ipn/ipnlocal, wgengine/wgcfg/nmcfg: stop building peer lists on delta path

Processing a peer add/remove delta still materialized the full netmap
(an O(n) slicesx.MapValues plus sort over all peers, at 10k+
peers in a large tailnet) twice per delta: once in UpdateNetmapDelta
purely to hand the self node to Engine.SetSelfNode, and once in
authReconfigLocked.

Neither needs peers anymore. SetSelfNode gets the self node from the
existing nodeBackend.Self accessor. authReconfigLocked only reads
self-node fields (SelfNode, NodeKey, GetAddresses, HasCap) now that
WireGuard peers ride the incremental route manager and per-peer config
source, so it can use the peers-free NetMap accessor.

That also makes nmcfg.WGCfg vestigial: since wgcfg.Config lost its
Peers field, its peer walk existed only to emit the [v1] skip logs
(expired peers, unselected exit nodes, unaccepted subnet routes),
duplicating filtering the route manager already does. Delete the
package and construct the two-field wgcfg.Config inline. The skip
logs go away; if they're missed, the route manager can log them
incrementally at upsert time instead of rescanning every peer on
every reconfig.

With this, the runtime.DidRange analysis (see the ts_rangehook test)
shows a delta netmap update performing no O(n) range loops except
updateRouteManagerExtras, and the delta phase of that test drops from
1.09s to 0.14s for 400 deltas at n=10000 (from 4.79s at the
start of this effort, before the incremental route manager work).

Updates #12542

Signed-off-by: Brad Fitzpatrick <bradfitz@tailscale.com>
Change-Id: Ia0e03ef9db0c988790b2c29de1f0505305e93f58

cmd/k8s-operator/depaware.txt

@@ -928,7 +928,6 @@ tailscale.com/cmd/k8s-operator dependencies: (generated by github.com/tailscale/
         tailscale.com/wgengine/netstack/gro                          from tailscale.com/net/tstun+
         tailscale.com/wgengine/router                                from tailscale.com/ipn/ipnlocal+
         tailscale.com/wgengine/wgcfg                                 from tailscale.com/ipn/ipnlocal+
-        tailscale.com/wgengine/wgcfg/nmcfg                           from tailscale.com/ipn/ipnlocal
      💣 tailscale.com/wgengine/wgint                                 from tailscale.com/wgengine+
         tailscale.com/wgengine/wglog                                 from tailscale.com/wgengine
         golang.org/x/crypto/argon2                                   from tailscale.com/tka

cmd/tailscaled/depaware-min.txt

@@ -200,7 +200,6 @@ tailscale.com/cmd/tailscaled dependencies: (generated by github.com/tailscale/de
         tailscale.com/wgengine/netstack/gro                          from tailscale.com/net/tstun+
         tailscale.com/wgengine/router                                from tailscale.com/cmd/tailscaled+
         tailscale.com/wgengine/wgcfg                                 from tailscale.com/ipn/ipnlocal+
-        tailscale.com/wgengine/wgcfg/nmcfg                           from tailscale.com/ipn/ipnlocal
      💣 tailscale.com/wgengine/wgint                                 from tailscale.com/wgengine+
         tailscale.com/wgengine/wglog                                 from tailscale.com/wgengine
         golang.org/x/crypto/blake2b                                  from golang.org/x/crypto/nacl/box

cmd/tailscaled/depaware-minbox.txt

@@ -221,7 +221,6 @@ tailscale.com/cmd/tailscaled dependencies: (generated by github.com/tailscale/de
         tailscale.com/wgengine/netstack/gro                          from tailscale.com/net/tstun+
         tailscale.com/wgengine/router                                from tailscale.com/cmd/tailscaled+
         tailscale.com/wgengine/wgcfg                                 from tailscale.com/ipn/ipnlocal+
-        tailscale.com/wgengine/wgcfg/nmcfg                           from tailscale.com/ipn/ipnlocal
      💣 tailscale.com/wgengine/wgint                                 from tailscale.com/wgengine+
         tailscale.com/wgengine/wglog                                 from tailscale.com/wgengine
         golang.org/x/crypto/blake2b                                  from golang.org/x/crypto/nacl/box

