr,
+					)
 					continue
 				}
 				var relays []uint32
@@ -916,8 +966,10 @@ func (lh *LightHouse) SendUpdate() {
 
 				v1Update, err = msg.Marshal()
 				if err != nil {
-					lh.l.WithError(err).WithField("lighthouseAddr", lhVpnAddr).
+					lh.l.Error("Error while marshaling for lighthouse v1 update",
-						Error("Error while marshaling for lighthouse v1 update")
+						"error", err,
+						"lighthouseAddr", lhVpnAddr,
+					)
 					continue
 				}
 			}
@@ -943,8 +995,10 @@ func (lh *LightHouse) SendUpdate() {
 
 				v2Update, err = msg.Marshal()
 				if err != nil {
-					lh.l.WithError(err).WithField("lighthouseAddr", lhVpnAddr).
+					lh.l.Error("Error while marshaling for lighthouse v2 update",
-						Error("Error while marshaling for lighthouse v2 update")
+						"error", err,
+						"lighthouseAddr", lhVpnAddr,
+					)
 					continue
 				}
 			}
@@ -953,7 +1007,10 @@ func (lh *LightHouse) SendUpdate() {
 			updated++
 
 		} else {
-			lh.l.Debugf("Can not update lighthouse using unknown protocol version: %v", v)
+			lh.l.Debug("unsupported protocol version",
+				"op", "update",
+				"version", v,
+			)
 			continue
 		}
 	}
@@ -967,7 +1024,7 @@ type LightHouseHandler struct {
 	out  []byte
 	pb   []byte
 	meta *NebulaMeta
-	l    *logrus.Logger
+	l    *slog.Logger
 }
 
 func (lh *LightHouse) NewRequestHandler() *LightHouseHandler {
@@ -1016,14 +1073,19 @@ func (lhh *LightHouseHandler) HandleRequest(rAddr netip.AddrPort, fromVpnAddrs [
 	n := lhh.resetMeta()
 	err := n.Unmarshal(p)
 	if err != nil {
-		lhh.l.WithError(err).WithField("vpnAddrs", fromVpnAddrs).WithField("udpAddr", rAddr).
+		lhh.l.Error("Failed to unmarshal lighthouse packet",
-			Error("Failed to unmarshal lighthouse packet")
+			"error", err,
+			"vpnAddrs", fromVpnAddrs,
+			"udpAddr", rAddr,
+		)
 		return
 	}
 
 	if n.Details == nil {
-		lhh.l.WithField("vpnAddrs", fromVpnAddrs).WithField("udpAddr", rAddr).
+		lhh.l.Error("Invalid lighthouse update",
-			Error("Invalid lighthouse update")
+			"vpnAddrs", fromVpnAddrs,
+			"udpAddr", rAddr,
+		)
 		return
 	}
 
@@ -1051,25 +1113,29 @@ func (lhh *LightHouseHandler) HandleRequest(rAddr netip.AddrPort, fromVpnAddrs [
 func (lhh *LightHouseHandler) handleHostQuery(n *NebulaMeta, fromVpnAddrs []netip.Addr, addr netip.AddrPort, w EncWriter) {
 	// Exit if we don't answer queries
 	if !lhh.lh.amLighthouse {
-		if lhh.l.Level >= logrus.DebugLevel {
+		if lhh.l.Enabled(context.Background(), slog.LevelDebug) {
-			lhh.l.Debugln("I don't answer queries, but received from: ", addr)
+			lhh.l.Debug("I don't answer queries, but received one", "from", addr)
 		}
 		return
 	}
 
 	queryVpnAddr, useVersion, err := n.Details.GetVpnAddrAndVersion()
 	if err != nil {
-		if lhh.l.Level >= logrus.DebugLevel {
+		if lhh.l.Enabled(context.Background(), slog.LevelDebug) {
-			lhh.l.WithField("from", fromVpnAddrs).WithField("details", n.Details).
+			lhh.l.Debug("Dropping malformed HostQuery",
-				Debugln("Dropping malformed HostQuery")
+				"from", fromVpnAddrs,
+				"details", n.Details,
+			)
 		}
 		return
 	}
 	if useVersion == cert.Version1 && queryVpnAddr.Is6() {
 		// this case really shouldn't be possible to represent, but reject it anyway.
-		if lhh.l.Level >= logrus.DebugLevel {
+		if lhh.l.Enabled(context.Background(), slog.LevelDebug) {
-			lhh.l.WithField("vpnAddrs", fromVpnAddrs).WithField("queryVpnAddr", queryVpnAddr).
+			lhh.l.Debug("invalid vpn addr for v1 handleHostQuery",
-				Debugln("invalid vpn addr for v1 handleHostQuery")
+				"vpnAddrs", fromVpnAddrs,
+				"queryVpnAddr", queryVpnAddr,
+			)
 		}
 		return
 	}
@@ -1094,7 +1160,10 @@ func (lhh *LightHouseHandler) handleHostQuery(n *NebulaMeta, fromVpnAddrs []neti
 	}
 
 	if err != nil {
-		lhh.l.WithError(err).WithField("vpnAddrs", fromVpnAddrs).Error("Failed to marshal lighthouse host query reply")
+		lhh.l.Error("Failed to marshal lighthouse host query reply",
+			"error", err,
+			"vpnAddrs", fromVpnAddrs,
+		)
 		return
 	}
 
@@ -1122,8 +1191,10 @@ func (lhh *LightHouseHandler) sendHostPunchNotification(n *NebulaMeta, fromVpnAd
 			if ok {
 				whereToPunch = newDest
 			} else {
-				if lhh.l.Level >= logrus.DebugLevel {
+				if lhh.l.Enabled(context.Background(), slog.LevelDebug) {
-					lhh.l.WithField("to", crt.Networks()).Debugln("unable to punch to host, no addresses in common")
+					lhh.l.Debug("unable to punch to host, no addresses in common",
+						"to", crt.Networks(),
+					)
 				}
 			}
 		}
@@ -1149,7 +1220,10 @@ func (lhh *LightHouseHandler) sendHostPunchNotification(n *NebulaMeta, fromVpnAd
 	}
 
 	if err != nil {
-		lhh.l.WithError(err).WithField("vpnAddrs", fromVpnAddrs).Error("Failed to marshal lighthouse host was queried for")
+		lhh.l.Error("Failed to marshal lighthouse host was queried for",
+			"error", err,
+			"vpnAddrs", fromVpnAddrs,
+		)
 		return
 	}
 
@@ -1191,8 +1265,11 @@ func (lhh *LightHouseHandler) coalesceAnswers(v cert.Version, c *cache, n *Nebul
 				n.Details.RelayVpnAddrs = append(n.Details.RelayVpnAddrs, netAddrToProtoAddr(r))
 			}
 		} else {
-			if lhh.l.Level >= logrus.DebugLevel {
+			if lhh.l.Enabled(context.Background(), slog.LevelDebug) {
-				lhh.l.WithField("version", v).Debug("unsupported protocol version")
+				lhh.l.Debug("unsupported protocol version",
+					"op", "coalesceAnswers",
+					"version", v,
+				)
 			}
 		}
 	}
@@ -1205,8 +1282,11 @@ func (lhh *LightHouseHandler) handleHostQueryReply(n *NebulaMeta, fromVpnAddrs [
 
 	certVpnAddr, _, err := n.Details.GetVpnAddrAndVersion()
 	if err != nil {
-		if lhh.l.Level >= logrus.DebugLevel {
+		if lhh.l.Enabled(context.Background(), slog.LevelDebug) {
-			lhh.l.WithError(err).WithField("vpnAddrs", fromVpnAddrs).Error("dropping malformed HostQueryReply")
+			lhh.l.Error("dropping malformed HostQueryReply",
+				"error", err,
+				"vpnAddrs", fromVpnAddrs,
+			)
 		}
 		return
 	}
@@ -1231,8 +1311,8 @@ func (lhh *LightHouseHandler) handleHostQueryReply(n *NebulaMeta, fromVpnAddrs [
 
 func (lhh *LightHouseHandler) handleHostUpdateNotification(n *NebulaMeta, fromVpnAddrs []netip.Addr, w EncWriter) {
 	if !lhh.lh.amLighthouse {
-		if lhh.l.Level >= logrus.DebugLevel {
+		if lhh.l.Enabled(context.Background(), slog.LevelDebug) {
-			lhh.l.Debugln("I am not a lighthouse, do not take host updates: ", fromVpnAddrs)
+			lhh.l.Debug("I am not a lighthouse, do not take host updates", "from", fromVpnAddrs)
 		}
 		return
 	}
@@ -1255,8 +1335,11 @@ func (lhh *LightHouseHandler) handleHostUpdateNotification(n *NebulaMeta, fromVp
 
 	//Simple check that the host sent this not someone else, if detailsVpnAddr is filled
 	if detailsVpnAddr.IsValid() && !slices.Contains(fromVpnAddrs, detailsVpnAddr) {
-		if lhh.l.Level >= logrus.DebugLevel {
+		if lhh.l.Enabled(context.Background(), slog.LevelDebug) {
-			lhh.l.WithField("vpnAddrs", fromVpnAddrs).WithField("answer", detailsVpnAddr).Debugln("Host sent invalid update")
+			lhh.l.Debug("Host sent invalid update",
+				"vpnAddrs", fromVpnAddrs,
+				"answer", detailsVpnAddr,
+			)
 		}
 		return
 	}
@@ -1278,7 +1361,9 @@ func (lhh *LightHouseHandler) handleHostUpdateNotification(n *NebulaMeta, fromVp
 	switch useVersion {
 	case cert.Version1:
 		if !fromVpnAddrs[0].Is4() {
-			lhh.l.WithField("vpnAddrs", fromVpnAddrs).Error("Can not send HostUpdateNotificationAck for a ipv6 vpn ip in a v1 message")
+			lhh.l.Error("Can not send HostUpdateNotificationAck for a ipv6 vpn ip in a v1 message",
+				"vpnAddrs", fromVpnAddrs,
+			)
 			return
 		}
 		vpnAddrB := fromVpnAddrs[0].As4()
@@ -1286,13 +1371,16 @@ func (lhh *LightHouseHandler) handleHostUpdateNotification(n *NebulaMeta, fromVp
 	case cert.Version2:
 		// do nothing, we want to send a blank message
 	default:
-		lhh.l.WithField("useVersion", useVersion).Error("invalid protocol version")
+		lhh.l.Error("invalid protocol version", "useVersion", useVersion)
 		return
 	}
 
 	ln, err := n.MarshalTo(lhh.pb)
 	if err != nil {
-		lhh.l.WithError(err).WithField("vpnAddrs", fromVpnAddrs).Error("Failed to marshal lighthouse host update ack")
+		lhh.l.Error("Failed to marshal lighthouse host update ack",
+			"error", err,
+			"vpnAddrs", fromVpnAddrs,
+		)
 		return
 	}
 
@@ -1309,8 +1397,11 @@ func (lhh *LightHouseHandler) handleHostPunchNotification(n *NebulaMeta, fromVpn
 
 	detailsVpnAddr, _, err := n.Details.GetVpnAddrAndVersion()
 	if err != nil {
-		if lhh.l.Level >= logrus.DebugLevel {
+		if lhh.l.Enabled(context.Background(), slog.LevelDebug) {
-			lhh.l.WithField("details", n.Details).WithError(err).Debugln("dropping invalid HostPunchNotification")
+			lhh.l.Debug("dropping invalid HostPunchNotification",
+				"details", n.Details,
+				"error", err,
+			)
 		}
 		return
 	}
@@ -1327,8 +1418,11 @@ func (lhh *LightHouseHandler) handleHostPunchNotification(n *NebulaMeta, fromVpn
 			lhh.lh.punchConn.WriteTo(empty, vpnPeer)
 		}()
 
-		if lhh.l.Level >= logrus.DebugLevel {
+		if lhh.l.Enabled(context.Background(), slog.LevelDebug) {
-			lhh.l.Debugf("Punching on %v for %v", vpnPeer, logVpnAddr)
+			lhh.l.Debug("Punching",
+				"vpnPeer", vpnPeer,
+				"logVpnAddr", logVpnAddr,
+			)
 		}
 	}
 
@@ -1353,8 +1447,10 @@ func (lhh *LightHouseHandler) handleHostPunchNotification(n *NebulaMeta, fromVpn
 	if lhh.lh.punchy.GetRespond() {
 		go func() {
 			time.Sleep(lhh.lh.punchy.GetRespondDelay())
-			if lhh.l.Level >= logrus.DebugLevel {
+			if lhh.l.Enabled(context.Background(), slog.LevelDebug) {
-				lhh.l.Debugf("Sending a nebula test packet to vpn addr %s", detailsVpnAddr)
+				lhh.l.Debug("Sending a nebula test packet",
+					"vpnAddr", detailsVpnAddr,
+				)
 			}
 			//NOTE: we have to allocate a new output buffer here since we are spawning a new goroutine
 			// for each punchBack packet. We should move this into a timerwheel or a single goroutine

logger.go

@@ -1,45 +0,0 @@
-package nebula
-
-import (
-	"fmt"
-	"strings"
-	"time"
-
-	"github.com/sirupsen/logrus"
-	"github.com/slackhq/nebula/config"
-)
-
-func configLogger(l *logrus.Logger, c *config.C) error {
-	// set up our logging level
-	logLevel, err := logrus.ParseLevel(strings.ToLower(c.GetString("logging.level", "info")))
-	if err != nil {
-		return fmt.Errorf("%s; possible levels: %s", err, logrus.AllLevels)
-	}
-	l.SetLevel(logLevel)
-
-	disableTimestamp := c.GetBool("logging.disable_timestamp", false)
-	timestampFormat := c.GetString("logging.timestamp_format", "")
-	fullTimestamp := (timestampFormat != "")
-	if timestampFormat == "" {
-		timestampFormat = time.RFC3339
-	}
-
-	logFormat := strings.ToLower(c.GetString("logging.format", "text"))
-	switch logFormat {
-	case "text":
-		l.Formatter = &logrus.TextFormatter{
-			TimestampFormat:  timestampFormat,
-			FullTimestamp:    fullTimestamp,
-			DisableTimestamp: disableTimestamp,
-		}
-	case "json":
-		l.Formatter = &logrus.JSONFormatter{
-			TimestampFormat:  timestampFormat,
-			DisableTimestamp: disableTimestamp,
-		}
-	default:
-		return fmt.Errorf("unknown log format `%s`. possible formats: %s", logFormat, []string{"text", "json"})
-	}
-
-	return nil
-}

logging/logger.go

@@ -0,0 +1,233 @@
+// Package logging wires the nebula runtime-reconfigurable slog handler used
+// by nebula.Main and the nebula CLI binaries. Callers build a logger with
+// NewLogger, then call ApplyConfig at startup and from a config reload
+// callback to push logging.level, logging.format, and
+// logging.disable_timestamp changes onto the logger without rebuilding it.
+package logging
+
+import (
+	"context"
+	"fmt"
+	"io"
+	"log/slog"
+	"strings"
+	"sync/atomic"
+	"time"
+)
+
+// Config is the subset of *config.C that ApplyConfig reads. Declaring it
+// here keeps the logging package from depending on config directly, which
+// would cycle through the shared test helpers (test.NewLogger imports
+// logging, and config's tests import test). *config.C satisfies this
+// interface structurally with no adapter.
+type Config interface {
+	GetString(key, def string) string
+	GetBool(key string, def bool) bool
+}
+
+// LevelTrace is a custom slog level below Debug, used when logging.level is
+// "trace". slog has no builtin trace level; the value is one step below
+// slog.LevelDebug in slog's 4-point spacing.
+const LevelTrace = slog.Level(-8)
+
+// NewLogger returns a *slog.Logger whose level, format, and timestamp
+// emission can be reconfigured at runtime via ApplyConfig and the SSH debug
+// commands. The default configuration is info-level text output so log
+// calls made before ApplyConfig runs still produce output. Timestamps
+// follow slog's default RFC3339Nano format; set logging.disable_timestamp
+// in config to suppress them.
+//
+// ApplyConfig and the SSH commands discover the reconfig surface via
+// structural type-assertion on l.Handler(), so replacement implementations
+// (tests, platform-specific sinks) need only implement the subset of
+// {SetLevel(slog.Level), SetFormat(string) error, SetDisableTimestamp(bool)}
+// they care about. Callers that pass a plain *slog.Logger without these
+// methods get a silent no-op; reconfiguration is always opt-in.
+func NewLogger(w io.Writer) *slog.Logger {
+	return slog.New(NewHandler(w))
+}
+
+// NewHandler builds the *Handler that NewLogger wraps. Exported for
+// platform-specific sinks (notably cmd/nebula-service/logs_windows.go)
+// that want to wrap the handler with extra behavior, such as tagging each
+// record with its Event Log severity, while still benefiting from all the
+// level / format / timestamp / WithAttrs machinery implemented here.
+func NewHandler(w io.Writer) *Handler {
+	root := &handlerRoot{}
+	root.level.Set(slog.LevelInfo)
+	opts := &slog.HandlerOptions{Level: &root.level}
+	return &Handler{
+		root: root,
+		text: slog.NewTextHandler(w, opts),
+		json: slog.NewJSONHandler(w, opts),
+	}
+}
+
+// handlerRoot carries the reconfiguration state shared by every logger
+// derived from a NewHandler call. All fields are consulted on the log
+// path and updated lock-free.
+type handlerRoot struct {
+	level            slog.LevelVar
+	disableTimestamp atomic.Bool
+	// jsonMode picks which of the pre-derived inner handlers Handler.Handle
+	// dispatches to. Flipping it propagates instantly to every derived logger
+	// without rebuilding or chain-replaying anything.
+	jsonMode atomic.Bool
+}
+
+// Handler is the slog.Handler returned by NewHandler. It holds two
+// pre-derived slog handlers -- one text, one json -- both built from the
+// same accumulated WithAttrs/WithGroup state. Handle picks which one to
+// dispatch to based on handlerRoot.jsonMode, so a SetFormat call takes
+// effect immediately across the whole process without having to rebuild
+// any derived loggers.
+type Handler struct {
+	root *handlerRoot
+	text slog.Handler
+	json slog.Handler
+}
+
+func (h *Handler) Enabled(_ context.Context, l slog.Level) bool {
+	return h.root.level.Level() <= l
+}
+
+func (h *Handler) Handle(ctx context.Context, r slog.Record) error {
+	if h.root.disableTimestamp.Load() {
+		r.Time = time.Time{}
+	}
+	if h.root.jsonMode.Load() {
+		return h.json.Handle(ctx, r)
+	}
+	return h.text.Handle(ctx, r)
+}
+
+func (h *Handler) WithAttrs(attrs []slog.Attr) slog.Handler {
+	if len(attrs) == 0 {
+		return h
+	}
+	return &Handler{
+		root: h.root,
+		text: h.text.WithAttrs(attrs),
+		json: h.json.WithAttrs(attrs),
+	}
+}
+
+func (h *Handler) WithGroup(name string) slog.Handler {
+	if name == "" {
+		return h
+	}
+	return &Handler{
+		root: h.root,
+		text: h.text.WithGroup(name),
+		json: h.json.WithGroup(name),
+	}
+}
+
+// SetLevel updates the effective log level. Propagates to every derived
+// logger via the shared LevelVar.
+func (h *Handler) SetLevel(level slog.Level) { h.root.level.Set(level) }
+
+// GetLevel reports the current log level.
+func (h *Handler) GetLevel() slog.Level { return h.root.level.Level() }
+
+// SetFormat flips the output format atomically. Valid formats are "text"
+// and "json". Every derived logger sees the new format on its next Handle
+// call; no rebuild or registration is required.
+func (h *Handler) SetFormat(format string) error {
+	switch format {
+	case "text":
+		h.root.jsonMode.Store(false)
+	case "json":
+		h.root.jsonMode.Store(true)
+	default:
+		return fmt.Errorf("unknown log format `%s`. possible formats: %s", format, []string{"text", "json"})
+	}
+	return nil
+}
+
+// GetFormat reports the currently selected format name.
+func (h *Handler) GetFormat() string {
+	if h.root.jsonMode.Load() {
+		return "json"
+	}
+	return "text"
+}
+
+// SetDisableTimestamp toggles whether Handle zeroes r.Time before
+// dispatching (slog's builtin text/json handlers skip emitting the time
+// attribute on a zero time).
+func (h *Handler) SetDisableTimestamp(v bool) { h.root.disableTimestamp.Store(v) }
+
+// ApplyConfig reads logging.level, logging.format, and (optionally)
+// logging.disable_timestamp from c and applies them to l. The reconfig
+// surface is discovered via structural type-assertion on l.Handler(), so
+// foreign handlers silently opt out of whichever capabilities they do not
+// implement.
+//
+// nebula.Main does NOT call this function on your behalf; callers that want
+// config-driven log level / format / timestamp updates invoke it at
+// startup and register it as a reload callback themselves. This keeps the
+// library from mutating an embedder's logger without their say-so.
+func ApplyConfig(l *slog.Logger, c Config) error {
+	h := l.Handler()
+
+	lvl, err := ParseLevel(strings.ToLower(c.GetString("logging.level", "info")))
+	if err != nil {
+		return err
+	}
+	if ls, ok := h.(interface{ SetLevel(slog.Level) }); ok {
+		ls.SetLevel(lvl)
+	}
+
+	format := strings.ToLower(c.GetString("logging.format", "text"))
+	if fs, ok := h.(interface{ SetFormat(string) error }); ok {
+		if err := fs.SetFormat(format); err != nil {
+			return err
+		}
+	}
+
+	if ts, ok := h.(interface{ SetDisableTimestamp(bool) }); ok {
+		ts.SetDisableTimestamp(c.GetBool("logging.disable_timestamp", false))
+	}
+	return nil
+}
+
+// ParseLevel converts a config-string level name ("trace", "debug", "info",
+// "warn"/"warning", "error", "fatal"/"panic") to a slog.Level. "fatal" and
+// "panic" are accepted for backwards compatibility with pre-slog configs
+// and both map to slog.LevelError.
+func ParseLevel(s string) (slog.Level, error) {
+	switch s {
+	case "trace":
+		return LevelTrace, nil
+	case "debug":
+		return slog.LevelDebug, nil
+	case "info":
+		return slog.LevelInfo, nil
+	case "warn", "warning":
+		return slog.LevelWarn, nil
+	case "error":
+		return slog.LevelError, nil
+	case "fatal", "panic":
+		return slog.LevelError, nil
+	default:
+		return 0, fmt.Errorf("not a valid logging level: %q", s)
+	}
+}
+
+// LevelName returns a human-readable name for a slog.Level matching the
+// strings accepted by ParseLevel.
+func LevelName(l slog.Level) string {
+	switch {
+	case l <= LevelTrace:
+		return "trace"
+	case l <= slog.LevelDebug:
+		return "debug"
+	case l <= slog.LevelInfo:
+		return "info"
+	case l <= slog.LevelWarn:
+		return "warn"
+	default:
+		return "error"
+	}
+}

logging/logger_bench_test.go

@@ -0,0 +1,90 @@
+package logging
+
+import (
+	"context"
+	"io"
+	"log/slog"
+	"testing"
+)
+
+// BenchmarkLogger_* compare the handler returned by NewLogger against a
+// stock slog text handler. The key thing we care about is the per-log
+// cost on a logger that has been derived via .With(), because that is the
+// shape subsystems store on their structs (HostInfo.logger(),
+// lh.l.With("subsystem", ...), etc.) and call from hot paths.
+
+func BenchmarkLogger_Stock_RootInfo(b *testing.B) {
+	l := slog.New(slog.DiscardHandler)
+	b.ReportAllocs()
+	b.ResetTimer()
+	for i := 0; i < b.N; i++ {
+		l.Info("hello", "i", i)
+	}
+}
+
+func BenchmarkLogger_Nebula_RootInfo(b *testing.B) {
+	l := NewLogger(io.Discard)
+	b.ReportAllocs()
+	b.ResetTimer()
+	for i := 0; i < b.N; i++ {
+		l.Info("hello", "i", i)
+	}
+}
+
+func BenchmarkLogger_Stock_DerivedInfo(b *testing.B) {
+	l := slog.New(slog.DiscardHandler).With(
+		"subsystem", "bench",
+		"localIndex", 1234,
+	)
+	b.ReportAllocs()
+	b.ResetTimer()
+	for i := 0; i < b.N; i++ {
+		l.Info("hello", "i", i)
+	}
+}
+
+func BenchmarkLogger_Nebula_DerivedInfo(b *testing.B) {
+	l := NewLogger(io.Discard).With(
+		"subsystem", "bench",
+		"localIndex", 1234,
+	)
+	b.ReportAllocs()
+	b.ResetTimer()
+	for i := 0; i < b.N; i++ {
+		l.Info("hello", "i", i)
+	}
+}
+
+// Gated-off-path benchmarks: mimic the typical hot-path shape
+// `if l.Enabled(ctx, slog.LevelDebug) { ... }` where the log is gated below
+// the active level. This is the dominant pattern in inside.go/outside.go and
+// what we pay on every packet.
+func BenchmarkLogger_Stock_DerivedEnabledGateMiss(b *testing.B) {
+	l := slog.New(slog.DiscardHandler).With(
+		"subsystem", "bench",
+		"localIndex", 1234,
+	)
+	ctx := context.Background()
+	b.ReportAllocs()
+	b.ResetTimer()
+	for i := 0; i < b.N; i++ {
+		if l.Enabled(ctx, slog.LevelDebug) {
+			l.Debug("hello", "i", i)
+		}
+	}
+}
+
+func BenchmarkLogger_Nebula_DerivedEnabledGateMiss(b *testing.B) {
+	l := NewLogger(io.Discard).With(
+		"subsystem", "bench",
+		"localIndex", 1234,
+	)
+	ctx := context.Background()
+	b.ReportAllocs()
+	b.ResetTimer()
+	for i := 0; i < b.N; i++ {
+		if l.Enabled(ctx, slog.LevelDebug) {
+			l.Debug("hello", "i", i)
+		}
+	}
+}

main.go

@@ -3,16 +3,13 @@ package nebula
 import (
 	"context"
 	"fmt"
-	"log"
+	"log/slog"
 	"net"
-	"net/http"
-	_ "net/http/pprof"
 	"net/netip"
 	"runtime/debug"
 	"strings"
 	"time"
 
-	"github.com/sirupsen/logrus"
 	"github.com/slackhq/nebula/config"
 	"github.com/slackhq/nebula/overlay"
 	"github.com/slackhq/nebula/sshd"
@@ -23,7 +20,7 @@ import (
 
 type m = map[string]any
 
-func Main(c *config.C, configTest bool, buildVersion string, logger *logrus.Logger, deviceFactory overlay.DeviceFactory) (retcon *Control, reterr error) {
+func Main(c *config.C, configTest bool, buildVersion string, l *slog.Logger, deviceFactory overlay.DeviceFactory) (retcon *Control, reterr error) {
 	ctx, cancel := context.WithCancel(context.Background())
 	// Automatically cancel the context if Main returns an error, to signal all created goroutines to quit.
 	defer func() {
@@ -36,11 +33,6 @@ func Main(c *config.C, configTest bool, buildVersion string, logger *logrus.Logg
 		buildVersion = moduleVersion()
 	}
 
-	l := logger
-	l.Formatter = &logrus.TextFormatter{
-		FullTimestamp: true,
-	}
-
 	// Print the config if in test, the exit comes later
 	if configTest {
 		b, err := yaml.Marshal(c.Settings)
@@ -49,26 +41,9 @@ func Main(c *config.C, configTest bool, buildVersion string, logger *logrus.Logg
 		}
 
 		// Print the final config
-		l.Println(string(b))
+		l.Info(string(b))
 	}
 
-	//todo!!!
-	go func() {
-		log.Println(http.ListenAndServe("0.0.0.0:6060", nil))
-	}()
-
-	err := configLogger(l, c)
-	if err != nil {
-		return nil, util.ContextualizeIfNeeded("Failed to configure the logger", err)
-	}
-
-	c.RegisterReloadCallback(func(c *config.C) {
-		err := configLogger(l, c)
-		if err != nil {
-			l.WithError(err).Error("Failed to configure the logger")
-		}
-	})
-
 	pki, err := NewPKIFromConfig(l, c)
 	if err != nil {
 		return nil, util.ContextualizeIfNeeded("Failed to load PKI from config", err)
@@ -78,9 +53,9 @@ func Main(c *config.C, configTest bool, buildVersion string, logger *logrus.Logg
 	if err != nil {
 		return nil, util.ContextualizeIfNeeded("Error while loading firewall rules", err)
 	}
-	l.WithField("firewallHashes", fw.GetRuleHashes()).Info("Firewall started")
+	l.Info("Firewall started", "firewallHashes", fw.GetRuleHashes())
 
-	ssh, err := sshd.NewSSHServer(l.WithField("subsystem", "sshd"))
+	ssh, err := sshd.NewSSHServer(l.With("subsystem", "sshd"))
 	if err != nil {
 		return nil, util.ContextualizeIfNeeded("Error while creating SSH server", err)
 	}
@@ -89,7 +64,7 @@ func Main(c *config.C, configTest bool, buildVersion string, logger *logrus.Logg
 	if c.GetBool("sshd.enabled", false) {
 		sshStart, err = configSSH(l, ssh, c)
 		if err != nil {
-			l.WithError(err).Warn("Failed to configure sshd, ssh debugging will not be available")
+			l.Warn("Failed to configure sshd, ssh debugging will not be available", "error", err)
 			sshStart = nil
 		}
 	}
@@ -107,7 +82,7 @@ func Main(c *config.C, configTest bool, buildVersion string, logger *logrus.Logg
 			routines = 1
 		}
 		if routines > 1 {
-			l.WithField("routines", routines).Info("Using multiple routines")
+			l.Info("Using multiple routines", "routines", routines)
 		}
 	} else {
 		// deprecated and undocumented
@@ -115,7 +90,7 @@ func Main(c *config.C, configTest bool, buildVersion string, logger *logrus.Logg
 		udpQueues := c.GetInt("listen.routines", 1)
 		routines = max(tunQueues, udpQueues)
 		if routines != 1 {
-			l.WithField("routines", routines).Warn("Setting tun.routines and listen.routines is deprecated. Use `routines` instead")
+			l.Warn("Setting tun.routines and listen.routines is deprecated. Use `routines` instead", "routines", routines)
 		}
 	}
 
@@ -128,7 +103,7 @@ func Main(c *config.C, configTest bool, buildVersion string, logger *logrus.Logg
 		conntrackCacheTimeout = 1 * time.Second
 	}
 	if conntrackCacheTimeout > 0 {
-		l.WithField("duration", conntrackCacheTimeout).Info("Using routine-local conntrack cache")
+		l.Info("Using routine-local conntrack cache", "duration", conntrackCacheTimeout)
 	}
 
 	var tun overlay.Device
@@ -174,7 +149,7 @@ func Main(c *config.C, configTest bool, buildVersion string, logger *logrus.Logg
 		}
 
 		for i := 0; i < routines; i++ {
-			l.Infof("listening on %v", netip.AddrPortFrom(listenHost, uint16(port)))
+			l.Info("listening", "addr", netip.AddrPortFrom(listenHost, uint16(port)))
 			udpServer, err := udp.NewListener(l, listenHost, port, routines > 1, c.GetInt("listen.batch", 64))
 			if err != nil {
 				return nil, util.NewContextualError("Failed to open udp listener", m{"queue": i}, err)
@@ -223,13 +198,9 @@ func Main(c *config.C, configTest bool, buildVersion string, logger *logrus.Logg
 	handshakeManager := NewHandshakeManager(l, hostMap, lightHouse, udpConns[0], handshakeConfig)
 	lightHouse.handshakeTrigger = handshakeManager.trigger
 
-	serveDns := false
+	ds, err := newDnsServerFromConfig(ctx, l, pki.getCertState(), hostMap, c)
-	if c.GetBool("lighthouse.serve_dns", false) {
+	if err != nil {
-		if c.GetBool("lighthouse.am_lighthouse", false) {
+		l.Warn("Failed to start DNS responder", "error", err)
-			serveDns = true
-		} else {
-			l.Warn("DNS server refusing to run because this host is not a lighthouse.")
-		}
 	}
 
 	ifConfig := &InterfaceConfig{
@@ -238,7 +209,7 @@ func Main(c *config.C, configTest bool, buildVersion string, logger *logrus.Logg
 		Outside:               udpConns[0],
 		pki:                   pki,
 		Firewall:              fw,
-		ServeDns:              serveDns,
+		DnsServer:             ds,
 		HandshakeManager:      handshakeManager,
 		connectionManager:     connManager,
 		lightHouse:            lightHouse,
@@ -275,7 +246,7 @@ func Main(c *config.C, configTest bool, buildVersion string, logger *logrus.Logg
 		go handshakeManager.Run(ctx)
 	}
 
-	statsStart, err := startStats(l, c, buildVersion, configTest)
+	stats, err := newStatsServerFromConfig(ctx, l, c, buildVersion, configTest)
 	if err != nil {
 		return nil, util.ContextualizeIfNeeded("Failed to start stats emitter", err)
 	}
@@ -288,13 +259,6 @@ func Main(c *config.C, configTest bool, buildVersion string, logger *logrus.Logg
 
 	attachCommands(l, c, ssh, ifce)
 
-	// Start DNS server last to allow using the nebula IP as lighthouse.dns.host
-	var dnsStart func()
-	if lightHouse.amLighthouse && serveDns {
-		l.Debugln("Starting dns server")
-		dnsStart = dnsMain(l, pki.getCertState(), hostMap, c)
-	}
-
 	return &Control{
 		state:                  StateReady,
 		f:                      ifce,
@@ -302,8 +266,8 @@ func Main(c *config.C, configTest bool, buildVersion string, logger *logrus.Logg
 		ctx:                    ctx,
 		cancel:                 cancel,
 		sshStart:               sshStart,
-		statsStart:             statsStart,
+		statsStart:             stats.Start,
-		dnsStart:               dnsStart,
+		dnsStart:               ds.Start,
 		lighthouseStart:        lightHouse.StartUpdateWorker,
 		connectionManagerStart: connManager.Start,
 	}, nil

noise.go

@@ -16,8 +16,6 @@ var noiseEndianness endianness = binary.BigEndian
 
 type NebulaCipherState struct {
 	c cipher.AEAD
-	//k [32]byte
-	//n uint64
 }
 
 func NewNebulaCipherState(s *noise.CipherState) *NebulaCipherState {

outside.go

@@ -1,15 +1,16 @@
 package nebula
 
 import (
+	"context"
 	"encoding/binary"
 	"errors"
+	"log/slog"
 	"net/netip"
 	"time"
 
