Merge remote-tracking branch 'origin/main' into embedded-vnc

# Conflicts:
#	client/cmd/up.go
#	client/internal/engine.go
#	client/internal/profilemanager/config.go
#	client/proto/daemon.pb.go
#	client/server/mdm.go
#	client/server/server.go
#	client/status/status.go
#	client/ui/main.go
#	client/ui/services/windowmanager.go
#	management/internals/shared/grpc/conversion.go
#	management/server/store/sql_store.go
#	management/server/store/sql_store_test.go
#	shared/management/networkmap/envelope.go
#	shared/management/proto/management.pb.go

client/server/forwardingrules.go

@@ -1,54 +0,0 @@
-package server
-
-import (
-	"context"
-
-	firewall "github.com/netbirdio/netbird/client/firewall/manager"
-	"github.com/netbirdio/netbird/client/proto"
-)
-
-func (s *Server) ForwardingRules(context.Context, *proto.EmptyRequest) (*proto.ForwardingRulesResponse, error) {
-	s.mutex.Lock()
-	defer s.mutex.Unlock()
-
-	rules := s.statusRecorder.ForwardingRules()
-	responseRules := make([]*proto.ForwardingRule, 0, len(rules))
-	for _, rule := range rules {
-		respRule := &proto.ForwardingRule{
-			Protocol:           string(rule.Protocol),
-			DestinationPort:    portToProto(rule.DestinationPort),
-			TranslatedAddress:  rule.TranslatedAddress.String(),
-			TranslatedHostname: s.hostNameByTranslateAddress(rule.TranslatedAddress.String()),
-			TranslatedPort:     portToProto(rule.TranslatedPort),
-		}
-		responseRules = append(responseRules, respRule)
-
-	}
-
-	return &proto.ForwardingRulesResponse{Rules: responseRules}, nil
-}
-
-func (s *Server) hostNameByTranslateAddress(ip string) string {
-	hostName, ok := s.statusRecorder.PeerByIP(ip)
-	if !ok {
-		return ip
-	}
-
-	return hostName
-}
-
-func portToProto(port firewall.Port) *proto.PortInfo {
-	var portInfo proto.PortInfo
-
-	if !port.IsRange {
-		portInfo.PortSelection = &proto.PortInfo_Port{Port: uint32(port.Values[0])}
-	} else {
-		portInfo.PortSelection = &proto.PortInfo_Range_{
-			Range: &proto.PortInfo_Range{
-				Start: uint32(port.Values[0]),
-				End:   uint32(port.Values[1]),
-			},
-		}
-	}
-	return &portInfo
-}

