Merge remote-tracking branch 'origin/main' into peer-acl-multi-source

# Conflicts:
#	.github/workflows/golang-test-linux.yml

client/cmd/version.go

@@ -12,7 +12,13 @@ var (
 		Short: "Print the NetBird's client application version",
 		Run: func(cmd *cobra.Command, args []string) {
 			cmd.SetOut(cmd.OutOrStdout())
-			cmd.Println(version.NetbirdVersion())
+			out := version.NetbirdVersion()
+			if version.IsDevelopmentVersion(out) {
+				if commit := version.NetbirdCommit(); commit != "" {
+					out += "-" + commit
+				}
+			}
+			cmd.Println(out)
 		},
 	}
 )

client/firewall/uspfilter/forwarder/icmp.go

@@ -362,6 +362,10 @@ func (f *Forwarder) injectICMPv6Reply(id stack.TransportEndpointID, icmpPayload
 		return 0
 	}
 
+	if pc := f.endpoint.capture.Load(); pc != nil {
+		(*pc).Offer(fullPacket, true)
+	}
+
 	return len(fullPacket)
 }
 

client/internal/connect.go

@@ -6,6 +6,7 @@ import (
 	"fmt"
 	"net"
 	"net/netip"
+	"path/filepath"
 	"runtime"
 	"runtime/debug"
 	"strings"
@@ -346,6 +347,11 @@ func (c *ConnectClient) run(mobileDependency MobileDependency, runningChan chan
 			return wrapErr(err)
 		}
 		engineConfig.TempDir = mobileDependency.TempDir
+		// Leave StateDir empty when there is no state path so a disk-backed
+		// syncstore falls back to os.TempDir() instead of filepath.Dir("") == ".".
+		if path != "" {
+			engineConfig.StateDir = filepath.Dir(path)
+		}
 
 		relayManager := relayClient.NewManager(engineCtx, relayURLs, myPrivateKey.PublicKey().String(), engineConfig.MTU)
 		c.statusRecorder.SetRelayMgr(relayManager)

client/internal/engine.go

@@ -22,7 +22,6 @@ import (
 	log "github.com/sirupsen/logrus"
 	"golang.zx2c4.com/wireguard/tun/netstack"
 	"golang.zx2c4.com/wireguard/wgctrl/wgtypes"
-	"google.golang.org/protobuf/proto"
 
 	nberrors "github.com/netbirdio/netbird/client/errors"
 	"github.com/netbirdio/netbird/client/firewall"
@@ -56,6 +55,7 @@ import (
 	"github.com/netbirdio/netbird/client/internal/routemanager"
 	"github.com/netbirdio/netbird/client/internal/routemanager/systemops"
 	"github.com/netbirdio/netbird/client/internal/statemanager"
+	"github.com/netbirdio/netbird/client/internal/syncstore"
 	"github.com/netbirdio/netbird/client/internal/updater"
 	"github.com/netbirdio/netbird/client/jobexec"
 	cProto "github.com/netbirdio/netbird/client/proto"
@@ -148,6 +148,10 @@ type EngineConfig struct {
 
 	LogPath string
 	TempDir string
+
+	// StateDir is the directory holding the state file. The sync response
+	// (network map) is serialized here on platforms that persist it to disk.
+	StateDir string
 }
 
 // EngineServices holds the external service dependencies required by the Engine.
@@ -226,10 +230,15 @@ type Engine struct {
 
 	afpacketCapture *capture.AFPacketCapture
 
-	// Sync response persistence (protected by syncRespMux)
-	syncRespMux         sync.RWMutex
-	persistSyncResponse bool
-	latestSyncResponse  *mgmProto.SyncResponse
+	// Sync response persistence (protected by syncRespMux).
+	// syncStore is nil unless persistence has been enabled; its presence is
+	// what marks persistence as active. The backend (disk or memory) is
+	// selected per-platform; see the syncstore package. syncStoreDir is where
+	// a disk-backed store serializes to.
+	syncRespMux  sync.RWMutex
+	syncStore    syncstore.Store
+	syncStoreDir string
+
 	flowManager         nftypes.FlowManager
 