 	"github.com/google/gopacket/layers"
 	"golang.org/x/net/ipv6"
 
-	"github.com/sirupsen/logrus"
 	"github.com/slackhq/nebula/firewall"
 	"github.com/slackhq/nebula/header"
 	"golang.org/x/net/ipv4"
@@ -24,7 +25,11 @@ func (f *Interface) readOutsidePackets(via ViaSender, out []byte, packet []byte,
 	if err != nil {
 		// Hole punch packets are 0 or 1 byte big, so lets ignore printing those errors
 		if len(packet) > 1 {
-			f.l.WithField("packet", packet).Infof("Error while parsing inbound packet from %s: %s", via, err)
+			f.l.Info("Error while parsing inbound packet",
+				"from", via,
+				"error", err,
+				"packet", packet,
+			)
 		}
 		return
 	}
@@ -32,8 +37,8 @@ func (f *Interface) readOutsidePackets(via ViaSender, out []byte, packet []byte,
 	//l.Error("in packet ", header, packet[HeaderLen:])
 	if !via.IsRelayed {
 		if f.myVpnNetworksTable.Contains(via.UdpAddr.Addr()) {
-			if f.l.Level >= logrus.DebugLevel {
+			if f.l.Enabled(context.Background(), slog.LevelDebug) {
-				f.l.WithField("from", via).Debug("Refusing to process double encrypted packet")
+				f.l.Debug("Refusing to process double encrypted packet", "from", via)
 			}
 			return
 		}
@@ -87,7 +92,10 @@ func (f *Interface) readOutsidePackets(via ViaSender, out []byte, packet []byte,
 			if !ok {
 				// The only way this happens is if hostmap has an index to the correct HostInfo, but the HostInfo is missing
 				// its internal mapping. This should never happen.
-				hostinfo.logger(f.l).WithFields(logrus.Fields{"vpnAddrs": hostinfo.vpnAddrs, "remoteIndex": h.RemoteIndex}).Error("HostInfo missing remote relay index")
+				hostinfo.logger(f.l).Error("HostInfo missing remote relay index",
+					"vpnAddrs", hostinfo.vpnAddrs,
+					"remoteIndex", h.RemoteIndex,
+				)
 				return
 			}
 
@@ -108,7 +116,11 @@ func (f *Interface) readOutsidePackets(via ViaSender, out []byte, packet []byte,
 				// Find the target HostInfo relay object
 				targetHI, targetRelay, err := f.hostMap.QueryVpnAddrsRelayFor(hostinfo.vpnAddrs, relay.PeerAddr)
 				if err != nil {
-					hostinfo.logger(f.l).WithField("relayTo", relay.PeerAddr).WithError(err).WithField("hostinfo.vpnAddrs", hostinfo.vpnAddrs).Info("Failed to find target host info by ip")
+					hostinfo.logger(f.l).Info("Failed to find target host info by ip",
+						"relayTo", relay.PeerAddr,
+						"error", err,
+						"hostinfo.vpnAddrs", hostinfo.vpnAddrs,
+					)
 					return
 				}
 
@@ -124,7 +136,11 @@ func (f *Interface) readOutsidePackets(via ViaSender, out []byte, packet []byte,
 						hostinfo.logger(f.l).Error("Unexpected Relay Type of Terminal")
 					}
 				} else {
-					hostinfo.logger(f.l).WithFields(logrus.Fields{"relayTo": relay.PeerAddr, "relayFrom": hostinfo.vpnAddrs[0], "targetRelayState": targetRelay.State}).Info("Unexpected target relay state")
+					hostinfo.logger(f.l).Info("Unexpected target relay state",
+						"relayTo", relay.PeerAddr,
+						"relayFrom", hostinfo.vpnAddrs[0],
+						"targetRelayState", targetRelay.State,
+					)
 					return
 				}
 			}
@@ -138,9 +154,11 @@ func (f *Interface) readOutsidePackets(via ViaSender, out []byte, packet []byte,
 
 		d, err := f.decrypt(hostinfo, h.MessageCounter, out, packet, h, nb)
 		if err != nil {
-			hostinfo.logger(f.l).WithError(err).WithField("from", via).
+			hostinfo.logger(f.l).Error("Failed to decrypt lighthouse packet",
-				WithField("packet", packet).
+				"error", err,
-				Error("Failed to decrypt lighthouse packet")
+				"from", via,
+				"packet", packet,
+			)
 			return
 		}
 
@@ -157,9 +175,11 @@ func (f *Interface) readOutsidePackets(via ViaSender, out []byte, packet []byte,
 
 		d, err := f.decrypt(hostinfo, h.MessageCounter, out, packet, h, nb)
 		if err != nil {
-			hostinfo.logger(f.l).WithError(err).WithField("from", via).
+			hostinfo.logger(f.l).Error("Failed to decrypt test packet",
-				WithField("packet", packet).
+				"error", err,
-				Error("Failed to decrypt test packet")
+				"from", via,
+				"packet", packet,
+			)
 			return
 		}
 
@@ -192,14 +212,15 @@ func (f *Interface) readOutsidePackets(via ViaSender, out []byte, packet []byte,
 		}
 		_, err = f.decrypt(hostinfo, h.MessageCounter, out, packet, h, nb)
 		if err != nil {
-			hostinfo.logger(f.l).WithError(err).WithField("from", via).
+			hostinfo.logger(f.l).Error("Failed to decrypt CloseTunnel packet",
-				WithField("packet", packet).
+				"error", err,
-				Error("Failed to decrypt CloseTunnel packet")
+				"from", via,
+				"packet", packet,
+			)
 			return
 		}
 
-		hostinfo.logger(f.l).WithField("from", via).
+		hostinfo.logger(f.l).Info("Close tunnel received, tearing down.", "from", via)
-			Info("Close tunnel received, tearing down.")
 
 		f.closeTunnel(hostinfo)
 		return
@@ -211,9 +232,11 @@ func (f *Interface) readOutsidePackets(via ViaSender, out []byte, packet []byte,
 
 		d, err := f.decrypt(hostinfo, h.MessageCounter, out, packet, h, nb)
 		if err != nil {
-			hostinfo.logger(f.l).WithError(err).WithField("from", via).
+			hostinfo.logger(f.l).Error("Failed to decrypt Control packet",
-				WithField("packet", packet).
+				"error", err,
-				Error("Failed to decrypt Control packet")
+				"from", via,
+				"packet", packet,
+			)
 			return
 		}
 
@@ -221,7 +244,9 @@ func (f *Interface) readOutsidePackets(via ViaSender, out []byte, packet []byte,
 
 	default:
 		f.messageMetrics.Rx(h.Type, h.Subtype, 1)
-		hostinfo.logger(f.l).Debugf("Unexpected packet received from %s", via)
+		if f.l.Enabled(context.Background(), slog.LevelDebug) {
+			hostinfo.logger(f.l).Debug("Unexpected packet received", "from", via)
+		}
 		return
 	}
 
@@ -247,20 +272,27 @@ func (f *Interface) sendCloseTunnel(h *HostInfo) {
 func (f *Interface) handleHostRoaming(hostinfo *HostInfo, via ViaSender) {
 	if !via.IsRelayed && hostinfo.remote != via.UdpAddr {
 		if !f.lightHouse.GetRemoteAllowList().AllowAll(hostinfo.vpnAddrs, via.UdpAddr.Addr()) {
-			hostinfo.logger(f.l).WithField("newAddr", via.UdpAddr).Debug("lighthouse.remote_allow_list denied roaming")
+			if f.l.Enabled(context.Background(), slog.LevelDebug) {
+				hostinfo.logger(f.l).Debug("lighthouse.remote_allow_list denied roaming", "newAddr", via.UdpAddr)
+			}
 			return
 		}
 
 		if !hostinfo.lastRoam.IsZero() && via.UdpAddr == hostinfo.lastRoamRemote && time.Since(hostinfo.lastRoam) < RoamingSuppressSeconds*time.Second {
-			if f.l.Level >= logrus.DebugLevel {
+			if f.l.Enabled(context.Background(), slog.LevelDebug) {
-				hostinfo.logger(f.l).WithField("udpAddr", hostinfo.remote).WithField("newAddr", via.UdpAddr).
+				hostinfo.logger(f.l).Debug("Suppressing roam back to previous remote",
-					Debugf("Suppressing roam back to previous remote for %d seconds", RoamingSuppressSeconds)
+					"suppressSeconds", RoamingSuppressSeconds,
+					"udpAddr", hostinfo.remote,
+					"newAddr", via.UdpAddr,
+				)
 			}
 			return
 		}
 
-		hostinfo.logger(f.l).WithField("udpAddr", hostinfo.remote).WithField("newAddr", via.UdpAddr).
+		hostinfo.logger(f.l).Info("Host roamed to new udp ip/port.",
-			Info("Host roamed to new udp ip/port.")
+			"udpAddr", hostinfo.remote,
+			"newAddr", via.UdpAddr,
+		)
 		hostinfo.lastRoam = time.Now()
 		hostinfo.lastRoamRemote = hostinfo.remote
 		hostinfo.SetRemote(via.UdpAddr)
@@ -491,8 +523,9 @@ func (f *Interface) decrypt(hostinfo *HostInfo, mc uint64, out []byte, packet []
 	}
 
 	if !hostinfo.ConnectionState.window.Update(f.l, mc) {
-		hostinfo.logger(f.l).WithField("header", h).
+		if f.l.Enabled(context.Background(), slog.LevelDebug) {
-			Debugln("dropping out of window packet")
+			hostinfo.logger(f.l).Debug("dropping out of window packet", "header", h)
+		}
 		return nil, errors.New("out of window packet")
 	}
 
@@ -504,20 +537,23 @@ func (f *Interface) decryptToTun(hostinfo *HostInfo, messageCounter uint64, out
 
 	out, err = hostinfo.ConnectionState.dKey.DecryptDanger(out, packet[:header.Len], packet[header.Len:], messageCounter, nb)
 	if err != nil {
-		hostinfo.logger(f.l).WithError(err).Error("Failed to decrypt packet")
+		hostinfo.logger(f.l).Error("Failed to decrypt packet", "error", err)
 		return false
 	}
 
 	err = newPacket(out, true, fwPacket)
 	if err != nil {
-		hostinfo.logger(f.l).WithError(err).WithField("packet", out).
+		hostinfo.logger(f.l).Warn("Error while validating inbound packet",
-			Warnf("Error while validating inbound packet")
+			"error", err,
+			"packet", out,
+		)
 		return false
 	}
 
 	if !hostinfo.ConnectionState.window.Update(f.l, messageCounter) {
-		hostinfo.logger(f.l).WithField("fwPacket", fwPacket).
+		if f.l.Enabled(context.Background(), slog.LevelDebug) {
-			Debugln("dropping out of window packet")
+			hostinfo.logger(f.l).Debug("dropping out of window packet", "fwPacket", fwPacket)
+		}
 		return false
 	}
 
@@ -526,18 +562,19 @@ func (f *Interface) decryptToTun(hostinfo *HostInfo, messageCounter uint64, out
 		// 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
 		f.rejectOutside(out, hostinfo.ConnectionState, hostinfo, nb, packet, q)
-		if f.l.Level >= logrus.DebugLevel {
+		if f.l.Enabled(context.Background(), slog.LevelDebug) {
-			hostinfo.logger(f.l).WithField("fwPacket", fwPacket).
+			hostinfo.logger(f.l).Debug("dropping inbound packet",
-				WithField("reason", dropReason).
+				"fwPacket", fwPacket,
-				Debugln("dropping inbound packet")
+				"reason", dropReason,
+			)
 		}
 		return false
 	}
 
 	f.connectionManager.In(hostinfo)
-	err = f.tunCoalescers[q].Add(out)
+	err = f.batchers[q].Commit(out)
 	if err != nil {
-		f.l.WithError(err).Error("Failed to write to tun")
+		f.l.Error("Failed to write to tun", "error", err)
 	}
 	return true
 }
@@ -553,35 +590,41 @@ func (f *Interface) sendRecvError(endpoint netip.AddrPort, index uint32) {
 
 	b := header.Encode(make([]byte, header.Len), header.Version, header.RecvError, 0, index, 0)
 	_ = f.outside.WriteTo(b, endpoint)
-	if f.l.Level >= logrus.DebugLevel {
+	if f.l.Enabled(context.Background(), slog.LevelDebug) {
-		f.l.WithField("index", index).
+		f.l.Debug("Recv error sent",
-			WithField("udpAddr", endpoint).
+			"index", index,
-			Debug("Recv error sent")
+			"udpAddr", endpoint,
+		)
 	}
 }
 
 func (f *Interface) handleRecvError(addr netip.AddrPort, h *header.H) {
 	if !f.acceptRecvErrorConfig.ShouldRecvError(addr) {
-		f.l.WithField("index", h.RemoteIndex).
+		f.l.Debug("Recv error received, ignoring",
-			WithField("udpAddr", addr).
+			"index", h.RemoteIndex,
-			Debug("Recv error received, ignoring")
+			"udpAddr", addr,
+		)
 		return
 	}
 
-	if f.l.Level >= logrus.DebugLevel {
+	if f.l.Enabled(context.Background(), slog.LevelDebug) {
-		f.l.WithField("index", h.RemoteIndex).
+		f.l.Debug("Recv error received",
-			WithField("udpAddr", addr).
+			"index", h.RemoteIndex,
-			Debug("Recv error received")
+			"udpAddr", addr,
+		)
 	}
 
 	hostinfo := f.hostMap.QueryReverseIndex(h.RemoteIndex)
 	if hostinfo == nil {
-		f.l.WithField("remoteIndex", h.RemoteIndex).Debugln("Did not find remote index in main hostmap")
+		f.l.Debug("Did not find remote index in main hostmap", "remoteIndex", h.RemoteIndex)
 		return
 	}
 
 	if hostinfo.remote.IsValid() && hostinfo.remote != addr {
-		f.l.Infoln("Someone spoofing recv_errors? ", addr, hostinfo.remote)
+		f.l.Info("Someone spoofing recv_errors?",
+			"addr", addr,
+			"hostinfoRemote", hostinfo.remote,
+		)
 		return
 	}
 

overlay/batch/batch.go

@@ -0,0 +1,33 @@
+package batch
+
+import "net/netip"
+
+type RxBatcher interface {
+	// Reserve creates a pkt to borrow
+	Reserve(sz int) []byte
+	// Commit borrows pkt. The caller must keep pkt valid until the next Flush
+	Commit(pkt []byte) error
+	// Flush emits every queued packet in arrival order. Returns the
+	// first error observed; keeps draining so one bad packet doesn't hold up
+	// the rest. After Flush returns, borrowed payload slices may be recycled.
+	Flush() error
+}
+
+type TxBatcher interface {
+	// Next returns a zero-length slice with slotCap capacity over the next unused
+	// slot's backing bytes. The caller writes into the returned slice and then
+	// calls Commit with the final length and destination. Next returns nil when
+	// the batch is full.
+	Next() []byte
+	// Commit records the slot just returned by Next as a packet of length n
+	// destined for dst.
+	Commit(n int, dst netip.AddrPort)
+	// Reset clears committed slots; backing storage is retained for reuse.
+	Reset()
+	// Len returns the number of committed packets.
+	Len() int
+	// Cap returns the maximum number of slots in the batch.
+	Cap() int
+	// Get returns the buffers needed to send the batch
+	Get() ([][]byte, []netip.AddrPort)
+}

overlay/batch/passthrough.go

@@ -0,0 +1,57 @@
+package batch
+
+import (
+	"io"
+
+	"github.com/slackhq/nebula/udp"
+)
+
+// Passthrough is a RxBatcher that doesn't batch anything, it just accumulates and then sends packets.
+type Passthrough struct {
+	out     io.Writer
+	slots   [][]byte
+	backing []byte
+	cursor  int
+}
+
+func NewPassthrough(w io.Writer) *Passthrough {
+	const baseNumSlots = 128
+	return &Passthrough{
+		out:     w,
+		slots:   make([][]byte, 0, baseNumSlots),
+		backing: make([]byte, 0, baseNumSlots*udp.MTU),
+	}
+}
+
+func (p *Passthrough) Reserve(sz int) []byte {
+	if len(p.backing)+sz > cap(p.backing) {
+		// Grow: allocate a fresh backing. Already-committed slices still
+		// reference the old array and remain valid until Flush drops them.
+		newCap := max(cap(p.backing)*2, sz)
+		p.backing = make([]byte, 0, newCap)
+	}
+	start := len(p.backing)
+	p.backing = p.backing[:start+sz]
+	return p.backing[start : start+sz : start+sz] //return zero length, sz-cap slice
+}
+
+func (p *Passthrough) Commit(pkt []byte) error {
+	p.slots = append(p.slots, pkt)
+	return nil
+}
+
+func (p *Passthrough) Flush() error {
+	var firstErr error
+	for _, s := range p.slots {
+		_, err := p.out.Write(s)
+		if err != nil && firstErr == nil {
+			firstErr = err
+		}
+	}
+	for i := range p.slots {
+		p.slots[i] = nil
+	}
+	p.slots = p.slots[:0]
+	p.backing = p.backing[:0]
+	return firstErr
+}

overlay/batch/tcp_coalesce.go

@@ -1,4 +1,4 @@
-package coalesce
+package batch
 
 import (
 	"bytes"
@@ -84,6 +84,8 @@ type TCPCoalescer struct {
 	// when a non-admissible packet for that flow arrives, or in Flush.
 	openSlots map[flowKey]*coalesceSlot
 	pool      []*coalesceSlot // free list for reuse
+
+	backing []byte
 }
 
 func NewTCPCoalescer(w io.Writer) *TCPCoalescer {
@@ -92,6 +94,7 @@ func NewTCPCoalescer(w io.Writer) *TCPCoalescer {
 		slots:     make([]*coalesceSlot, 0, initialSlots),
 		openSlots: make(map[flowKey]*coalesceSlot, initialSlots),
 		pool:      make([]*coalesceSlot, 0, initialSlots),
+		backing:   make([]byte, 0, initialSlots*65535),
 	}
 	if gw, ok := w.(tio.GSOWriter); ok && gw.GSOSupported() {
 		c.gsoW = gw
@@ -194,10 +197,22 @@ func (p parsedTCP) coalesceable() bool {
 	return p.payLen > 0
 }
 
-// Add borrows pkt. The caller must keep pkt valid until the next Flush,
+func (c *TCPCoalescer) Reserve(sz int) []byte {
+	if len(c.backing)+sz > cap(c.backing) {
+		// Grow: allocate a fresh backing. Already-committed slices still
+		// reference the old array and remain valid until Flush drops them.
+		newCap := max(cap(c.backing)*2, sz)
+		c.backing = make([]byte, 0, newCap)
+	}
+	start := len(c.backing)
+	c.backing = c.backing[:start+sz]
+	return c.backing[start : start+sz : start+sz] //return zero length, sz-cap slice
+}
+
+// Commit borrows pkt. The caller must keep pkt valid until the next Flush,
 // whether or not the packet was coalesced — passthrough (non-admissible)
 // packets are queued and written at Flush time, not synchronously.
-func (c *TCPCoalescer) Add(pkt []byte) error {
+func (c *TCPCoalescer) Commit(pkt []byte) error {
 	if c.gsoW == nil {
 		c.addPassthrough(pkt)
 		return nil
@@ -258,6 +273,8 @@ func (c *TCPCoalescer) Flush() error {
 	for k := range c.openSlots {
 		delete(c.openSlots, k)
 	}
+
+	c.backing = c.backing[:0]
 	return first
 }
 
@@ -327,6 +344,11 @@ func (c *TCPCoalescer) appendPayload(s *coalesceSlot, pkt []byte, info parsedTCP
 	s.numSeg++
 	s.totalPay += info.payLen
 	s.nextSeq = info.seq + uint32(info.payLen)
+	if info.flags&0x08 != 0 {
+		// Propagate PSH into the seed header so kernel TSO sets it on the
+		// last segment. Without this the sender's push signal is dropped.
+		s.hdrBuf[s.ipHdrLen+13] |= 0x08
+	}
 	if info.payLen < s.gsoSize || info.flags&0x08 != 0 {
 		s.psh = true
 	}
@@ -380,7 +402,7 @@ func (c *TCPCoalescer) flushSlot(s *coalesceSlot) error {
 	tcsum := s.ipHdrLen + 16
 	binary.BigEndian.PutUint16(hdr[tcsum:tcsum+2], foldOnceNoInvert(psum))
 
-	return c.gsoW.WriteGSO(hdr, s.payIovs, uint16(s.gsoSize), s.isV6, uint16(s.ipHdrLen))
+	return c.gsoW.WriteGSO(hdr[:s.ipHdrLen], hdr[s.ipHdrLen:], s.payIovs)
 }
 
 // headersMatch compares two IP+TCP header prefixes for byte-for-byte

overlay/batch/tcp_coalesce_test.go

@@ -1,4 +1,4 @@
-package coalesce
+package batch
 
 import (
 	"encoding/binary"
@@ -6,14 +6,17 @@ import (
 )
 
 // fakeTunWriter records plain Writes and WriteGSO calls without touching a
-// real TUN fd. WriteGSO preserves the split between hdr and borrowed pays
+// real TUN fd. WriteGSO records the IP header, transport header, and
-// so tests can inspect each independently.
+// borrowed payload fragments separately so tests can inspect each.
 type fakeTunWriter struct {
 	gsoEnabled bool
 	writes     [][]byte
 	gsoWrites  []fakeGSOWrite
 }
 
+// fakeGSOWrite captures one WriteGSO call. hdr is the concatenation of the
+// IP and transport headers (in that order), gsoSize / isV6 / csumStart are
+// derived from the call so existing assertions keep working unchanged.
 type fakeGSOWrite struct {
 	hdr       []byte
 	pays      [][]byte
@@ -47,21 +50,27 @@ func (w *fakeTunWriter) Write(p []byte) (int, error) {
 	return len(p), nil
 }
 
-func (w *fakeTunWriter) WriteGSO(hdr []byte, pays [][]byte, gsoSize uint16, isV6 bool, csumStart uint16) error {
+func (w *fakeTunWriter) WriteGSO(hdr []byte, transportHdr []byte, pays [][]byte) error {
-	hcopy := make([]byte, len(hdr))
+	hcopy := make([]byte, len(hdr)+len(transportHdr))
 	copy(hcopy, hdr)
+	copy(hcopy[len(hdr):], transportHdr)
 	paysCopy := make([][]byte, len(pays))
 	for i, p := range pays {
 		pc := make([]byte, len(p))
 		copy(pc, p)
 		paysCopy[i] = pc
 	}
+	var gsoSize uint16
+	if len(pays) > 1 {
+		gsoSize = uint16(len(pays[0]))
+	}
+	isV6 := len(hdr) > 0 && hdr[0]>>4 == 6
 	w.gsoWrites = append(w.gsoWrites, fakeGSOWrite{
 		hdr:       hcopy,
 		pays:      paysCopy,
 		gsoSize:   gsoSize,
 		isV6:      isV6,
-		csumStart: csumStart,
+		csumStart: uint16(len(hdr)),
 	})
 	return nil
 }
@@ -116,7 +125,7 @@ func TestCoalescerPassthroughWhenGSOUnavailable(t *testing.T) {
 	w := &fakeTunWriter{gsoEnabled: false}
 	c := NewTCPCoalescer(w)
 	pkt := buildTCPv4(1000, tcpAck, []byte("hello"))
-	if err := c.Add(pkt); err != nil {
+	if err := c.Commit(pkt); err != nil {
 		t.Fatal(err)
 	}
 	// No sync write — passthrough is deferred to Flush.
@@ -140,7 +149,7 @@ func TestCoalescerNonTCPPassthrough(t *testing.T) {
 	pkt[9] = 1
 	copy(pkt[12:16], []byte{10, 0, 0, 1})
 	copy(pkt[16:20], []byte{10, 0, 0, 2})
-	if err := c.Add(pkt); err != nil {
+	if err := c.Commit(pkt); err != nil {
 		t.Fatal(err)
 	}
 	if err := c.Flush(); err != nil {
@@ -155,7 +164,7 @@ func TestCoalescerSeedThenFlushAlone(t *testing.T) {
 	w := &fakeTunWriter{gsoEnabled: true}
 	c := NewTCPCoalescer(w)
 	pkt := buildTCPv4(1000, tcpAck, make([]byte, 1000))
-	if err := c.Add(pkt); err != nil {
+	if err := c.Commit(pkt); err != nil {
 		t.Fatal(err)
 	}
 	if len(w.writes) != 0 || len(w.gsoWrites) != 0 {
@@ -182,13 +191,13 @@ func TestCoalescerCoalescesAdjacentACKs(t *testing.T) {
 	w := &fakeTunWriter{gsoEnabled: true}
 	c := NewTCPCoalescer(w)
 	pay := make([]byte, 1200)
-	if err := c.Add(buildTCPv4(1000, tcpAck, pay)); err != nil {
+	if err := c.Commit(buildTCPv4(1000, tcpAck, pay)); err != nil {
 		t.Fatal(err)
 	}
-	if err := c.Add(buildTCPv4(2200, tcpAck, pay)); err != nil {
+	if err := c.Commit(buildTCPv4(2200, tcpAck, pay)); err != nil {
 		t.Fatal(err)
 	}
-	if err := c.Add(buildTCPv4(3400, tcpAck, pay)); err != nil {
+	if err := c.Commit(buildTCPv4(3400, tcpAck, pay)); err != nil {
 		t.Fatal(err)
 	}
 	if err := c.Flush(); err != nil {
@@ -222,10 +231,10 @@ func TestCoalescerRejectsSeqGap(t *testing.T) {
 	w := &fakeTunWriter{gsoEnabled: true}
 	c := NewTCPCoalescer(w)
 	pay := make([]byte, 1200)
-	if err := c.Add(buildTCPv4(1000, tcpAck, pay)); err != nil {
+	if err := c.Commit(buildTCPv4(1000, tcpAck, pay)); err != nil {
 		t.Fatal(err)
 	}
-	if err := c.Add(buildTCPv4(3000, tcpAck, pay)); err != nil {
+	if err := c.Commit(buildTCPv4(3000, tcpAck, pay)); err != nil {
 		t.Fatal(err)
 	}
 	if err := c.Flush(); err != nil {
@@ -241,13 +250,13 @@ func TestCoalescerRejectsFlagMismatch(t *testing.T) {
 	w := &fakeTunWriter{gsoEnabled: true}
 	c := NewTCPCoalescer(w)
 	pay := make([]byte, 1200)
-	if err := c.Add(buildTCPv4(1000, tcpAck, pay)); err != nil {
+	if err := c.Commit(buildTCPv4(1000, tcpAck, pay)); err != nil {
 		t.Fatal(err)
 	}
 	// SYN|ACK is non-admissible. Must flush matching flow's slot (gso)
 	// and then plain-write the SYN packet itself.
 	syn := buildTCPv4(2200, tcpSyn|tcpAck, pay)
-	if err := c.Add(syn); err != nil {
+	if err := c.Commit(syn); err != nil {
 		t.Fatal(err)
 	}
 	if err := c.Flush(); err != nil {
@@ -262,7 +271,7 @@ func TestCoalescerRejectsFIN(t *testing.T) {
 	w := &fakeTunWriter{gsoEnabled: true}
 	c := NewTCPCoalescer(w)
 	fin := buildTCPv4(1000, tcpAck|tcpFin, []byte("x"))
-	if err := c.Add(fin); err != nil {
+	if err := c.Commit(fin); err != nil {
 		t.Fatal(err)
 	}
 	if err := c.Flush(); err != nil {
@@ -279,15 +288,15 @@ func TestCoalescerShortLastSegmentClosesChain(t *testing.T) {
 	c := NewTCPCoalescer(w)
 	full := make([]byte, 1200)
 	half := make([]byte, 500)
-	if err := c.Add(buildTCPv4(1000, tcpAck, full)); err != nil {
+	if err := c.Commit(buildTCPv4(1000, tcpAck, full)); err != nil {
 		t.Fatal(err)
 	}
-	if err := c.Add(buildTCPv4(2200, tcpAck, half)); err != nil {
+	if err := c.Commit(buildTCPv4(2200, tcpAck, half)); err != nil {
 		t.Fatal(err)
 	}
 	// Chain now closed; next packet seeds a new slot on the same flow
 	// after flushing the old one.
-	if err := c.Add(buildTCPv4(2700, tcpAck, full)); err != nil {
+	if err := c.Commit(buildTCPv4(2700, tcpAck, full)); err != nil {
 		t.Fatal(err)
 	}
 	if err := c.Flush(); err != nil {
@@ -313,13 +322,13 @@ func TestCoalescerPSHFinalizesChain(t *testing.T) {
 	w := &fakeTunWriter{gsoEnabled: true}
 	c := NewTCPCoalescer(w)
 	pay := make([]byte, 1200)
-	if err := c.Add(buildTCPv4(1000, tcpAck, pay)); err != nil {
+	if err := c.Commit(buildTCPv4(1000, tcpAck, pay)); err != nil {
 		t.Fatal(err)
 	}
-	if err := c.Add(buildTCPv4(2200, tcpAckPsh, pay)); err != nil {
+	if err := c.Commit(buildTCPv4(2200, tcpAckPsh, pay)); err != nil {
 		t.Fatal(err)
 	}
-	if err := c.Add(buildTCPv4(3400, tcpAck, pay)); err != nil {
+	if err := c.Commit(buildTCPv4(3400, tcpAck, pay)); err != nil {
 		t.Fatal(err)
 	}
 	if err := c.Flush(); err != nil {
@@ -334,6 +343,39 @@ func TestCoalescerPSHFinalizesChain(t *testing.T) {
 	}
 }
 
+// TestCoalescerPropagatesPSHFromAppended ensures that when an appended
+// segment carries PSH (or is short, sealing the chain), the PSH bit ends
+// up in the emitted superpacket's TCP flags. The kernel TSO path keeps
+// PSH only on the last segment iff the input header has it set; if the
+// coalescer drops it the sender's push signal never reaches the receiver.
+func TestCoalescerPropagatesPSHFromAppended(t *testing.T) {
+	w := &fakeTunWriter{gsoEnabled: true}
+	c := NewTCPCoalescer(w)
+	pay := make([]byte, 1200)
+	// Seed has no PSH; second segment carries PSH and seals the chain.
+	if err := c.Commit(buildTCPv4(1000, tcpAck, pay)); err != nil {
+		t.Fatal(err)
+	}
+	if err := c.Commit(buildTCPv4(2200, tcpAckPsh, pay)); err != nil {
+		t.Fatal(err)
+	}
+	if err := c.Flush(0); err != nil {
+		t.Fatal(err)
+	}
+	if len(w.gsoWrites) != 1 {
+		t.Fatalf("want 1 gso write got %d", len(w.gsoWrites))
+	}
+	g := w.gsoWrites[0]
+	const ipHdrLen = 20
+	flags := g.hdr[ipHdrLen+13]
+	if flags&tcpPsh == 0 {
+		t.Fatalf("PSH lost from coalesced superpacket: flags=0x%02x", flags)
+	}
+	if flags&tcpAck == 0 {
+		t.Fatalf("ACK missing from coalesced superpacket: flags=0x%02x", flags)
+	}
+}
+
 func TestCoalescerRejectsDifferentFlow(t *testing.T) {
 	w := &fakeTunWriter{gsoEnabled: true}
 	c := NewTCPCoalescer(w)
@@ -341,10 +383,10 @@ func TestCoalescerRejectsDifferentFlow(t *testing.T) {
 	p1 := buildTCPv4(1000, tcpAck, pay)
 	p2 := buildTCPv4(2200, tcpAck, pay)
 	binary.BigEndian.PutUint16(p2[20:22], 9999)
-	if err := c.Add(p1); err != nil {
+	if err := c.Commit(p1); err != nil {
 		t.Fatal(err)
 	}
-	if err := c.Add(p2); err != nil {
+	if err := c.Commit(p2); err != nil {
 		t.Fatal(err)
 	}
 	if err := c.Flush(); err != nil {
@@ -364,7 +406,7 @@ func TestCoalescerRejectsIPOptions(t *testing.T) {
 	// Bump IHL to 6 to simulate 4 bytes of IP options. Don't actually add
 	// bytes — parser should bail before it matters.
 	pkt[0] = 0x46
-	if err := c.Add(pkt); err != nil {
+	if err := c.Commit(pkt); err != nil {
 		t.Fatal(err)
 	}
 	if err := c.Flush(); err != nil {
@@ -382,7 +424,7 @@ func TestCoalescerCapBySegments(t *testing.T) {
 	pay := make([]byte, 512)
 	seq := uint32(1000)
 	for i := 0; i < tcpCoalesceMaxSegs+5; i++ {
-		if err := c.Add(buildTCPv4(seq, tcpAck, pay)); err != nil {
+		if err := c.Commit(buildTCPv4(seq, tcpAck, pay)); err != nil {
 			t.Fatal(err)
 		}
 		seq += uint32(len(pay))
@@ -406,22 +448,22 @@ func TestCoalescerMultipleFlowsInSameBatch(t *testing.T) {
 	pay := make([]byte, 1200)
 