client/server/login_account_test.go

@@ -0,0 +1,332 @@
+package server
+
+import (
+	"context"
+	"errors"
+	"path/filepath"
+	"testing"
+	"time"
+
+	"github.com/stretchr/testify/require"
+
+	"github.com/netbirdio/netbird/client/internal"
+	"github.com/netbirdio/netbird/client/internal/auth"
+	"github.com/netbirdio/netbird/client/internal/ipcauth"
+	"github.com/netbirdio/netbird/client/internal/profilemanager"
+	"github.com/netbirdio/netbird/client/mdm"
+	"github.com/netbirdio/netbird/client/proto"
+)
+
+type jwtStoringMDMFetcher struct {
+	cache      *jwtCache
+	caller     ipcauth.Identity
+	generation uint64
+}
+
+func (f *jwtStoringMDMFetcher) Fetch() map[string]any {
+	f.generation = f.cache.currentGeneration()
+	f.cache.store("previous-profile-token", f.caller, time.Minute, f.generation)
+	return nil
+}
+
+type stubOAuthFlow struct {
+	token  auth.TokenInfo
+	onWait func()
+}
+
+func (f *stubOAuthFlow) RequestAuthInfo(context.Context) (auth.AuthFlowInfo, error) {
+	return auth.AuthFlowInfo{}, nil
+}
+
+func (f *stubOAuthFlow) WaitToken(context.Context, auth.AuthFlowInfo) (auth.TokenInfo, error) {
+	if f.onWait != nil {
+		f.onWait()
+	}
+	return f.token, nil
+}
+
+func (f *stubOAuthFlow) GetClientID(context.Context) string {
+	return "stub-client"
+}
+
+func TestWaitSSOLogin_WrongAccountArmsPromptAndFails(t *testing.T) {
+	s := newSSOTestServer(t, "user@example.com", false, "other@example.com")
+	attempts := 0
+	s.loginAttemptFn = func(context.Context, string, string) (internal.StatusType, error) {
+		attempts++
+		return "", nil
+	}
+
+	resp, err := s.WaitSSOLogin(callerCtx(t), &proto.WaitSSOLoginRequest{UserCode: "code"})
+	require.Error(t, err)
+	require.Nil(t, resp)
+	require.Equal(t, 0, attempts, "the wrong account's token reached the management login")
+	require.True(t, s.forceAccountPrompt, "the next login was not armed to ask for the account")
+	require.Nil(t, s.oauthAuthFlow.flow, "the mismatched flow stayed cached for reuse")
+
+	status, stateErr := internal.CtxGetState(s.rootCtx).Status()
+	require.NoError(t, stateErr)
+	require.Equal(t, internal.StatusNeedsLogin, status, "the mismatch must stay retryable")
+}
+
+func TestWaitSSOLogin_WrongAccountAfterPromptProceeds(t *testing.T) {
+	s := newSSOTestServer(t, "user@example.com", true, "other@example.com")
+	attempts := 0
+	s.loginAttemptFn = func(context.Context, string, string) (internal.StatusType, error) {
+		attempts++
+		return "", nil
+	}
+
+	resp, err := s.WaitSSOLogin(callerCtx(t), &proto.WaitSSOLoginRequest{UserCode: "code"})
+	require.NoError(t, err, "a prompted round must not error again on a mismatch")
+	require.NotNil(t, resp)
+	require.Equal(t, "other@example.com", resp.Email)
+	require.Equal(t, 1, attempts)
+	require.False(t, s.forceAccountPrompt)
+}
+
+func TestWaitSSOLogin_MatchingAccountProceeds(t *testing.T) {
+	s := newSSOTestServer(t, "user@example.com", false, "User@Example.com")
+	attempts := 0
+	s.loginAttemptFn = func(context.Context, string, string) (internal.StatusType, error) {
+		attempts++
+		return "", nil
+	}
+
+	resp, err := s.WaitSSOLogin(callerCtx(t), &proto.WaitSSOLoginRequest{UserCode: "code"})
+	require.NoError(t, err)
+	require.NotNil(t, resp)
+	require.Equal(t, 1, attempts)
+	require.False(t, s.forceAccountPrompt)
+}
+
+func TestWaitSSOLogin_NoHintIsNotJudged(t *testing.T) {
+	s := newSSOTestServer(t, "", false, "whoever@example.com")
+	attempts := 0
+	s.loginAttemptFn = func(context.Context, string, string) (internal.StatusType, error) {
+		attempts++
+		return "", nil
+	}
+
+	_, err := s.WaitSSOLogin(callerCtx(t), &proto.WaitSSOLoginRequest{UserCode: "code"})
+	require.NoError(t, err)
+	require.Equal(t, 1, attempts)
+	require.False(t, s.forceAccountPrompt)
+}
+
+func TestSwitchProfile_DropsAccountPromptAndPendingFlow(t *testing.T) {
+	s, ctx, _, _, _ := setupServerWithProfile(t)
+	s.forceAccountPrompt = true
+	cancelled := false
+	s.oauthAuthFlow = oauthAuthFlow{
+		flow:       &stubOAuthFlow{},
+		hint:       "user@example.com",
+		waitCancel: func() { cancelled = true },
+	}
+
+	extendCancelled := false
+	s.extendAuthSessionFlow.Set(&stubOAuthFlow{}, auth.AuthFlowInfo{DeviceCode: "device"})
+	s.extendAuthSessionFlow.SetWaitCancel(func() { extendCancelled = true })
+
+	_, err := s.SwitchProfile(ctx, nil)
+	require.NoError(t, err)