 	// auto-update
@@ -292,6 +301,7 @@ func NewEngine(
 		jobExecutor:        jobexec.NewExecutor(),
 		clientMetrics:      services.ClientMetrics,
 		updateManager:      services.UpdateManager,
+		syncStoreDir:       config.StateDir,
 	}
 
 	log.Infof("I am: %s", config.WgPrivateKey.PublicKey().String())
@@ -913,19 +923,18 @@ func (e *Engine) handleSync(update *mgmProto.SyncResponse) error {
 	}
 
 	// Persist sync response under the dedicated lock (syncRespMux), not under syncMsgMux.
-	// Read the storage-enabled flag under the syncRespMux too.
+	// A non-nil syncStore is what marks persistence as enabled. Hold the lock for
+	// the whole Set so the store cannot be cleared (disabled / engine close)
+	// mid-call and have this write resurrect a file that was just removed.
 	e.syncRespMux.RLock()
-	enabled := e.persistSyncResponse
-	e.syncRespMux.RUnlock()
-
-	// Store sync response if persistence is enabled
-	if enabled {
-		e.syncRespMux.Lock()
-		e.latestSyncResponse = update
-		e.syncRespMux.Unlock()
-
-		log.Debugf("sync response persisted with serial %d", nm.GetSerial())
+	if e.syncStore != nil {
+		if err := e.syncStore.Set(update); err != nil {
+			log.Errorf("failed to persist sync response: %v", err)
+		} else {
+			log.Debugf("sync response persisted with serial %d", nm.GetSerial())
+		}
 	}
+	e.syncRespMux.RUnlock()
 
 	// only apply new changes and ignore old ones
 	if err := e.updateNetworkMap(nm); err != nil {
@@ -1813,6 +1822,18 @@ func (e *Engine) close() {
 	if err := e.portForwardManager.GracefullyStop(ctx); err != nil {
 		log.Warnf("failed to gracefully stop port forwarding manager: %s", err)
 	}
+
+	// Drop any persisted sync response so its network map does not linger on
+	// disk after the engine stops (and cannot leak into a later run).
+	e.syncRespMux.Lock()
+	store := e.syncStore
+	e.syncStore = nil
+	e.syncRespMux.Unlock()
+	if store != nil {
+		if err := store.Clear(); err != nil {
+			log.Warnf("failed to clear persisted sync response on close: %v", err)
+		}
+	}
 }
 
 func (e *Engine) readInitialSettings() ([]*route.Route, *nbdns.Config, bool, error) {
@@ -2142,45 +2163,42 @@ func (e *Engine) stopDNSServer() {
 	e.statusRecorder.UpdateDNSStates(nsGroupStates)
 }
 
-// SetSyncResponsePersistence enables or disables sync response persistence
+// SetSyncResponsePersistence enables or disables sync response persistence.
+// The store is only instantiated while persistence is enabled; construction
+// itself drops any stale data left over from an earlier run (see syncstore).
 func (e *Engine) SetSyncResponsePersistence(enabled bool) {
 	e.syncRespMux.Lock()
 	defer e.syncRespMux.Unlock()
 
-	if enabled == e.persistSyncResponse {
+	if enabled == (e.syncStore != nil) {
 		return
 	}
-	e.persistSyncResponse = enabled
 	log.Debugf("Sync response persistence is set to %t", enabled)
 
 	if !enabled {
-		e.latestSyncResponse = nil
+		if err := e.syncStore.Clear(); err != nil {
+			log.Warnf("failed to clear persisted sync response: %v", err)
+		}
+		e.syncStore = nil
+		return
 	}
+
+	e.syncStore = syncstore.New(e.syncStoreDir)
 }
 
 // GetLatestSyncResponse returns the stored sync response if persistence is enabled
 func (e *Engine) GetLatestSyncResponse() (*mgmProto.SyncResponse, error) {
+	// Hold the lock for the whole Get so the store cannot be cleared
+	// (disabled / engine close) mid-call.
 	e.syncRespMux.RLock()
-	enabled := e.persistSyncResponse
-	latest := e.latestSyncResponse
-	e.syncRespMux.RUnlock()
+	defer e.syncRespMux.RUnlock()
 