 	// Flow A: sport 1000. Flow B: sport 3000.
-	if err := c.Add(buildTCPv4Ports(1000, 2000, 100, tcpAck, pay)); err != nil {
+	if err := c.Commit(buildTCPv4Ports(1000, 2000, 100, tcpAck, pay)); err != nil {
 		t.Fatal(err)
 	}
-	if err := c.Add(buildTCPv4Ports(3000, 2000, 500, tcpAck, pay)); err != nil {
+	if err := c.Commit(buildTCPv4Ports(3000, 2000, 500, tcpAck, pay)); err != nil {
 		t.Fatal(err)
 	}
-	if err := c.Add(buildTCPv4Ports(1000, 2000, 1300, tcpAck, pay)); err != nil {
+	if err := c.Commit(buildTCPv4Ports(1000, 2000, 1300, tcpAck, pay)); err != nil {
 		t.Fatal(err)
 	}
-	if err := c.Add(buildTCPv4Ports(3000, 2000, 1700, tcpAck, pay)); err != nil {
+	if err := c.Commit(buildTCPv4Ports(3000, 2000, 1700, tcpAck, pay)); err != nil {
 		t.Fatal(err)
 	}
-	if err := c.Add(buildTCPv4Ports(1000, 2000, 2500, tcpAck, pay)); err != nil {
+	if err := c.Commit(buildTCPv4Ports(1000, 2000, 2500, tcpAck, pay)); err != nil {
 		t.Fatal(err)
 	}
-	if err := c.Add(buildTCPv4Ports(3000, 2000, 2900, tcpAck, pay)); err != nil {
+	if err := c.Commit(buildTCPv4Ports(3000, 2000, 2900, tcpAck, pay)); err != nil {
 		t.Fatal(err)
 	}
 	if err := c.Flush(); err != nil {
@@ -463,7 +505,7 @@ func TestCoalescerPreservesArrivalOrder(t *testing.T) {
 	// Sequence: coalesceable TCP, ICMP (passthrough), coalesceable TCP on
 	// a different flow. Expected emit order: gso(X), plain(ICMP), gso(Y).
 	pay := make([]byte, 1200)
-	if err := c.Add(buildTCPv4Ports(1000, 2000, 100, tcpAck, pay)); err != nil {
+	if err := c.Commit(buildTCPv4Ports(1000, 2000, 100, tcpAck, pay)); err != nil {
 		t.Fatal(err)
 	}
 	icmp := make([]byte, 28)
@@ -472,10 +514,10 @@ func TestCoalescerPreservesArrivalOrder(t *testing.T) {
 	icmp[9] = 1
 	copy(icmp[12:16], []byte{10, 0, 0, 1})
 	copy(icmp[16:20], []byte{10, 0, 0, 3})
-	if err := c.Add(icmp); err != nil {
+	if err := c.Commit(icmp); err != nil {
 		t.Fatal(err)
 	}
-	if err := c.Add(buildTCPv4Ports(3000, 2000, 500, tcpAck, pay)); err != nil {
+	if err := c.Commit(buildTCPv4Ports(3000, 2000, 500, tcpAck, pay)); err != nil {
 		t.Fatal(err)
 	}
 	// Nothing should have hit the writer synchronously.
@@ -502,7 +544,7 @@ func (w *orderedFakeWriter) Write(p []byte) (int, error) {
 	return len(p), nil
 }
 
-func (w *orderedFakeWriter) WriteGSO(hdr []byte, pays [][]byte, gsoSize uint16, isV6 bool, csumStart uint16) error {
+func (w *orderedFakeWriter) WriteGSO(hdr []byte, transportHdr []byte, pays [][]byte) error {
 	w.events = append(w.events, "gso")
 	return nil
 }
@@ -529,26 +571,26 @@ func TestCoalescerInterleavedFlowsPreserveOrdering(t *testing.T) {
 	pay := make([]byte, 1200)
 
 	// Flow A two segments.
-	if err := c.Add(buildTCPv4Ports(1000, 2000, 100, tcpAck, pay)); err != nil {
+	if err := c.Commit(buildTCPv4Ports(1000, 2000, 100, tcpAck, pay)); err != nil {
 		t.Fatal(err)
 	}
-	if err := c.Add(buildTCPv4Ports(1000, 2000, 1300, tcpAck, pay)); err != nil {
+	if err := c.Commit(buildTCPv4Ports(1000, 2000, 1300, tcpAck, pay)); err != nil {
 		t.Fatal(err)
 	}
 	// Flow B two segments.
-	if err := c.Add(buildTCPv4Ports(3000, 2000, 500, tcpAck, pay)); err != nil {
+	if err := c.Commit(buildTCPv4Ports(3000, 2000, 500, tcpAck, pay)); err != nil {
 		t.Fatal(err)
 	}
-	if err := c.Add(buildTCPv4Ports(3000, 2000, 1700, tcpAck, pay)); err != nil {
+	if err := c.Commit(buildTCPv4Ports(3000, 2000, 1700, tcpAck, pay)); err != nil {
 		t.Fatal(err)
 	}
 	// Flow A SYN (non-admissible) — must flush only flow A's slot.
 	syn := buildTCPv4Ports(1000, 2000, 9999, tcpSyn|tcpAck, pay)
-	if err := c.Add(syn); err != nil {
+	if err := c.Commit(syn); err != nil {
 		t.Fatal(err)
 	}
 	// Flow B continues — should still be coalesced with its seed.
-	if err := c.Add(buildTCPv4Ports(3000, 2000, 2900, tcpAck, pay)); err != nil {
+	if err := c.Commit(buildTCPv4Ports(3000, 2000, 2900, tcpAck, pay)); err != nil {
 		t.Fatal(err)
 	}
 	if err := c.Flush(); err != nil {

overlay/batch/tx_batch.go

@@ -0,0 +1,61 @@
+package batch
+
+import "net/netip"
+
+const SendBatchCap = 128
+
+// SendBatch accumulates encrypted UDP packets for potential TX offloading.
+// One SendBatch is owned by each listenIn goroutine; no locking is needed.
+// The backing storage holds up to batchCap packets of slotCap bytes each;
+// bufs and dsts are parallel slices of committed slots.
+type SendBatch struct {
+	bufs     [][]byte
+	dsts     []netip.AddrPort
+	backing  []byte
+	slotCap  int
+	batchCap int
+	nextSlot int
+}
+
+func NewSendBatch(batchCap, slotCap int) *SendBatch {
+	return &SendBatch{
+		bufs:     make([][]byte, 0, batchCap),
+		dsts:     make([]netip.AddrPort, 0, batchCap),
+		backing:  make([]byte, batchCap*slotCap),
+		slotCap:  slotCap,
+		batchCap: batchCap,
+	}
+}
+
+func (b *SendBatch) Next() []byte {
+	if b.nextSlot >= b.batchCap {
+		return nil
+	}
+	start := b.nextSlot * b.slotCap
+	return b.backing[start : start : start+b.slotCap] //set len to 0 but cap to slotCap
+}
+
+func (b *SendBatch) Commit(n int, dst netip.AddrPort) {
+	start := b.nextSlot * b.slotCap
+	b.bufs = append(b.bufs, b.backing[start:start+n])
+	b.dsts = append(b.dsts, dst)
+	b.nextSlot++
+}
+
+func (b *SendBatch) Reset() {
+	b.bufs = b.bufs[:0]
+	b.dsts = b.dsts[:0]
+	b.nextSlot = 0
+}
+
+func (b *SendBatch) Len() int {
+	return len(b.bufs)
+}
+
+func (b *SendBatch) Cap() int {
+	return b.batchCap
+}
+
+func (b *SendBatch) Get() ([][]byte, []netip.AddrPort) {
+	return b.bufs, b.dsts
+}

overlay/batch/tx_batch_test.go

@@ -1,4 +1,4 @@
-package nebula
+package batch
 
 import (
 	"net/netip"
@@ -6,7 +6,7 @@ import (
 )
 
 func TestSendBatchBookkeeping(t *testing.T) {
-	b := newSendBatch(4, 32)
+	b := NewSendBatch(4, 32)
 	if b.Len() != 0 || b.Cap() != 4 {
 		t.Fatalf("fresh batch: len=%d cap=%d", b.Len(), b.Cap())
 	}
@@ -51,7 +51,7 @@ func TestSendBatchBookkeeping(t *testing.T) {
 }
 
 func TestSendBatchSlotsDoNotOverlap(t *testing.T) {
-	b := newSendBatch(3, 8)
+	b := NewSendBatch(3, 8)
 	ap := netip.MustParseAddrPort("10.0.0.1:80")
 
 	// Fill three slots, each with its own sentinel byte.

overlay/overlaytest/noop.go

@@ -1,4 +1,6 @@
-package overlay
+// Package overlaytest provides fakes of overlay.Device for tests that do
+// not want to touch a real tun device or route table.
+package overlaytest
 
 import (
 	"errors"
@@ -8,6 +10,9 @@ import (
 	"github.com/slackhq/nebula/routing"
 )
 
+// NoopTun is an overlay.Device that silently discards every read and write.
+// Useful in tests that need to construct a nebula Interface but do not
+// exercise the datapath.
 type NoopTun struct{}
 
 func (NoopTun) RoutesFor(addr netip.Addr) routing.Gateways {
@@ -34,10 +39,6 @@ func (NoopTun) Write([]byte) (int, error) {
 	return 0, nil
 }
 
-func (NoopTun) WriteReject(p []byte) (int, error) {
-	return 0, nil
-}
-
 func (NoopTun) SupportsMultiqueue() bool {
 	return false
 }

overlay/route.go

@@ -2,6 +2,7 @@ package overlay
 
 import (
 	"fmt"
+	"log/slog"
 	"math"
 	"net"
 	"net/netip"
@@ -9,7 +10,6 @@ import (
 	"strconv"
 
 	"github.com/gaissmai/bart"
-	"github.com/sirupsen/logrus"
 	"github.com/slackhq/nebula/config"
 	"github.com/slackhq/nebula/routing"
 )
@@ -48,11 +48,14 @@ func (r Route) String() string {
 	return s
 }
 
-func makeRouteTree(l *logrus.Logger, routes []Route, allowMTU bool) (*bart.Table[routing.Gateways], error) {
+func makeRouteTree(l *slog.Logger, routes []Route, allowMTU bool) (*bart.Table[routing.Gateways], error) {
 	routeTree := new(bart.Table[routing.Gateways])
 	for _, r := range routes {
 		if !allowMTU && r.MTU > 0 {
-			l.WithField("route", r).Warnf("route MTU is not supported in %s", runtime.GOOS)
+			l.Warn("route MTU is not supported on this platform",
+				"goos", runtime.GOOS,
+				"route", r,
+			)
 		}
 
 		gateways := r.Via

overlay/route_test.go

@@ -295,7 +295,7 @@ func Test_makeRouteTree(t *testing.T) {
 	routes, err := parseUnsafeRoutes(c, []netip.Prefix{n})
 	require.NoError(t, err)
 	assert.Len(t, routes, 2)
-	routeTree, err := makeRouteTree(l, routes, true)
+	routeTree, err := makeRouteTree(test.NewLogger(), routes, true)
 	require.NoError(t, err)
 
 	ip, err := netip.ParseAddr("1.0.0.2")
@@ -367,7 +367,7 @@ func Test_makeMultipathUnsafeRouteTree(t *testing.T) {
 	routes, err := parseUnsafeRoutes(c, []netip.Prefix{n})
 	require.NoError(t, err)
 	assert.Len(t, routes, 3)
-	routeTree, err := makeRouteTree(l, routes, true)
+	routeTree, err := makeRouteTree(test.NewLogger(), routes, true)
 	require.NoError(t, err)
 
 	ip, err := netip.ParseAddr("192.168.86.1")

overlay/tio/tio.go

@@ -1,13 +1,19 @@
 package tio
 
-import "io"
+import (
+	"io"
+)
 
 // defaultBatchBufSize is the per-Queue scratch size for Read on backends
 // that don't do TSO segmentation. 65535 covers any single IP packet.
 const defaultBatchBufSize = 65535
 
+// Container holds one or many Queue objects and helps close them in an orderly way
 type Container interface {
+	io.Closer
 	Queues() []Queue
+
+	// Add takes a tun fd, adds it to the container, and prepares it for use as a Queue
 	Add(fd int) error
 
 	io.Closer
@@ -20,22 +26,13 @@ type Queue interface {
 
 	// Read returns one or more packets. The returned slices are borrowed
 	// from the Queue's internal buffer and are only valid until the next
-	// Read or Close on this Queue — callers must encrypt or copy each
+	// Read or Close on this Queue - callers must encrypt or copy each
-	// slice before the next call. Not safe for concurrent Reads; exactly
+	// slice before the next call. Not safe for concurrent Reads.
-	// one goroutine per Queue reads.
 	Read() ([][]byte, error)
 
 	// Write emits a single packet on the plaintext (outside→inside)
-	// delivery path. May run concurrently with WriteReject on the same
+	// delivery path. Not safe for concurrent Writes.
-	// Queue, but not with itself.
 	Write(p []byte) (int, error)
-
-	// WriteReject writes a single packet that originated from the inside
-	// path (reject replies or self-forward) using scratch state distinct
-	// from Write, so it can run concurrently with Write on the same Queue
-	// without a data race. On backends without a shared-scratch Write, a
-	// trivial delegation to Write is acceptable.
-	WriteReject(p []byte) (int, error)
 }
 
 // GSOWriter is implemented by Queues that can emit a TCP TSO superpacket
@@ -45,7 +42,7 @@ type Queue interface {
 // between the caller's decrypt buffer and the TUN. Backends without GSO
 // support return false from GSOSupported and coalescing is skipped.
 //
-// hdr contains the IPv4/IPv6 + TCP header prefix (mutable — callers will
+// hdr contains the IPv4/IPv6 + TCP header prefix (mutable - callers will
 // have filled in total length and pseudo-header partial). pays are
 // non-overlapping payload fragments whose concatenation is the full
 // superpacket payload; they are read-only from the writer's perspective
@@ -58,6 +55,13 @@ type Queue interface {
 // hdr's TCP checksum field must already hold the pseudo-header partial
 // sum (single-fold, not inverted), per virtio NEEDS_CSUM semantics.
 type GSOWriter interface {
-	WriteGSO(hdr []byte, pays [][]byte, gsoSize uint16, isV6 bool, csumStart uint16) error
+	// WriteGSO emits a TCP TSO superpacket in a single writev. hdr is the
+	// IPv4/IPv6 + TCP header prefix (already finalized — total length, IP csum,
+	// and TCP pseudo-header partial set by the caller). pays are payload
+	// fragments whose concatenation forms the full coalesced payload; each
+	// slice is read-only and must stay valid until return.
+	// every segment in pays except possibly the last is exactly the same size.
+	// csumStart is the byte offset where the TCP header begins within hdr.
+	WriteGSO(hdr []byte, transportHdr []byte, pays [][]byte) error
 	GSOSupported() bool
 }

overlay/tio/tio_gso_linux.go

@@ -4,6 +4,7 @@ import (
 	"fmt"
 	"io"
 	"os"
+	"sync"
 	"sync/atomic"
 	"syscall"
 	"unsafe"
@@ -32,19 +33,10 @@ const tunDrainCap = 64 //256
 // any reallocations.
 const gsoInitialPayIovs = 66
 
-// gsoWriteBufCap is the initial per-queue coalesce scratch capacity used by
-// WriteGSO to assemble [virtio_hdr || IP/TCP hdr || pays...] into a single
-// contiguous buffer so we can emit the superpacket via a single write()
-// instead of writev(). One worst-case TSO superpacket is bounded by the
-// virtio spec at 64KiB; 128KiB gives comfortable slack for the 10-byte
-// virtio header, the IP/TCP header, and any future size bumps. Grown on
-// demand if a superpacket exceeds this.
-const gsoWriteBufCap = tunSegBufSize
-
 // validVnetHdr is the 10-byte virtio_net_hdr we prepend to every non-GSO TUN
 // write. Only flag set is VIRTIO_NET_HDR_F_DATA_VALID, which marks the skb
-// CHECKSUM_UNNECESSARY so the receiving network stack skips L4 checksum
+// CHECKSUM_UNNECESSARY so the receiving network stack skips L4 checks
-// verification. All packets that reach the plain Write / WriteReject paths
+// verification. All packets that reach the plain Write paths
 // already carry a valid L4 checksum (either supplied by a remote peer whose
 // ciphertext we AEAD-authenticated, or produced by finishChecksum during TSO
 // segmentation, or built locally by CreateRejectPacket), so trusting them is
@@ -58,36 +50,21 @@ type Offload struct {
 	shutdownFd int
 	readPoll   [2]unix.PollFd
 	writePoll  [2]unix.PollFd
+	writeLock  sync.Mutex //there's more than one potential write source per-routine, so we need this to protect writePoll
 	closed     atomic.Bool
 	readBuf    []byte   // scratch for a single raw read (virtio hdr + superpacket)
 	segBuf     []byte   // backing store for segmented output
 	segOff     int      // cursor into segBuf for the current Read drain
 	pending    [][]byte // segments returned from the most recent Read
-	writeIovs  [2]unix.Iovec // preallocated iovecs for Write (coalescer passthrough); iovs[0] is fixed to validVnetHdr
-	// rejectIovs is a second preallocated iovec scratch used exclusively by
-	// WriteReject (reject + self-forward from the inside path). It mirrors
-	// writeIovs but lets listenIn goroutines emit reject packets without
-	// racing with the listenOut coalescer that owns writeIovs.
-	rejectIovs [2]unix.Iovec
 
 	// gsoHdrBuf is a per-queue 10-byte scratch for the virtio_net_hdr emitted
 	// by WriteGSO. Separate from validVnetHdr so a concurrent non-GSO Write on
 	// another queue never observes a half-written header.
 	gsoHdrBuf [virtioNetHdrLen]byte
-	// gsoIovs is a legacy writev iovec scratch. No longer used by the
+	// gsoIovs is the writev iovec scratch for WriteGSO. Sized to hold the
-	// WriteGSO path (which coalesces into gsoWriteBuf and uses a single
+	// virtio header + IP/TCP header + up to gsoInitialPayIovs payload
-	// write()) but retained for any other iovec-based path that may use it.
+	// fragments; grown on demand if a coalescer pushes more.
 	gsoIovs []unix.Iovec
-
-	// gsoWriteBuf is a per-queue scratch used by WriteGSO to coalesce the
-	// virtio_net_hdr + IP/TCP header + payload fragments into a single
-	// contiguous buffer, which is then written to the TUN fd with one
-	// write() syscall. This mirrors wireguard-go's approach and avoids
-	// triggering a kernel refcount use-after-free in skb_set_owner_w /
-	// sock_wfree observed on Linux 4.19 TUN when scatter-gather writev is
-	// combined with GSO-flagged virtio_net_hdr in the tun_chr_write_iter
-	// path. Grown on demand if a superpacket exceeds the initial cap.
-	gsoWriteBuf []byte
 }
 
 func newOffload(fd int, shutdownFd int) (*Offload, error) {
@@ -99,7 +76,7 @@ func newOffload(fd int, shutdownFd int) (*Offload, error) {
 		fd:         fd,
 		shutdownFd: shutdownFd,
 		closed:     atomic.Bool{},
-		readBuf:    make([]byte, tunReadBufSize),
+		readBuf:    make([]byte, virtioNetHdrLen+tunReadBufSize),
 		readPoll: [2]unix.PollFd{
 			{Fd: int32(fd), Events: unix.POLLIN},
 			{Fd: int32(shutdownFd), Events: unix.POLLIN},
@@ -108,16 +85,12 @@ func newOffload(fd int, shutdownFd int) (*Offload, error) {
 			{Fd: int32(fd), Events: unix.POLLOUT},
 			{Fd: int32(shutdownFd), Events: unix.POLLIN},
 		},
+		writeLock: sync.Mutex{},
 
 		segBuf:  make([]byte, tunSegBufCap),
 		gsoIovs: make([]unix.Iovec, 2, 2+gsoInitialPayIovs),
-		gsoWriteBuf: make([]byte, 0, gsoWriteBufCap),
 	}
 
-	out.writeIovs[0].Base = &validVnetHdr[0]
-	out.writeIovs[0].SetLen(virtioNetHdrLen)
-	out.rejectIovs[0].Base = &validVnetHdr[0]
-	out.rejectIovs[0].SetLen(virtioNetHdrLen)
 	out.gsoIovs[0].Base = &out.gsoHdrBuf[0]
 	out.gsoIovs[0].SetLen(virtioNetHdrLen)
 
@@ -160,10 +133,12 @@ func (r *Offload) blockOnWrite() error {
 		}
 	}
 	//always reset these!
+	r.writeLock.Lock()
 	tunEvents := r.writePoll[0].Revents
 	shutdownEvents := r.writePoll[1].Revents
 	r.writePoll[0].Revents = 0
 	r.writePoll[1].Revents = 0
+	r.writeLock.Unlock()
 	//do the err check before trusting the potentially bogus bits we just got
 	if err != nil {
 		return err
@@ -245,7 +220,6 @@ func (r *Offload) Read() ([][]byte, error) {
 
 // decodeRead decodes the virtio header plus payload in r.readBuf[:n], appends
 // the segments to r.pending, and advances r.segOff by the total scratch used.
-// Caller must have already ensured r.vnetHdr is true.
 func (r *Offload) decodeRead(n int) error {
 	if n < virtioNetHdrLen {
 		return fmt.Errorf("short tun read: %d < %d", n, virtioNetHdrLen)
@@ -263,16 +237,13 @@ func (r *Offload) decodeRead(n int) error {
 }
 
 func (r *Offload) Write(buf []byte) (int, error) {
-	return r.writeWithScratch(buf, &r.writeIovs)
+	iovs := [2]unix.Iovec{
+		{Base: &validVnetHdr[0]},
+		{Base: &buf[0]},
 	}
-
+	iovs[0].SetLen(virtioNetHdrLen)
-// WriteReject emits a packet using a dedicated iovec scratch (rejectIovs)
+	iovs[1].SetLen(len(buf))
-// distinct from the one used by the coalescer's Write path. This avoids a
+	return r.writeWithScratch(buf, &iovs)
-// data race between the inside (listenIn) goroutine emitting reject or
-// self-forward packets and the outside (listenOut) goroutine flushing TCP
-// coalescer passthroughs on the same Offload.
-func (r *Offload) WriteReject(buf []byte) (int, error) {
-	return r.writeWithScratch(buf, &r.rejectIovs)
 }
 
 func (r *Offload) writeWithScratch(buf []byte, iovs *[2]unix.Iovec) (int, error) {
@@ -311,83 +282,31 @@ func (r *Offload) rawWrite(iovs []unix.Iovec) (int, error) {
 	}
 }
 
-// rawWriteSingle writes buf to the TUN fd with a single write() syscall.
-// Unlike rawWrite (which uses writev), this avoids the kernel
-// scatter-gather path that triggers a use-after-free in
-// tun_chr_write_iter → sock_alloc_send_pskb → skb_set_owner_w on Linux
-// 4.19 TUN when the virtio_net_hdr requests TSO segmentation. The caller
-// is responsible for including the virtio_net_hdr prefix in buf.
-func (r *Offload) rawWriteSingle(buf []byte) (int, error) {
-	for {
-		n, err := unix.Write(r.fd, buf)
-		if err == nil {
-			if n < virtioNetHdrLen {
-				return 0, io.ErrShortWrite
-			}
-			return n - virtioNetHdrLen, nil
-		}
-		if err == unix.EAGAIN {
-			if werr := r.blockOnWrite(); werr != nil {
-				return 0, werr
-			}
-			continue
-		}
-		if err == unix.EINTR {
-			continue
-		}
-		if err == unix.EBADF {
-			return 0, os.ErrClosed
-		}
-		return 0, err
-	}
-}
-
 // GSOSupported reports whether this queue was opened with IFF_VNET_HDR and
 // can accept WriteGSO. When false, callers should fall back to per-segment
 // Write calls.
 func (r *Offload) GSOSupported() bool { return true }
 
-// WriteGSO emits a TCP TSO superpacket. hdr is the IPv4/IPv6 + TCP header
+func (r *Offload) WriteGSO(hdr []byte, transportHdr []byte, pays [][]byte) error {
-// prefix (already finalized — total length, IP csum, and TCP pseudo-header
+	if len(hdr) == 0 || len(pays) == 0 || len(transportHdr) == 0 {
-// partial set by the caller). pays are payload fragments whose concatenation
-// forms the full coalesced payload. gsoSize is the MSS; every segment except
-// possibly the last is exactly gsoSize bytes. csumStart is the byte offset
-// where the TCP header begins within hdr.
-//
-// Implementation note: this path coalesces [virtio_hdr || hdr || pays...]
-// into a single contiguous scratch buffer (r.gsoWriteBuf) and emits it via
-// one write() syscall rather than writev() with a scatter-gather iovec.
-// The scatter-gather path triggered a kernel-side use-after-free on Linux
-// 4.19 TUN where tun_chr_write_iter → sock_alloc_send_pskb →
-// skb_set_owner_w could be invoked with a zero sk_wmem_alloc, crashing
-// the router. The single-write path mirrors wireguard-go's design (see
-// golang.zx2c4.com/wireguard/tun/tun_linux.go Write — it always coalesces
-// GRO-merged data into a single contiguous buffer before calling
-// tunFile.Write) and has no equivalent failure mode.
-func (r *Offload) WriteGSO(hdr []byte, pays [][]byte, gsoSize uint16, isV6 bool, csumStart uint16) error {
-	if len(hdr) == 0 || len(pays) == 0 {
 		return nil
 	}
-
-	// Build the virtio_net_hdr. When pays total to <= gsoSize the kernel
-	// would produce a single segment; keep NEEDS_CSUM semantics but skip
-	// the GSO type so the kernel doesn't spuriously mark this as TSO.
 	vhdr := VirtioNetHdr{
 		Flags:      unix.VIRTIO_NET_HDR_F_NEEDS_CSUM,
-		HdrLen:     uint16(len(hdr)),
+		HdrLen:     uint16(len(hdr) + len(transportHdr)),
-		GSOSize:    gsoSize,
+		GSOSize:    uint16(len(pays[0])),
-		CsumStart:  csumStart,
+		CsumStart:  uint16(len(hdr)),
 		CsumOffset: 16, // TCP checksum field lives 16 bytes into the TCP header
 	}
-	var totalPay int
+	if len(pays) > 1 {
-	for _, p := range pays {
+		ipVer := hdr[0] >> 4
-		totalPay += len(p)
+		if ipVer == 6 {
-	}
-	if totalPay > int(gsoSize) {
-		if isV6 {
 			vhdr.GSOType = unix.VIRTIO_NET_HDR_GSO_TCPV6
-		} else {
+		} else if ipVer == 4 {
 			vhdr.GSOType = unix.VIRTIO_NET_HDR_GSO_TCPV4
+		} else {
+			vhdr.GSOType = unix.VIRTIO_NET_HDR_GSO_NONE
+			vhdr.GSOSize = 0
 		}
 	} else {
 		vhdr.GSOType = unix.VIRTIO_NET_HDR_GSO_NONE
@@ -395,26 +314,26 @@ func (r *Offload) WriteGSO(hdr []byte, pays [][]byte, gsoSize uint16, isV6 bool,
 	}
 	vhdr.encode(r.gsoHdrBuf[:])
 
-	// Coalesce [virtio_hdr || hdr || pays...] into a single contiguous
+	// Build the iovec array: [virtio_hdr, hdr, transportHdr, pays...]. r.gsoIovs[0] is
-	// buffer. This avoids the kernel scatter-gather write path entirely.
+	// wired to gsoHdrBuf at construction and never changes.
-	need := virtioNetHdrLen + len(hdr) + totalPay
+	need := 3 + len(pays)
-	if cap(r.gsoWriteBuf) < need {
+	if cap(r.gsoIovs) < need {
-		// Grow geometrically to amortize reallocs.
+		grown := make([]unix.Iovec, need)
-		newCap := cap(r.gsoWriteBuf) * 2
+		grown[0] = r.gsoIovs[0]
-		if newCap < need {
+		r.gsoIovs = grown
-			newCap = need
-		}
-		r.gsoWriteBuf = make([]byte, 0, newCap)
 	} else {
-		r.gsoWriteBuf = r.gsoWriteBuf[:0]
+		r.gsoIovs = r.gsoIovs[:need]
 	}
-	r.gsoWriteBuf = append(r.gsoWriteBuf, r.gsoHdrBuf[:]...)
+	r.gsoIovs[1].Base = &hdr[0]
-	r.gsoWriteBuf = append(r.gsoWriteBuf, hdr...)
+	r.gsoIovs[1].SetLen(len(hdr))
-	for _, p := range pays {
+	r.gsoIovs[2].Base = &transportHdr[0]
-		r.gsoWriteBuf = append(r.gsoWriteBuf, p...)
+	r.gsoIovs[2].SetLen(len(transportHdr))
+	for i, p := range pays {
+		r.gsoIovs[3+i].Base = &p[0]
+		r.gsoIovs[3+i].SetLen(len(p))
 	}
 
-	_, err := r.rawWriteSingle(r.gsoWriteBuf)
+	_, err := r.rawWrite(r.gsoIovs)
 	return err
 }
 

overlay/tio/tio_poll_linux.go

@@ -4,8 +4,6 @@ import (
 	"fmt"
 	"os"
 	"sync/atomic"
-	"syscall"
-	"unsafe"
 
 	"golang.org/x/sys/unix"
 )
@@ -100,9 +98,6 @@ func (t *Poll) blockOnWrite() error {
 }
 
 func (t *Poll) Read() ([][]byte, error) {
-	if t.readBuf == nil {
-		t.readBuf = make([]byte, defaultBatchBufSize)
-	}
 	n, err := t.readOne(t.readBuf)
 	if err != nil {
 		return nil, err
@@ -112,20 +107,10 @@ func (t *Poll) Read() ([][]byte, error) {
 }
 
 func (t *Poll) readOne(to []byte) (int, error) {
-	// first 4 bytes is protocol family, in network byte order
-	var head [4]byte
-	iovecs := [2]syscall.Iovec{ //todo plat-specific
-		{&head[0], 4},
-		{&to[0], uint64(len(to))},
-	}
 	for {
-		n, _, errno := syscall.Syscall(syscall.SYS_READV, uintptr(t.fd), uintptr(unsafe.Pointer(&iovecs[0])), 2)
+		n, errno := unix.Read(t.fd, to)
-		if errno == 0 {
+		if errno == nil {
-			bytesRead := int(n)
+			return n, nil
-			if bytesRead < 4 {
-				return 0, nil
-			}
-			return bytesRead - 4, nil
 		}
 		switch errno {
 		case unix.EAGAIN:
@@ -144,30 +129,10 @@ func (t *Poll) readOne(to []byte) (int, error) {
 
 // Write is only valid for single threaded use
 func (t *Poll) Write(from []byte) (int, error) {
-	if len(from) <= 1 {
-		return 0, syscall.EIO
-	}
-
-	ipVer := from[0] >> 4
-	var head [4]byte
-	// first 4 bytes is protocol family, in network byte order
-	switch ipVer {
-	case 4:
-		head[3] = syscall.AF_INET
-	case 6:
-		head[3] = syscall.AF_INET6
-	default:
-		return 0, fmt.Errorf("unable to determine IP version from packet")
-	}
-
-	iovecs := [2]syscall.Iovec{ //todo plat specific
-		{&head[0], 4},
-		{&from[0], uint64(len(from))},
-	}
 	for {
-		n, _, errno := syscall.Syscall(syscall.SYS_WRITEV, uintptr(t.fd), uintptr(unsafe.Pointer(&iovecs[0])), 2)
+		n, errno := unix.Write(t.fd, from)
-		if errno == 0 {
+		if errno == nil {
-			return int(n) - 4, nil
+			return n, nil
 		}
 		switch errno {
 		case unix.EAGAIN:
@@ -188,9 +153,7 @@ func (t *Poll) Close() error {
 	if t.closed.Swap(true) {
 		return nil
 	}
-
 	//shutdownFd is owned by the container, so we should not close it
-
 	var err error
 	if t.fd >= 0 {
 		err = unix.Close(t.fd)
@@ -199,7 +162,3 @@ func (t *Poll) Close() error {
 
 	return err
 }
-
-func (t *Poll) WriteReject(p []byte) (int, error) {
-	return t.Write(p)
-}

overlay/tio/tun_file_linux_test.go

@@ -26,13 +26,11 @@ func newReadPipe(t *testing.T) int {
 	return fds[0]
 }
 
-func TestOffload_WakeForShutdown_WakesFriends(t *testing.T) {
+func TestPoll_WakeForShutdown_WakesFriends(t *testing.T) {
 	pipe1 := newReadPipe(t)
 	pipe2 := newReadPipe(t)
-	parent, err := NewOffloadContainer()
+	parent, err := NewPollContainer()
-	if err != nil {
+	require.NoError(t, err)
-		t.Fatalf("newOffload: %v", err)
-	}
 	require.NoError(t, parent.Add(pipe1))
 	require.NoError(t, parent.Add(pipe2))
 	t.Cleanup(func() {
@@ -72,11 +70,9 @@ func TestOffload_WakeForShutdown_WakesFriends(t *testing.T) {
 	}
 }
 
-func TestTunFile_Close_Idempotent(t *testing.T) {
+func TestPoll_Close_Idempotent(t *testing.T) {
-	tf, err := newOffload(newReadPipe(t), 1)
+	tf, err := newPoll(newReadPipe(t), 1)
-	if err != nil {
+	require.NoError(t, err)
-		t.Fatalf("newOffload: %v", err)
-	}
 	if err := tf.Close(); err != nil {
 		t.Fatalf("first Close: %v", err)
 	}

overlay/tio/tun_linux_offload.go

@@ -5,6 +5,7 @@ package tio
 
 import (
 	"encoding/binary"
+	"errors"
 	"fmt"
 
 	"golang.org/x/sys/unix"
@@ -47,10 +48,7 @@ const (
 // tcpFinPshMask is cleared on every segment except the last of a TSO burst.
 const tcpFinPshMask = 0x09 // FIN(0x01) | PSH(0x08)
 