+	require.False(t, s.forceAccountPrompt, "the prompt flag leaked across a profile switch")
+	require.Nil(t, s.oauthAuthFlow.flow, "the previous profile's flow leaked across a profile switch")
+	require.Empty(t, s.oauthAuthFlow.hint)
+	require.True(t, cancelled, "the pending wait was not cancelled")
+
+	require.True(t, extendCancelled, "the pending extend wait was not cancelled")
+	_, _, pending := s.extendAuthSessionFlow.Get()
+	require.False(t, pending, "the previous profile's extend flow leaked across a profile switch")
+}
+
+func TestLogin_ProfileSwitchDropsAccountPromptAndPendingFlow(t *testing.T) {
+	s, _, _, username, cfgPath := setupServerWithProfile(t)
+	s.rootCtx = internal.CtxInitState(context.Background())
+	s.isLoginRequiredFn = func(context.Context) (bool, error) {
+		return true, nil
+	}
+
+	other := "other-profile"
+	otherPath := filepath.Join(filepath.Dir(cfgPath), other+".json")
+	_, err := profilemanager.UpdateOrCreateConfig(profilemanager.ConfigInput{
+		ConfigPath:    otherPath,
+		ManagementURL: "https://api.netbird.io:443",
+	})
+	require.NoError(t, err)
+	breakProfilePrivateKey(t, otherPath)
+
+	s.forceAccountPrompt = true
+	cancelled := false
+	s.oauthAuthFlow = oauthAuthFlow{
+		flow:       &stubOAuthFlow{},
+		hint:       "user@example.com",
+		waitCancel: func() { cancelled = true },
+	}
+
+	extendCancelled := false
+	s.extendAuthSessionFlow.Set(&stubOAuthFlow{}, auth.AuthFlowInfo{DeviceCode: "device"})
+	s.extendAuthSessionFlow.SetWaitCancel(func() { extendCancelled = true })
+
+	generation := s.jwtCache.currentGeneration()
+
+	_, err = s.Login(userCtx(), &proto.LoginRequest{ProfileName: &other, Username: &username})
+	require.Error(t, err, "the broken key must stop the login before a flow is built")
+
+	active, err := s.profileManager.GetActiveProfileState()
+	require.NoError(t, err)
+	require.Equal(t, profilemanager.ID(other), active.ID, "the login did not switch the profile")
+
+	require.False(t, s.forceAccountPrompt, "the prompt flag leaked across a login-driven profile switch")
+	require.Nil(t, s.oauthAuthFlow.flow, "the previous profile's flow leaked across a login-driven profile switch")
+	require.Empty(t, s.oauthAuthFlow.hint)
+	require.True(t, cancelled, "the pending wait was not cancelled")
+
+	require.True(t, extendCancelled, "the pending extend wait was not cancelled")
+	_, _, pending := s.extendAuthSessionFlow.Get()
+	require.False(t, pending, "the previous profile's extend flow leaked across a login-driven profile switch")
+
+	require.Greater(t, s.jwtCache.currentGeneration(), generation, "the previous profile's JWT cache survived a login-driven profile switch")
+}
+
+func TestLogin_ProfileSwitchRejectsJWTObtainedUnderPreviousConfig(t *testing.T) {
+	s, _, _, username, cfgPath := setupServerWithProfile(t)
+	s.rootCtx = internal.CtxInitState(context.Background())
+	s.isLoginRequiredFn = func(context.Context) (bool, error) {
+		return false, errors.New("stop once the config is swapped")
+	}
+
+	other := "other-profile"
+	_, err := profilemanager.UpdateOrCreateConfig(profilemanager.ConfigInput{
+		ConfigPath:    filepath.Join(filepath.Dir(cfgPath), other+".json"),
+		ManagementURL: "https://api.netbird.io:443",
+	})
+	require.NoError(t, err)
+
+	// getConfig loads the MDM policy right before Login swaps s.config, so the
+	// fetcher runs where a RequestJWTAuth racing the login would land: after the
+	// switch dropped the previous profile's state, while s.config still belongs
+	// to the previous profile. It caches a token under the generation current at
+	// that point, as WaitJWTToken would.
+	generation := s.jwtCache.currentGeneration()
+	fetcher := &jwtStoringMDMFetcher{cache: s.jwtCache, caller: unprivilegedIdentity()}
+	s.mdmLoader = mdm.NewLoader(fetcher)
+
+	_, err = s.Login(userCtx(), &proto.LoginRequest{ProfileName: &other, Username: &username})
+	require.Error(t, err)
+
+	require.Greater(t, fetcher.generation, generation, "the token was not cached after the switch dropped the previous profile's state")
+	_, found := s.jwtCache.get(fetcher.caller)
+	require.False(t, found, "a JWT obtained under the previous profile's config survived the login-driven switch")
+}
+
+func TestLogin_SameProfileKeepsPendingFlow(t *testing.T) {
+	s, _, profName, username, cfgPath := setupServerWithProfile(t)
+	s.rootCtx = internal.CtxInitState(context.Background())