-	if !enabled {
+	if e.syncStore == nil {
 		return nil, errors.New("sync response persistence is disabled")
 	}
 
-	if latest == nil {
-		//nolint:nilnil
-		return nil, nil
-	}
-
-	log.Debugf("Retrieving latest sync response with size %d bytes", proto.Size(latest))
-	sr, ok := proto.Clone(latest).(*mgmProto.SyncResponse)
-	if !ok {
-		return nil, fmt.Errorf("failed to clone sync response")
-	}
-
-	return sr, nil
+	//nolint:nilnil
+	return e.syncStore.Get()
 }
 
 // GetWgAddr returns the wireguard address
@@ -2216,7 +2234,7 @@ func (e *Engine) updateDNSForwarder(
 	enabled bool,
 	fwdEntries []*dnsfwd.ForwarderEntry,
 ) {
-	if e.config.DisableServerRoutes {
+	if e.config.DisableServerRoutes || e.config.BlockInbound {
 		return
 	}
 

client/internal/lazyconn/support.go

@@ -4,6 +4,8 @@ import (
 	"strings"
 
 	"github.com/hashicorp/go-version"
+
+	nbversion "github.com/netbirdio/netbird/version"
 )
 
 var (
@@ -11,7 +13,7 @@ var (
 )
 
 func IsSupported(agentVersion string) bool {
-	if agentVersion == "development" {
+	if nbversion.IsDevelopmentVersion(agentVersion) {
 		return true
 	}
 

client/internal/peer/status.go

@@ -310,8 +310,12 @@ func (d *Status) PeerByIP(ip string) (string, bool) {
 
 // PeerStateByIP returns the full peer State for the given tunnel IP.
 // Matches against either the IPv4 (State.IP) or IPv6 (State.IPv6) tunnel
-// address so dual-stack peers are reachable on either family. Returns the
-// zero State and false when no peer matches or the input is empty.
+// address so dual-stack peers are reachable on either family. Searches
+// both d.peers and d.offlinePeers — peers that have been moved into
+// the offline slice by ReplaceOfflinePeers are still part of the
+// account's roster and callers (DNS filter, embed.Client.IdentityForIP)
+// need to recognise them rather than treating them as unknown. Returns
+// the zero State and false when no peer matches or the input is empty.
 func (d *Status) PeerStateByIP(ip string) (State, bool) {
 	if ip == "" {
 		return State{}, false
@@ -324,6 +328,11 @@ func (d *Status) PeerStateByIP(ip string) (State, bool) {
 			return state, true
 		}
 	}
+	for _, state := range d.offlinePeers {
+		if (state.IP != "" && state.IP == ip) || (state.IPv6 != "" && state.IPv6 == ip) {
+			return state, true
+		}
+	}
 	return State{}, false
 }
 

client/internal/peer/status_test.go

@@ -90,6 +90,28 @@ func TestStatus_PeerStateByIP_MatchesIPv6(t *testing.T) {
 	req.Equal("pk-1", state.PubKey, "matching state must carry the right pub key")
 }
 
+// TestStatus_PeerStateByIP_MatchesOfflinePeers covers peers that have
+// been moved into the offline slice via ReplaceOfflinePeers. Callers
+// (DNS filter, embed.Client.IdentityForIP) need to treat them as known
+// rather than unknown — otherwise authentication / DNS filtering treats
+// known-but-offline peers as foreign IPs.
+func TestStatus_PeerStateByIP_MatchesOfflinePeers(t *testing.T) {
+	status := NewRecorder("https://mgm")
+	req := require.New(t)
+
+	status.ReplaceOfflinePeers([]State{
+		{PubKey: "pk-offline", FQDN: "offline.netbird", IP: "100.64.0.20", IPv6: "fd00::20"},
+	})
+
+	state, ok := status.PeerStateByIP("100.64.0.20")
+	req.True(ok, "offline peer must resolve by IPv4 tunnel address")
+	req.Equal("pk-offline", state.PubKey, "matching state must carry the offline peer's pub key")
+
+	state, ok = status.PeerStateByIP("fd00::20")
+	req.True(ok, "offline peer must resolve by IPv6 tunnel address")
+	req.Equal("pk-offline", state.PubKey, "IPv6 match must carry the offline peer's pub key")
+}
+
 func TestStatus_UpdatePeerFQDN(t *testing.T) {
 	key := "abc"
 	fqdn := "peer-a.netbird.local"