-// segmentInto splits a TUN-side packet described by hdr into one or more
+func checkVirtioValid(pkt []byte, hdr VirtioNetHdr) error {
-// IP packets, each appended to *out as a slice of scratch. scratch must be
-// sized to hold every segment (including replicated headers).
-func segmentInto(pkt []byte, hdr VirtioNetHdr, out *[][]byte, scratch []byte) error {
 	// When RSC_INFO is set the csum_start/csum_offset fields are repurposed to
 	// carry coalescing info rather than checksum offsets. A TUN writing via
 	// IFF_VNET_HDR should never emit this, but if it did we would silently
@@ -58,9 +56,29 @@ func segmentInto(pkt []byte, hdr VirtioNetHdr, out *[][]byte, scratch []byte) er
 	if hdr.Flags&unix.VIRTIO_NET_HDR_F_RSC_INFO != 0 {
 		return fmt.Errorf("virtio RSC_INFO flag not supported on TUN reads")
 	}
-
+	if len(pkt) < ipv4HeaderMinLen {
+		return fmt.Errorf("packet too short")
+	}
+	ipVersion := pkt[0] >> 4
 	switch hdr.GSOType {
-	case unix.VIRTIO_NET_HDR_GSO_NONE:
+	case unix.VIRTIO_NET_HDR_GSO_TCPV4:
+		if ipVersion != 4 {
+			return fmt.Errorf("invalid IP version %d for GSO type %d", ipVersion, hdr.GSOType)
+		}
+	case unix.VIRTIO_NET_HDR_GSO_TCPV6:
+		if ipVersion != 6 {
+			return fmt.Errorf("invalid IP version %d for GSO type %d", ipVersion, hdr.GSOType)
+		}
+	default:
+		if !(ipVersion == 6 || ipVersion == 4) {
+			return fmt.Errorf("invalid IP version %d for GSO type %d", ipVersion, hdr.GSOType)
+		}
+	}
+
+	return nil
+}
+
+func handleGSONone(pkt []byte, hdr VirtioNetHdr, out *[][]byte, scratch []byte) error {
 	if len(pkt) > len(scratch) {
 		return fmt.Errorf("packet larger than segment buffer: %d > %d", len(pkt), len(scratch))
 	}
@@ -73,7 +91,59 @@ func segmentInto(pkt []byte, hdr VirtioNetHdr, out *[][]byte, scratch []byte) er
 	}
 	*out = append(*out, seg)
 	return nil
+}
 
+func correctHdrLen(pkt []byte, hdr *VirtioNetHdr) error {
+	// Thank you wireguard-go for documenting these edge-cases
+	// Don't trust hdr.hdrLen from the kernel as it can be equal to the length
+	// of the entire first packet when the kernel is handling it as part of a
+	// FORWARD path. Instead, parse the transport header length and add it onto
+	// csumStart, which is synonymous for IP header length.
+	const tcpDataOffset = 12
+
+	if hdr.GSOType == unix.VIRTIO_NET_HDR_GSO_UDP_L4 {
+		hdr.HdrLen = hdr.CsumStart + 8
+	} else {
+		if len(pkt) <= int(hdr.CsumStart+tcpDataOffset) {
+			return errors.New("packet is too short")
+		}
+
+		tcpHLen := uint16(pkt[hdr.CsumStart+tcpDataOffset] >> 4 * 4)
+		if tcpHLen < 20 || tcpHLen > 60 {
+			// A TCP header must be between 20 and 60 bytes in length.
+			return fmt.Errorf("tcp header len is invalid: %d", tcpHLen)
+		}
+		hdr.HdrLen = hdr.CsumStart + tcpHLen
+	}
+
+	if len(pkt) < int(hdr.HdrLen) {
+		return fmt.Errorf("length of packet (%d) < virtioNetHdr.HdrLen (%d)", len(pkt), hdr.HdrLen)
+	}
+
+	if hdr.HdrLen < hdr.CsumStart {
+		return fmt.Errorf("virtioNetHdr.HdrLen (%d) < virtioNetHdr.CsumStart (%d)", hdr.HdrLen, hdr.CsumStart)
+	}
+	cSumAt := int(hdr.CsumStart + hdr.CsumStart)
+	if cSumAt+1 >= len(pkt) {
+		return fmt.Errorf("end of checksum offset (%d) exceeds packet length (%d)", cSumAt+1, len(pkt))
+	}
+	return nil
+}
+
+// segmentInto splits a TUN-side packet described by hdr into one or more
+// IP packets, each appended to *out as a slice of scratch. scratch must be
+// sized to hold every segment (including replicated headers).
+func segmentInto(pkt []byte, hdr VirtioNetHdr, out *[][]byte, scratch []byte) error {
+	if err := checkVirtioValid(pkt, hdr); err != nil {
+		return err
+	}
+	if hdr.GSOType == unix.VIRTIO_NET_HDR_GSO_NONE {
+		return handleGSONone(pkt, hdr, out, scratch)
+	}
+	if err := correctHdrLen(pkt, &hdr); err != nil {
+		return err
+	}
+	switch hdr.GSOType {
 	case unix.VIRTIO_NET_HDR_GSO_TCPV4, unix.VIRTIO_NET_HDR_GSO_TCPV6:
 		return segmentTCP(pkt, hdr, out, scratch)
 
@@ -118,35 +188,15 @@ func segmentTCP(pkt []byte, hdr VirtioNetHdr, out *[][]byte, scratch []byte) err
 	}
 
 	isV4 := hdr.GSOType == unix.VIRTIO_NET_HDR_GSO_TCPV4
+	headerLen := int(hdr.HdrLen) // already corrected by the caller
 	csumStart := int(hdr.CsumStart)
 
-	if isV4 && csumStart < ipv4HeaderMinLen {
-		return fmt.Errorf("csum_start %d too small for IPv4", csumStart)
-	}
-	if !isV4 && csumStart < ipv6FixedLen {
-		return fmt.Errorf("csum_start %d too small for IPv6", csumStart)
-	}
-
-	// Don't trust hdr.HdrLen from the kernel: on some paths it can be set
-	// to the full length of the first packet rather than the true L3+L4 header length.
-	// Instead, read the TCP data-offset field from the packet itself and derive
-	// headerLen = csum_start + tcpHdrLen. Matches wireguard-go's approach.
-	if csumStart+tcpFlagsOff+1 > len(pkt) {
-		return fmt.Errorf("packet too short for tcp header at csum_start=%d (pkt %d)", csumStart, len(pkt))
-	}
 	tcpHdrLen := int(pkt[csumStart+tcpDataOffOff]>>4) * 4
-	if tcpHdrLen < tcpHeaderMinLen || tcpHdrLen > tcpHeaderMaxLen {
-		return fmt.Errorf("tcp data-offset out of range: %d", tcpHdrLen)
-	}
-	headerLen := csumStart + tcpHdrLen
-	if headerLen > len(pkt) {
-		return fmt.Errorf("derived hdr_len %d > pkt %d", headerLen, len(pkt))
-	}
 
 	payload := pkt[headerLen:]
 	payLen := len(payload)
-	gso := int(hdr.GSOSize)
+	gsoSize := int(hdr.GSOSize)
-	numSeg := (payLen + gso - 1) / gso
+	numSeg := (payLen + gsoSize - 1) / gsoSize
 	if numSeg == 0 {
 		numSeg = 1
 	}
@@ -197,8 +247,8 @@ func segmentTCP(pkt []byte, hdr VirtioNetHdr, out *[][]byte, scratch []byte) err
 
 	off := 0
 	for i := 0; i < numSeg; i++ {
-		segStart := i * gso
+		segStart := i * gsoSize
-		segEnd := segStart + gso
+		segEnd := segStart + gsoSize
 		if segEnd > payLen {
 			segEnd = payLen
 		}

overlay/tio/tun_linux_offload_test.go

@@ -310,8 +310,6 @@ func TestTunFileWriteVnetHdrNoAlloc(t *testing.T) {
 	t.Cleanup(func() { _ = unix.Close(fd) })
 
 	tf := &Offload{fd: fd}
-	tf.writeIovs[0].Base = &validVnetHdr[0]
-	tf.writeIovs[0].SetLen(virtioNetHdrLen)
 
 	payload := make([]byte, 1400)
 	// Warm up (first call may trigger one-time internal allocations elsewhere).

overlay/tun.go

@@ -2,10 +2,10 @@ package overlay
 
 import (
 	"fmt"
+	"log/slog"
 	"net"
 	"net/netip"
 
-	"github.com/sirupsen/logrus"
 	"github.com/slackhq/nebula/config"
 	"github.com/slackhq/nebula/util"
 )
@@ -22,9 +22,9 @@ func (e *NameError) Error() string {
 }
 
 // TODO: We may be able to remove routines
-type DeviceFactory func(c *config.C, l *logrus.Logger, vpnNetworks []netip.Prefix, routines int) (Device, error)
+type DeviceFactory func(c *config.C, l *slog.Logger, vpnNetworks []netip.Prefix, routines int) (Device, error)
 
-func NewDeviceFromConfig(c *config.C, l *logrus.Logger, vpnNetworks []netip.Prefix, routines int) (Device, error) {
+func NewDeviceFromConfig(c *config.C, l *slog.Logger, vpnNetworks []netip.Prefix, routines int) (Device, error) {
 	switch {
 	case c.GetBool("tun.disabled", false):
 		tun := newDisabledTun(vpnNetworks, c.GetInt("tun.tx_queue", 500), c.GetBool("stats.message_metrics", false), l)
@@ -36,7 +36,7 @@ func NewDeviceFromConfig(c *config.C, l *logrus.Logger, vpnNetworks []netip.Pref
 }
 
 func NewFdDeviceFromConfig(fd *int) DeviceFactory {
-	return func(c *config.C, l *logrus.Logger, vpnNetworks []netip.Prefix, routines int) (Device, error) {
+	return func(c *config.C, l *slog.Logger, vpnNetworks []netip.Prefix, routines int) (Device, error) {
 		return newTunFromFd(c, l, *fd, vpnNetworks)
 	}
 }

overlay/tun_android.go

@@ -6,12 +6,12 @@ package overlay
 import (
 	"fmt"
 	"io"
+	"log/slog"
 	"net/netip"
 	"os"
 	"sync/atomic"
 
 	"github.com/gaissmai/bart"
-	"github.com/sirupsen/logrus"
 	"github.com/slackhq/nebula/config"
 	"github.com/slackhq/nebula/overlay/tio"
 	"github.com/slackhq/nebula/routing"
@@ -24,16 +24,13 @@ type tun struct {
 	vpnNetworks []netip.Prefix
 	Routes      atomic.Pointer[[]Route]
 	routeTree   atomic.Pointer[bart.Table[routing.Gateways]]
-	l           *logrus.Logger
+	l           *slog.Logger
 
 	readBuf  []byte
 	batchRet [1][]byte
 }
 
 func (t *tun) Read() ([][]byte, error) {
-	if t.readBuf == nil {
-		t.readBuf = make([]byte, defaultBatchBufSize)
-	}
 	n, err := t.rwc.Read(t.readBuf)
 	if err != nil {
 		return nil, err
@@ -46,15 +43,11 @@ func (t *tun) Write(p []byte) (int, error) {
 	return t.rwc.Write(p)
 }
 
-func (t *tun) WriteReject(p []byte) (int, error) {
-	return t.rwc.Write(p)
-}
-
 func (t *tun) Close() error {
 	return t.rwc.Close()
 }
 
-func newTunFromFd(c *config.C, l *logrus.Logger, deviceFd int, vpnNetworks []netip.Prefix) (*tun, error) {
+func newTunFromFd(c *config.C, l *slog.Logger, deviceFd int, vpnNetworks []netip.Prefix) (*tun, error) {
 	// XXX Android returns an fd in non-blocking mode which is necessary for shutdown to work properly.
 	// Be sure not to call file.Fd() as it will set the fd to blocking mode.
 	file := os.NewFile(uintptr(deviceFd), "/dev/net/tun")
@@ -64,6 +57,7 @@ func newTunFromFd(c *config.C, l *logrus.Logger, deviceFd int, vpnNetworks []net
 		fd:          deviceFd,
 		vpnNetworks: vpnNetworks,
 		l:           l,
+		readBuf:     make([]byte, defaultBatchBufSize),
 	}
 
 	err := t.reload(c, true)
@@ -81,7 +75,7 @@ func newTunFromFd(c *config.C, l *logrus.Logger, deviceFd int, vpnNetworks []net
 	return t, nil
 }
 
-func newTun(_ *config.C, _ *logrus.Logger, _ []netip.Prefix, _ bool) (*tun, error) {
+func newTun(_ *config.C, _ *slog.Logger, _ []netip.Prefix, _ bool) (*tun, error) {
 	return nil, fmt.Errorf("newTun not supported in Android")
 }
 
@@ -90,7 +84,7 @@ func (t *tun) RoutesFor(ip netip.Addr) routing.Gateways {
 	return r
 }
 
-func (t tun) Activate() error {
+func (t *tun) Activate() error {
 	return nil
 }
 
@@ -127,6 +121,10 @@ func (t *tun) SupportsMultiqueue() bool {
 	return false
 }
 
-func (t *tun) NewMultiQueueReader() (tio.Queue, error) {
+func (t *tun) NewMultiQueueReader() error {
-	return nil, fmt.Errorf("TODO: multiqueue not implemented for android")
+	return fmt.Errorf("TODO: multiqueue not implemented for android")
+}
+
+func (t *tun) Readers() []tio.Queue {
+	return []tio.Queue{t}
 }

overlay/tun_darwin.go

@@ -7,6 +7,7 @@ import (
 	"errors"
 	"fmt"
 	"io"
+	"log/slog"
 	"net/netip"
 	"os"
 	"sync/atomic"
@@ -14,7 +15,6 @@ import (
 	"unsafe"
 
 	"github.com/gaissmai/bart"
-	"github.com/sirupsen/logrus"
 	"github.com/slackhq/nebula/config"
 	"github.com/slackhq/nebula/overlay/tio"
 	"github.com/slackhq/nebula/routing"
@@ -31,7 +31,7 @@ type tun struct {
 	Routes      atomic.Pointer[[]Route]
 	routeTree   atomic.Pointer[bart.Table[routing.Gateways]]
 	linkAddr    *netroute.LinkAddr
-	l           *logrus.Logger
+	l           *slog.Logger
 
 	// cache out buffer since we need to prepend 4 bytes for tun metadata
 	out []byte
@@ -83,7 +83,7 @@ type ifreqAlias6 struct {
 	Lifetime   addrLifetime
 }
 
-func newTun(c *config.C, l *logrus.Logger, vpnNetworks []netip.Prefix, _ bool) (*tun, error) {
+func newTun(c *config.C, l *slog.Logger, vpnNetworks []netip.Prefix, _ bool) (*tun, error) {
 	name := c.GetString("tun.dev", "")
 	ifIndex := -1
 	if name != "" && name != "utun" {
@@ -133,6 +133,7 @@ func newTun(c *config.C, l *logrus.Logger, vpnNetworks []netip.Prefix, _ bool) (
 		vpnNetworks: vpnNetworks,
 		DefaultMTU:  c.GetInt("tun.mtu", DefaultMTU),
 		l:           l,
+		readBuf:     make([]byte, defaultBatchBufSize),
 	}
 
 	err = t.reload(c, true)
@@ -157,7 +158,7 @@ func (t *tun) deviceBytes() (o [16]byte) {
 	return
 }
 
-func newTunFromFd(_ *config.C, _ *logrus.Logger, _ int, _ []netip.Prefix) (*tun, error) {
+func newTunFromFd(_ *config.C, _ *slog.Logger, _ int, _ []netip.Prefix) (*tun, error) {
 	return nil, fmt.Errorf("newTunFromFd not supported in Darwin")
 }
 
@@ -393,8 +394,7 @@ func (t *tun) addRoutes(logErrors bool) error {
 		err := addRoute(r.Cidr, t.linkAddr)
 		if err != nil {
 			if errors.Is(err, unix.EEXIST) {
-				t.l.WithField("route", r.Cidr).
+				t.l.Warn("unable to add unsafe_route, identical route already exists", "route", r.Cidr)
-					Warnf("unable to add unsafe_route, identical route already exists")
 			} else {
 				retErr := util.NewContextualError("Failed to add route", map[string]any{"route": r}, err)
 				if logErrors {
@@ -404,7 +404,7 @@ func (t *tun) addRoutes(logErrors bool) error {
 				}
 			}
 		} else {
-			t.l.WithField("route", r).Info("Added route")
+			t.l.Info("Added route", "route", r)
 		}
 	}
 
@@ -419,9 +419,9 @@ func (t *tun) removeRoutes(routes []Route) error {
 
 		err := delRoute(r.Cidr, t.linkAddr)
 		if err != nil {
-			t.l.WithError(err).WithField("route", r).Error("Failed to remove route")
+			t.l.Error("Failed to remove route", "error", err, "route", r)
 		} else {
-			t.l.WithField("route", r).Info("Removed route")
+			t.l.Info("Removed route", "route", r)
 		}
 	}
 	return nil
@@ -517,9 +517,6 @@ func (t *tun) readOne(to []byte) (int, error) {
 }
 
 func (t *tun) Read() ([][]byte, error) {
-	if t.readBuf == nil {
-		t.readBuf = make([]byte, defaultBatchBufSize)
-	}
 	n, err := t.readOne(t.readBuf)
 	if err != nil {
 		return nil, err
@@ -528,10 +525,6 @@ func (t *tun) Read() ([][]byte, error) {
 	return t.batchRet[:], nil
 }
 
-func (t *tun) WriteReject(p []byte) (int, error) {
-	return t.Write(p)
-}
-
 // Write is only valid for single threaded use
 func (t *tun) Write(from []byte) (int, error) {
 	buf := t.out
@@ -573,6 +566,10 @@ func (t *tun) SupportsMultiqueue() bool {
 	return false
 }
 
-func (t *tun) NewMultiQueueReader() (tio.Queue, error) {
+func (t *tun) NewMultiQueueReader() error {
-	return nil, fmt.Errorf("TODO: multiqueue not implemented for darwin")
+	return fmt.Errorf("TODO: multiqueue not implemented for darwin")
+}
+
+func (t *tun) Readers() []tio.Queue {
+	return []tio.Queue{t}
 }

overlay/tun_disabled.go

@@ -1,13 +1,14 @@
 package overlay
 
 import (
+	"context"
 	"fmt"
 	"io"
+	"log/slog"
 	"net/netip"
 	"strings"
 
 	"github.com/rcrowley/go-metrics"
-	"github.com/sirupsen/logrus"
 	"github.com/slackhq/nebula/iputil"
 	"github.com/slackhq/nebula/overlay/tio"
 	"github.com/slackhq/nebula/routing"
@@ -20,7 +21,7 @@ type disabledTun struct {
 	// Track these metrics since we don't have the tun device to do it for us
 	tx         metrics.Counter
 	rx         metrics.Counter
-	l          *logrus.Logger
+	l          *slog.Logger
 	numReaders int
 
 	batchRet [1][]byte
@@ -33,15 +34,15 @@ func (t *disabledTun) Read() ([][]byte, error) {
 	}
 
 	t.tx.Inc(1)
-	if t.l.Level >= logrus.DebugLevel {
+	if t.l.Enabled(context.Background(), slog.LevelDebug) {
-		t.l.WithField("raw", prettyPacket(r)).Debugf("Write payload")
+		t.l.Debug("Write payload", "raw", prettyPacket(r))
 	}
 
 	t.batchRet[0] = r
 	return t.batchRet[:], nil
 }
 
-func newDisabledTun(vpnNetworks []netip.Prefix, queueLen int, metricsEnabled bool, l *logrus.Logger) *disabledTun {
+func newDisabledTun(vpnNetworks []netip.Prefix, queueLen int, metricsEnabled bool, l *slog.Logger) *disabledTun {
 	tun := &disabledTun{
 		vpnNetworks: vpnNetworks,
 		read:        make(chan []byte, queueLen),
@@ -87,7 +88,7 @@ func (t *disabledTun) handleICMPEchoRequest(b []byte) bool {
 	select {
 	case t.read <- out:
 	default:
-		t.l.Debugf("tun_disabled: dropped ICMP Echo Reply response")
+		t.l.Debug("tun_disabled: dropped ICMP Echo Reply response")
 	}
 
 	return true
@@ -98,19 +99,15 @@ func (t *disabledTun) Write(b []byte) (int, error) {
 
 	// Check for ICMP Echo Request before spending time doing the full parsing
 	if t.handleICMPEchoRequest(b) {
-		if t.l.Level >= logrus.DebugLevel {
+		if t.l.Enabled(context.Background(), slog.LevelDebug) {
-			t.l.WithField("raw", prettyPacket(b)).Debugf("Disabled tun responded to ICMP Echo Request")
+			t.l.Debug("Disabled tun responded to ICMP Echo Request", "raw", prettyPacket(b))
 		}
-	} else if t.l.Level >= logrus.DebugLevel {
+	} else if t.l.Enabled(context.Background(), slog.LevelDebug) {
-		t.l.WithField("raw", prettyPacket(b)).Debugf("Disabled tun received unexpected payload")
+		t.l.Debug("Disabled tun received unexpected payload", "raw", prettyPacket(b))
 	}
 	return len(b), nil
 }
 
-func (t *disabledTun) WriteReject(b []byte) (int, error) {
-	return t.Write(b)
-}
-
 func (t *disabledTun) SupportsMultiqueue() bool {
 	return true
 }

overlay/tun_freebsd.go

@@ -8,6 +8,7 @@ import (
 	"errors"
 	"fmt"
 	"io/fs"
+	"log/slog"
 	"net/netip"
 	"os"
 	"sync/atomic"
@@ -16,7 +17,7 @@ import (
 	"unsafe"
 
 	"github.com/gaissmai/bart"
-	"github.com/sirupsen/logrus"
+
 	"github.com/slackhq/nebula/config"
 	"github.com/slackhq/nebula/overlay/tio"
 	"github.com/slackhq/nebula/routing"
@@ -93,7 +94,7 @@ type tun struct {
 	Routes      atomic.Pointer[[]Route]
 	routeTree   atomic.Pointer[bart.Table[routing.Gateways]]
 	linkAddr    *netroute.LinkAddr
-	l           *logrus.Logger
+	l           *slog.Logger
 
 	fd        int
 	shutdownR int // read end of the shutdown pipe; closing the write end wakes blocked polls
@@ -159,9 +160,6 @@ func (t *tun) blockOnWrite() error {
 }
 
 func (t *tun) Read() ([][]byte, error) {
-	if t.readBuf == nil {
-		t.readBuf = make([]byte, defaultBatchBufSize)
-	}
 	n, err := t.readOne(t.readBuf)
 	if err != nil {
 		return nil, err
@@ -170,10 +168,6 @@ func (t *tun) Read() ([][]byte, error) {
 	return t.batchRet[:], nil
 }
 
-func (t *tun) WriteReject(p []byte) (int, error) {
-	return t.Write(p)
-}
-
 func (t *tun) readOne(to []byte) (int, error) {
 	// first 4 bytes is protocol family, in network byte order
 	var head [4]byte
@@ -262,7 +256,7 @@ func (t *tun) Close() error {
 
 	if t.fd >= 0 {
 		if err := unix.Close(t.fd); err != nil {
-			t.l.WithError(err).Error("Error closing device")
+			t.l.Error("Error closing device", "error", err)
 		}
 		t.fd = -1
 	}
@@ -283,7 +277,7 @@ func (t *tun) Close() error {
 			err = ioctl(uintptr(s), syscall.SIOCIFDESTROY, uintptr(unsafe.Pointer(&ifreq)))
 		}
 		if err != nil {
-			t.l.WithError(err).Error("Error destroying tunnel")
+			t.l.Error("Error destroying tunnel", "error", err)
 		}
 	}()
 
@@ -296,11 +290,11 @@ func (t *tun) Close() error {
 	return nil
 }
 
-func newTunFromFd(_ *config.C, _ *logrus.Logger, _ int, _ []netip.Prefix) (*tun, error) {
+func newTunFromFd(_ *config.C, _ *slog.Logger, _ int, _ []netip.Prefix) (*tun, error) {
 	return nil, fmt.Errorf("newTunFromFd not supported in FreeBSD")
 }
 
-func newTun(c *config.C, l *logrus.Logger, vpnNetworks []netip.Prefix, _ bool) (*tun, error) {
+func newTun(c *config.C, l *slog.Logger, vpnNetworks []netip.Prefix, _ bool) (*tun, error) {
 	// Try to open existing tun device
 	var fd int
 	var err error
@@ -392,6 +386,7 @@ func newTun(c *config.C, l *logrus.Logger, vpnNetworks []netip.Prefix, _ bool) (
 		MTU:         c.GetInt("tun.mtu", DefaultMTU),
 		l:           l,
 		fd:          fd,
+		readBuf:     make([]byte, defaultBatchBufSize),
 		shutdownR:   shutdownR,
 		shutdownW:   shutdownW,
 		readPoll: [2]unix.PollFd{
@@ -582,8 +577,8 @@ func (t *tun) SupportsMultiqueue() bool {
 	return false
 }
 
-func (t *tun) NewMultiQueueReader() (tio.Queue, error) {
+func (t *tun) NewMultiQueueReader() error {
-	return nil, fmt.Errorf("TODO: multiqueue not implemented for freebsd")
+	return fmt.Errorf("TODO: multiqueue not implemented for freebsd")
 }
 
 func (t *tun) addRoutes(logErrors bool) error {
@@ -603,13 +598,17 @@ func (t *tun) addRoutes(logErrors bool) error {
 				return retErr
 			}
 		} else {
-			t.l.WithField("route", r).Info("Added route")
+			t.l.Info("Added route", "route", r)
 		}
 	}
 
 	return nil
 }
 
+func (t *tun) Readers() []tio.Queue {
+	return []tio.Queue{t}
+}
+
 func (t *tun) removeRoutes(routes []Route) error {
 	for _, r := range routes {
 		if !r.Install {
@@ -618,9 +617,9 @@ func (t *tun) removeRoutes(routes []Route) error {
 
 		err := delRoute(r.Cidr, t.linkAddr)
 		if err != nil {
-			t.l.WithError(err).WithField("route", r).Error("Failed to remove route")
+			t.l.Error("Failed to remove route", "error", err, "route", r)
 		} else {
-			t.l.WithField("route", r).Info("Removed route")
+			t.l.Info("Removed route", "route", r)
 		}
 	}
 	return nil

overlay/tun_ios.go

@@ -7,6 +7,7 @@ import (
 	"errors"
 	"fmt"
 	"io"
+	"log/slog"
 	"net/netip"
 	"os"
 	"sync"
@@ -14,7 +15,6 @@ import (
 	"syscall"
 
 	"github.com/gaissmai/bart"
-	"github.com/sirupsen/logrus"
 	"github.com/slackhq/nebula/config"
 	"github.com/slackhq/nebula/overlay/tio"
 	"github.com/slackhq/nebula/routing"
@@ -26,16 +26,13 @@ type tun struct {
 	vpnNetworks []netip.Prefix
 	Routes      atomic.Pointer[[]Route]
 	routeTree   atomic.Pointer[bart.Table[routing.Gateways]]
-	l           *logrus.Logger
+	l           *slog.Logger
 
 	readBuf  []byte
 	batchRet [1][]byte
 }
 
 func (t *tun) Read() ([][]byte, error) {
-	if t.readBuf == nil {
-		t.readBuf = make([]byte, defaultBatchBufSize)
-	}
 	n, err := t.rwc.Read(t.readBuf)
 	if err != nil {
 		return nil, err
@@ -48,24 +45,21 @@ func (t *tun) Write(p []byte) (int, error) {
 	return t.rwc.Write(p)
 }
 
-func (t *tun) WriteReject(p []byte) (int, error) {
-	return t.rwc.Write(p)
-}
-
 func (t *tun) Close() error {
 	return t.rwc.Close()
 }
 
-func newTun(_ *config.C, _ *logrus.Logger, _ []netip.Prefix, _ bool) (*tun, error) {
+func newTun(_ *config.C, _ *slog.Logger, _ []netip.Prefix, _ bool) (*tun, error) {
 	return nil, fmt.Errorf("newTun not supported in iOS")
 }
 
-func newTunFromFd(c *config.C, l *logrus.Logger, deviceFd int, vpnNetworks []netip.Prefix) (*tun, error) {
+func newTunFromFd(c *config.C, l *slog.Logger, deviceFd int, vpnNetworks []netip.Prefix) (*tun, error) {
 	file := os.NewFile(uintptr(deviceFd), "/dev/tun")
 	t := &tun{
 		vpnNetworks: vpnNetworks,
 		rwc:         &tunReadCloser{f: file},
 		l:           l,
+		readBuf:     make([]byte, defaultBatchBufSize),
 	}
 
 	err := t.reload(c, true)
@@ -183,6 +177,10 @@ func (t *tun) SupportsMultiqueue() bool {
 	return false
 }
 
-func (t *tun) NewMultiQueueReader() (tio.Queue, error) {
+func (t *tun) NewMultiQueueReader() error {
-	return nil, fmt.Errorf("TODO: multiqueue not implemented for ios")
+	return fmt.Errorf("TODO: multiqueue not implemented for ios")
+}
+
+func (t *tun) Readers() []tio.Queue {
+	return []tio.Queue{t}
 }

overlay/tun_linux.go

@@ -5,6 +5,7 @@ package overlay
 
 import (
 	"fmt"
+	"log/slog"
 	"net"
 	"net/netip"
 	"os"
@@ -15,7 +16,6 @@ import (
 	"unsafe"
 
 	"github.com/gaissmai/bart"
-	"github.com/sirupsen/logrus"
 	"github.com/slackhq/nebula/config"
 	"github.com/slackhq/nebula/overlay/tio"
 	"github.com/slackhq/nebula/routing"
@@ -47,7 +47,7 @@ type tun struct {
 	routesFromSystem     map[netip.Prefix]routing.Gateways
 	routesFromSystemLock sync.Mutex
 
-	l *logrus.Logger
+	l *slog.Logger
 }
 
 func (t *tun) Networks() []netip.Prefix {
@@ -72,7 +72,7 @@ type ifreqQLEN struct {
 	pad   [8]byte
 }
 
-func newTunFromFd(c *config.C, l *logrus.Logger, deviceFd int, vpnNetworks []netip.Prefix) (*tun, error) {
+func newTunFromFd(c *config.C, l *slog.Logger, deviceFd int, vpnNetworks []netip.Prefix) (*tun, error) {
 	// We don't know what flags the caller opened this fd with and can't turn
 	// on IFF_VNET_HDR after TUNSETIFF, so skip offload on inherited fds.
 	t, err := newTunGeneric(c, l, deviceFd, false, vpnNetworks)
@@ -124,7 +124,7 @@ func tunSetIff(fd int, name string, flags uint16) (string, error) {
 // TSO-capable TUN is available. CSUM is required as a prerequisite for TSO.
 const tsoOffloadFlags = unix.TUN_F_CSUM | unix.TUN_F_TSO4 | unix.TUN_F_TSO6
 
-func newTun(c *config.C, l *logrus.Logger, vpnNetworks []netip.Prefix, multiqueue bool) (*tun, error) {
+func newTun(c *config.C, l *slog.Logger, vpnNetworks []netip.Prefix, multiqueue bool) (*tun, error) {
 	baseFlags := uint16(unix.IFF_TUN | unix.IFF_NO_PI)
 	if multiqueue {
 		baseFlags |= unix.IFF_MULTI_QUEUE
@@ -139,12 +139,12 @@ func newTun(c *config.C, l *logrus.Logger, vpnNetworks []netip.Prefix, multiqueu
 		return nil, err
 	}
 	vnetHdr := true
-	name, err := tunSetIff(fd, nameStr, baseFlags|unix.IFF_VNET_HDR|unix.IFF_NAPI)
+	name, err := tunSetIff(fd, nameStr, baseFlags|unix.IFF_VNET_HDR)
 	if err != nil {
 		_ = unix.Close(fd)
 		vnetHdr = false
 	} else if err = ioctl(uintptr(fd), unix.TUNSETOFFLOAD, uintptr(tsoOffloadFlags)); err != nil {
-		l.WithError(err).Warn("Failed to enable TUN offload (TSO); proceeding without virtio headers")
+		l.Warn("Failed to enable TUN offload (TSO); proceeding without virtio headers", "error", err)
 		_ = unix.Close(fd)
 		vnetHdr = false
 	}
@@ -172,7 +172,7 @@ func newTun(c *config.C, l *logrus.Logger, vpnNetworks []netip.Prefix, multiqueu
 }
 
 // newTunGeneric does all the stuff common to different tun initialization paths. It will close your files on error.
-func newTunGeneric(c *config.C, l *logrus.Logger, fd int, vnetHdr bool, vpnNetworks []netip.Prefix) (*tun, error) {
+func newTunGeneric(c *config.C, l *slog.Logger, fd int, vnetHdr bool, vpnNetworks []netip.Prefix) (*tun, error) {
 	var container tio.Container
 	var err error
 	if vnetHdr {
@@ -264,16 +264,16 @@ func (t *tun) reload(c *config.C, initial bool) error {
 	if !initial {
 		if oldMaxMTU != newMaxMTU {
 			t.setMTU()
-			t.l.Infof("Set max MTU to %v was %v", t.MaxMTU, oldMaxMTU)
+			t.l.Info("Set max MTU", "mtu", t.MaxMTU, "oldMTU", oldMaxMTU)
 		}
 
 		if oldDefaultMTU != newDefaultMTU {
 			for i := range t.vpnNetworks {
 				err := t.setDefaultRoute(t.vpnNetworks[i])
 				if err != nil {
-					t.l.Warn(err)
+					t.l.Warn(err.Error())
 				} else {
-					t.l.Infof("Set default MTU to %v was %v", t.DefaultMTU, oldDefaultMTU)
+					t.l.Info("Set default MTU", "mtu", t.DefaultMTU, "oldMTU", oldDefaultMTU)
 				}
 			}
 		}
@@ -307,7 +307,7 @@ func (t *tun) NewMultiQueueReader() error {
 
 	flags := uint16(unix.IFF_TUN | unix.IFF_NO_PI | unix.IFF_MULTI_QUEUE)
 	if t.vnetHdr {
-		flags |= unix.IFF_VNET_HDR | unix.IFF_NAPI
+		flags |= unix.IFF_VNET_HDR
 	}
 	if _, err = tunSetIff(fd, t.Device, flags); err != nil {
 		_ = unix.Close(fd)
@@ -384,9 +384,9 @@ func (t *tun) addIPs(link netlink.Link) error {
 		}
 		err = netlink.AddrDel(link, &al[i])
 		if err != nil {
-			t.l.WithError(err).Error("failed to remove address from tun address list")
+			t.l.Error("failed to remove address from tun address list", "error", err)
 		} else {
-			t.l.WithField("removed", al[i].String()).Info("removed address not listed in cert(s)")
+			t.l.Info("removed address not listed in cert(s)", "removed", al[i].String())
 		}
 	}
 