+	s.isLoginRequiredFn = func(context.Context) (bool, error) {
+		return true, nil
+	}
+	breakProfilePrivateKey(t, cfgPath)
+
+	cancelled := false
+	flow := &stubOAuthFlow{}
+	s.oauthAuthFlow = oauthAuthFlow{
+		flow:       flow,
+		hint:       "user@example.com",
+		waitCancel: func() { cancelled = true },
+	}
+
+	_, err := s.Login(userCtx(), &proto.LoginRequest{ProfileName: &profName, Username: &username})
+	require.Error(t, err, "the broken key must stop the login before a flow is built")
+
+	require.Equal(t, flow, s.oauthAuthFlow.flow, "a login on the same profile dropped the flow a second client could join")
+	require.Equal(t, "user@example.com", s.oauthAuthFlow.hint)
+	require.False(t, cancelled, "a login on the same profile cancelled the pending wait")
+}
+
+func TestWaitSSOLogin_JudgesTheFlowThatProducedTheToken(t *testing.T) {
+	s := newSSOTestServer(t, "user@example.com", false, "user@example.com")
+	attempts := 0
+	s.loginAttemptFn = func(context.Context, string, string) (internal.StatusType, error) {
+		attempts++
+		return "", nil
+	}
+
+	flow := s.oauthAuthFlow.flow.(*stubOAuthFlow)
+	flow.onWait = func() {
+		s.mutex.Lock()
+		defer s.mutex.Unlock()
+		s.oauthAuthFlow.hint = "someone-else@example.com"
+	}
+
+	resp, err := s.WaitSSOLogin(callerCtx(t), &proto.WaitSSOLoginRequest{UserCode: "code"})
+	require.NoError(t, err, "a flow replaced mid-wait must not decide this wait's verdict")
+	require.NotNil(t, resp)
+	require.Equal(t, 1, attempts)
+	require.False(t, s.forceAccountPrompt, "the prompt was armed off another flow's hint")
+}
+
+func TestReuseOAuthFlow_ForcedPromptRefusesTheCachedFlow(t *testing.T) {
+	s := New(internal.CtxInitState(context.Background()), "console", "", false, false, false, false)
+	cancelled := false
+	s.oauthAuthFlow = oauthAuthFlow{
+		flow:       &stubOAuthFlow{},
+		info:       auth.AuthFlowInfo{UserCode: "code"},
+		expiresAt:  time.Now().Add(time.Hour),
+		waitCancel: func() { cancelled = true },
+	}
+
+	state := internal.CtxGetState(s.rootCtx)
+	resp := s.reuseOAuthFlow(context.Background(), &stubOAuthFlow{}, state, true)
+	require.Nil(t, resp, "a forced account prompt reused the flow that skipped it")
+	require.True(t, cancelled, "the predecessor wait was orphaned")
+
+	resp = s.reuseOAuthFlow(context.Background(), &stubOAuthFlow{}, state, false)
+	require.NotNil(t, resp, "an unforced login stopped reusing a live flow")
+	require.Equal(t, "code", resp.UserCode)
+}
+
+func TestReplaceOAuthFlow_CancelsTheDisplacedWait(t *testing.T) {
+	s := New(internal.CtxInitState(context.Background()), "console", "", false, false, false, false)
+	cancelled := false
+	s.oauthAuthFlow = oauthAuthFlow{
+		flow:            &stubOAuthFlow{},
+		info:            auth.AuthFlowInfo{UserCode: "code"},
+		expiresAt:       time.Now().Add(time.Hour),
+		hint:            "user@example.com",
+		accountPrompted: true,
+		waitCancel:      func() { cancelled = true },
+	}
+
+	next := &stubOAuthFlow{}
+	s.replaceOAuthFlow(oauthAuthFlow{flow: next, info: auth.AuthFlowInfo{UserCode: "next"}})
+
+	require.True(t, cancelled, "the displaced wait was left without an owner")
+	require.Equal(t, next, s.oauthAuthFlow.flow)
+	require.Equal(t, "next", s.oauthAuthFlow.info.UserCode)
+	require.Empty(t, s.oauthAuthFlow.hint, "the previous flow's hint survived the replacement")
+	require.False(t, s.oauthAuthFlow.accountPrompted)
+	require.Nil(t, s.oauthAuthFlow.waitCancel, "the consumed cancel stayed on the record")
+}
+
+func newSSOTestServer(t *testing.T, hint string, accountPrompted bool, tokenEmail string) *Server {
+	t.Helper()
+	s := New(internal.CtxInitState(context.Background()), "console", "", false, false, false, false)
+	s.oauthAuthFlow = oauthAuthFlow{
+		flow:            &stubOAuthFlow{token: auth.TokenInfo{Email: tokenEmail, EmailClaim: tokenEmail}},
+		info:            auth.AuthFlowInfo{UserCode: "code"},
+		expiresAt:       time.Now().Add(time.Minute),
+		hint:            hint,
+		accountPrompted: accountPrompted,
+	}
+	return s
+}
+
+// callerCtx is the gRPC caller's context. WaitSSOLogin parks a goroutine on it
+// for the whole browser leg, so a test that never cancels leaks one.
+func callerCtx(t *testing.T) context.Context {
+	t.Helper()
+	ctx, cancel := context.WithCancel(context.Background())
+	t.Cleanup(cancel)
+	return ctx
+}