cmd/tailscaled/depaware.txt

@@ -521,7 +521,6 @@ tailscale.com/cmd/tailscaled dependencies: (generated by github.com/tailscale/de
         tailscale.com/wgengine/router                                from tailscale.com/cmd/tailscaled+
         tailscale.com/wgengine/router/osrouter                       from tailscale.com/feature/condregister
         tailscale.com/wgengine/wgcfg                                 from tailscale.com/ipn/ipnlocal+
-        tailscale.com/wgengine/wgcfg/nmcfg                           from tailscale.com/ipn/ipnlocal
      💣 tailscale.com/wgengine/wgint                                 from tailscale.com/wgengine+
         tailscale.com/wgengine/wglog                                 from tailscale.com/wgengine
    W 💣 tailscale.com/wgengine/winnet                                from tailscale.com/wgengine/router/osrouter

cmd/tsidp/depaware.txt

@@ -326,7 +326,6 @@ tailscale.com/cmd/tsidp dependencies: (generated by github.com/tailscale/depawar
         tailscale.com/wgengine/netstack/gro                          from tailscale.com/net/tstun+
         tailscale.com/wgengine/router                                from tailscale.com/ipn/ipnlocal+
         tailscale.com/wgengine/wgcfg                                 from tailscale.com/ipn/ipnlocal+
-        tailscale.com/wgengine/wgcfg/nmcfg                           from tailscale.com/ipn/ipnlocal
      💣 tailscale.com/wgengine/wgint                                 from tailscale.com/wgengine+
         tailscale.com/wgengine/wglog                                 from tailscale.com/wgengine
         golang.org/x/crypto/argon2                                   from tailscale.com/tka

ipn/ipnlocal/local.go

@@ -107,7 +107,6 @@ import (
 	"tailscale.com/wgengine/magicsock"
 	"tailscale.com/wgengine/router"
 	"tailscale.com/wgengine/wgcfg"
-	"tailscale.com/wgengine/wgcfg/nmcfg"
 )
 
 var controlDebugFlags = getControlDebugFlags()
@@ -2502,8 +2501,8 @@ func (b *LocalBackend) UpdateNetmapDelta(muts []netmap.NodeMutation) (handled bo
 	needsAuthReconfig = needsAuthReconfig || peersUpsertedOrRemoved
 	if needsAuthReconfig {
 		if peersUpsertedOrRemoved {
-			if nm := cn.netMapWithPeers(); nm != nil {
+			if self := cn.Self(); self.Valid() {
-				b.e.SetSelfNode(nm.SelfNode)
+				b.e.SetSelfNode(self)
 			}
 		}
 		b.authReconfigLocked()
@@ -6054,7 +6053,10 @@ func (b *LocalBackend) authReconfigLocked() {
 
 	cn := b.currentNode()
 
-	nm := cn.netMapWithPeers()
+	// Note this netmap does not have its Peers populated. Nothing
+	// below needs them; per-peer work rides the incremental route
+	// manager and engine paths instead.
+	nm := cn.NetMap()
 	if nm == nil {
 		b.logf("[v1] authReconfig: netmap not yet valid. Skipping.")
 		return
@@ -6100,10 +6102,12 @@ func (b *LocalBackend) authReconfigLocked() {
 		priv = key.NodePrivate{}
 	}
 
-	cfg, err := nmcfg.WGCfg(priv, nm, b.logf, flags, prefs.ExitNodeID())
+	// The config carries no peers; wireguard-go gets those from the
-	if err != nil {
+	// live per-peer config source installed via
-		b.logf("wgcfg: %v", err)
+	// [wgengine.Engine.SetPeerConfigFunc], fed by the route manager.
-		return
+	cfg := &wgcfg.Config{
+		PrivateKey: priv,
+		Addresses:  nm.GetAddresses().AsSlice(),
 	}
 
 	// Note: b.goos (set only by tests) speaks runtime.GOOS while
@@ -6146,7 +6150,7 @@ func (b *LocalBackend) authReconfigLocked() {
 	// the new config never see a stale peer table.
 	b.setDataPlanePeerRoutes()
 