@@ -430,12 +430,12 @@ func (t *tun) Activate() error {
 	ifrq := ifreqQLEN{Name: devName, Value: int32(t.TXQueueLen)}
 	if err = ioctl(t.ioctlFd, unix.SIOCSIFTXQLEN, uintptr(unsafe.Pointer(&ifrq))); err != nil {
 		// If we can't set the queue length nebula will still work but it may lead to packet loss
-		t.l.WithError(err).Error("Failed to set tun tx queue length")
+		t.l.Error("Failed to set tun tx queue length", "error", err)
 	}
 
 	const modeNone = 1
 	if err = netlink.LinkSetIP6AddrGenMode(link, modeNone); err != nil {
-		t.l.WithError(err).Warn("Failed to disable link local address generation")
+		t.l.Warn("Failed to disable link local address generation", "error", err)
 	}
 
 	if err = t.addIPs(link); err != nil {
@@ -474,7 +474,7 @@ func (t *tun) setMTU() {
 	ifm := ifreqMTU{Name: t.deviceBytes(), MTU: int32(t.MaxMTU)}
 	if err := ioctl(t.ioctlFd, unix.SIOCSIFMTU, uintptr(unsafe.Pointer(&ifm))); err != nil {
 		// This is currently a non fatal condition because the route table must have the MTU set appropriately as well
-		t.l.WithError(err).Error("Failed to set tun mtu")
+		t.l.Error("Failed to set tun mtu", "error", err)
 	}
 }
 
@@ -497,7 +497,7 @@ func (t *tun) setDefaultRoute(cidr netip.Prefix) error {
 	}
 	err := netlink.RouteReplace(&nr)
 	if err != nil {
-		t.l.WithError(err).WithField("cidr", cidr).Warn("Failed to set default route MTU, retrying")
+		t.l.Warn("Failed to set default route MTU, retrying", "error", err, "cidr", cidr)
 		//retry twice more -- on some systems there appears to be a race condition where if we set routes too soon, netlink says `invalid argument`
 		for i := 0; i < 2; i++ {
 			time.Sleep(100 * time.Millisecond)
@@ -505,7 +505,11 @@ func (t *tun) setDefaultRoute(cidr netip.Prefix) error {
 			if err == nil {
 				break
 			} else {
-				t.l.WithError(err).WithField("cidr", cidr).WithField("mtu", t.DefaultMTU).Warn("Failed to set default route MTU, retrying")
+				t.l.Warn("Failed to set default route MTU, retrying",
+					"error", err,
+					"cidr", cidr,
+					"mtu", t.DefaultMTU,
+				)
 			}
 		}
 		if err != nil {
@@ -550,7 +554,7 @@ func (t *tun) addRoutes(logErrors bool) error {
 				return retErr
 			}
 		} else {
-			t.l.WithField("route", r).Info("Added route")
+			t.l.Info("Added route", "route", r)
 		}
 	}
 
@@ -582,9 +586,9 @@ func (t *tun) removeRoutes(routes []Route) {
 
 		err := netlink.RouteDel(&nr)
 		if err != nil {
-			t.l.WithError(err).WithField("route", r).Error("Failed to remove route")
+			t.l.Error("Failed to remove route", "error", err, "route", r)
 		} else {
-			t.l.WithField("route", r).Info("Removed route")
+			t.l.Info("Removed route", "route", r)
 		}
 	}
 }
@@ -613,11 +617,11 @@ func (t *tun) watchRoutes() {
 	netlinkOptions := netlink.RouteSubscribeOptions{
 		ReceiveBufferSize:      t.useSystemRoutesBufferSize,
 		ReceiveBufferForceSize: t.useSystemRoutesBufferSize != 0,
-		ErrorCallback:          func(e error) { t.l.WithError(e).Errorf("netlink error") },
+		ErrorCallback:          func(e error) { t.l.Error("netlink error", "error", e) },
 	}
 
 	if err := netlink.RouteSubscribeWithOptions(rch, doneChan, netlinkOptions); err != nil {
-		t.l.WithError(err).Errorf("failed to subscribe to system route changes")
+		t.l.Error("failed to subscribe to system route changes", "error", err)
 		return
 	}
 
@@ -659,7 +663,7 @@ func (t *tun) getGatewaysFromRoute(r *netlink.Route) routing.Gateways {
 
 	link, err := netlink.LinkByName(t.Device)
 	if err != nil {
-		t.l.WithField("deviceName", t.Device).Error("Ignoring route update: failed to get link by name")
+		t.l.Error("Ignoring route update: failed to get link by name", "deviceName", t.Device)
 		return gateways
 	}
 
@@ -671,10 +675,10 @@ func (t *tun) getGatewaysFromRoute(r *netlink.Route) routing.Gateways {
 				gateways = append(gateways, routing.NewGateway(gwAddr, 1))
 			} else {
 				// Gateway isn't in our overlay network, ignore
-				t.l.WithField("route", r).Debug("Ignoring route update, gateway is not in our network")
+				t.l.Debug("Ignoring route update, gateway is not in our network", "route", r)
 			}
 		} else {
-			t.l.WithField("route", r).Debug("Ignoring route update, invalid gateway or via address")
+			t.l.Debug("Ignoring route update, invalid gateway or via address", "route", r)
 		}
 	}
 
@@ -687,10 +691,10 @@ func (t *tun) getGatewaysFromRoute(r *netlink.Route) routing.Gateways {
 					gateways = append(gateways, routing.NewGateway(gwAddr, p.Hops+1))
 				} else {
 					// Gateway isn't in our overlay network, ignore
-					t.l.WithField("route", r).Debug("Ignoring route update, gateway is not in our network")
+					t.l.Debug("Ignoring route update, gateway is not in our network", "route", r)
 				}
 			} else {
-				t.l.WithField("route", r).Debug("Ignoring route update, invalid gateway or via address")
+				t.l.Debug("Ignoring route update, invalid gateway or via address", "route", r)
 			}
 		}
 	}
@@ -722,18 +726,18 @@ func (t *tun) updateRoutes(r netlink.RouteUpdate) {
 	gateways := t.getGatewaysFromRoute(&r.Route)
 	if len(gateways) == 0 {
 		// No gateways relevant to our network, no routing changes required.
-		t.l.WithField("route", r).Debug("Ignoring route update, no gateways")
+		t.l.Debug("Ignoring route update, no gateways", "route", r)
 		return
 	}
 
 	if r.Dst == nil {
-		t.l.WithField("route", r).Debug("Ignoring route update, no destination address")
+		t.l.Debug("Ignoring route update, no destination address", "route", r)
 		return
 	}
 
 	dstAddr, ok := netip.AddrFromSlice(r.Dst.IP)
 	if !ok {
-		t.l.WithField("route", r).Debug("Ignoring route update, invalid destination address")
+		t.l.Debug("Ignoring route update, invalid destination address", "route", r)
 		return
 	}
 
@@ -744,12 +748,12 @@ func (t *tun) updateRoutes(r netlink.RouteUpdate) {
 
 	t.routesFromSystemLock.Lock()
 	if r.Type == unix.RTM_NEWROUTE {
-		t.l.WithField("destination", dst).WithField("via", gateways).Info("Adding route")
+		t.l.Info("Adding route", "destination", dst, "via", gateways)
 		t.routesFromSystem[dst] = gateways
 		newTree.Insert(dst, gateways)
 
 	} else {
-		t.l.WithField("destination", dst).WithField("via", gateways).Info("Removing route")
+		t.l.Info("Removing route", "destination", dst, "via", gateways)
 		delete(t.routesFromSystem, dst)
 		newTree.Delete(dst)
 	}

overlay/tun_netbsd.go

@@ -6,6 +6,7 @@ package overlay
 import (
 	"errors"
 	"fmt"
+	"log/slog"
 	"net/netip"
 	"os"
 	"regexp"
@@ -14,7 +15,6 @@ import (
 	"unsafe"
 
 	"github.com/gaissmai/bart"
-	"github.com/sirupsen/logrus"
 	"github.com/slackhq/nebula/config"
 	"github.com/slackhq/nebula/overlay/tio"
 	"github.com/slackhq/nebula/routing"
@@ -63,7 +63,7 @@ type tun struct {
 	MTU         int
 	Routes      atomic.Pointer[[]Route]
 	routeTree   atomic.Pointer[bart.Table[routing.Gateways]]
-	l           *logrus.Logger
+	l           *slog.Logger
 	f           *os.File
 	fd          int
 
@@ -72,9 +72,6 @@ type tun struct {
 }
 
 func (t *tun) Read() ([][]byte, error) {
-	if t.readBuf == nil {
-		t.readBuf = make([]byte, defaultBatchBufSize)
-	}
 	n, err := t.readOne(t.readBuf)
 	if err != nil {
 		return nil, err
@@ -83,17 +80,17 @@ func (t *tun) Read() ([][]byte, error) {
 	return t.batchRet[:], nil
 }
 
-func (t *tun) WriteReject(p []byte) (int, error) {
+func (t *tun) Readers() []tio.Queue {
-	return t.Write(p)
+	return []tio.Queue{t}
 }
 
 var deviceNameRE = regexp.MustCompile(`^tun[0-9]+$`)
 
-func newTunFromFd(_ *config.C, _ *logrus.Logger, _ int, _ []netip.Prefix) (*tun, error) {
+func newTunFromFd(_ *config.C, _ *slog.Logger, _ int, _ []netip.Prefix) (*tun, error) {
 	return nil, fmt.Errorf("newTunFromFd not supported in NetBSD")
 }
 
-func newTun(c *config.C, l *logrus.Logger, vpnNetworks []netip.Prefix, _ bool) (*tun, error) {
+func newTun(c *config.C, l *slog.Logger, vpnNetworks []netip.Prefix, _ bool) (*tun, error) {
 	// Try to open tun device
 	var err error
 	deviceName := c.GetString("tun.dev", "")
@@ -111,7 +108,7 @@ func newTun(c *config.C, l *logrus.Logger, vpnNetworks []netip.Prefix, _ bool) (
 
 	err = unix.SetNonblock(fd, true)
 	if err != nil {
-		l.WithError(err).Warn("Failed to set the tun device as nonblocking")
+		l.Warn("Failed to set the tun device as nonblocking", "error", err)
 	}
 
 	t := &tun{
@@ -121,6 +118,7 @@ func newTun(c *config.C, l *logrus.Logger, vpnNetworks []netip.Prefix, _ bool) (
 		vpnNetworks: vpnNetworks,
 		MTU:         c.GetInt("tun.mtu", DefaultMTU),
 		l:           l,
+		readBuf:     make([]byte, defaultBatchBufSize),
 	}
 
 	err = t.reload(c, true)
@@ -413,8 +411,8 @@ func (t *tun) SupportsMultiqueue() bool {
 	return false
 }
 
-func (t *tun) NewMultiQueueReader() (tio.Queue, error) {
+func (t *tun) NewMultiQueueReader() error {
-	return nil, fmt.Errorf("TODO: multiqueue not implemented for netbsd")
+	return fmt.Errorf("TODO: multiqueue not implemented for netbsd")
 }
 
 func (t *tun) addRoutes(logErrors bool) error {
@@ -435,7 +433,7 @@ func (t *tun) addRoutes(logErrors bool) error {
 				return retErr
 			}
 		} else {
-			t.l.WithField("route", r).Info("Added route")
+			t.l.Info("Added route", "route", r)
 		}
 	}
 
@@ -450,9 +448,9 @@ func (t *tun) removeRoutes(routes []Route) error {
 
 		err := delRoute(r.Cidr, t.vpnNetworks)
 		if err != nil {
-			t.l.WithError(err).WithField("route", r).Error("Failed to remove route")
+			t.l.Error("Failed to remove route", "error", err, "route", r)
 		} else {
-			t.l.WithField("route", r).Info("Removed route")
+			t.l.Info("Removed route", "route", r)
 		}
 	}
 	return nil

overlay/tun_openbsd.go

@@ -6,6 +6,7 @@ package overlay
 import (
 	"errors"
 	"fmt"
+	"log/slog"
 	"net/netip"
 	"os"
 	"regexp"
@@ -14,7 +15,6 @@ import (
 	"unsafe"
 
 	"github.com/gaissmai/bart"
-	"github.com/sirupsen/logrus"
 	"github.com/slackhq/nebula/config"
 	"github.com/slackhq/nebula/overlay/tio"
 	"github.com/slackhq/nebula/routing"
@@ -54,7 +54,7 @@ type tun struct {
 	MTU         int
 	Routes      atomic.Pointer[[]Route]
 	routeTree   atomic.Pointer[bart.Table[routing.Gateways]]
-	l           *logrus.Logger
+	l           *slog.Logger
 	f           *os.File
 	fd          int
 	// cache out buffer since we need to prepend 4 bytes for tun metadata
@@ -65,9 +65,6 @@ type tun struct {
 }
 
 func (t *tun) Read() ([][]byte, error) {
-	if t.readBuf == nil {
-		t.readBuf = make([]byte, defaultBatchBufSize)
-	}
 	n, err := t.readOne(t.readBuf)
 	if err != nil {
 		return nil, err
@@ -76,17 +73,13 @@ func (t *tun) Read() ([][]byte, error) {
 	return t.batchRet[:], nil
 }
 
-func (t *tun) WriteReject(p []byte) (int, error) {
-	return t.Write(p)
-}
-
 var deviceNameRE = regexp.MustCompile(`^tun[0-9]+$`)
 
-func newTunFromFd(_ *config.C, _ *logrus.Logger, _ int, _ []netip.Prefix) (*tun, error) {
+func newTunFromFd(_ *config.C, _ *slog.Logger, _ int, _ []netip.Prefix) (*tun, error) {
 	return nil, fmt.Errorf("newTunFromFd not supported in openbsd")
 }
 
-func newTun(c *config.C, l *logrus.Logger, vpnNetworks []netip.Prefix, _ bool) (*tun, error) {
+func newTun(c *config.C, l *slog.Logger, vpnNetworks []netip.Prefix, _ bool) (*tun, error) {
 	// Try to open tun device
 	var err error
 	deviceName := c.GetString("tun.dev", "")
@@ -104,7 +97,7 @@ func newTun(c *config.C, l *logrus.Logger, vpnNetworks []netip.Prefix, _ bool) (
 
 	err = unix.SetNonblock(fd, true)
 	if err != nil {
-		l.WithError(err).Warn("Failed to set the tun device as nonblocking")
+		l.Warn("Failed to set the tun device as nonblocking", "error", err)
 	}
 
 	t := &tun{
@@ -114,6 +107,7 @@ func newTun(c *config.C, l *logrus.Logger, vpnNetworks []netip.Prefix, _ bool) (
 		vpnNetworks: vpnNetworks,
 		MTU:         c.GetInt("tun.mtu", DefaultMTU),
 		l:           l,
+		readBuf:     make([]byte, defaultBatchBufSize),
 	}
 
 	err = t.reload(c, true)
@@ -333,8 +327,8 @@ func (t *tun) SupportsMultiqueue() bool {
 	return false
 }
 
-func (t *tun) NewMultiQueueReader() (tio.Queue, error) {
+func (t *tun) NewMultiQueueReader() error {
-	return nil, fmt.Errorf("TODO: multiqueue not implemented for openbsd")
+	return fmt.Errorf("TODO: multiqueue not implemented for openbsd")
 }
 
 func (t *tun) addRoutes(logErrors bool) error {
@@ -355,7 +349,7 @@ func (t *tun) addRoutes(logErrors bool) error {
 				return retErr
 			}
 		} else {
-			t.l.WithField("route", r).Info("Added route")
+			t.l.Info("Added route", "route", r)
 		}
 	}
 
@@ -370,9 +364,9 @@ func (t *tun) removeRoutes(routes []Route) error {
 
 		err := delRoute(r.Cidr, t.vpnNetworks)
 		if err != nil {
-			t.l.WithError(err).WithField("route", r).Error("Failed to remove route")
+			t.l.Error("Failed to remove route", "error", err, "route", r)
 		} else {
-			t.l.WithField("route", r).Info("Removed route")
+			t.l.Info("Removed route", "route", r)
 		}
 	}
 	return nil
@@ -385,6 +379,10 @@ func (t *tun) deviceBytes() (o [16]byte) {
 	return
 }
 
+func (t *tun) Readers() []tio.Queue {
+	return []tio.Queue{t}
+}
+
 func addRoute(prefix netip.Prefix, gateways []netip.Prefix) error {
 	sock, err := unix.Socket(unix.AF_ROUTE, unix.SOCK_RAW, unix.AF_UNSPEC)
 	if err != nil {

overlay/tun_tester.go

@@ -4,14 +4,15 @@
 package overlay
 
 import (
+	"context"
 	"fmt"
 	"io"
+	"log/slog"
 	"net/netip"
 	"os"
 	"sync/atomic"
 
 	"github.com/gaissmai/bart"
-	"github.com/sirupsen/logrus"
 	"github.com/slackhq/nebula/config"
 	"github.com/slackhq/nebula/overlay/tio"
 	"github.com/slackhq/nebula/routing"
@@ -22,7 +23,7 @@ type TestTun struct {
 	vpnNetworks []netip.Prefix
 	Routes      []Route
 	routeTree   *bart.Table[routing.Gateways]
-	l           *logrus.Logger
+	l           *slog.Logger
 
 	closed    atomic.Bool
 	rxPackets chan []byte // Packets to receive into nebula
@@ -40,7 +41,7 @@ func (t *TestTun) Read() ([][]byte, error) {
 	return t.batchRet[:], nil
 }
 
-func newTun(c *config.C, l *logrus.Logger, vpnNetworks []netip.Prefix, _ bool) (*TestTun, error) {
+func newTun(c *config.C, l *slog.Logger, vpnNetworks []netip.Prefix, _ bool) (*TestTun, error) {
 	_, routes, err := getAllRoutesFromConfig(c, vpnNetworks, true)
 	if err != nil {
 		return nil, err
@@ -61,7 +62,7 @@ func newTun(c *config.C, l *logrus.Logger, vpnNetworks []netip.Prefix, _ bool) (
 	}, nil
 }
 
-func newTunFromFd(_ *config.C, _ *logrus.Logger, _ int, _ []netip.Prefix) (*TestTun, error) {
+func newTunFromFd(_ *config.C, _ *slog.Logger, _ int, _ []netip.Prefix) (*TestTun, error) {
 	return nil, fmt.Errorf("newTunFromFd not supported")
 }
 
@@ -73,8 +74,8 @@ func (t *TestTun) Send(packet []byte) {
 		return
 	}
 
-	if t.l.Level >= logrus.DebugLevel {
+	if t.l.Enabled(context.Background(), slog.LevelDebug) {
-		t.l.WithField("dataLen", len(packet)).Debug("Tun receiving injected packet")
+		t.l.Debug("Tun receiving injected packet", "dataLen", len(packet))
 	}
 	t.rxPackets <- packet
 }
@@ -127,10 +128,6 @@ func (t *TestTun) Write(b []byte) (n int, err error) {
 	return len(b), nil
 }
 
-func (t *TestTun) WriteReject(b []byte) (int, error) {
-	return t.Write(b)
-}
-
 func (t *TestTun) Close() error {
 	if t.closed.CompareAndSwap(false, true) {
 		close(t.rxPackets)
@@ -139,10 +136,14 @@ func (t *TestTun) Close() error {
 	return nil
 }
 
+func (t *TestTun) Readers() []tio.Queue {
+	return []tio.Queue{t}
+}
+
 func (t *TestTun) SupportsMultiqueue() bool {
 	return false
 }
 
-func (t *TestTun) NewMultiQueueReader() (tio.Queue, error) {
+func (t *TestTun) NewMultiQueueReader() error {
-	return nil, fmt.Errorf("TODO: multiqueue not implemented")
+	return fmt.Errorf("TODO: multiqueue not implemented")
 }

overlay/tun_windows.go

@@ -6,6 +6,7 @@ package overlay
 import (
 	"crypto"
 	"fmt"
+	"log/slog"
 	"net/netip"
 	"os"
 	"path/filepath"
@@ -15,7 +16,6 @@ import (
 	"unsafe"
 
 	"github.com/gaissmai/bart"
-	"github.com/sirupsen/logrus"
 	"github.com/slackhq/nebula/config"
 	"github.com/slackhq/nebula/overlay/tio"
 	"github.com/slackhq/nebula/routing"
@@ -33,7 +33,7 @@ type winTun struct {
 	MTU         int
 	Routes      atomic.Pointer[[]Route]
 	routeTree   atomic.Pointer[bart.Table[routing.Gateways]]
-	l           *logrus.Logger
+	l           *slog.Logger
 
 	tun *wintun.NativeTun
 
@@ -42,9 +42,6 @@ type winTun struct {
 }
 
 func (t *winTun) Read() ([][]byte, error) {
-	if t.readBuf == nil {
-		t.readBuf = make([]byte, defaultBatchBufSize)
-	}
 	n, err := t.tun.Read(t.readBuf, 0)
 	if err != nil {
 		return nil, err
@@ -53,15 +50,11 @@ func (t *winTun) Read() ([][]byte, error) {
 	return t.batchRet[:], nil
 }
 
-func (t *winTun) WriteReject(p []byte) (int, error) {
+func newTunFromFd(_ *config.C, _ *slog.Logger, _ int, _ []netip.Prefix) (Device, error) {
-	return t.Write(p)
-}
-
-func newTunFromFd(_ *config.C, _ *logrus.Logger, _ int, _ []netip.Prefix) (Device, error) {
 	return nil, fmt.Errorf("newTunFromFd not supported in Windows")
 }
 
-func newTun(c *config.C, l *logrus.Logger, vpnNetworks []netip.Prefix, _ bool) (*winTun, error) {
+func newTun(c *config.C, l *slog.Logger, vpnNetworks []netip.Prefix, _ bool) (*winTun, error) {
 	err := checkWinTunExists()
 	if err != nil {
 		return nil, fmt.Errorf("can not load the wintun driver: %w", err)
@@ -74,6 +67,7 @@ func newTun(c *config.C, l *logrus.Logger, vpnNetworks []netip.Prefix, _ bool) (
 	}
 
 	t := &winTun{
+		readBuf:     make([]byte, defaultBatchBufSize),
 		Device:      deviceName,
 		vpnNetworks: vpnNetworks,
 		MTU:         c.GetInt("tun.mtu", DefaultMTU),
@@ -90,7 +84,7 @@ func newTun(c *config.C, l *logrus.Logger, vpnNetworks []netip.Prefix, _ bool) (
 	if err != nil {
 		// Windows 10 has an issue with unclean shutdowns not fully cleaning up the wintun device.
 		// Trying a second time resolves the issue.
-		l.WithError(err).Debug("Failed to create wintun device, retrying")
+		l.Debug("Failed to create wintun device, retrying", "error", err)
 		tunDevice, err = wintun.CreateTUNWithRequestedGUID(deviceName, guid, t.MTU)
 		if err != nil {
 			return nil, &NameError{
@@ -189,7 +183,7 @@ func (t *winTun) addRoutes(logErrors bool) error {
 				return retErr
 			}
 		} else {
-			t.l.WithField("route", r).Info("Added route")
+			t.l.Info("Added route", "route", r)
 		}
 
 		if !foundDefault4 {
@@ -227,9 +221,9 @@ func (t *winTun) removeRoutes(routes []Route) error {
 		// See comment on luid.AddRoute
 		err := luid.DeleteRoute(r.Cidr, r.Via[0].Addr())
 		if err != nil {
-			t.l.WithError(err).WithField("route", r).Error("Failed to remove route")
+			t.l.Error("Failed to remove route", "error", err, "route", r)
 		} else {
-			t.l.WithField("route", r).Info("Removed route")
+			t.l.Info("Removed route", "route", r)
 		}
 	}
 	return nil
@@ -256,8 +250,12 @@ func (t *winTun) SupportsMultiqueue() bool {
 	return false
 }
 
-func (t *winTun) NewMultiQueueReader() (tio.Queue, error) {
+func (t *winTun) NewMultiQueueReader() error {
-	return nil, fmt.Errorf("TODO: multiqueue not implemented for windows")
+	return fmt.Errorf("TODO: multiqueue not implemented for windows")
+}
+
+func (t *winTun) Readers() []tio.Queue {
+	return []tio.Queue{t}
 }
 
 func (t *winTun) Close() error {

overlay/user.go

@@ -2,15 +2,15 @@ package overlay
 
 import (
 	"io"
+	"log/slog"
 	"net/netip"
 
-	"github.com/sirupsen/logrus"
 	"github.com/slackhq/nebula/config"
 	"github.com/slackhq/nebula/overlay/tio"
 	"github.com/slackhq/nebula/routing"
 )
 
-func NewUserDeviceFromConfig(c *config.C, l *logrus.Logger, vpnNetworks []netip.Prefix, routines int) (Device, error) {
+func NewUserDeviceFromConfig(c *config.C, l *slog.Logger, vpnNetworks []netip.Prefix, routines int) (Device, error) {
 	return NewUserDevice(vpnNetworks)
 }
 
@@ -88,9 +88,7 @@ func (d *UserDevice) Pipe() (*io.PipeReader, *io.PipeWriter) {
 func (d *UserDevice) Write(p []byte) (n int, err error) {
 	return d.inboundWriter.Write(p)
 }
-func (d *UserDevice) WriteReject(p []byte) (n int, err error) {
+
-	return d.Write(p)
-}
 func (d *UserDevice) Close() error {
 	d.inboundWriter.Close()
 	d.outboundWriter.Close()

pki.go

@@ -6,6 +6,7 @@ import (
 	"errors"
 	"fmt"
 	"io"
+	"log/slog"
 	"net"
 	"net/netip"
 	"os"
@@ -15,7 +16,6 @@ import (
 	"time"
 
 	"github.com/gaissmai/bart"
-	"github.com/sirupsen/logrus"
 	"github.com/slackhq/nebula/cert"
 	"github.com/slackhq/nebula/config"
 	"github.com/slackhq/nebula/util"
@@ -24,7 +24,7 @@ import (
 type PKI struct {
 	cs     atomic.Pointer[CertState]
 	caPool atomic.Pointer[cert.CAPool]
-	l      *logrus.Logger
+	l      *slog.Logger
 }
 
 type CertState struct {
@@ -46,7 +46,7 @@ type CertState struct {
 	myVpnBroadcastAddrsTable *bart.Lite
 }
 
-func NewPKIFromConfig(l *logrus.Logger, c *config.C) (*PKI, error) {
+func NewPKIFromConfig(l *slog.Logger, c *config.C) (*PKI, error) {
 	pki := &PKI{l: l}
 	err := pki.reload(c, true)
 	if err != nil {
@@ -182,9 +182,9 @@ func (p *PKI) reloadCerts(c *config.C, initial bool) *util.ContextualError {
 	p.cs.Store(newState)
 
 	if initial {
-		p.l.WithField("cert", newState).Debug("Client nebula certificate(s)")
+		p.l.Debug("Client nebula certificate(s)", "cert", newState)
 	} else {
-		p.l.WithField("cert", newState).Info("Client certificate(s) refreshed from disk")
+		p.l.Info("Client certificate(s) refreshed from disk", "cert", newState)
 	}
 	return nil
 }
@@ -196,7 +196,7 @@ func (p *PKI) reloadCAPool(c *config.C) *util.ContextualError {
 	}
 
 	p.caPool.Store(caPool)
-	p.l.WithField("fingerprints", caPool.GetFingerprints()).Debug("Trusted CA fingerprints")
+	p.l.Debug("Trusted CA fingerprints", "fingerprints", caPool.GetFingerprints())
 	return nil
 }
 
@@ -487,7 +487,7 @@ func loadCertificate(b []byte) (cert.Certificate, []byte, error) {
 	return c, b, nil
 }
 
-func loadCAPoolFromConfig(l *logrus.Logger, c *config.C) (*cert.CAPool, error) {
+func loadCAPoolFromConfig(l *slog.Logger, c *config.C) (*cert.CAPool, error) {
 	caPathOrPEM := c.GetString("pki.ca", "")
 	if caPathOrPEM == "" {
 		return nil, errors.New("no pki.ca path or PEM data provided")
@@ -512,7 +512,7 @@ func loadCAPoolFromConfig(l *logrus.Logger, c *config.C) (*cert.CAPool, error) {
 		for _, crt := range caPool.CAs {
 			if crt.Certificate.Expired(time.Now()) {
 				expired++
-				l.WithField("cert", crt).Warn("expired certificate present in CA pool")
+				l.Warn("expired certificate present in CA pool", "cert", crt)
 			}
 		}
 
@@ -530,7 +530,7 @@ func loadCAPoolFromConfig(l *logrus.Logger, c *config.C) (*cert.CAPool, error) {
 			caPool.BlocklistFingerprint(fp)
 		}
 
-		l.WithField("fingerprintCount", len(bl)).Info("Blocklisted certificates")
+		l.Info("Blocklisted certificates", "fingerprintCount", len(bl))
 	}
 
 	return caPool, nil

pki_hup_benchmark_test.go

@@ -41,7 +41,7 @@ func BenchmarkReloadConfigWithCAs(b *testing.B) {
 			c := config.NewC(l)
 			require.NoError(b, c.Load(dir))
 
-			_, err := NewPKIFromConfig(l, c)
+			_, err := NewPKIFromConfig(test.NewLogger(), c)
 			require.NoError(b, err)
 
 			b.ReportAllocs()

punchy.go

@@ -1,10 +1,10 @@
 package nebula
 
 import (
+	"log/slog"
 	"sync/atomic"
 	"time"
 
-	"github.com/sirupsen/logrus"
 	"github.com/slackhq/nebula/config"
 )
 
@@ -14,10 +14,10 @@ type Punchy struct {
 	delay           atomic.Int64
 	respondDelay    atomic.Int64
 	punchEverything atomic.Bool
-	l               *logrus.Logger
+	l               *slog.Logger
 }
 
-func NewPunchyFromConfig(l *logrus.Logger, c *config.C) *Punchy {
+func NewPunchyFromConfig(l *slog.Logger, c *config.C) *Punchy {
 	p := &Punchy{l: l}
 
 	p.reload(c, true)
@@ -62,7 +62,7 @@ func (p *Punchy) reload(c *config.C, initial bool) {
 		p.respond.Store(yes)
 
 		if !initial {
-			p.l.Infof("punchy.respond changed to %v", p.GetRespond())
+			p.l.Info("punchy.respond changed", "respond", p.GetRespond())
 		}
 	}
 
@@ -70,21 +70,21 @@ func (p *Punchy) reload(c *config.C, initial bool) {
 	if initial || c.HasChanged("punchy.delay") {
 		p.delay.Store((int64)(c.GetDuration("punchy.delay", time.Second)))
 		if !initial {
-			p.l.Infof("punchy.delay changed to %s", p.GetDelay())
+			p.l.Info("punchy.delay changed", "delay", p.GetDelay())
 		}
 	}
 
 	if initial || c.HasChanged("punchy.target_all_remotes") {
 		p.punchEverything.Store(c.GetBool("punchy.target_all_remotes", false))
 		if !initial {
-			p.l.WithField("target_all_remotes", p.GetTargetEverything()).Info("punchy.target_all_remotes changed")
+			p.l.Info("punchy.target_all_remotes changed", "target_all_remotes", p.GetTargetEverything())
 		}
 	}
 
 	if initial || c.HasChanged("punchy.respond_delay") {
 		p.respondDelay.Store((int64)(c.GetDuration("punchy.respond_delay", 5*time.Second)))
 		if !initial {
-			p.l.Infof("punchy.respond_delay changed to %s", p.GetRespondDelay())
+			p.l.Info("punchy.respond_delay changed", "respond_delay", p.GetRespondDelay())
 		}
 	}
 }

punchy_test.go

@@ -1,6 +1,8 @@
 package nebula
 
 import (
+	"context"
+	"log/slog"
 	"testing"
 	"time"
 
@@ -15,7 +17,7 @@ func TestNewPunchyFromConfig(t *testing.T) {
 	c := config.NewC(l)
 
 	// Test defaults
-	p := NewPunchyFromConfig(l, c)
+	p := NewPunchyFromConfig(test.NewLogger(), c)
 	assert.False(t, p.GetPunch())
 	assert.False(t, p.GetRespond())
 	assert.Equal(t, time.Second, p.GetDelay())
@@ -23,33 +25,33 @@ func TestNewPunchyFromConfig(t *testing.T) {
 
 	// punchy deprecation
 	c.Settings["punchy"] = true
-	p = NewPunchyFromConfig(l, c)
+	p = NewPunchyFromConfig(test.NewLogger(), c)
 	assert.True(t, p.GetPunch())
 
 	// punchy.punch
 	c.Settings["punchy"] = map[string]any{"punch": true}
-	p = NewPunchyFromConfig(l, c)
+	p = NewPunchyFromConfig(test.NewLogger(), c)
 	assert.True(t, p.GetPunch())
 
 	// punch_back deprecation
 	c.Settings["punch_back"] = true
-	p = NewPunchyFromConfig(l, c)
+	p = NewPunchyFromConfig(test.NewLogger(), c)
 	assert.True(t, p.GetRespond())
 
 	// punchy.respond
 	c.Settings["punchy"] = map[string]any{"respond": true}
 	c.Settings["punch_back"] = false
-	p = NewPunchyFromConfig(l, c)
+	p = NewPunchyFromConfig(test.NewLogger(), c)
 	assert.True(t, p.GetRespond())
 
 	// punchy.delay
 	c.Settings["punchy"] = map[string]any{"delay": "1m"}
-	p = NewPunchyFromConfig(l, c)
+	p = NewPunchyFromConfig(test.NewLogger(), c)
 	assert.Equal(t, time.Minute, p.GetDelay())
 
 	// punchy.respond_delay
 	c.Settings["punchy"] = map[string]any{"respond_delay": "1m"}
-	p = NewPunchyFromConfig(l, c)
+	p = NewPunchyFromConfig(test.NewLogger(), c)
 	assert.Equal(t, time.Minute, p.GetRespondDelay())
 }
 
@@ -62,7 +64,7 @@ punchy:
   delay: 1m
   respond: false
 `))
-	p := NewPunchyFromConfig(l, c)
+	p := NewPunchyFromConfig(test.NewLogger(), c)
 	assert.Equal(t, delay, p.GetDelay())
 	assert.False(t, p.GetRespond())
 