client/server/login_gate_test.go

@@ -93,7 +93,7 @@ func TestLogin_ChangeThatBecomesPrivilegedMidRequestHasNoSideEffects(t *testing.
 	require.NoError(t, err)
 	require.Equal(t, profilemanager.ID(activeProfile), active.ID, "the refused login switched the active profile anyway")
 
-	stored, err := profilemanager.ReadConfig(targetPath)
+	stored, err := profilemanager.GetExistingConfig(targetPath)
 	require.NoError(t, err)
 	require.Equal(t, "https://api.netbird.io:443", stored.ManagementURL.String(), "the refused login moved the management URL")
 }

client/server/login_overrides_test.go

@@ -7,6 +7,7 @@ import (
 	"github.com/stretchr/testify/require"
 
 	"github.com/netbirdio/netbird/client/internal/profilemanager"
+	"github.com/netbirdio/netbird/client/proto"
 )
 
 func TestPersistLoginOverrides(t *testing.T) {
@@ -80,10 +81,13 @@ func TestPersistLoginOverrides(t *testing.T) {
 			require.NoError(t, err, "seed config")
 
 			activeProf := &profilemanager.ActiveProfileState{ID: "default"}
-			err = persistLoginOverrides(activeProf, tt.newMgmtURL, tt.newPSK)
+			err = persistLoginOverrides(activeProf, &proto.LoginRequest{
+				ManagementUrl:        tt.newMgmtURL,
+				OptionalPreSharedKey: tt.newPSK,
+			})
 			require.NoError(t, err, "persistLoginOverrides")
 
-			cfg, err := profilemanager.ReadConfig(profilemanager.DefaultConfigPath)
+			cfg, err := profilemanager.ReadConfigOrDefault(profilemanager.DefaultConfigPath)
 			require.NoError(t, err, "read back config")
 
 			require.Equal(t, tt.wantMgmtURL, cfg.ManagementURL.String(), "management URL")

client/server/logout_gate_test.go

@@ -129,7 +129,7 @@ func TestLogout_ForeignUserProfileDoesNotUseTheRunningConfig(t *testing.T) {
 	// refused with PermissionDenied. The namesake profile does not, so the
 	// correct path gets as far as dialing its own unreachable management URL.
 	enableSSHOnProfile(t, cfgPath)
-	running, err := profilemanager.GetConfig(cfgPath)
+	running, err := profilemanager.GetExistingConfig(cfgPath)
 	require.NoError(t, err)
 	s.config = running
 	s.connectClient = newDummyConnectClient(context.Background())

client/server/mdm.go

@@ -182,96 +182,6 @@ func mdmManagedFieldConflicts(msg *proto.SetConfigRequest, policy *mdm.Policy) [
 	})
 }
 