client/internal/syncstore/disk.go

@@ -0,0 +1,99 @@
+package syncstore
+
+import (
+	"context"
+	"errors"
+	"fmt"
+	"os"
+	"path/filepath"
+	"sync"
+
+	log "github.com/sirupsen/logrus"
+	"google.golang.org/protobuf/proto"
+
+	mgmProto "github.com/netbirdio/netbird/shared/management/proto"
+	"github.com/netbirdio/netbird/util"
+)
+
+// syncResponseFileName is the name of the file the sync response is serialized
+// to, placed inside the configured directory (the state directory).
+const syncResponseFileName = "networkmap.pb"
+
+// diskStore serializes the latest sync response to a file on disk instead of
+// keeping it in memory. This trades disk I/O for a much smaller memory
+// footprint, which matters on memory-constrained platforms (iOS).
+type diskStore struct {
+	mu   sync.Mutex
+	path string
+}
+
+// NewDiskStore returns a Store that serializes the sync response to a file in
+// the given directory. If dir is empty it falls back to the OS temp directory.
+//
+// Any file left over from a previous run is removed on construction so a fresh
+// store never reads stale data (e.g. another profile's network map).
+func NewDiskStore(dir string) Store {
+	if dir == "" {
+		dir = os.TempDir()
+	}
+	s := &diskStore{
+		path: filepath.Join(dir, syncResponseFileName),
+	}
+	if err := s.Clear(); err != nil {
+		log.Warnf("failed to clear stale sync response file: %v", err)
+	}
+	return s
+}
+
+func (s *diskStore) Set(resp *mgmProto.SyncResponse) error {
+	if resp == nil {
+		return s.Clear()
+	}
+
+	bs, err := proto.Marshal(resp)
+	if err != nil {
+		return fmt.Errorf("marshal sync response: %w", err)
+	}
+
+	s.mu.Lock()
+	defer s.mu.Unlock()
+
+	if err := util.WriteBytesWithRestrictedPermission(context.Background(), s.path, bs); err != nil {
+		return fmt.Errorf("write sync response to %s: %w", s.path, err)
+	}
+
+	log.Debugf("sync response persisted to %s (%d bytes)", s.path, len(bs))
+	return nil
+}
+
+func (s *diskStore) Get() (*mgmProto.SyncResponse, error) {
+	s.mu.Lock()
+	defer s.mu.Unlock()
+
+	bs, err := os.ReadFile(s.path)
+	if err != nil {
+		if errors.Is(err, os.ErrNotExist) {
+			//nolint:nilnil // nil,nil means "nothing stored", per the Store contract; preserve the original behaviour
+			return nil, nil
+		}
+		return nil, fmt.Errorf("read sync response from %s: %w", s.path, err)
+	}
+
+	resp := &mgmProto.SyncResponse{}
+	if err := proto.Unmarshal(bs, resp); err != nil {
+		return nil, fmt.Errorf("unmarshal sync response: %w", err)
+	}
+
+	log.Debugf("retrieving latest sync response from %s (%d bytes)", s.path, len(bs))
+	return resp, nil
+}
+
+func (s *diskStore) Clear() error {
+	s.mu.Lock()
+	defer s.mu.Unlock()
+
+	if err := os.Remove(s.path); err != nil && !errors.Is(err, os.ErrNotExist) {
+		return fmt.Errorf("remove sync response file %s: %w", s.path, err)
+	}
+	return nil
+}