@@ -76,3 +78,158 @@ punchy:
 	assert.Equal(t, newDelay, p.GetDelay())
 	assert.True(t, p.GetRespond())
 }
+
+// The tests below pin the shape of each log line Punchy produces so changes
+// cannot silently break whatever operators are grepping for. The assertions
+// are on the structured message + attrs (e.g. "punchy.respond changed" with
+// a respond=true field) rather than a formatted string.
+//
+// Punchy.reload also emits a spurious "Changing punchy.punch with reload is
+// not supported" warning whenever any key under punchy changes, because of
+// the c.HasChanged("punchy") fallback kept for the deprecated top-level
+// punchy form. The tests filter by message rather than asserting total
+// entry counts so that warning is tolerated without being locked into
+// the format.
+
+type capturedEntry struct {
+	Level slog.Level
+	Msg   string
+	Attrs map[string]any
+}
+
+// capturingHandler is a slog.Handler that records each Record it receives so
+// tests can assert on the level, message, and attribute map of individual log
+// lines without coupling to any specific text format.
+type capturingHandler struct {
+	entries []capturedEntry
+}
+
+func (h *capturingHandler) Enabled(_ context.Context, _ slog.Level) bool { return true }
+
+func (h *capturingHandler) Handle(_ context.Context, r slog.Record) error {
+	e := capturedEntry{
+		Level: r.Level,
+		Msg:   r.Message,
+		Attrs: make(map[string]any),
+	}
+	r.Attrs(func(a slog.Attr) bool {
+		e.Attrs[a.Key] = a.Value.Resolve().Any()
+		return true
+	})
+	h.entries = append(h.entries, e)
+	return nil
+}
+
+func (h *capturingHandler) WithAttrs(_ []slog.Attr) slog.Handler { return h }
+func (h *capturingHandler) WithGroup(_ string) slog.Handler      { return h }
+
+func newCapturingPunchyLogger(t *testing.T) (*slog.Logger, *capturingHandler) {
+	t.Helper()
+	hook := &capturingHandler{}
+	return slog.New(hook), hook
+}
+
+func findEntry(t *testing.T, entries []capturedEntry, msg string) capturedEntry {
+	t.Helper()
+	for _, e := range entries {
+		if e.Msg == msg {
+			return e
+		}
+	}
+	t.Fatalf("no entry with message %q among %d entries", msg, len(entries))
+	return capturedEntry{}
+}
+
+func TestPunchy_LogFormat_InitialEnabled(t *testing.T) {
+	l, hook := newCapturingPunchyLogger(t)
+	c := config.NewC(test.NewLogger())
+	require.NoError(t, c.LoadString(`punchy: {punch: true}`))
+
+	NewPunchyFromConfig(l, c)
+
+	entry := findEntry(t, hook.entries, "punchy enabled")
+	assert.Equal(t, slog.LevelInfo, entry.Level)
+	assert.Empty(t, entry.Attrs)
+}
+
+func TestPunchy_LogFormat_InitialDisabled(t *testing.T) {
+	l, hook := newCapturingPunchyLogger(t)
+	c := config.NewC(test.NewLogger())
+	require.NoError(t, c.LoadString(`punchy: {punch: false}`))
+
+	NewPunchyFromConfig(l, c)
+
+	entry := findEntry(t, hook.entries, "punchy disabled")
+	assert.Equal(t, slog.LevelInfo, entry.Level)
+	assert.Empty(t, entry.Attrs)
+}
+
+func TestPunchy_LogFormat_ReloadPunchUnsupported(t *testing.T) {
+	l, hook := newCapturingPunchyLogger(t)
+	c := config.NewC(test.NewLogger())
+	require.NoError(t, c.LoadString(`punchy: {punch: false}`))
+	NewPunchyFromConfig(l, c)
+	hook.entries = nil
+
+	require.NoError(t, c.ReloadConfigString(`punchy: {punch: true}`))
+
+	entry := findEntry(t, hook.entries, "Changing punchy.punch with reload is not supported, ignoring.")
+	assert.Equal(t, slog.LevelWarn, entry.Level)
+	assert.Empty(t, entry.Attrs)
+}
+
+func TestPunchy_LogFormat_ReloadRespond(t *testing.T) {
+	l, hook := newCapturingPunchyLogger(t)
+	c := config.NewC(test.NewLogger())
+	require.NoError(t, c.LoadString(`punchy: {respond: false}`))
+	NewPunchyFromConfig(l, c)
+	hook.entries = nil
+
+	require.NoError(t, c.ReloadConfigString(`punchy: {respond: true}`))
+
+	entry := findEntry(t, hook.entries, "punchy.respond changed")
+	assert.Equal(t, slog.LevelInfo, entry.Level)
+	assert.Equal(t, map[string]any{"respond": true}, entry.Attrs)
+}
+
+func TestPunchy_LogFormat_ReloadDelay(t *testing.T) {
+	l, hook := newCapturingPunchyLogger(t)
+	c := config.NewC(test.NewLogger())
+	require.NoError(t, c.LoadString(`punchy: {delay: 1s}`))
+	NewPunchyFromConfig(l, c)
+	hook.entries = nil
+
+	require.NoError(t, c.ReloadConfigString(`punchy: {delay: 10s}`))
+
+	entry := findEntry(t, hook.entries, "punchy.delay changed")
+	assert.Equal(t, slog.LevelInfo, entry.Level)
+	assert.Equal(t, map[string]any{"delay": 10 * time.Second}, entry.Attrs)
+}
+
+func TestPunchy_LogFormat_ReloadTargetAllRemotes(t *testing.T) {
+	l, hook := newCapturingPunchyLogger(t)
+	c := config.NewC(test.NewLogger())
+	require.NoError(t, c.LoadString(`punchy: {target_all_remotes: false}`))
+	NewPunchyFromConfig(l, c)
+	hook.entries = nil
+
+	require.NoError(t, c.ReloadConfigString(`punchy: {target_all_remotes: true}`))
+
+	entry := findEntry(t, hook.entries, "punchy.target_all_remotes changed")
+	assert.Equal(t, slog.LevelInfo, entry.Level)
+	assert.Equal(t, map[string]any{"target_all_remotes": true}, entry.Attrs)
+}
+
+func TestPunchy_LogFormat_ReloadRespondDelay(t *testing.T) {
+	l, hook := newCapturingPunchyLogger(t)
+	c := config.NewC(test.NewLogger())
+	require.NoError(t, c.LoadString(`punchy: {respond_delay: 5s}`))
+	NewPunchyFromConfig(l, c)
+	hook.entries = nil
+
+	require.NoError(t, c.ReloadConfigString(`punchy: {respond_delay: 15s}`))
+
+	entry := findEntry(t, hook.entries, "punchy.respond_delay changed")
+	assert.Equal(t, slog.LevelInfo, entry.Level)
+	assert.Equal(t, map[string]any{"respond_delay": 15 * time.Second}, entry.Attrs)
+}

relay_manager.go

@@ -5,22 +5,22 @@ import (
 	"encoding/binary"
 	"errors"
 	"fmt"
+	"log/slog"
 	"net/netip"
 	"sync/atomic"
 
-	"github.com/sirupsen/logrus"
 	"github.com/slackhq/nebula/cert"
 	"github.com/slackhq/nebula/config"
 	"github.com/slackhq/nebula/header"
 )
 
 type relayManager struct {
-	l       *logrus.Logger
+	l       *slog.Logger
 	hostmap *HostMap
 	amRelay atomic.Bool
 }
 
-func NewRelayManager(ctx context.Context, l *logrus.Logger, hostmap *HostMap, c *config.C) *relayManager {
+func NewRelayManager(ctx context.Context, l *slog.Logger, hostmap *HostMap, c *config.C) *relayManager {
 	rm := &relayManager{
 		l:       l,
 		hostmap: hostmap,
@@ -29,7 +29,7 @@ func NewRelayManager(ctx context.Context, l *logrus.Logger, hostmap *HostMap, c
 	c.RegisterReloadCallback(func(c *config.C) {
 		err := rm.reload(c, false)
 		if err != nil {
-			l.WithError(err).Error("Failed to reload relay_manager")
+			rm.l.Error("Failed to reload relay_manager", "error", err)
 		}
 	})
 	return rm
@@ -52,7 +52,7 @@ func (rm *relayManager) setAmRelay(v bool) {
 
 // AddRelay finds an available relay index on the hostmap, and associates the relay info with it.
 // relayHostInfo is the Nebula peer which can be used as a relay to access the target vpnIp.
-func AddRelay(l *logrus.Logger, relayHostInfo *HostInfo, hm *HostMap, vpnIp netip.Addr, remoteIdx *uint32, relayType int, state int) (uint32, error) {
+func AddRelay(l *slog.Logger, relayHostInfo *HostInfo, hm *HostMap, vpnIp netip.Addr, remoteIdx *uint32, relayType int, state int) (uint32, error) {
 	hm.Lock()
 	defer hm.Unlock()
 	for range 32 {
@@ -92,24 +92,24 @@ func AddRelay(l *logrus.Logger, relayHostInfo *HostInfo, hm *HostMap, vpnIp neti
 func (rm *relayManager) EstablishRelay(relayHostInfo *HostInfo, m *NebulaControl) (*Relay, error) {
 	relay, ok := relayHostInfo.relayState.CompleteRelayByIdx(m.InitiatorRelayIndex, m.ResponderRelayIndex)
 	if !ok {
-		fields := logrus.Fields{
+		var relayFrom, relayTo any
-			"relay":               relayHostInfo.vpnAddrs[0],
-			"initiatorRelayIndex": m.InitiatorRelayIndex,
-		}
-
 		if m.RelayFromAddr == nil {
-			fields["relayFrom"] = m.OldRelayFromAddr
+			relayFrom = m.OldRelayFromAddr
 		} else {
-			fields["relayFrom"] = m.RelayFromAddr
+			relayFrom = m.RelayFromAddr
 		}
-
 		if m.RelayToAddr == nil {
-			fields["relayTo"] = m.OldRelayToAddr
+			relayTo = m.OldRelayToAddr
 		} else {
-			fields["relayTo"] = m.RelayToAddr
+			relayTo = m.RelayToAddr
 		}
 
-		rm.l.WithFields(fields).Info("relayManager failed to update relay")
+		rm.l.Info("relayManager failed to update relay",
+			"relay", relayHostInfo.vpnAddrs[0],
+			"initiatorRelayIndex", m.InitiatorRelayIndex,
+			"relayFrom", relayFrom,
+			"relayTo", relayTo,
+		)
 		return nil, fmt.Errorf("unknown relay")
 	}
 
@@ -120,7 +120,7 @@ func (rm *relayManager) HandleControlMsg(h *HostInfo, d []byte, f *Interface) {
 	msg := &NebulaControl{}
 	err := msg.Unmarshal(d)
 	if err != nil {
-		h.logger(f.l).WithError(err).Error("Failed to unmarshal control message")
+		h.logger(f.l).Error("Failed to unmarshal control message", "error", err)
 		return
 	}
 
@@ -147,20 +147,20 @@ func (rm *relayManager) HandleControlMsg(h *HostInfo, d []byte, f *Interface) {
 }
 
 func (rm *relayManager) handleCreateRelayResponse(v cert.Version, h *HostInfo, f *Interface, m *NebulaControl) {
-	rm.l.WithFields(logrus.Fields{
+	rm.l.Info("handleCreateRelayResponse",
-		"relayFrom":           protoAddrToNetAddr(m.RelayFromAddr),
+		"relayFrom", protoAddrToNetAddr(m.RelayFromAddr),
-		"relayTo":             protoAddrToNetAddr(m.RelayToAddr),
+		"relayTo", protoAddrToNetAddr(m.RelayToAddr),
-		"initiatorRelayIndex": m.InitiatorRelayIndex,
+		"initiatorRelayIndex", m.InitiatorRelayIndex,
-		"responderRelayIndex": m.ResponderRelayIndex,
+		"responderRelayIndex", m.ResponderRelayIndex,
-		"vpnAddrs":            h.vpnAddrs}).
+		"vpnAddrs", h.vpnAddrs,
-		Info("handleCreateRelayResponse")
+	)
 
 	target := m.RelayToAddr
 	targetAddr := protoAddrToNetAddr(target)
 
 	relay, err := rm.EstablishRelay(h, m)
 	if err != nil {
-		rm.l.WithError(err).Error("Failed to update relay for relayTo")
+		rm.l.Error("Failed to update relay for relayTo", "error", err)
 		return
 	}
 	// Do I need to complete the relays now?
@@ -170,12 +170,12 @@ func (rm *relayManager) handleCreateRelayResponse(v cert.Version, h *HostInfo, f
 	// I'm the middle man. Let the initiator know that the I've established the relay they requested.
 	peerHostInfo := rm.hostmap.QueryVpnAddr(relay.PeerAddr)
 	if peerHostInfo == nil {
-		rm.l.WithField("relayTo", relay.PeerAddr).Error("Can't find a HostInfo for peer")
+		rm.l.Error("Can't find a HostInfo for peer", "relayTo", relay.PeerAddr)
 		return
 	}
 	peerRelay, ok := peerHostInfo.relayState.QueryRelayForByIp(targetAddr)
 	if !ok {
-		rm.l.WithField("relayTo", peerHostInfo.vpnAddrs[0]).Error("peerRelay does not have Relay state for relayTo")
+		rm.l.Error("peerRelay does not have Relay state for relayTo", "relayTo", peerHostInfo.vpnAddrs[0])
 		return
 	}
 	switch peerRelay.State {
@@ -193,12 +193,13 @@ func (rm *relayManager) handleCreateRelayResponse(v cert.Version, h *HostInfo, f
 		if v == cert.Version1 {
 			peer := peerHostInfo.vpnAddrs[0]
 			if !peer.Is4() {
-				rm.l.WithField("relayFrom", peer).
+				rm.l.Error("Refusing to CreateRelayResponse for a v1 relay with an ipv6 address",
-					WithField("relayTo", target).
+					"relayFrom", peer,
-					WithField("initiatorRelayIndex", resp.InitiatorRelayIndex).
+					"relayTo", target,
-					WithField("responderRelayIndex", resp.ResponderRelayIndex).
+					"initiatorRelayIndex", resp.InitiatorRelayIndex,
-					WithField("vpnAddrs", peerHostInfo.vpnAddrs).
+					"responderRelayIndex", resp.ResponderRelayIndex,
-					Error("Refusing to CreateRelayResponse for a v1 relay with an ipv6 address")
+					"vpnAddrs", peerHostInfo.vpnAddrs,
+				)
 				return
 			}
 
@@ -213,17 +214,16 @@ func (rm *relayManager) handleCreateRelayResponse(v cert.Version, h *HostInfo, f
 
 		msg, err := resp.Marshal()
 		if err != nil {
-			rm.l.WithError(err).
+			rm.l.Error("relayManager Failed to marshal Control CreateRelayResponse message to create relay", "error", err)
-				Error("relayManager Failed to marshal Control CreateRelayResponse message to create relay")
 		} else {
 			f.SendMessageToHostInfo(header.Control, 0, peerHostInfo, msg, make([]byte, 12), make([]byte, mtu))
-			rm.l.WithFields(logrus.Fields{
+			rm.l.Info("send CreateRelayResponse",
-				"relayFrom":           resp.RelayFromAddr,
+				"relayFrom", resp.RelayFromAddr,
-				"relayTo":             resp.RelayToAddr,
+				"relayTo", resp.RelayToAddr,
-				"initiatorRelayIndex": resp.InitiatorRelayIndex,
+				"initiatorRelayIndex", resp.InitiatorRelayIndex,
-				"responderRelayIndex": resp.ResponderRelayIndex,
+				"responderRelayIndex", resp.ResponderRelayIndex,
-				"vpnAddrs":            peerHostInfo.vpnAddrs}).
+				"vpnAddrs", peerHostInfo.vpnAddrs,
-				Info("send CreateRelayResponse")
+			)
 		}
 	}
 }
@@ -232,17 +232,18 @@ func (rm *relayManager) handleCreateRelayRequest(v cert.Version, h *HostInfo, f
 	from := protoAddrToNetAddr(m.RelayFromAddr)
 	target := protoAddrToNetAddr(m.RelayToAddr)
 
-	logMsg := rm.l.WithFields(logrus.Fields{
+	logMsg := rm.l.With(
-		"relayFrom":           from,
+		"relayFrom", from,
-		"relayTo":             target,
+		"relayTo", target,
-		"initiatorRelayIndex": m.InitiatorRelayIndex,
+		"initiatorRelayIndex", m.InitiatorRelayIndex,
-		"vpnAddrs":            h.vpnAddrs})
+		"vpnAddrs", h.vpnAddrs,
+	)
 
 	logMsg.Info("handleCreateRelayRequest")
 	// Is the source of the relay me? This should never happen, but did happen due to
 	// an issue migrating relays over to newly re-handshaked host info objects.
 	if f.myVpnAddrsTable.Contains(from) {
-		logMsg.WithField("myIP", from).Error("Discarding relay request from myself")
+		logMsg.Error("Discarding relay request from myself", "myIP", from)
 		return
 	}
 
@@ -261,37 +262,37 @@ func (rm *relayManager) handleCreateRelayRequest(v cert.Version, h *HostInfo, f
 				if existingRelay.RemoteIndex != m.InitiatorRelayIndex {
 					// We got a brand new Relay request, because its index is different than what we saw before.
 					// This should never happen. The peer should never change an index, once created.
-					logMsg.WithFields(logrus.Fields{
+					logMsg.Error("Existing relay mismatch with CreateRelayRequest",
-						"existingRemoteIndex": existingRelay.RemoteIndex}).Error("Existing relay mismatch with CreateRelayRequest")
+						"existingRemoteIndex", existingRelay.RemoteIndex)
 					return
 				}
 			case Disestablished:
 				if existingRelay.RemoteIndex != m.InitiatorRelayIndex {
 					// We got a brand new Relay request, because its index is different than what we saw before.
 					// This should never happen. The peer should never change an index, once created.
-					logMsg.WithFields(logrus.Fields{
+					logMsg.Error("Existing relay mismatch with CreateRelayRequest",
-						"existingRemoteIndex": existingRelay.RemoteIndex}).Error("Existing relay mismatch with CreateRelayRequest")
+						"existingRemoteIndex", existingRelay.RemoteIndex)
 					return
 				}
 				// Mark the relay as 'Established' because it's safe to use again
 				h.relayState.UpdateRelayForByIpState(from, Established)
 			case PeerRequested:
 				// I should never be in this state, because I am terminal, not forwarding.
-				logMsg.WithFields(logrus.Fields{
+				logMsg.Error("Unexpected Relay State found",
-					"existingRemoteIndex": existingRelay.RemoteIndex,
+					"existingRemoteIndex", existingRelay.RemoteIndex,
-					"state":               existingRelay.State}).Error("Unexpected Relay State found")
+					"state", existingRelay.State)
 			}
 		} else {
 			_, err := AddRelay(rm.l, h, f.hostMap, from, &m.InitiatorRelayIndex, TerminalType, Established)
 			if err != nil {
-				logMsg.WithError(err).Error("Failed to add relay")
+				logMsg.Error("Failed to add relay", "error", err)
 				return
 			}
 		}
 
 		relay, ok := h.relayState.QueryRelayForByIp(from)
 		if !ok {
-			logMsg.WithField("from", from).Error("Relay State not found")
+			logMsg.Error("Relay State not found", "from", from)
 			return
 		}
 
@@ -313,17 +314,16 @@ func (rm *relayManager) handleCreateRelayRequest(v cert.Version, h *HostInfo, f
 
 		msg, err := resp.Marshal()
 		if err != nil {
-			logMsg.
+			logMsg.Error("relayManager Failed to marshal Control CreateRelayResponse message to create relay", "error", err)
-				WithError(err).Error("relayManager Failed to marshal Control CreateRelayResponse message to create relay")
 		} else {
 			f.SendMessageToHostInfo(header.Control, 0, h, msg, make([]byte, 12), make([]byte, mtu))
-			rm.l.WithFields(logrus.Fields{
+			rm.l.Info("send CreateRelayResponse",
-				"relayFrom":           from,
+				"relayFrom", from,
-				"relayTo":             target,
+				"relayTo", target,
-				"initiatorRelayIndex": resp.InitiatorRelayIndex,
+				"initiatorRelayIndex", resp.InitiatorRelayIndex,
-				"responderRelayIndex": resp.ResponderRelayIndex,
+				"responderRelayIndex", resp.ResponderRelayIndex,
-				"vpnAddrs":            h.vpnAddrs}).
+				"vpnAddrs", h.vpnAddrs,
-				Info("send CreateRelayResponse")
+			)
 		}
 		return
 	} else {
@@ -363,12 +363,13 @@ func (rm *relayManager) handleCreateRelayRequest(v cert.Version, h *HostInfo, f
 
 		if v == cert.Version1 {
 			if !h.vpnAddrs[0].Is4() {
-				rm.l.WithField("relayFrom", h.vpnAddrs[0]).
+				rm.l.Error("Refusing to CreateRelayRequest for a v1 relay with an ipv6 address",
-					WithField("relayTo", target).
+					"relayFrom", h.vpnAddrs[0],
-					WithField("initiatorRelayIndex", req.InitiatorRelayIndex).
+					"relayTo", target,
-					WithField("responderRelayIndex", req.ResponderRelayIndex).
+					"initiatorRelayIndex", req.InitiatorRelayIndex,
-					WithField("vpnAddr", target).
+					"responderRelayIndex", req.ResponderRelayIndex,
-					Error("Refusing to CreateRelayRequest for a v1 relay with an ipv6 address")
+					"vpnAddr", target,
+				)
 				return
 			}
 
@@ -383,17 +384,16 @@ func (rm *relayManager) handleCreateRelayRequest(v cert.Version, h *HostInfo, f
 
 		msg, err := req.Marshal()
 		if err != nil {
-			logMsg.
+			logMsg.Error("relayManager Failed to marshal Control message to create relay", "error", err)
-				WithError(err).Error("relayManager Failed to marshal Control message to create relay")
 		} else {
 			f.SendMessageToHostInfo(header.Control, 0, peer, msg, make([]byte, 12), make([]byte, mtu))
-			rm.l.WithFields(logrus.Fields{
+			rm.l.Info("send CreateRelayRequest",
-				"relayFrom":           h.vpnAddrs[0],
+				"relayFrom", h.vpnAddrs[0],
-				"relayTo":             target,
+				"relayTo", target,
-				"initiatorRelayIndex": req.InitiatorRelayIndex,
+				"initiatorRelayIndex", req.InitiatorRelayIndex,
-				"responderRelayIndex": req.ResponderRelayIndex,
+				"responderRelayIndex", req.ResponderRelayIndex,
-				"vpnAddr":             target}).
+				"vpnAddr", target,
-				Info("send CreateRelayRequest")
+			)
 		}
 
 		// Also track the half-created Relay state just received
@@ -401,8 +401,7 @@ func (rm *relayManager) handleCreateRelayRequest(v cert.Version, h *HostInfo, f
 		if !ok {
 			_, err := AddRelay(rm.l, h, f.hostMap, target, &m.InitiatorRelayIndex, ForwardingType, PeerRequested)
 			if err != nil {
-				logMsg.
+				logMsg.Error("relayManager Failed to allocate a local index for relay", "error", err)
-					WithError(err).Error("relayManager Failed to allocate a local index for relay")
 				return
 			}
 		}

remote_list.go

@@ -2,6 +2,7 @@ package nebula
 
 import (
 	"context"
+	"log/slog"
 	"net"
 	"net/netip"
 	"slices"
@@ -10,8 +11,6 @@ import (
 	"sync"
 	"sync/atomic"
 	"time"
-
-	"github.com/sirupsen/logrus"
 )
 
 // forEachFunc is used to benefit folks that want to do work inside the lock
@@ -66,11 +65,11 @@ type hostnamesResults struct {
 	network       string
 	lookupTimeout time.Duration
 	cancelFn      func()
-	l             *logrus.Logger
+	l             *slog.Logger
 	ips           atomic.Pointer[map[netip.AddrPort]struct{}]
 }
 
-func NewHostnameResults(ctx context.Context, l *logrus.Logger, d time.Duration, network string, timeout time.Duration, hostPorts []string, onUpdate func()) (*hostnamesResults, error) {
+func NewHostnameResults(ctx context.Context, l *slog.Logger, d time.Duration, network string, timeout time.Duration, hostPorts []string, onUpdate func()) (*hostnamesResults, error) {
 	r := &hostnamesResults{
 		hostnames:     make([]hostnamePort, len(hostPorts)),
 		network:       network,
@@ -121,7 +120,11 @@ func NewHostnameResults(ctx context.Context, l *logrus.Logger, d time.Duration,
 					addrs, err := net.DefaultResolver.LookupNetIP(timeoutCtx, r.network, hostPort.name)
 					timeoutCancel()
 					if err != nil {
-						l.WithFields(logrus.Fields{"hostname": hostPort.name, "network": r.network}).WithError(err).Error("DNS resolution failed for static_map host")
+						l.Error("DNS resolution failed for static_map host",
+							"hostname", hostPort.name,
+							"network", r.network,
+							"error", err,
+						)
 						continue
 					}
 					for _, a := range addrs {
@@ -145,7 +148,10 @@ func NewHostnameResults(ctx context.Context, l *logrus.Logger, d time.Duration,
 					}
 				}
 				if different {
-					l.WithFields(logrus.Fields{"origSet": origSet, "newSet": netipAddrs}).Info("DNS results changed for host list")
+					l.Info("DNS results changed for host list",
+						"origSet", origSet,
+						"newSet", netipAddrs,
+					)
 					r.ips.Store(&netipAddrs)
 					onUpdate()
 				}

service/service_test.go

@@ -10,11 +10,11 @@ import (
 	"time"
 
 	"dario.cat/mergo"
-	"github.com/sirupsen/logrus"
 	"github.com/slackhq/nebula"
 	"github.com/slackhq/nebula/cert"
 	"github.com/slackhq/nebula/cert_test"
 	"github.com/slackhq/nebula/config"
+	"github.com/slackhq/nebula/logging"
 	"github.com/slackhq/nebula/overlay"
 	"go.yaml.in/yaml/v3"
 	"golang.org/x/sync/errgroup"
@@ -75,8 +75,7 @@ func newSimpleService(caCrt cert.Certificate, caKey []byte, name string, udpIp n
 		panic(err)
 	}
 
-	logger := logrus.New()
+	logger := logging.NewLogger(os.Stdout)
-	logger.Out = os.Stdout
 
 	control, err := nebula.Main(&c, false, "custom-app", logger, overlay.NewUserDeviceFromConfig)
 	if err != nil {

ssh.go

@@ -6,21 +6,21 @@ import (
 	"errors"
 	"flag"
 	"fmt"
+	"log/slog"
 	"maps"
 	"net"
 	"net/netip"
 	"os"
 	"path/filepath"
-	"reflect"
 	"runtime"
 	"runtime/pprof"
 	"sort"
 	"strconv"
 	"strings"
 
-	"github.com/sirupsen/logrus"
 	"github.com/slackhq/nebula/config"
 	"github.com/slackhq/nebula/header"
+	"github.com/slackhq/nebula/logging"
 	"github.com/slackhq/nebula/sshd"
 )
 
@@ -57,12 +57,12 @@ type sshDeviceInfoFlags struct {
 	Pretty bool
 }
 
-func wireSSHReload(l *logrus.Logger, ssh *sshd.SSHServer, c *config.C) {
+func wireSSHReload(l *slog.Logger, ssh *sshd.SSHServer, c *config.C) {
 	c.RegisterReloadCallback(func(c *config.C) {
 		if c.GetBool("sshd.enabled", false) {
 			sshRun, err := configSSH(l, ssh, c)
 			if err != nil {
-				l.WithError(err).Error("Failed to reconfigure the sshd")
+				l.Error("Failed to reconfigure the sshd", "error", err)
 				ssh.Stop()
 			}
 			if sshRun != nil {
@@ -78,7 +78,7 @@ func wireSSHReload(l *logrus.Logger, ssh *sshd.SSHServer, c *config.C) {
 // updates the passed-in SSHServer. On success, it returns a function
 // that callers may invoke to run the configured ssh server. On
 // failure, it returns nil, error.
-func configSSH(l *logrus.Logger, ssh *sshd.SSHServer, c *config.C) (func(), error) {
+func configSSH(l *slog.Logger, ssh *sshd.SSHServer, c *config.C) (func(), error) {
 	listen := c.GetString("sshd.listen", "")
 	if listen == "" {
 		return nil, fmt.Errorf("sshd.listen must be provided")
@@ -120,7 +120,7 @@ func configSSH(l *logrus.Logger, ssh *sshd.SSHServer, c *config.C) (func(), erro
 	for _, caAuthorizedKey := range rawCAs {
 		err := ssh.AddTrustedCA(caAuthorizedKey)
 		if err != nil {
-			l.WithError(err).WithField("sshCA", caAuthorizedKey).Warn("SSH CA had an error, ignoring")
+			l.Warn("SSH CA had an error, ignoring", "error", err, "sshCA", caAuthorizedKey)
 			continue
 		}
 	}
@@ -131,13 +131,13 @@ func configSSH(l *logrus.Logger, ssh *sshd.SSHServer, c *config.C) (func(), erro
 		for _, rk := range keys {
 			kDef, ok := rk.(map[string]any)
 			if !ok {
-				l.WithField("sshKeyConfig", rk).Warn("Authorized user had an error, ignoring")
+				l.Warn("Authorized user had an error, ignoring", "sshKeyConfig", rk)
 				continue
 			}
 
 			user, ok := kDef["user"].(string)
 			if !ok {
-				l.WithField("sshKeyConfig", rk).Warn("Authorized user is missing the user field")
+				l.Warn("Authorized user is missing the user field", "sshKeyConfig", rk)
 				continue
 			}
 
@@ -146,7 +146,11 @@ func configSSH(l *logrus.Logger, ssh *sshd.SSHServer, c *config.C) (func(), erro
 			case string:
 				err := ssh.AddAuthorizedKey(user, v)
 				if err != nil {
-					l.WithError(err).WithField("sshKeyConfig", rk).WithField("sshKey", v).Warn("Failed to authorize key")
+					l.Warn("Failed to authorize key",
+						"error", err,
+						"sshKeyConfig", rk,
+						"sshKey", v,
+					)
 					continue
 				}
 
@@ -154,19 +158,25 @@ func configSSH(l *logrus.Logger, ssh *sshd.SSHServer, c *config.C) (func(), erro
 				for _, subK := range v {
 					sk, ok := subK.(string)
 					if !ok {
-						l.WithField("sshKeyConfig", rk).WithField("sshKey", subK).Warn("Did not understand ssh key")
+						l.Warn("Did not understand ssh key",
+							"sshKeyConfig", rk,
+							"sshKey", subK,
+						)
 						continue
 					}
 
 					err := ssh.AddAuthorizedKey(user, sk)
 					if err != nil {
-						l.WithError(err).WithField("sshKeyConfig", sk).Warn("Failed to authorize key")
+						l.Warn("Failed to authorize key",
+							"error", err,
+							"sshKeyConfig", sk,
+						)
 						continue
 					}
 				}
 
 			default:
-				l.WithField("sshKeyConfig", rk).Warn("Authorized user is missing the keys field or was not understood")
+				l.Warn("Authorized user is missing the keys field or was not understood", "sshKeyConfig", rk)
 			}
 		}
 	} else {
@@ -178,7 +188,7 @@ func configSSH(l *logrus.Logger, ssh *sshd.SSHServer, c *config.C) (func(), erro
 		ssh.Stop()
 		runner = func() {
 			if err := ssh.Run(listen); err != nil {
-				l.WithField("err", err).Warn("Failed to run the SSH server")
+				l.Warn("Failed to run the SSH server", "error", err)
 			}
 		}
 	} else {
@@ -188,7 +198,7 @@ func configSSH(l *logrus.Logger, ssh *sshd.SSHServer, c *config.C) (func(), erro
 	return runner, nil
 }
 
-func attachCommands(l *logrus.Logger, c *config.C, ssh *sshd.SSHServer, f *Interface) {
+func attachCommands(l *slog.Logger, c *config.C, ssh *sshd.SSHServer, f *Interface) {
 	// sandboxDir defaults to a dir in temp. The intention is that end user will
 	// create this dir as needed. Overriding this config value to "" allows
 	// writing to anywhere in the system.
@@ -789,36 +799,45 @@ func sshGetMutexProfile(sandboxDir string, fs any, a []string, w sshd.StringWrit
 	return w.WriteLine(fmt.Sprintf("Mutex profile created at %s", a))
 }
 
-func sshLogLevel(l *logrus.Logger, fs any, a []string, w sshd.StringWriter) error {
+func sshLogLevel(l *slog.Logger, fs any, a []string, w sshd.StringWriter) error {
-	if len(a) == 0 {
+	ctrl, ok := l.Handler().(interface {
-		return w.WriteLine(fmt.Sprintf("Log level is: %s", l.Level))
+		GetLevel() slog.Level
+		SetLevel(slog.Level)
+	})
+	if !ok {
+		return w.WriteLine("Log level is not reconfigurable on this logger")
 	}
 
-	level, err := logrus.ParseLevel(a[0])
+	if len(a) == 0 {
+		return w.WriteLine(fmt.Sprintf("Log level is: %s", logging.LevelName(ctrl.GetLevel())))
+	}
+
+	level, err := logging.ParseLevel(strings.ToLower(a[0]))
 	if err != nil {
-		return w.WriteLine(fmt.Sprintf("Unknown log level %s. Possible log levels: %s", a, logrus.AllLevels))
+		return w.WriteLine(fmt.Sprintf("Unknown log level %s. Possible log levels: trace, debug, info, warn, error", a))
 	}
 
-	l.SetLevel(level)
+	ctrl.SetLevel(level)
-	return w.WriteLine(fmt.Sprintf("Log level is: %s", l.Level))
+	return w.WriteLine(fmt.Sprintf("Log level is: %s", logging.LevelName(ctrl.GetLevel())))
+}
+
+func sshLogFormat(l *slog.Logger, fs any, a []string, w sshd.StringWriter) error {
+	ctrl, ok := l.Handler().(interface {
+		GetFormat() string
+		SetFormat(string) error
+	})
+	if !ok {
+		return w.WriteLine("Log format is not reconfigurable on this logger")
 	}
 
-func sshLogFormat(l *logrus.Logger, fs any, a []string, w sshd.StringWriter) error {
 	if len(a) == 0 {
-		return w.WriteLine(fmt.Sprintf("Log format is: %s", reflect.TypeOf(l.Formatter)))
+		return w.WriteLine(fmt.Sprintf("Log format is: %s", ctrl.GetFormat()))
 	}
 