-// setConfigRequestHasConfigOverrides reports whether the SetConfigRequest
-// carries ANY field that would actually mutate the persisted config.
-// The CLI builds a SetConfigRequest unconditionally on every
-// `netbird up` (see setupSetConfigReq in cmd/up.go) — a plain
-// `netbird up` produces a request with every field at its zero value;
-// the gate must skip such no-op invocations or it would always fire
-// even when the user did not pass any --flag. Returns false on a nil
-// msg; true when any management/admin URL, PSK, DNS/NAT list+clean
-// flag, interface/port/MTU, or any optional bool/duration field is set.
-func setConfigRequestHasConfigOverrides(msg *proto.SetConfigRequest) bool {
-	if msg == nil {
-		return false
-	}
-	return msg.ManagementUrl != "" ||
-		msg.AdminURL != "" ||
-		msg.OptionalPreSharedKey != nil ||
-		len(msg.CustomDNSAddress) > 0 ||
-		len(msg.NatExternalIPs) > 0 || msg.CleanNATExternalIPs ||
-		len(msg.ExtraIFaceBlacklist) > 0 ||
-		len(msg.DnsLabels) > 0 || msg.CleanDNSLabels ||
-		msg.DnsRouteInterval != nil ||
-		msg.RosenpassEnabled != nil ||
-		msg.RosenpassPermissive != nil ||
-		msg.InterfaceName != nil ||
-		msg.WireguardPort != nil ||
-		msg.Mtu != nil ||
-		msg.DisableAutoConnect != nil ||
-		msg.ServerSSHAllowed != nil ||
-		msg.RemoteJobsAllowed != nil ||
-		msg.ServerVNCAllowed != nil ||
-		msg.DisableVNCApproval != nil ||
-		msg.NetworkMonitor != nil ||
-		msg.DisableClientRoutes != nil ||
-		msg.DisableServerRoutes != nil ||
-		msg.DisableDns != nil ||
-		msg.DisableFirewall != nil ||
-		msg.BlockLanAccess != nil ||
-		msg.DisableNotifications != nil ||
-		msg.BlockInbound != nil ||
-		msg.DisableIpv6 != nil ||
-		msg.EnableSSHRoot != nil ||
-		msg.EnableSSHSFTP != nil ||
-		msg.EnableSSHLocalPortForwarding != nil ||
-		msg.EnableSSHRemotePortForwarding != nil ||
-		msg.DisableSSHAuth != nil ||
-		msg.SshJWTCacheTTL != nil ||
-		msg.EnableLocalMetrics != nil ||
-		msg.LocalMetricsAddress != nil
-}
-
-// loginRequestHasConfigOverrides reports whether the LoginRequest
-// carries ANY field that would mutate persisted daemon configuration
-// (as opposed to pure-auth fields like setupKey, hostname, hint,
-// profileName, username). Used by the Login handler to decide whether
-// the `--disable-update-settings` / MDM gates must run: a re-auth that
-// changes nothing about the configuration is always allowed.
-func loginRequestHasConfigOverrides(msg *proto.LoginRequest) bool {
-	if msg == nil {
-		return false
-	}
-	return msg.ManagementUrl != "" ||
-		msg.AdminURL != "" ||
-		msg.PreSharedKey != "" || //nolint:staticcheck // SA1019: legacy proto field still accepted by Login
-		msg.OptionalPreSharedKey != nil ||
-		len(msg.CustomDNSAddress) > 0 ||
-		len(msg.NatExternalIPs) > 0 || msg.CleanNATExternalIPs ||
-		msg.RosenpassEnabled != nil ||
-		msg.InterfaceName != nil ||
-		msg.WireguardPort != nil ||
-		msg.DisableAutoConnect != nil ||
-		msg.ServerSSHAllowed != nil ||
-		msg.RemoteJobsAllowed != nil ||
-		msg.ServerVNCAllowed != nil ||
-		msg.DisableVNCApproval != nil ||
-		msg.RosenpassPermissive != nil ||
-		len(msg.ExtraIFaceBlacklist) > 0 ||
-		msg.NetworkMonitor != nil ||
-		msg.DnsRouteInterval != nil ||
-		msg.DisableClientRoutes != nil ||
-		msg.DisableServerRoutes != nil ||
-		msg.DisableDns != nil ||
-		msg.DisableFirewall != nil ||
-		msg.BlockLanAccess != nil ||
-		msg.DisableNotifications != nil ||
-		len(msg.DnsLabels) > 0 || msg.CleanDNSLabels ||
-		msg.BlockInbound != nil ||
-		msg.EnableLocalMetrics != nil ||
-		msg.LocalMetricsAddress != nil
-}
-
 // loginRequestMDMConflicts mirrors mdmManagedFieldConflicts but for the
 // LoginRequest surface. Same value-aware semantics: a field set to the
 // MDM-enforced value is a no-op echo, not a conflict; only a divergent

client/server/provision_identity_test.go

@@ -0,0 +1,59 @@
+package server
+
+import (
+	"os"
+	"path/filepath"
+	"testing"
+
+	"github.com/stretchr/testify/require"
+
+	"github.com/netbirdio/netbird/client/internal/profilemanager"
+)
+
+// The daemon provisions the peer's identity and persists it, because a key that
+// stayed in memory would come back different on the next start and register a
+// second peer. Provisioning is idempotent: a profile that already has an
+// identity keeps the one on disk.
+func TestProvisionProfileIdentity(t *testing.T) {
+	origDir := profilemanager.DefaultConfigPathDir
+	origPath := profilemanager.DefaultConfigPath
+	t.Cleanup(func() {
+		profilemanager.DefaultConfigPathDir = origDir
+		profilemanager.DefaultConfigPath = origPath
+	})
+
+	dir := t.TempDir()
+	profilemanager.DefaultConfigPathDir = dir
+	profilemanager.DefaultConfigPath = filepath.Join(dir, "default.json")
+
+	activeProf := &profilemanager.ActiveProfileState{ID: "default"}
+
+	t.Run("a profile with no file is provisioned and written", func(t *testing.T) {
+		_, err := os.Stat(profilemanager.DefaultConfigPath)
+		require.True(t, os.IsNotExist(err), "the fixture starts without a config file")
+
+		config, existed, err := provisionProfileIdentity(activeProf)
+		require.NoError(t, err)
+		require.False(t, existed, "the file was reported as pre-existing")
+		require.NotEmpty(t, config.PrivateKey)
+
+		stored, err := profilemanager.GetExistingConfig(profilemanager.DefaultConfigPath)
+		require.NoError(t, err, "provisioning did not write the config out")
+		require.Equal(t, config.PrivateKey, stored.PrivateKey, "the persisted identity is not the one returned")
+		require.NotEmpty(t, stored.SSHKey)
+	})
+
+	t.Run("a second call keeps the identity on disk", func(t *testing.T) {
+		before, err := profilemanager.GetExistingConfig(profilemanager.DefaultConfigPath)
+		require.NoError(t, err)
+
+		config, existed, err := provisionProfileIdentity(activeProf)
+		require.NoError(t, err)
+		require.True(t, existed)
+		require.Equal(t, before.PrivateKey, config.PrivateKey, "provisioning minted a second identity")
+
+		after, err := profilemanager.GetExistingConfig(profilemanager.DefaultConfigPath)
+		require.NoError(t, err)
+		require.Equal(t, before.PrivateKey, after.PrivateKey, "provisioning rewrote the stored identity")
+	})
+}