-	err = b.e.Reconfig(cfg, rcfg, dcfg)
+	err := b.e.Reconfig(cfg, rcfg, dcfg)
 	if err == wgengine.ErrNoChanges {
 		return
 	}

net/routemanager/routemanager.go

@@ -397,12 +397,11 @@ func (m *Mutation) UpsertPeer(n tailcfg.NodeView) {
 // peerViewOf reduces a tailcfg.NodeView to the routing-relevant
 // peerView.
 //
-// It mirrors the peer and prefix filtering in nmcfg.WGCfg: peers we
+// Peers we cannot communicate with (expired, or predating both DERP
-// cannot communicate with (expired, or predating both DERP and disco)
+// and disco) contribute no prefixes. They remain tracked by ID and key
-// contribute no prefixes. They remain tracked by ID and key so that a
+// so that a later update can make them routable again. AllowedIPs is
-// later update can make them routable again. AllowedIPs is the sole
+// the sole source of prefixes; an address absent from AllowedIPs is
-// source of prefixes; an address absent from AllowedIPs is not
+// not routable. For the self-vs-route split,
-// routable. The self-vs-route split mirrors nmcfg's cidrIsSubnet:
 // single Tailscale IPs are never subnets, so a VIP service address
 // hosted by the peer lands in SelfAddrs and stays routable without
 // Prefs.RouteAll.
@@ -662,8 +661,7 @@ func normalizePrefix(p netip.Prefix) netip.Prefix {
 // non-address bits set, mirroring the defensive check in
 // ipnlocal.peerRoutes. It includes the peer's extra allowed IPs,
 // except for peers that contribute no addresses or routes of their
-// own (expired or otherwise non-communicable peers, which mirrors
+// own (expired or otherwise non-communicable peers).
-// nmcfg.WGCfg dropping such peers entirely).
 func (rm *RouteManager) contribs(p peerView) map[netip.Prefix]contribKind {
 	c := make(map[netip.Prefix]contribKind, len(p.SelfAddrs)+len(p.Routes))
 	add := func(pfx netip.Prefix, kind contribKind) {

tsnet/depaware.txt

@@ -321,7 +321,6 @@ tailscale.com/tsnet dependencies: (generated by github.com/tailscale/depaware)
         tailscale.com/wgengine/netstack/gro                          from tailscale.com/net/tstun+
         tailscale.com/wgengine/router                                from tailscale.com/ipn/ipnlocal+
         tailscale.com/wgengine/wgcfg                                 from tailscale.com/ipn/ipnlocal+
-        tailscale.com/wgengine/wgcfg/nmcfg                           from tailscale.com/ipn/ipnlocal
      💣 tailscale.com/wgengine/wgint                                 from tailscale.com/wgengine+
         tailscale.com/wgengine/wglog                                 from tailscale.com/wgengine
         golang.org/x/crypto/argon2                                   from tailscale.com/tka

wgengine/magicsock/magicsock_test.go

@@ -81,7 +81,6 @@ import (
 	"tailscale.com/wgengine/filter"
 	"tailscale.com/wgengine/filter/filtertype"
 	"tailscale.com/wgengine/wgcfg"
-	"tailscale.com/wgengine/wgcfg/nmcfg"
 	"tailscale.com/wgengine/wglog"
 )
 
@@ -246,6 +245,17 @@ func newMagicStackWithKey(t testing.TB, logf logger.Logf, ln nettype.PacketListe
 	}
 }
 
+// wgCfgOf builds a test node's wgcfg.Config from its netmap the same
+// way ipnlocal.authReconfigLocked does in production: just the private
+// key and self addresses. Peers ride the per-peer config source
+// instead; see [magicStack.Reconfig].
+func wgCfgOf(pk key.NodePrivate, nm *netmap.NetworkMap) *wgcfg.Config {
+	return &wgcfg.Config{
+		PrivateKey: pk,
+		Addresses:  nm.GetAddresses().AsSlice(),
+	}
+}
+
 // Reconfig applies cfg and peers to the stack's WireGuard device and
 // tun-layer data plane. In production these flow from LocalBackend
 // (via [tailscale.com/wgengine.Engine.Reconfig] and the live per-peer
@@ -259,8 +269,8 @@ func (s *magicStack) Reconfig(cfg *wgcfg.Config, peers []tailcfg.NodeView) error
 