-	logFormat := strings.ToLower(a[0])
+	if err := ctrl.SetFormat(strings.ToLower(a[0])); err != nil {
-	switch logFormat {
+		return err
-	case "text":
-		l.Formatter = &logrus.TextFormatter{}
-	case "json":
-		l.Formatter = &logrus.JSONFormatter{}
-	default:
-		return fmt.Errorf("unknown log format `%s`. possible formats: %s", logFormat, []string{"text", "json"})
 	}
-
+	return w.WriteLine(fmt.Sprintf("Log format is: %s", ctrl.GetFormat()))
-	return w.WriteLine(fmt.Sprintf("Log format is: %s", reflect.TypeOf(l.Formatter)))
 }
 
 func sshPrintCert(ifce *Interface, fs any, a []string, w sshd.StringWriter) error {

sshd/server.go

@@ -2,19 +2,19 @@ package sshd
 
 import (
 	"bytes"
+	"context"
 	"errors"
 	"fmt"
+	"log/slog"
 	"net"
-	"sync"
 
 	"github.com/armon/go-radix"
-	"github.com/sirupsen/logrus"
 	"golang.org/x/crypto/ssh"
 )
 
 type SSHServer struct {
 	config *ssh.ServerConfig
-	l      *logrus.Entry
+	l      *slog.Logger
 
 	certChecker *ssh.CertChecker
 
@@ -27,20 +27,21 @@ type SSHServer struct {
 	commands    *radix.Tree
 	listener    net.Listener
 
-	// Locks the conns/counter to avoid concurrent map access
+	// Call the cancel() function to stop all active sessions
-	connsLock sync.Mutex
+	ctx    context.Context
-	conns     map[int]*session
+	cancel func()
-	counter   int
 }
 
 // NewSSHServer creates a new ssh server rigged with default commands and prepares to listen
-func NewSSHServer(l *logrus.Entry) (*SSHServer, error) {
+func NewSSHServer(l *slog.Logger) (*SSHServer, error) {
 
+	ctx, cancel := context.WithCancel(context.Background())
 	s := &SSHServer{
 		trustedKeys: make(map[string]map[string]bool),
 		l:           l,
 		commands:    radix.New(),
-		conns:       make(map[int]*session),
+		ctx:         ctx,
+		cancel:      cancel,
 	}
 
 	cc := ssh.CertChecker{
@@ -120,7 +121,7 @@ func (s *SSHServer) AddTrustedCA(pubKey string) error {
 	}
 
 	s.trustedCAs = append(s.trustedCAs, pk)
-	s.l.WithField("sshKey", pubKey).Info("Trusted CA key")
+	s.l.Info("Trusted CA key", "sshKey", pubKey)
 	return nil
 }
 
@@ -138,7 +139,10 @@ func (s *SSHServer) AddAuthorizedKey(user, pubKey string) error {
 	}
 
 	tk[string(pk.Marshal())] = true
-	s.l.WithField("sshKey", pubKey).WithField("sshUser", user).Info("Authorized ssh key")
+	s.l.Info("Authorized ssh key",
+		"sshKey", pubKey,
+		"sshUser", user,
+	)
 	return nil
 }
 
@@ -155,7 +159,7 @@ func (s *SSHServer) Run(addr string) error {
 		return err
 	}
 
-	s.l.WithField("sshListener", addr).Info("SSH server is listening")
+	s.l.Info("SSH server is listening", "sshListener", addr)
 
 	// Run loops until there is an error
 	s.run()
@@ -171,11 +175,20 @@ func (s *SSHServer) run() {
 		c, err := s.listener.Accept()
 		if err != nil {
 			if !errors.Is(err, net.ErrClosed) {
-				s.l.WithError(err).Warn("Error in listener, shutting down")
+				s.l.Warn("Error in listener, shutting down", "error", err)
 			}
 			return
 		}
-
+		go func(c net.Conn) {
+			// NewServerConn may block while waiting for the client to complete the handshake.
+			// Ensure that a bad client doesn't hurt us by checking for the parent context
+			// cancellation before calling NewServerConn, and forcing the socket to close when
+			// the context is cancelled.
+			sessionContext, sessionCancel := context.WithCancel(s.ctx)
+			go func() {
+				<-sessionContext.Done()
+				c.Close()
+			}()
 			conn, chans, reqs, err := ssh.NewServerConn(c, s.config)
 			fp := ""
 			if conn != nil {
@@ -183,36 +196,33 @@ func (s *SSHServer) run() {
 			}
 
 			if err != nil {
-			l := s.l.WithError(err).WithField("remoteAddress", c.RemoteAddr())
+				l := s.l.With(
+					"error", err,
+					"remoteAddress", c.RemoteAddr(),
+				)
 				if conn != nil {
-				l = l.WithField("sshUser", conn.User())
+					l = l.With("sshUser", conn.User())
 					conn.Close()
 				}
 				if fp != "" {
-				l = l.WithField("sshFingerprint", fp)
+					l = l.With("sshFingerprint", fp)
 				}
 				l.Warn("failed to handshake")
-			continue
+				sessionCancel()
+				return
 			}
 
-		l := s.l.WithField("sshUser", conn.User())
+			l := s.l.With("sshUser", conn.User())
-		l.WithField("remoteAddress", c.RemoteAddr()).WithField("sshFingerprint", fp).Info("ssh user logged in")
+			l.Info("ssh user logged in",
+				"remoteAddress", c.RemoteAddr(),
+				"sshFingerprint", fp,
+			)
 
-		session := NewSession(s.commands, conn, chans, l.WithField("subsystem", "sshd.session"))
+			NewSession(s.commands, conn, chans, sessionCancel, l.With("subsystem", "sshd.session"))
-		s.connsLock.Lock()
-		s.counter++
-		counter := s.counter
-		s.conns[counter] = session
-		s.connsLock.Unlock()
 
 			go ssh.DiscardRequests(reqs)
-		go func() {
+
-			<-session.exitChan
+		}(c)
-			s.l.WithField("id", counter).Debug("closing conn")
-			s.connsLock.Lock()
-			delete(s.conns, counter)
-			s.connsLock.Unlock()
-		}()
 	}
 }
 
@@ -220,15 +230,11 @@ func (s *SSHServer) Stop() {
 	// Close the listener, this will cause all session to terminate as well, see SSHServer.Run
 	if s.listener != nil {
 		if err := s.listener.Close(); err != nil {
-			s.l.WithError(err).Warn("Failed to close the sshd listener")
+			s.l.Warn("Failed to close the sshd listener", "error", err)
 		}
 	}
 }
 
 func (s *SSHServer) closeSessions() {
-	s.connsLock.Lock()
+	s.cancel()
-	for _, c := range s.conns {
-		c.Close()
-	}
-	s.connsLock.Unlock()
 }

sshd/session.go

@@ -2,30 +2,30 @@ package sshd
 
 import (
 	"fmt"
+	"log/slog"
 	"sort"
 	"strings"
 
 	"github.com/anmitsu/go-shlex"
 	"github.com/armon/go-radix"
-	"github.com/sirupsen/logrus"
 	"golang.org/x/crypto/ssh"
 	"golang.org/x/term"
 )
 
 type session struct {
-	l        *logrus.Entry
+	l        *slog.Logger
 	c        *ssh.ServerConn
 	term     *term.Terminal
 	commands *radix.Tree
-	exitChan chan bool
+	cancel   func()
 }
 
-func NewSession(commands *radix.Tree, conn *ssh.ServerConn, chans <-chan ssh.NewChannel, l *logrus.Entry) *session {
+func NewSession(commands *radix.Tree, conn *ssh.ServerConn, chans <-chan ssh.NewChannel, cancel func(), l *slog.Logger) *session {
 	s := &session{
 		commands: radix.NewFromMap(commands.ToMap()),
 		l:        l,
 		c:        conn,
-		exitChan: make(chan bool),
+		cancel:   cancel,
 	}
 
 	s.commands.Insert("logout", &Command{
@@ -42,16 +42,17 @@ func NewSession(commands *radix.Tree, conn *ssh.ServerConn, chans <-chan ssh.New
 }
 
 func (s *session) handleChannels(chans <-chan ssh.NewChannel) {
+	defer s.Close()
 	for newChannel := range chans {
 		if newChannel.ChannelType() != "session" {
-			s.l.WithField("sshChannelType", newChannel.ChannelType()).Error("unknown channel type")
+			s.l.Error("unknown channel type", "sshChannelType", newChannel.ChannelType())
 			newChannel.Reject(ssh.UnknownChannelType, "unknown channel type")
 			continue
 		}
 
 		channel, requests, err := newChannel.Accept()
 		if err != nil {
-			s.l.WithError(err).Warn("could not accept channel")
+			s.l.Warn("could not accept channel", "error", err)
 			continue
 		}
 
@@ -94,13 +95,12 @@ func (s *session) handleRequests(in <-chan *ssh.Request, channel ssh.Channel) {
 			return
 
 		default:
-			s.l.WithField("sshRequest", req.Type).Debug("Rejected unknown request")
+			s.l.Debug("Rejected unknown request", "sshRequest", req.Type)
 			err = req.Reply(false, nil)
 		}
 
 		if err != nil {
-			s.l.WithError(err).Info("Error handling ssh session requests")
+			s.l.Info("Error handling ssh session requests", "error", err)
-			s.Close()
 			return
 		}
 	}
@@ -123,12 +123,11 @@ func (s *session) createTerm(channel ssh.Channel) *term.Terminal {
 		return "", 0, false
 	}
 
-	go s.handleInput(channel)
+	go s.handleInput()
 	return term
 }
 
-func (s *session) handleInput(channel ssh.Channel) {
+func (s *session) handleInput() {
-	defer s.Close()
 	w := &stringWriter{w: s.term}
 	for {
 		line, err := s.term.ReadLine()
@@ -170,10 +169,9 @@ func (s *session) dispatchCommand(line string, w StringWriter) {
 	}
 
 	_ = execCommand(c, args[1:], w)
-	return
 }
 
 func (s *session) Close() {
 	s.c.Close()
-	s.exitChan <- true
+	s.cancel()
 }

stats.go

@@ -1,13 +1,16 @@
 package nebula
 
 import (
+	"context"
 	"errors"
 	"fmt"
-	"log"
+	"log/slog"
 	"net"
 	"net/http"
 	"runtime"
 	"strconv"
+	"sync"
+	"sync/atomic"
 	"time"
 
 	graphite "github.com/cyberdelia/go-metrics-graphite"
@@ -15,98 +18,277 @@ import (
 	"github.com/prometheus/client_golang/prometheus"
 	"github.com/prometheus/client_golang/prometheus/promhttp"
 	"github.com/rcrowley/go-metrics"
-	"github.com/sirupsen/logrus"
 	"github.com/slackhq/nebula/config"
 )
 
-// startStats initializes stats from config. On success, if any further work
+// statsServer owns nebula's stats subsystem: the periodic metric capture
-// is needed to serve stats, it returns a func to handle that work. If no
+// goroutine and (for prometheus) an HTTP listener. It mirrors the lifecycle
-// work is needed, it'll return nil. On failure, it returns nil, error.
+// shape of dnsServer: constructor wires the reload callback, reload records
-func startStats(l *logrus.Logger, c *config.C, buildVersion string, configTest bool) (func(), error) {
+// config, Start builds and runs the runtime, Stop tears it down.
-	mType := c.GetString("stats.type", "")
+type statsServer struct {
-	if mType == "" || mType == "none" {
+	l            *slog.Logger
-		return nil, nil
+	ctx          context.Context
+	buildVersion string
+	configTest   bool
+
+	// enabled mirrors "stats configured to a real backend". Start consults
+	// it so callers don't need to know the gating rules.
+	enabled atomic.Bool
+
+	runMu  sync.Mutex
+	runCfg *statsConfig
+	run    *statsRuntime // non-nil while a runtime is live
 }
 
-	interval := c.GetDuration("stats.interval", 0)
+// statsRuntime is the live state owned by a single Start invocation. Start
-	if interval == 0 {
+// stashes a pointer under runMu; Stop and Start's own exit path use pointer
-		return nil, fmt.Errorf("stats.interval was an invalid duration: %s", c.GetString("stats.interval", ""))
+// equality to tell "my runtime" apart from one that replaced it after a
+// reload.
+type statsRuntime struct {
+	cancel   context.CancelFunc
+	listener *http.Server // nil for graphite
 }
 
-	var startFn func()
+// statsConfig is the snapshot of stats-related config that drives the runtime.
-	switch mType {
+// It is comparable with == so reload can detect "no change" cheaply.
-	case "graphite":
+type statsConfig struct {
-		err := startGraphiteStats(l, interval, c, configTest)
+	typ      string
+	interval time.Duration
+	graphite graphiteConfig
+	prom     promConfig
+}
+
+type graphiteConfig struct {
+	protocol string
+	host     string
+	// resolvedAddr is the string form of host resolved at config-load time.
+	// Including it in the struct means a SIGHUP picks up DNS changes even
+	// when stats.host hasn't been edited.
+	resolvedAddr string
+	prefix       string
+}
+
+type promConfig struct {
+	listen    string
+	path      string
+	namespace string
+	subsystem string
+}
+
+// newStatsServerFromConfig builds a statsServer, applies the initial config,
+// and registers a reload callback. The reload callback is registered before
+// the initial config is applied so a SIGHUP can later enable, fix, or disable
+// stats even if the initial application failed.
+//
+// Start is safe to call unconditionally: it no-ops when stats are disabled.
+// The returned pointer is always non-nil, even on error.
+func newStatsServerFromConfig(ctx context.Context, l *slog.Logger, c *config.C, buildVersion string, configTest bool) (*statsServer, error) {
+	s := &statsServer{
+		l:            l,
+		ctx:          ctx,
+		buildVersion: buildVersion,
+		configTest:   configTest,
+	}
+
+	c.RegisterReloadCallback(func(c *config.C) {
+		if err := s.reload(c, false); err != nil {
+			s.l.Error("Failed to reload stats from config", "error", err)
+		}
+	})
+
+	if err := s.reload(c, true); err != nil {
+		return s, err
+	}
+	return s, nil
+}
+
+// reload records the latest config. On the initial call it only records it;
+// Control.Start is what launches the first runtime via statsStart. On later
+// calls it reconciles the running runtime with the new config:
+//
+//   - newly enabled -> spawn Start
+//   - newly disabled -> Stop the runtime
+//   - config changed (still enabled) -> Stop the old, Start the new
+//   - no change -> no-op
+func (s *statsServer) reload(c *config.C, initial bool) error {
+	newCfg, err := loadStatsConfig(c)
 	if err != nil {
-			return nil, err
+		return err
 	}
-	case "prometheus":
+	enabled := newCfg.typ != "" && newCfg.typ != "none"
-		var err error
+
-		startFn, err = startPrometheusStats(l, interval, c, buildVersion, configTest)
+	s.runMu.Lock()
-		if err != nil {
+	sameCfg := s.runCfg != nil && *s.runCfg == newCfg
-			return nil, err
+	s.runCfg = &newCfg
-		}
+	running := s.run != nil
-	default:
+	s.runMu.Unlock()
-		return nil, fmt.Errorf("stats.type was not understood: %s", mType)
+
+	s.enabled.Store(enabled)
+
+	if initial || sameCfg {
+		return nil
 	}
 
-	metrics.RegisterDebugGCStats(metrics.DefaultRegistry)
+	if running {
-	metrics.RegisterRuntimeMemStats(metrics.DefaultRegistry)
+		s.Stop()
-
-	go metrics.CaptureDebugGCStats(metrics.DefaultRegistry, interval)
-	go metrics.CaptureRuntimeMemStats(metrics.DefaultRegistry, interval)
-
-	return startFn, nil
 	}
-
+	if enabled && !s.configTest {
-func startGraphiteStats(l *logrus.Logger, i time.Duration, c *config.C, configTest bool) error {
+		go s.Start()
-	proto := c.GetString("stats.protocol", "tcp")
-	host := c.GetString("stats.host", "")
-	if host == "" {
-		return errors.New("stats.host can not be empty")
-	}
-
-	prefix := c.GetString("stats.prefix", "nebula")
-	addr, err := net.ResolveTCPAddr(proto, host)
-	if err != nil {
-		return fmt.Errorf("error while setting up graphite sink: %s", err)
-	}
-
-	if !configTest {
-		l.Infof("Starting graphite. Interval: %s, prefix: %s, addr: %s", i, prefix, addr)
-		go graphite.Graphite(metrics.DefaultRegistry, i, prefix, addr)
 	}
 	return nil
 }
 
-func startPrometheusStats(l *logrus.Logger, i time.Duration, c *config.C, buildVersion string, configTest bool) (func(), error) {
+// Start builds the runtime from the latest config, spawns the capture loop,
-	namespace := c.GetString("stats.namespace", "")
+// and blocks until Stop is called or ctx fires. For prometheus it also serves
-	subsystem := c.GetString("stats.subsystem", "")
+// the HTTP listener. For graphite it blocks on the capture loop's context.
-
+// Safe to call when stats are disabled or already running (both no-op).
-	listen := c.GetString("stats.listen", "")
+func (s *statsServer) Start() {
-	if listen == "" {
+	if !s.enabled.Load() || s.configTest {
-		return nil, fmt.Errorf("stats.listen should not be empty")
+		return
 	}
 
-	path := c.GetString("stats.path", "")
+	s.runMu.Lock()
-	if path == "" {
+	if s.ctx.Err() != nil || s.run != nil || s.runCfg == nil {
-		return nil, fmt.Errorf("stats.path should not be empty")
+		s.runMu.Unlock()
+		return
+	}
+	cfg := *s.runCfg
+	captureFns, listener := s.buildRuntime(cfg)
+	runCtx, cancel := context.WithCancel(s.ctx)
+	rt := &statsRuntime{cancel: cancel, listener: listener}
+	s.run = rt
+	s.runMu.Unlock()
+
+	go captureStatsLoop(runCtx, cfg.interval, captureFns)
+
+	cleanExit := true
+	if listener == nil {
+		// Graphite: no HTTP listener to serve; block until teardown.
+		<-runCtx.Done()
+	} else {
+		cleanExit = s.serveListener(listener)
 	}
 
+	// Clear our runtime only if nothing has replaced it. Stop races through
+	// here too but leaves s.run == nil, so the pointer check skips.
+	s.runMu.Lock()
+	if s.run == rt {
+		rt.cancel()
+		s.run = nil
+		// A listener that exited with an error (e.g., bind conflict) leaves
+		// runCfg cached as if it were applied. Drop it so a SIGHUP with the
+		// same config re-triggers Start once the user fixes the underlying
+		// problem.
+		if !cleanExit {
+			s.runCfg = nil
+		}
+	}
+	s.runMu.Unlock()
+}
+
+// serveListener runs ListenAndServe and ensures ctx cancellation unblocks it.
+// Returns true if the listener exited cleanly (Stop, ctx cancellation, or any
+// other http.ErrServerClosed path), false on an unexpected error.
+func (s *statsServer) serveListener(listener *http.Server) bool {
+	// Per-invocation watcher: ctx cancellation triggers a listener shutdown
+	// which in turn unblocks ListenAndServe. Closing `done` on exit keeps
+	// the watcher from outliving this call.
+	done := make(chan struct{})
+	go func() {
+		select {
+		case <-s.ctx.Done():
+			shutdownCtx, cancel := context.WithTimeout(context.Background(), 5*time.Second)
+			defer cancel()
+			if err := listener.Shutdown(shutdownCtx); err != nil {
+				s.l.Warn("Failed to shut down prometheus stats listener", "error", err)
+			}
+		case <-done:
+		}
+	}()
+	defer close(done)
+
+	s.l.Info("Starting prometheus stats listener", "addr", listener.Addr)
+	err := listener.ListenAndServe()
+	if err == nil || errors.Is(err, http.ErrServerClosed) {
+		return true
+	}
+	s.l.Error("Prometheus stats listener exited", "error", err)
+	return false
+}
+
+// Stop tears down the active runtime, if any. Idempotent.
+func (s *statsServer) Stop() {
+	s.runMu.Lock()
+	rt := s.run
+	s.run = nil
+	s.runMu.Unlock()
+	if rt == nil {
+		return
+	}
+	rt.cancel()
+	if rt.listener != nil {
+		shutdownCtx, cancel := context.WithTimeout(context.Background(), 5*time.Second)
+		if err := rt.listener.Shutdown(shutdownCtx); err != nil {
+			s.l.Warn("Failed to shut down prometheus stats listener", "error", err)
+		}
+		cancel()
+	}
+}
+
+// buildRuntime produces the capture functions and, for prometheus, an un-served
+// http.Server from cfg. cfg has already been validated by loadStatsConfig.
+func (s *statsServer) buildRuntime(cfg statsConfig) ([]func(), *http.Server) {
+	// rcrowley/go-metrics guards these registrations with a private sync.Once,
+	// so subsequent reloads are no-ops.
+	metrics.RegisterDebugGCStats(metrics.DefaultRegistry)
+	metrics.RegisterRuntimeMemStats(metrics.DefaultRegistry)
+
+	captureFns := []func(){
+		func() { metrics.CaptureDebugGCStatsOnce(metrics.DefaultRegistry) },
+		func() { metrics.CaptureRuntimeMemStatsOnce(metrics.DefaultRegistry) },
+	}
+
+	switch cfg.typ {
+	case "graphite":
+		// loadStatsConfig already resolved and validated the address; re-parse
+		// the resolved form (no DNS lookup) to get a *net.TCPAddr.
+		addr, _ := net.ResolveTCPAddr(cfg.graphite.protocol, cfg.graphite.resolvedAddr)
+		gcfg := graphite.Config{
+			Addr:          addr,
+			Registry:      metrics.DefaultRegistry,
+			FlushInterval: cfg.interval,
+			DurationUnit:  time.Nanosecond,
+			Prefix:        cfg.graphite.prefix,
+			Percentiles:   []float64{0.5, 0.75, 0.95, 0.99, 0.999},
+		}
+		captureFns = append(captureFns, func() {
+			if err := graphite.Once(gcfg); err != nil {
+				s.l.Error("Graphite export failed", "error", err)
+			}
+		})
+		s.l.Info("Starting graphite stats",
+			"interval", cfg.interval,
+			"prefix", cfg.graphite.prefix,
+			"addr", addr,
+		)
+		return captureFns, nil
+
+	case "prometheus":
 		pr := prometheus.NewRegistry()
-	pClient := mp.NewPrometheusProvider(metrics.DefaultRegistry, namespace, subsystem, pr, i)
+		pClient := mp.NewPrometheusProvider(metrics.DefaultRegistry, cfg.prom.namespace, cfg.prom.subsystem, pr, cfg.interval)
-	if !configTest {
+		captureFns = append(captureFns, func() {
-		go pClient.UpdatePrometheusMetrics()
+			if err := pClient.UpdatePrometheusMetricsOnce(); err != nil {
+				s.l.Error("Prometheus metrics update failed", "error", err)
 			}
+		})
 
-	// Export our version information as labels on a static gauge
 		g := prometheus.NewGauge(prometheus.GaugeOpts{
-		Namespace: namespace,
+			Namespace: cfg.prom.namespace,
-		Subsystem: subsystem,
+			Subsystem: cfg.prom.subsystem,
 			Name:      "info",
 			Help:      "Version information for the Nebula binary",
 			ConstLabels: prometheus.Labels{
-			"version":      buildVersion,
+				"version":      s.buildVersion,
 				"goversion":    runtime.Version(),
 				"boringcrypto": strconv.FormatBool(boringEnabled()),
 			},
@@ -114,14 +296,72 @@ func startPrometheusStats(l *logrus.Logger, i time.Duration, c *config.C, buildV
 		pr.MustRegister(g)
 		g.Set(1)
 
-	var startFn func()
+		// promhttp.HandlerOpts.ErrorLog needs a stdlib-shaped Println logger,
-	if !configTest {
+		// so bridge our slog.Logger back to a *log.Logger that emits at Error.
-		startFn = func() {
+		errLog := slog.NewLogLogger(s.l.Handler(), slog.LevelError)
-			l.Infof("Prometheus stats listening on %s at %s", listen, path)
+		mux := http.NewServeMux()
-			http.Handle(path, promhttp.HandlerFor(pr, promhttp.HandlerOpts{ErrorLog: l}))
+		mux.Handle(cfg.prom.path, promhttp.HandlerFor(pr, promhttp.HandlerOpts{ErrorLog: errLog}))
-			log.Fatal(http.ListenAndServe(listen, nil))
+		return captureFns, &http.Server{Addr: cfg.prom.listen, Handler: mux}
+	}
+	return captureFns, nil
+}
+
+// captureStatsLoop runs each fn on every tick of d until ctx is cancelled.
+func captureStatsLoop(ctx context.Context, d time.Duration, fns []func()) {
+	t := time.NewTicker(d)
+	defer t.Stop()
+	for {
+		select {
+		case <-ctx.Done():
+			return
+		case <-t.C:
+			for _, fn := range fns {
+				fn()
+			}
+		}
 	}
 }
 
-	return startFn, nil
+func loadStatsConfig(c *config.C) (statsConfig, error) {
+	cfg := statsConfig{
+		typ: c.GetString("stats.type", ""),
+	}
+	if cfg.typ == "" || cfg.typ == "none" {
+		return cfg, nil
+	}
+
+	cfg.interval = c.GetDuration("stats.interval", 0)
+	if cfg.interval == 0 {
+		return cfg, fmt.Errorf("stats.interval was an invalid duration: %s", c.GetString("stats.interval", ""))
+	}
+
+	switch cfg.typ {
+	case "graphite":
+		cfg.graphite.protocol = c.GetString("stats.protocol", "tcp")
+		cfg.graphite.host = c.GetString("stats.host", "")
+		if cfg.graphite.host == "" {
+			return cfg, errors.New("stats.host can not be empty")
+		}
+		addr, err := net.ResolveTCPAddr(cfg.graphite.protocol, cfg.graphite.host)
+		if err != nil {
+			return cfg, fmt.Errorf("error while setting up graphite sink: %s", err)
+		}
+		cfg.graphite.resolvedAddr = addr.String()
+		cfg.graphite.prefix = c.GetString("stats.prefix", "nebula")
+	case "prometheus":
+		cfg.prom.listen = c.GetString("stats.listen", "")
+		if cfg.prom.listen == "" {
+			return cfg, errors.New("stats.listen should not be empty")
+		}
+		cfg.prom.path = c.GetString("stats.path", "")
+		if cfg.prom.path == "" {
+			return cfg, errors.New("stats.path should not be empty")
+		}
+		cfg.prom.namespace = c.GetString("stats.namespace", "")
+		cfg.prom.subsystem = c.GetString("stats.subsystem", "")
+	default:
+		return cfg, fmt.Errorf("stats.type was not understood: %s", cfg.typ)
+	}
+
+	return cfg, nil
 }

stats_test.go

@@ -0,0 +1,410 @@
+package nebula
+
+import (
+	"context"
+	"io"
+	"log/slog"
+	"net"
+	"strconv"
+	"testing"
+	"time"
+
+	"github.com/slackhq/nebula/config"
+	"github.com/stretchr/testify/assert"
+	"github.com/stretchr/testify/require"
+)
+
+func newTestStatsServer(t *testing.T) (*statsServer, *config.C) {
+	t.Helper()
+	l := slog.New(slog.DiscardHandler)
+	ctx, cancel := context.WithCancel(context.Background())
+	t.Cleanup(cancel)
+	return &statsServer{
+		l:   l,
+		ctx: ctx,
+	}, config.NewC(l)
+}
+
+func setStatsConfig(c *config.C, m map[string]any) {
+	c.Settings["stats"] = m
+}
+
+func currentRuntime(s *statsServer) *statsRuntime {
+	s.runMu.Lock()
+	defer s.runMu.Unlock()
+	return s.run
+}
+
+func TestStatsServer_reload_initial_disabled(t *testing.T) {
+	s, c := newTestStatsServer(t)
+	setStatsConfig(c, map[string]any{"type": "none"})
+
+	require.NoError(t, s.reload(c, true))
+	assert.False(t, s.enabled.Load())
+	assert.Nil(t, currentRuntime(s))
+}
+
+func TestStatsServer_reload_initial_invalidInterval(t *testing.T) {
+	s, c := newTestStatsServer(t)
+	setStatsConfig(c, map[string]any{
+		"type":   "graphite",
+		"host":   "127.0.0.1:0",
+		"prefix": "test",
+	})
+
+	err := s.reload(c, true)
+	require.Error(t, err)
+	assert.False(t, s.enabled.Load())
+}
+
+func TestStatsServer_reload_initial_unknownType(t *testing.T) {
+	s, c := newTestStatsServer(t)
+	setStatsConfig(c, map[string]any{
+		"type":     "carbon",
+		"interval": "1s",
+	})
+
+	err := s.reload(c, true)
+	require.Error(t, err)
+	assert.False(t, s.enabled.Load())
+}
+
+func TestStatsServer_reload_unchanged_noOp(t *testing.T) {
+	s, c := newTestStatsServer(t)
+	setStatsConfig(c, map[string]any{"type": "none"})
+
+	require.NoError(t, s.reload(c, true))
+	require.NoError(t, s.reload(c, false))
+	assert.False(t, s.enabled.Load())
+}
+
+func TestStatsServer_reload_initial_graphite(t *testing.T) {
+	s, c := newTestStatsServer(t)
+	setStatsConfig(c, map[string]any{
+		"type":     "graphite",
+		"interval": "1s",
+		"protocol": "tcp",
+		"host":     "127.0.0.1:2003",
+		"prefix":   "test",
+	})
+
+	require.NoError(t, s.reload(c, true))
+	assert.True(t, s.enabled.Load())
+	// reload only records config; Start builds the runtime.
+	assert.Nil(t, currentRuntime(s))
+}
+
+func TestStatsServer_reload_initial_prometheus(t *testing.T) {
+	s, c := newTestStatsServer(t)
+	setStatsConfig(c, map[string]any{
+		"type":     "prometheus",
+		"interval": "1s",
+		"listen":   "127.0.0.1:0",
+		"path":     "/metrics",
+	})
+
+	require.NoError(t, s.reload(c, true))
+	assert.True(t, s.enabled.Load())
+	// reload only records config; Start builds the runtime.
+	assert.Nil(t, currentRuntime(s))
+}
+
+func TestStatsServer_Start_graphite_blocksUntilStop(t *testing.T) {
+	sink := newGraphiteSink(t)
+	defer sink.Close()
+
+	s, c := newTestStatsServer(t)
+	setStatsConfig(c, map[string]any{
+		"type":     "graphite",
+		"interval": "1s",
+		"protocol": "tcp",
+		"host":     sink.Addr(),
+		"prefix":   "test",
+	})
+	require.NoError(t, s.reload(c, true))
+
+	done := make(chan struct{})
+	go func() {
+		s.Start()
+		close(done)
+	}()
+
+	// Wait for Start to publish runtime state.
+	waitFor(t, func() bool { return currentRuntime(s) != nil })
+	rt := currentRuntime(s)
+	require.NotNil(t, rt)
+	assert.Nil(t, rt.listener, "graphite has no listener")
+
+	s.Stop()
+	select {
+	case <-done:
+	case <-time.After(5 * time.Second):
+		t.Fatal("graphite Start did not return after Stop")
+	}
+	assert.Nil(t, currentRuntime(s))
+}
+
+func TestStatsServer_StartStop_lifecycle(t *testing.T) {
+	port := freeTCPPort(t)
+	s, c := newTestStatsServer(t)
+	setStatsConfig(c, map[string]any{
+		"type":     "prometheus",
+		"interval": "1s",
+		"listen":   "127.0.0.1:" + port,
+		"path":     "/metrics",
+	})
+	require.NoError(t, s.reload(c, true))
+
+	done := make(chan struct{})
+	go func() {
+		s.Start()
+		close(done)
+	}()
+
+	waitForListening(t, "127.0.0.1:"+port)
+	rt := currentRuntime(s)
+	require.NotNil(t, rt)
+	require.NotNil(t, rt.listener)
+
+	s.Stop()
+	select {
+	case <-done:
+	case <-time.After(5 * time.Second):
+		t.Fatal("Start did not return after Stop")
+	}
+	assert.Nil(t, currentRuntime(s))
+}
+
+func TestStatsServer_reload_disable_stopsRunningRuntime(t *testing.T) {
+	port := freeTCPPort(t)
+	s, c := newTestStatsServer(t)
+	setStatsConfig(c, map[string]any{
+		"type":     "prometheus",
+		"interval": "1s",
+		"listen":   "127.0.0.1:" + port,
+		"path":     "/metrics",
+	})
+	require.NoError(t, s.reload(c, true))
+
+	done := make(chan struct{})
+	go func() {
+		s.Start()
+		close(done)
+	}()
+	waitForListening(t, "127.0.0.1:"+port)
+
+	setStatsConfig(c, map[string]any{"type": "none"})
+	require.NoError(t, s.reload(c, false))
+
+	select {
+	case <-done:
+	case <-time.After(5 * time.Second):
+		t.Fatal("Start did not return after reload disabled stats")
+	}
+	assert.False(t, s.enabled.Load())
+	assert.Nil(t, currentRuntime(s))
+}
+
+func TestStatsServer_reload_changeListener_restartsListener(t *testing.T) {
+	port1 := freeTCPPort(t)
+	s, c := newTestStatsServer(t)
+	setStatsConfig(c, map[string]any{
+		"type":     "prometheus",
+		"interval": "1s",
+		"listen":   "127.0.0.1:" + port1,
+		"path":     "/metrics",
+	})
+	require.NoError(t, s.reload(c, true))
+
+	firstDone := make(chan struct{})
+	go func() {
+		s.Start()
+		close(firstDone)
+	}()
+	waitForListening(t, "127.0.0.1:"+port1)
+	first := currentRuntime(s)
+	require.NotNil(t, first)
+
+	port2 := freeTCPPort(t)
+	setStatsConfig(c, map[string]any{
+		"type":     "prometheus",
+		"interval": "1s",
+		"listen":   "127.0.0.1:" + port2,
+		"path":     "/metrics",
+	})
+	require.NoError(t, s.reload(c, false))
+
+	select {
+	case <-firstDone:
+	case <-time.After(5 * time.Second):
+		t.Fatal("old Start did not return after reload")
+	}
+
+	waitForListening(t, "127.0.0.1:"+port2)
+	second := currentRuntime(s)
+	require.NotNil(t, second)
+	assert.NotSame(t, first, second, "expected a new runtime after listen address change")
+
+	s.Stop()
+}
+
+func TestStatsServer_Stop_beforeStart_doesNotBlock(t *testing.T) {
+	s, c := newTestStatsServer(t)
+	setStatsConfig(c, map[string]any{
+		"type":     "prometheus",
+		"interval": "1s",
+		"listen":   "127.0.0.1:0",
+		"path":     "/metrics",
+	})
+	require.NoError(t, s.reload(c, true))
+
+	stopped := make(chan struct{})
+	go func() {
+		s.Stop()
+		close(stopped)
+	}()
+	select {
+	case <-stopped:
+	case <-time.After(time.Second):
+		t.Fatal("Stop hung with no runtime started")
+	}
+}
+
+func TestStatsServer_configTest_validatesWithoutSpawning(t *testing.T) {
+	s, c := newTestStatsServer(t)
+	s.configTest = true
+	setStatsConfig(c, map[string]any{
+		"type":     "prometheus",
+		"interval": "1s",
+		"listen":   "127.0.0.1:0",
+		"path":     "/metrics",
+	})
+
+	require.NoError(t, s.reload(c, true))
+	s.Start()
+	assert.Nil(t, currentRuntime(s))
+}
+
+func TestStatsServer_ctxCancel_unblocksStart(t *testing.T) {
+	// Ensures ctx cancellation alone (no explicit Stop) tears down both
+	// graphite and prom Start invocations.
+	port := freeTCPPort(t)
+	l := slog.New(slog.DiscardHandler)
+	ctx, cancel := context.WithCancel(context.Background())
+	defer cancel()
+	s := &statsServer{l: l, ctx: ctx}
+	c := config.NewC(l)
+	setStatsConfig(c, map[string]any{
+		"type":     "prometheus",
+		"interval": "1s",
+		"listen":   "127.0.0.1:" + port,
+		"path":     "/metrics",
+	})
+	require.NoError(t, s.reload(c, true))
+
+	done := make(chan struct{})
+	go func() {
+		s.Start()
+		close(done)
+	}()
+	waitForListening(t, "127.0.0.1:"+port)
+
+	cancel()
+	select {
+	case <-done:
+	case <-time.After(5 * time.Second):
+		t.Fatal("Start did not return after ctx cancel")
+	}
+}
+
+func TestStatsServer_listenerBindFailure_sameCfgReloadRetries(t *testing.T) {
+	// Hold the port so ListenAndServe will fail on first Start.
+	blocker, err := net.Listen("tcp", "127.0.0.1:0")
+	require.NoError(t, err)
+	port := strconv.Itoa(blocker.Addr().(*net.TCPAddr).Port)
+
+	s, c := newTestStatsServer(t)
+	setStatsConfig(c, map[string]any{
+		"type":     "prometheus",
+		"interval": "1s",
+		"listen":   "127.0.0.1:" + port,
+		"path":     "/metrics",
+	})
+	require.NoError(t, s.reload(c, true))
+
+	done := make(chan struct{})
+	go func() {
+		s.Start()
+		close(done)
+	}()
+	select {
+	case <-done:
+	case <-time.After(5 * time.Second):
+		t.Fatal("Start did not return after bind failure")
+	}
+	// Bind failure should have dropped the cached config so a same-cfg
+	// SIGHUP can retry.
+	s.runMu.Lock()
+	cfgAfterFailure := s.runCfg
+	s.runMu.Unlock()
+	assert.Nil(t, cfgAfterFailure)
+
+	// Free the port and reload with the same config; Start should fire again.
+	require.NoError(t, blocker.Close())
+	require.NoError(t, s.reload(c, false))
+
+	waitForListening(t, "127.0.0.1:"+port)
+	require.NotNil(t, currentRuntime(s))
+
+	s.Stop()
+}
+
+func waitForListening(t *testing.T, addr string) {
+	t.Helper()
+	waitFor(t, func() bool {
+		conn, err := net.DialTimeout("tcp", addr, 200*time.Millisecond)
+		if err != nil {
+			return false
+		}
+		_ = conn.Close()
+		return true
+	})
+}
+
+// graphiteSink is a minimal TCP accept-and-discard server so graphite.Once
+// calls in tests don't spam error logs or wedge on connection refused.
+type graphiteSink struct {
+	ln net.Listener
+}
+
+func newGraphiteSink(t *testing.T) *graphiteSink {
+	t.Helper()
+	ln, err := net.Listen("tcp", "127.0.0.1:0")
+	require.NoError(t, err)
+	g := &graphiteSink{ln: ln}
+	go func() {
+		for {
+			conn, err := ln.Accept()
+			if err != nil {
+				return
+			}
+			go func(c net.Conn) {
+				_, _ = io.Copy(io.Discard, c)
+				_ = c.Close()
+			}(conn)
+		}
+	}()
+	return g
+}
+
+func (g *graphiteSink) Addr() string { return g.ln.Addr().String() }
+func (g *graphiteSink) Close()       { _ = g.ln.Close() }
+
+func freeTCPPort(t *testing.T) string {
+	t.Helper()
+	ln, err := net.Listen("tcp", "127.0.0.1:0")
+	require.NoError(t, err)
+	port := ln.Addr().(*net.TCPAddr).Port
+	require.NoError(t, ln.Close())
+	return strconv.Itoa(port)
+}