client/server/server.go

@@ -58,14 +58,23 @@ const (
 	// JWT token cache TTL for the client daemon (disabled by default)
 	defaultJWTCacheTTL = 0
 
-	errRestoreResidualState   = "failed to restore residual state: %v"
-	errProfilesDisabled       = "profiles are disabled, you cannot use this feature without profiles enabled"
+	errRestoreResidualState = "failed to restore residual state: %v"
+	errProfilesDisabled     = "profiles are disabled, you cannot use this feature without profiles enabled"
+	// errUpdateSettingsDisabled is returned with codes.FailedPrecondition, not
+	// codes.Unavailable: the daemon answered, and it refused. Unavailable means
+	// "the daemon cannot serve this", which is why the CLI downgrades it to a
+	// warning and the GUI reads it as an unreachable daemon — both wrong for a
+	// refusal the caller has to act on.
 	errUpdateSettingsDisabled = "update settings are disabled, you cannot use this feature without update settings enabled"
 	errNetworksDisabled       = "network selection is disabled by the administrator"
 )
 
 var ErrServiceNotUp = errors.New("service is not up")
 
+type statusSetter interface {
+	Set(update internal.StatusType)
+}
+
 // Server for service control.
 type Server struct {
 	rootCtx   context.Context
@@ -83,6 +92,12 @@ type Server struct {
 	uiLogPath string
 
 	oauthAuthFlow oauthAuthFlow
+	// forceAccountPrompt makes the next startSSOLogin build its flow with a
+	// forced account prompt. Armed when a login came back for an account other
+	// than the hinted one: that flow's browser is gone, so the correction has to
+	// ride on the user's next connect. Guarded by mutex; deliberately not
+	// persisted — a lost flag only costs one more mismatch round.
+	forceAccountPrompt bool
 	// extendAuthSessionFlow holds the pending PKCE flow created by
 	// RequestExtendAuthSession until WaitExtendAuthSession resolves it.
 	// Kept separate from oauthAuthFlow (which is reserved for the SSH
@@ -176,6 +191,14 @@ type oauthAuthFlow struct {
 	cacheGeneration uint64
 
 	waitCancel context.CancelFunc
+	// hint is the account the flow was asked to sign in (login_hint). The token
+	// that comes back is compared against it; empty means nothing to compare.
+	hint string
+	// accountPrompted records that this flow already asked the IdP to re-decide
+	// the account (or could not ask — the device flow has no way to). A token
+	// for the wrong account is then let through with a warning instead of
+	// erroring again, so the flow cannot loop.
+	accountPrompted bool
 }
 
 // New server instance constructor.
@@ -496,16 +519,27 @@ func (s *Server) SetConfig(callerCtx context.Context, msg *proto.SetConfigReques
 	s.mutex.Lock()
 	defer s.mutex.Unlock()
 
-	// Skip the update-settings gate when the request carries no actual
-	// overrides: the CLI builds a SetConfigRequest unconditionally on
-	// every `netbird up` (setupSetConfigReq in cmd/up.go), so a plain
-	// `netbird up` would otherwise always trip the gate and surface a
-	// misleading "setConfig method is not available" warning, even when
-	// the user did not pass any config flag.
-	if setConfigRequestHasConfigOverrides(msg) {
-		if s.checkUpdateSettingsDisabled() {
-			return nil, gstatus.Errorf(codes.Unavailable, errUpdateSettingsDisabled)
-		}
+	stored, err := s.storedProfileConfig(msg.ProfileName, msg.Username)
+	if err != nil {
+		return nil, err
+	}
+
+	config, err := s.setConfigInputFromRequest(msg)
+	if err != nil {
+		return nil, err
+	}
+
+	// Update-settings gate: refuse the request only when it would actually
+	// change a persisted setting. The CLI builds a SetConfigRequest
+	// unconditionally on every `netbird up` (setupSetConfigReq in
+	// cmd/up.go) and fills it from its flags and environment, so a service
+	// or container that restates the configuration it already runs with
+	// must pass the gate. Deciding this on field presence alone refused
+	// those callers, and — through the identical gate in Login — refused
+	// their login too, which left a client configured by environment
+	// (NB_MANAGEMENT_URL and friends) unable to come up at all.
+	if s.checkUpdateSettingsDisabled() && configChangeRequested(stored, config) {
+		return nil, gstatus.Errorf(codes.FailedPrecondition, errUpdateSettingsDisabled)
 	}
 