 	rm := routemanager.New(nil)
 	mut := rm.Begin()
-	// Mirror the netmap.AllowSubnetRoutes flag the tests pass to
+	// Tests want subnet routes accepted, matching what production
-	// nmcfg.WGCfg.
+	// gets from Prefs.RouteAll.
 	mut.SetPrefs(routemanager.Prefs{RouteAll: true})
 	// idByKey stands in for nodeBackend's public-key-to-node-ID
 	// index, which LocalBackend uses to serve the engine's per-peer
@@ -412,12 +422,7 @@ func meshStacks(logf logger.Logf, mutateNetmap func(idx int, nm *netmap.NetworkM
 		for i, m := range ms {
 			nm := buildNetmapLocked(i)
 			m.conn.SetNetworkMap(nm.SelfNode, nm.Peers)
-			wg, err := nmcfg.WGCfg(ms[i].privateKey, nm, logf, 0, "")
+			wg := wgCfgOf(ms[i].privateKey, nm)
-			if err != nil {
-				// We're too far from the *testing.T to be graceful,
-				// blow up. Shouldn't happen anyway.
-				panic(fmt.Sprintf("failed to construct wgcfg from netmap: %v", err))
-			}
 			if err := m.Reconfig(wg, nm.Peers); err != nil {
 				if ctx.Err() != nil || errors.Is(err, errConnClosed) {
 					// shutdown race, don't care.
@@ -2568,10 +2573,7 @@ func TestIsWireGuardOnlyPeer(t *testing.T) {
 	}
 	m.conn.SetNetworkMap(nm.SelfNode, nm.Peers)
 
-	cfg, err := nmcfg.WGCfg(m.privateKey, nm, t.Logf, netmap.AllowSubnetRoutes, "")
+	cfg := wgCfgOf(m.privateKey, nm)
-	if err != nil {
-		t.Fatal(err)
-	}
 	m.Reconfig(cfg, nm.Peers)
 
 	pbuf := tuntest.Ping(wgaip.Addr(), tsaip.Addr())
@@ -2629,10 +2631,7 @@ func TestIsWireGuardOnlyPeerWithMasquerade(t *testing.T) {
 	}
 	m.conn.SetNetworkMap(nm.SelfNode, nm.Peers)
 
-	cfg, err := nmcfg.WGCfg(m.privateKey, nm, t.Logf, netmap.AllowSubnetRoutes, "")
+	cfg := wgCfgOf(m.privateKey, nm)
-	if err != nil {
-		t.Fatal(err)
-	}
 	m.Reconfig(cfg, nm.Peers)
 
 	pbuf := tuntest.Ping(wgaip.Addr(), tsaip.Addr())
@@ -2669,10 +2668,7 @@ func applyNetworkMap(t *testing.T, m *magicStack, nm *netmap.NetworkMap) {
 	m.conn.noV6.Store(true)
 
 	// Turn the network map into a wireguard config (for the tailscale internal wireguard device).
-	cfg, err := nmcfg.WGCfg(m.privateKey, nm, t.Logf, netmap.AllowSubnetRoutes, "")
+	cfg := wgCfgOf(m.privateKey, nm)
-	if err != nil {
-		t.Fatal(err)
-	}
 	// Apply the wireguard config to the tailscale internal wireguard device.
 	if err := m.Reconfig(cfg, nm.Peers); err != nil {
 		t.Fatal(err)