test/logger.go

@@ -1,29 +1,73 @@
 package test
 
 import (
+	"context"
 	"io"
+	"log/slog"
 	"os"
+	"time"
 
-	"github.com/sirupsen/logrus"
+	"github.com/slackhq/nebula/logging"
 )
 
-func NewLogger() *logrus.Logger {
+// NewLogger returns a *slog.Logger suitable for use in tests. Output goes to
-	l := logrus.New()
+// io.Discard by default; set TEST_LOGS=1 (info), 2 (debug), or 3 (trace) to
-
+// stream output to stderr for local debugging.
+func NewLogger() *slog.Logger {
 	v := os.Getenv("TEST_LOGS")
 	if v == "" {
-		l.SetOutput(io.Discard)
+		return slog.New(slog.DiscardHandler)
-		return l
 	}
 
+	level := slog.LevelInfo
 	switch v {
 	case "2":
-		l.SetLevel(logrus.DebugLevel)
+		level = slog.LevelDebug
 	case "3":
-		l.SetLevel(logrus.TraceLevel)
+		level = logging.LevelTrace
-	default:
+	}
-		l.SetLevel(logrus.InfoLevel)
+	return slog.New(slog.NewTextHandler(os.Stderr, &slog.HandlerOptions{Level: level}))
 }
 
-	return l
+// NewLoggerWithOutput returns a *slog.Logger whose text output is captured by
+// w. Timestamps are suppressed so tests can assert on exact output without
+// baking the current time into expected strings.
+func NewLoggerWithOutput(w io.Writer) *slog.Logger {
+	return slog.New(&stripTimeHandler{inner: slog.NewTextHandler(w, nil)})
+}
+
+// NewLoggerWithOutputAndLevel is NewLoggerWithOutput with an explicit level
+// so tests can exercise Enabled-gated paths.
+func NewLoggerWithOutputAndLevel(w io.Writer, level slog.Level) *slog.Logger {
+	return slog.New(&stripTimeHandler{inner: slog.NewTextHandler(w, &slog.HandlerOptions{Level: level})})
+}
+
+// NewJSONLoggerWithOutput returns a *slog.Logger emitting JSON to w with
+// timestamps suppressed, for tests that pin the JSON shape.
+func NewJSONLoggerWithOutput(w io.Writer, level slog.Level) *slog.Logger {
+	return slog.New(&stripTimeHandler{inner: slog.NewJSONHandler(w, &slog.HandlerOptions{Level: level})})
+}
+
+// stripTimeHandler zeros each record's time before delegating so slog's
+// built-in handlers skip emitting the time attribute. Used to avoid
+// timestamp-dependent assertions in tests without resorting to ReplaceAttr.
+type stripTimeHandler struct {
+	inner slog.Handler
+}
+
+func (h *stripTimeHandler) Enabled(ctx context.Context, l slog.Level) bool {
+	return h.inner.Enabled(ctx, l)
+}
+
+func (h *stripTimeHandler) Handle(ctx context.Context, r slog.Record) error {
+	r.Time = time.Time{}
+	return h.inner.Handle(ctx, r)
+}
+
+func (h *stripTimeHandler) WithAttrs(attrs []slog.Attr) slog.Handler {
+	return &stripTimeHandler{inner: h.inner.WithAttrs(attrs)}
+}
+
+func (h *stripTimeHandler) WithGroup(name string) slog.Handler {
+	return &stripTimeHandler{inner: h.inner.WithGroup(name)}
 }

udp/udp_android.go

@@ -9,11 +9,12 @@ import (
 	"net/netip"
 	"syscall"
 
-	"github.com/sirupsen/logrus"
+	"log/slog"
+
 	"golang.org/x/sys/unix"
 )
 
-func NewListener(l *logrus.Logger, ip netip.Addr, port int, multi bool, batch int) (Conn, error) {
+func NewListener(l *slog.Logger, ip netip.Addr, port int, multi bool, batch int) (Conn, error) {
 	return NewGenericListener(l, ip, port, multi, batch)
 }
 

udp/udp_bsd.go

@@ -12,11 +12,12 @@ import (
 	"net/netip"
 	"syscall"
 
-	"github.com/sirupsen/logrus"
+	"log/slog"
+
 	"golang.org/x/sys/unix"
 )
 
-func NewListener(l *logrus.Logger, ip netip.Addr, port int, multi bool, batch int) (Conn, error) {
+func NewListener(l *slog.Logger, ip netip.Addr, port int, multi bool, batch int) (Conn, error) {
 	return NewGenericListener(l, ip, port, multi, batch)
 }
 

udp/udp_darwin.go

@@ -8,12 +8,12 @@ import (
 	"encoding/binary"
 	"errors"
 	"fmt"
+	"log/slog"
 	"net"
 	"net/netip"
 	"syscall"
 	"unsafe"
 
-	"github.com/sirupsen/logrus"
 	"github.com/slackhq/nebula/config"
 	"golang.org/x/sys/unix"
 )
@@ -22,12 +22,12 @@ type StdConn struct {
 	*net.UDPConn
 	isV4  bool
 	sysFd uintptr
-	l     *logrus.Logger
+	l     *slog.Logger
 }
 
 var _ Conn = &StdConn{}
 
-func NewListener(l *logrus.Logger, ip netip.Addr, port int, multi bool, batch int) (Conn, error) {
+func NewListener(l *slog.Logger, ip netip.Addr, port int, multi bool, batch int) (Conn, error) {
 	lc := NewListenConfig(multi)
 	pc, err := lc.ListenPacket(context.TODO(), "udp", net.JoinHostPort(ip.String(), fmt.Sprintf("%v", port)))
 	if err != nil {
@@ -149,17 +149,6 @@ func (u *StdConn) WriteBatch(bufs [][]byte, addrs []netip.AddrPort) error {
 	return nil
 }
 
-func (u *StdConn) WriteSegmented(bufs [][]byte, addr netip.AddrPort, _ int) error {
-	for _, b := range bufs {
-		if err := u.WriteTo(b, addr); err != nil {
-			return err
-		}
-	}
-	return nil
-}
-
-func (u *StdConn) SupportsGSO() bool { return false }
-
 func (u *StdConn) LocalAddr() (netip.AddrPort, error) {
 	a := u.UDPConn.LocalAddr()
 
@@ -196,7 +185,7 @@ func (u *StdConn) ListenOut(r EncReader, flush func()) error {
 				return err
 			}
 
-			u.l.WithError(err).Error("unexpected udp socket receive error")
+			u.l.Error("unexpected udp socket receive error", "error", err)
 		}
 
 		r(netip.AddrPortFrom(rua.Addr().Unmap(), rua.Port()), buffer[:n])
@@ -217,7 +206,7 @@ func (u *StdConn) Rebind() error {
 	}
 
 	if err != nil {
-		u.l.WithError(err).Error("Failed to rebind udp socket")
+		u.l.Error("Failed to rebind udp socket", "error", err)
 	}
 
 	return nil

udp/udp_generic.go

@@ -12,22 +12,22 @@ import (
 	"context"
 	"errors"
 	"fmt"
+	"log/slog"
 	"net"
 	"net/netip"
 	"time"
 
-	"github.com/sirupsen/logrus"
 	"github.com/slackhq/nebula/config"
 )
 
 type GenericConn struct {
 	*net.UDPConn
-	l *logrus.Logger
+	l *slog.Logger
 }
 
 var _ Conn = &GenericConn{}
 
-func NewGenericListener(l *logrus.Logger, ip netip.Addr, port int, multi bool, batch int) (Conn, error) {
+func NewGenericListener(l *slog.Logger, ip netip.Addr, port int, multi bool, batch int) (Conn, error) {
 	lc := NewListenConfig(multi)
 	pc, err := lc.ListenPacket(context.TODO(), "udp", net.JoinHostPort(ip.String(), fmt.Sprintf("%v", port)))
 	if err != nil {
@@ -53,17 +53,6 @@ func (u *GenericConn) WriteBatch(bufs [][]byte, addrs []netip.AddrPort) error {
 	return nil
 }
 
-func (u *GenericConn) WriteSegmented(bufs [][]byte, addr netip.AddrPort, _ int) error {
-	for _, b := range bufs {
-		if _, err := u.UDPConn.WriteToUDPAddrPort(b, addr); err != nil {
-			return err
-		}
-	}
-	return nil
-}
-
-func (u *GenericConn) SupportsGSO() bool { return false }
-
 func (u *GenericConn) LocalAddr() (netip.AddrPort, error) {
 	a := u.UDPConn.LocalAddr()
 
@@ -108,7 +97,7 @@ func (u *GenericConn) ListenOut(r EncReader, flush func()) error {
 			// Dampen unexpected message warns to once per minute
 			if lastRecvErr.IsZero() || time.Since(lastRecvErr) > time.Minute {
 				lastRecvErr = time.Now()
-				u.l.WithError(err).Warn("unexpected udp socket receive error")
+				u.l.Warn("unexpected udp socket receive error", "error", err)
 			}
 			continue
 		}

udp/udp_linux.go

@@ -7,13 +7,13 @@ import (
 	"context"
 	"encoding/binary"
 	"fmt"
+	"log/slog"
 	"net"
 	"net/netip"
 	"syscall"
 	"unsafe"
 
 	"github.com/rcrowley/go-metrics"
-	"github.com/sirupsen/logrus"
 	"github.com/slackhq/nebula/config"
 	"golang.org/x/sys/unix"
 )
@@ -22,7 +22,7 @@ type StdConn struct {
 	udpConn *net.UDPConn
 	rawConn syscall.RawConn
 	isV4    bool
-	l       *logrus.Logger
+	l       *slog.Logger
 	batch   int
 
 	// sendmmsg scratch. Each queue has its own StdConn, so no locking is
@@ -75,7 +75,7 @@ func setReusePort(network, address string, c syscall.RawConn) error {
 	return opErr
 }
 
-func NewListener(l *logrus.Logger, ip netip.Addr, port int, multi bool, batch int) (Conn, error) {
+func NewListener(l *slog.Logger, ip netip.Addr, port int, multi bool, batch int) (Conn, error) {
 	listen := netip.AddrPortFrom(ip, uint16(port))
 	lc := net.ListenConfig{}
 	if multi {
@@ -643,12 +643,12 @@ func (u *StdConn) ReloadConfig(c *config.C) {
 		if err == nil {
 			s, err := u.GetRecvBuffer()
 			if err == nil {
-				u.l.WithField("size", s).Info("listen.read_buffer was set")
+				u.l.Info("listen.read_buffer was set", "size", s)
 			} else {
-				u.l.WithError(err).Warn("Failed to get listen.read_buffer")
+				u.l.Warn("Failed to get listen.read_buffer", "error", err)
 			}
 		} else {
-			u.l.WithError(err).Error("Failed to set listen.read_buffer")
+			u.l.Error("Failed to set listen.read_buffer", "error", err)
 		}
 	}
 
@@ -658,12 +658,12 @@ func (u *StdConn) ReloadConfig(c *config.C) {
 		if err == nil {
 			s, err := u.GetSendBuffer()
 			if err == nil {
-				u.l.WithField("size", s).Info("listen.write_buffer was set")
+				u.l.Info("listen.write_buffer was set", "size", s)
 			} else {
-				u.l.WithError(err).Warn("Failed to get listen.write_buffer")
+				u.l.Warn("Failed to get listen.write_buffer", "error", err)
 			}
 		} else {
-			u.l.WithError(err).Error("Failed to set listen.write_buffer")
+			u.l.Error("Failed to set listen.write_buffer", "error", err)
 		}
 	}
 
@@ -674,12 +674,12 @@ func (u *StdConn) ReloadConfig(c *config.C) {
 		if err == nil {
 			s, err := u.GetSoMark()
 			if err == nil {
-				u.l.WithField("mark", s).Info("listen.so_mark was set")
+				u.l.Info("listen.so_mark was set", "mark", s)
 			} else {
-				u.l.WithError(err).Warn("Failed to get listen.so_mark")
+				u.l.Warn("Failed to get listen.so_mark", "error", err)
 			}
 		} else {
-			u.l.WithError(err).Error("Failed to set listen.so_mark")
+			u.l.Error("Failed to set listen.so_mark", "error", err)
 		}
 	}
 }

udp/udp_netbsd.go

@@ -11,11 +11,12 @@ import (
 	"net/netip"
 	"syscall"
 
-	"github.com/sirupsen/logrus"
+	"log/slog"
+
 	"golang.org/x/sys/unix"
 )
 
-func NewListener(l *logrus.Logger, ip netip.Addr, port int, multi bool, batch int) (Conn, error) {
+func NewListener(l *slog.Logger, ip netip.Addr, port int, multi bool, batch int) (Conn, error) {
 	return NewGenericListener(l, ip, port, multi, batch)
 }
 

udp/udp_rio_windows.go

@@ -9,6 +9,7 @@ import (
 	"errors"
 	"fmt"
 	"io"
+	"log/slog"
 	"net"
 	"net/netip"
 	"sync"
@@ -17,7 +18,6 @@ import (
 	"time"
 	"unsafe"
 
-	"github.com/sirupsen/logrus"
 	"github.com/slackhq/nebula/config"
 	"golang.org/x/sys/windows"
 	"golang.zx2c4.com/wireguard/conn/winrio"
@@ -53,14 +53,14 @@ type ringBuffer struct {
 
 type RIOConn struct {
 	isOpen  atomic.Bool
-	l       *logrus.Logger
+	l       *slog.Logger
 	sock    windows.Handle
 	rx, tx  ringBuffer
 	rq      winrio.Rq
 	results [packetsPerRing]winrio.Result
 }
 
-func NewRIOListener(l *logrus.Logger, addr netip.Addr, port int) (*RIOConn, error) {
+func NewRIOListener(l *slog.Logger, addr netip.Addr, port int) (*RIOConn, error) {
 	if !winrio.Initialize() {
 		return nil, errors.New("could not initialize winrio")
 	}
@@ -83,7 +83,7 @@ func NewRIOListener(l *logrus.Logger, addr netip.Addr, port int) (*RIOConn, erro
 	return u, nil
 }
 
-func (u *RIOConn) bind(l *logrus.Logger, sa windows.Sockaddr) error {
+func (u *RIOConn) bind(l *slog.Logger, sa windows.Sockaddr) error {
 	var err error
 	u.sock, err = winrio.Socket(windows.AF_INET6, windows.SOCK_DGRAM, windows.IPPROTO_UDP)
 	if err != nil {
@@ -103,7 +103,7 @@ func (u *RIOConn) bind(l *logrus.Logger, sa windows.Sockaddr) error {
 	if err != nil {
 		// This is a best-effort to prevent errors from being returned by the udp recv operation.
 		// Quietly log a failure and continue.
-		l.WithError(err).Debug("failed to set UDP_CONNRESET ioctl")
+		l.Debug("failed to set UDP_CONNRESET ioctl", "error", err)
 	}
 
 	ret = 0
@@ -114,7 +114,7 @@ func (u *RIOConn) bind(l *logrus.Logger, sa windows.Sockaddr) error {
 	if err != nil {
 		// This is a best-effort to prevent errors from being returned by the udp recv operation.
 		// Quietly log a failure and continue.
-		l.WithError(err).Debug("failed to set UDP_NETRESET ioctl")
+		l.Debug("failed to set UDP_NETRESET ioctl", "error", err)
 	}
 
 	err = u.rx.Open()
@@ -156,7 +156,7 @@ func (u *RIOConn) ListenOut(r EncReader, flush func()) error {
 			// Dampen unexpected message warns to once per minute
 			if lastRecvErr.IsZero() || time.Since(lastRecvErr) > time.Minute {
 				lastRecvErr = time.Now()
-				u.l.WithError(err).Warn("unexpected udp socket receive error")
+				u.l.Warn("unexpected udp socket receive error", "error", err)
 			}
 			continue
 		}
@@ -326,17 +326,6 @@ func (u *RIOConn) WriteBatch(bufs [][]byte, addrs []netip.AddrPort) error {
 	return nil
 }
 
-func (u *RIOConn) WriteSegmented(bufs [][]byte, addr netip.AddrPort, _ int) error {
-	for _, b := range bufs {
-		if err := u.WriteTo(b, addr); err != nil {
-			return err
-		}
-	}
-	return nil
-}
-
-func (u *RIOConn) SupportsGSO() bool { return false }
-
 func (u *RIOConn) LocalAddr() (netip.AddrPort, error) {
 	sa, err := windows.Getsockname(u.sock)
 	if err != nil {

udp/udp_tester.go

@@ -4,12 +4,13 @@
 package udp
 
 import (
+	"context"
 	"io"
+	"log/slog"
 	"net/netip"
 	"os"
-	"sync/atomic"
+	"sync"
 
-	"github.com/sirupsen/logrus"
 	"github.com/slackhq/nebula/config"
 	"github.com/slackhq/nebula/header"
 )
@@ -37,15 +38,24 @@ type TesterConn struct {
 	RxPackets chan *Packet // Packets to receive into nebula
 	TxPackets chan *Packet // Packets transmitted outside by nebula
 
-	closed atomic.Bool
+	// done is closed exactly once by Close. Senders select on it so they
-	l      *logrus.Logger
+	// never race with a channel close; readers exit when it fires. The
+	// packet channels are intentionally never closed - that was the source
+	// of `send on closed channel` panics when a WriteTo/Send from another
+	// goroutine passed the close check and reached the send just after
+	// Close ran.
+	done      chan struct{}
+	closeOnce sync.Once
+
+	l *slog.Logger
 }
 
-func NewListener(l *logrus.Logger, ip netip.Addr, port int, _ bool, _ int) (Conn, error) {
+func NewListener(l *slog.Logger, ip netip.Addr, port int, _ bool, _ int) (Conn, error) {
 	return &TesterConn{
 		Addr:      netip.AddrPortFrom(ip, uint16(port)),
 		RxPackets: make(chan *Packet, 10),
 		TxPackets: make(chan *Packet, 10),
+		done:      make(chan struct{}),
 		l:         l,
 	}, nil
 }
@@ -54,21 +64,21 @@ func NewListener(l *logrus.Logger, ip netip.Addr, port int, _ bool, _ int) (Conn
 // this is an encrypted packet or a handshake message in most cases
 // packets were transmitted from another nebula node, you can send them with Tun.Send
 func (u *TesterConn) Send(packet *Packet) {
-	if u.closed.Load() {
-		return
-	}
-
 	h := &header.H{}
 	if err := h.Parse(packet.Data); err != nil {
 		panic(err)
 	}
-	if u.l.Level >= logrus.DebugLevel {
+	if u.l.Enabled(context.Background(), slog.LevelDebug) {
-		u.l.WithField("header", h).
+		u.l.Debug("UDP receiving injected packet",
-			WithField("udpAddr", packet.From).
+			"header", h,
-			WithField("dataLen", len(packet.Data)).
+			"udpAddr", packet.From,
-			Debug("UDP receiving injected packet")
+			"dataLen", len(packet.Data),
+		)
+	}
+	select {
+	case <-u.done:
+	case u.RxPackets <- packet:
 	}
-	u.RxPackets <- packet
 }
 
 // Get will pull a UdpPacket from the transmit queue
@@ -76,7 +86,12 @@ func (u *TesterConn) Send(packet *Packet) {
 // packets were ingested from the tun side (in most cases), you can send them with Tun.Send
 func (u *TesterConn) Get(block bool) *Packet {
 	if block {
-		return <-u.TxPackets
+		select {
+		case <-u.done:
+			return nil
+		case p := <-u.TxPackets:
+			return p
+		}
 	}
 
 	select {
@@ -92,10 +107,6 @@ func (u *TesterConn) Get(block bool) *Packet {
 //********************************************************************************************************************//
 
 func (u *TesterConn) WriteTo(b []byte, addr netip.AddrPort) error {
-	if u.closed.Load() {
-		return io.ErrClosedPipe
-	}
-
 	p := &Packet{
 		Data: make([]byte, len(b), len(b)),
 		From: u.Addr,
@@ -103,9 +114,13 @@ func (u *TesterConn) WriteTo(b []byte, addr netip.AddrPort) error {
 	}
 
 	copy(p.Data, b)
-	u.TxPackets <- p
+	select {
+	case <-u.done:
+		return io.ErrClosedPipe
+	case u.TxPackets <- p:
 		return nil
 	}
+}
 
 func (u *TesterConn) WriteBatch(bufs [][]byte, addrs []netip.AddrPort) error {
 	for i, b := range bufs {
@@ -116,27 +131,17 @@ func (u *TesterConn) WriteBatch(bufs [][]byte, addrs []netip.AddrPort) error {
 	return nil
 }
 
-func (u *TesterConn) WriteSegmented(bufs [][]byte, addr netip.AddrPort, _ int) error {
-	for _, b := range bufs {
-		if err := u.WriteTo(b, addr); err != nil {
-			return err
-		}
-	}
-	return nil
-}
-
-func (u *TesterConn) SupportsGSO() bool { return false }
-
 func (u *TesterConn) ListenOut(r EncReader, flush func()) error {
 	for {
-		p, ok := <-u.RxPackets
+		select {
-		if !ok {
+		case <-u.done:
 			return os.ErrClosed
-		}
+		case p := <-u.RxPackets:
 			r(p.From, p.Data)
 			flush()
 		}
 	}
+}
 
 func (u *TesterConn) ReloadConfig(*config.C) {}
 
@@ -158,9 +163,8 @@ func (u *TesterConn) Rebind() error {
 }
 
 func (u *TesterConn) Close() error {
-	if u.closed.CompareAndSwap(false, true) {
+	u.closeOnce.Do(func() {
-		close(u.RxPackets)
+		close(u.done)
-		close(u.TxPackets)
+	})
-	}
 	return nil
 }

udp/udp_windows.go

@@ -5,14 +5,13 @@ package udp
 
 import (
 	"fmt"
+	"log/slog"
 	"net"
 	"net/netip"
 	"syscall"
-
-	"github.com/sirupsen/logrus"
 )
 
-func NewListener(l *logrus.Logger, ip netip.Addr, port int, multi bool, batch int) (Conn, error) {
+func NewListener(l *slog.Logger, ip netip.Addr, port int, multi bool, batch int) (Conn, error) {
 	if multi {
 		//NOTE: Technically we can support it with RIO but it wouldn't be at the socket level
 		// The udp stack would need to be reworked to hide away the implementation differences between
@@ -25,7 +24,7 @@ func NewListener(l *logrus.Logger, ip netip.Addr, port int, multi bool, batch in
 		return rc, nil
 	}
 
-	l.WithError(err).Error("Falling back to standard udp sockets")
+	l.Error("Falling back to standard udp sockets", "error", err)
 	return NewGenericListener(l, ip, port, multi, batch)
 }
 

util/error.go

@@ -1,10 +1,10 @@
 package util
 
 import (
+	"context"
 	"errors"
 	"fmt"
-
+	"log/slog"
-	"github.com/sirupsen/logrus"
 )
 
 type ContextualError struct {
@@ -28,12 +28,12 @@ func ContextualizeIfNeeded(msg string, err error) error {
 }
 
 // LogWithContextIfNeeded is a helper function to log an error line for an error or ContextualError
-func LogWithContextIfNeeded(msg string, err error, l *logrus.Logger) {
+func LogWithContextIfNeeded(msg string, err error, l *slog.Logger) {
 	switch v := err.(type) {
 	case *ContextualError:
 		v.Log(l)
 	default:
-		l.WithError(err).Error(msg)
+		l.Error(msg, "error", err)
 	}
 }
 
@@ -51,10 +51,19 @@ func (ce *ContextualError) Unwrap() error {
 	return ce.RealError
 }
 
-func (ce *ContextualError) Log(lr *logrus.Logger) {
+// Log emits ce as a single error-level log line with Fields and RealError
+// promoted to top-level attributes, producing a flat shape callers can grep
+// or parse without walking into a nested object.
+func (ce *ContextualError) Log(l *slog.Logger) {
+	attrs := make([]slog.Attr, 0, len(ce.Fields)+1)
+	for k, v := range ce.Fields {
+		attrs = append(attrs, slog.Any(k, v))
+	}
 	if ce.RealError != nil {
-		lr.WithFields(ce.Fields).WithError(ce.RealError).Error(ce.Context)
+		attrs = append(attrs, slog.Any("error", ce.RealError))
-	} else {
-		lr.WithFields(ce.Fields).Error(ce.Context)
 	}
+	// LogAttrs is intentional: attrs is built from a map[string]any so it has
+	// no pair-form equivalent.
+	//nolint:sloglint
+	l.LogAttrs(context.Background(), slog.LevelError, ce.Context, attrs...)
 }

util/error_test.go

@@ -1,95 +1,67 @@
 package util
 
 import (
+	"bytes"
 	"errors"
 	"fmt"
 	"testing"
 
-	"github.com/sirupsen/logrus"
+	"github.com/slackhq/nebula/test"
 	"github.com/stretchr/testify/assert"
 )
 
 type m = map[string]any
 
-type TestLogWriter struct {
-	Logs []string
-}
-
-func NewTestLogWriter() *TestLogWriter {
-	return &TestLogWriter{Logs: make([]string, 0)}
-}
-
-func (tl *TestLogWriter) Write(p []byte) (n int, err error) {
-	tl.Logs = append(tl.Logs, string(p))
-	return len(p), nil
-}
-
-func (tl *TestLogWriter) Reset() {
-	tl.Logs = tl.Logs[:0]
-}
-
 func TestContextualError_Log(t *testing.T) {
-	l := logrus.New()
+	buf := &bytes.Buffer{}
-	l.Formatter = &logrus.TextFormatter{
+	l := test.NewLoggerWithOutput(buf)
-		DisableTimestamp: true,
-		DisableColors:    true,
-	}
-
-	tl := NewTestLogWriter()
-	l.Out = tl
 
 	// Test a full context line
-	tl.Reset()
+	buf.Reset()
 	e := NewContextualError("test message", m{"field": "1"}, errors.New("error"))
 	e.Log(l)
-	assert.Equal(t, []string{"level=error msg=\"test message\" error=error field=1\n"}, tl.Logs)
+	assert.Equal(t, "level=ERROR msg=\"test message\" field=1 error=error\n", buf.String())
 
 	// Test a line with an error and msg but no fields
-	tl.Reset()
+	buf.Reset()
 	e = NewContextualError("test message", nil, errors.New("error"))
 	e.Log(l)
-	assert.Equal(t, []string{"level=error msg=\"test message\" error=error\n"}, tl.Logs)
+	assert.Equal(t, "level=ERROR msg=\"test message\" error=error\n", buf.String())
 
 	// Test just a context and fields
-	tl.Reset()
+	buf.Reset()
 	e = NewContextualError("test message", m{"field": "1"}, nil)
 	e.Log(l)
-	assert.Equal(t, []string{"level=error msg=\"test message\" field=1\n"}, tl.Logs)
+	assert.Equal(t, "level=ERROR msg=\"test message\" field=1\n", buf.String())
 
 	// Test just a context
-	tl.Reset()
+	buf.Reset()
 	e = NewContextualError("test message", nil, nil)
 	e.Log(l)
-	assert.Equal(t, []string{"level=error msg=\"test message\"\n"}, tl.Logs)
+	assert.Equal(t, "level=ERROR msg=\"test message\"\n", buf.String())
 
 	// Test just an error
-	tl.Reset()
+	buf.Reset()
 	e = NewContextualError("", nil, errors.New("error"))
 	e.Log(l)
-	assert.Equal(t, []string{"level=error error=error\n"}, tl.Logs)
+	assert.Equal(t, "level=ERROR msg=\"\" error=error\n", buf.String())
 }
 
 func TestLogWithContextIfNeeded(t *testing.T) {
-	l := logrus.New()
+	buf := &bytes.Buffer{}
-	l.Formatter = &logrus.TextFormatter{
+	l := test.NewLoggerWithOutput(buf)
-		DisableTimestamp: true,
-		DisableColors:    true,
-	}
-
-	tl := NewTestLogWriter()
-	l.Out = tl
 
 	// Test ignoring fallback context
-	tl.Reset()
+	buf.Reset()
 	e := NewContextualError("test message", m{"field": "1"}, errors.New("error"))
 	LogWithContextIfNeeded("This should get thrown away", e, l)
-	assert.Equal(t, []string{"level=error msg=\"test message\" error=error field=1\n"}, tl.Logs)
+	assert.Equal(t, "level=ERROR msg=\"test message\" field=1 error=error\n", buf.String())
 
 	// Test using fallback context
-	tl.Reset()
+	buf.Reset()
 	err := fmt.Errorf("this is a normal error")
 	LogWithContextIfNeeded("Fallback context woo", err, l)
-	assert.Equal(t, []string{"level=error msg=\"Fallback context woo\" error=\"this is a normal error\"\n"}, tl.Logs)
+	assert.Equal(t, "level=ERROR msg=\"Fallback context woo\" error=\"this is a normal error\"\n", buf.String())
 }
 
 func TestContextualizeIfNeeded(t *testing.T) {