client/internal/syncstore/factory_ios.go

@@ -0,0 +1,9 @@
+//go:build ios
+
+package syncstore
+
+// New returns the platform default store. On iOS the sync response is
+// serialized to disk (in dir) to keep it out of the constrained process memory.
+func New(dir string) Store {
+	return NewDiskStore(dir)
+}

client/internal/syncstore/factory_other.go

@@ -0,0 +1,9 @@
+//go:build !ios
+
+package syncstore
+
+// New returns the platform default store. On all non-iOS platforms the sync
+// response is kept in memory; dir is unused.
+func New(_ string) Store {
+	return NewMemoryStore()
+}

client/internal/syncstore/memory.go

@@ -0,0 +1,56 @@
+package syncstore
+
+import (
+	"fmt"
+	"sync"
+
+	log "github.com/sirupsen/logrus"
+	"google.golang.org/protobuf/proto"
+
+	mgmProto "github.com/netbirdio/netbird/shared/management/proto"
+)
+
+// memoryStore keeps the latest sync response in memory.
+type memoryStore struct {
+	mu     sync.RWMutex
+	latest *mgmProto.SyncResponse
+}
+
+// NewMemoryStore returns a Store that keeps the sync response in memory.
+func NewMemoryStore() Store {
+	return &memoryStore{}
+}
+
+func (s *memoryStore) Set(resp *mgmProto.SyncResponse) error {
+	s.mu.Lock()
+	defer s.mu.Unlock()
+
+	s.latest = resp
+	return nil
+}
+
+func (s *memoryStore) Get() (*mgmProto.SyncResponse, error) {
+	s.mu.RLock()
+	latest := s.latest
+	s.mu.RUnlock()
+
+	if latest == nil {
+		//nolint:nilnil // nil,nil means "nothing stored", per the Store contract; preserve the original behaviour
+		return nil, nil
+	}
+
+	log.Debugf("retrieving latest sync response with size %d bytes", proto.Size(latest))
+	sr, ok := proto.Clone(latest).(*mgmProto.SyncResponse)
+	if !ok {
+		return nil, fmt.Errorf("clone sync response")
+	}
+	return sr, nil
+}
+
+func (s *memoryStore) Clear() error {
+	s.mu.Lock()
+	defer s.mu.Unlock()
+
+	s.latest = nil
+	return nil
+}

client/internal/syncstore/syncstore.go

@@ -0,0 +1,29 @@
+// Package syncstore stores the latest Management sync response (which carries
+// the network map) for debug bundle generation.
+//
+// The storage backend is selected at build time per operating system: on iOS
+// the response is serialized to disk to keep it out of the (tightly
+// constrained) process memory, while on all other platforms it is kept in
+// memory. The backend is chosen by the New constructor; see factory_ios.go and
+// factory_other.go.
+package syncstore
+
+import (
+	mgmProto "github.com/netbirdio/netbird/shared/management/proto"
+)
+
+// Store persists the latest sync response and returns it on demand.
+//
+// Implementations must be safe for concurrent use.
+type Store interface {
+	// Set stores the given sync response, replacing any previously stored one.
+	Set(resp *mgmProto.SyncResponse) error
+
+	// Get returns the stored sync response, or nil if none is stored.
+	// The returned value is an independent copy that the caller may retain.
+	Get() (*mgmProto.SyncResponse, error)
+
+	// Clear removes any stored sync response. It is safe to call when nothing
+	// is stored.
+	Clear() error
+}

client/internal/updater/manager.go

@@ -19,8 +19,6 @@ import (
 
 const (
 	latestVersion = "latest"
-	// this version will be ignored
-	developmentVersion = "development"
 )
 
 var errNoUpdateState = errors.New("no update state found")
@@ -483,7 +481,7 @@ func (m *Manager) loadAndDeleteUpdateState(ctx context.Context) (*UpdateState, e
 }
 
 func (m *Manager) shouldUpdate(updateVersion *v.Version, forceUpdate bool) bool {
-	if m.currentVersion == developmentVersion {
+	if version.IsDevelopmentVersion(m.currentVersion) {
 		log.Debugf("skipping auto-update, running development version")
 		return false
 	}