wgengine/wgcfg/nmcfg/nmcfg.go

@@ -1,117 +0,0 @@
-// Copyright (c) Tailscale Inc & contributors
-// SPDX-License-Identifier: BSD-3-Clause
-
-// Package nmcfg converts a controlclient.NetMap into a wgcfg config.
-package nmcfg
-
-import (
-	"bufio"
-	"cmp"
-	"fmt"
-	"net/netip"
-	"strings"
-
-	"tailscale.com/net/tsaddr"
-	"tailscale.com/tailcfg"
-	"tailscale.com/types/key"
-	"tailscale.com/types/logger"
-	"tailscale.com/types/netmap"
-	"tailscale.com/wgengine/wgcfg"
-)
-
-func nodeDebugName(n tailcfg.NodeView) string {
-	name, _, _ := strings.Cut(cmp.Or(n.Name(), n.Hostinfo().Hostname()), ".")
-	return name
-}
-
-// cidrIsSubnet reports whether cidr is a non-default-route subnet
-// exported by node that is not one of its own self addresses.
-func cidrIsSubnet(node tailcfg.NodeView, cidr netip.Prefix) bool {
-	if cidr.Bits() == 0 {
-		return false
-	}
-	if !cidr.IsSingleIP() {
-		return true
-	}
-	if tsaddr.IsTailscaleIP(cidr.Addr()) {
-		return false
-	}
-	for _, selfCIDR := range node.Addresses().All() {
-		if cidr == selfCIDR {
-			return false
-		}
-	}
-	return true
-}
-
-// WGCfg returns the NetworkMaps's WireGuard configuration.
-//
-// The config does not include peers; wireguard-go gets those from the
-// live per-peer config source installed via
-// [tailscale.com/wgengine.Engine.SetPeerConfigFunc], fed by the route
-// manager. WGCfg still walks the peers to log which ones are not
-// routable and why, mirroring the route manager's filtering.
-func WGCfg(pk key.NodePrivate, nm *netmap.NetworkMap, logf logger.Logf, flags netmap.WGConfigFlags, exitNode tailcfg.StableNodeID) (*wgcfg.Config, error) {
-	cfg := &wgcfg.Config{
-		PrivateKey: pk,
-		Addresses:  nm.GetAddresses().AsSlice(),
-	}
-
-	var skippedExitNode, skippedSubnetRouter, skippedExpired []tailcfg.NodeView
-
-	for _, peer := range nm.Peers {
-		if peer.DiscoKey().IsZero() && peer.HomeDERP() == 0 && !peer.IsWireGuardOnly() {
-			// Peer predates both DERP and active discovery, we cannot
-			// communicate with it.
-			logf("[v1] wgcfg: skipped peer %s, doesn't offer DERP or disco", peer.Key().ShortString())
-			continue
-		}
-		// Skip expired peers; we'll end up failing to connect to them
-		// anyway, since control intentionally breaks node keys for
-		// expired peers so that we can't discover endpoints via DERP.
-		if peer.Expired() {
-			skippedExpired = append(skippedExpired, peer)
-			continue
-		}
-
-		didExitNodeLog := false
-		for _, allowedIP := range peer.AllowedIPs().All() {
-			if allowedIP.Bits() == 0 && peer.StableID() != exitNode {
-				if didExitNodeLog {
-					// Don't log about both the IPv4 /0 and IPv6 /0.
-					continue
-				}
-				didExitNodeLog = true
-				skippedExitNode = append(skippedExitNode, peer)
-			} else if cidrIsSubnet(peer, allowedIP) {
-				if (flags & netmap.AllowSubnetRoutes) == 0 {
-					skippedSubnetRouter = append(skippedSubnetRouter, peer)
-				}
-			}
-		}
-	}
-
-	logList := func(title string, nodes []tailcfg.NodeView) {
-		if len(nodes) == 0 {
-			return
-		}
-		logf("[v1] wgcfg: %s from %d nodes: %s", title, len(nodes), logger.ArgWriter(func(bw *bufio.Writer) {
-			const max = 5
-			for i, n := range nodes {
-				if i == max {
-					fmt.Fprintf(bw, "... +%d", len(nodes)-max)
-					return
-				}
-				if i > 0 {
-					bw.WriteString(", ")
-				}
-				fmt.Fprintf(bw, "%s (%s)", nodeDebugName(n), n.StableID())
-			}
-		}))
-	}
-	logList("skipped unselected exit nodes", skippedExitNode)
-	logList("did not accept subnet routes", skippedSubnetRouter)
-	logList("skipped expired peers", skippedExpired)
-
-	return cfg, nil
-}