 	// MDM gate: refuse the whole request if any of its fields is enforced
@@ -517,19 +551,10 @@ func (s *Server) SetConfig(callerCtx context.Context, msg *proto.SetConfigReques
 		return nil, err
 	}
 
-	stored, err := s.storedProfileConfig(msg.ProfileName, msg.Username)
-	if err != nil {
-		return nil, err
-	}
 	if err := requirePrivilegeForConfigChange(callerCtx, stored, privilegedChangeFromSetConfig(msg)); err != nil {
 		return nil, err
 	}
 
-	config, err := s.setConfigInputFromRequest(msg)
-	if err != nil {
-		return nil, err
-	}
-
 	updatedConf, err := profilemanager.UpdateConfig(config)
 	if err != nil {
 		log.Errorf("failed to update profile config: %v", err)
@@ -647,37 +672,45 @@ func (s *Server) setConfigInputFromRequest(msg *proto.SetConfigRequest) (profile
 
 // Login uses setup key to prepare configuration for the daemon.
 func (s *Server) Login(callerCtx context.Context, msg *proto.LoginRequest) (*proto.LoginResponse, error) {
+	activeProf, err := s.profileManager.GetActiveProfileState()
+	if err != nil {
+		return nil, fmt.Errorf("failed to get active profile state: %w", err)
+	}
+
+	// The stored config of the profile this request targets backs all three
+	// gates below. It is read before anything changes daemon state, so a
+	// refused login neither switches the profile nor cancels a login already
+	// in progress, and it is the profile the switch further down would
+	// activate.
+	stored, err := s.storedLoginConfig(activeProf, msg)
+	if err != nil {
+		return nil, err
+	}
+
 	// Config-override gates. LoginRequest carries the same surface as
 	// SetConfigRequest (managementUrl, PSK, ssh/rosenpass/port toggles,
 	// ...), so the same protections must apply. Without these the CLI
 	// command `netbird up --management-url=X` (which falls through to
 	// Login when SetConfig is rejected — see cmd/up.go) would silently
 	// bypass `--disable-update-settings` and any MDM policy.
-	if loginRequestHasConfigOverrides(msg) {
-		if s.checkUpdateSettingsDisabled() {
-			return nil, gstatus.Errorf(codes.Unavailable, errUpdateSettingsDisabled)
-		}
-		policy := s.mdmLoader.Load()
-		if err := rejectMDMManagedFieldConflicts(loginRequestMDMConflicts(msg, policy)); err != nil {
-			return nil, err
-		}
+	//
+	// The update-settings gate is value-aware, as in SetConfig: it looks at
+	// what a login would actually persist (loginOverridesInput) and refuses
+	// only a real divergence from the stored config. A login that restates
+	// the values already on disk changes nothing, so it must go through —
+	// that is what keeps a re-login, or a container restart carrying
+	// NB_MANAGEMENT_URL, working with the kill switch on.
+	if s.checkUpdateSettingsDisabled() && configChangeRequested(stored, loginOverridesInput(msg)) {
+		return nil, gstatus.Errorf(codes.FailedPrecondition, errUpdateSettingsDisabled)
 	}
 
-	activeProf, err := s.profileManager.GetActiveProfileState()
-	if err != nil {
-		log.Errorf("failed to get active profile state: %v", err)
-		return nil, fmt.Errorf("failed to get active profile state: %w", err)
+	policy := s.mdmLoader.Load()
+	if err := rejectMDMManagedFieldConflicts(loginRequestMDMConflicts(msg, policy)); err != nil {
+		return nil, err
 	}
 
 	// Privilege gate: same restrictions as SetConfig, since LoginRequest can carry
-	// the same fields. It runs before anything here changes daemon state, so a
-	// refused login neither switches the profile nor cancels a login already in
-	// progress, and it reads the profile the request targets, which is the one the
-	// switch below would activate.
-	stored, err := s.storedLoginConfig(activeProf, msg)
-	if err != nil {
-		return nil, err
-	}
+	// the same fields.
 	if err := requirePrivilegeForConfigChange(callerCtx, stored, privilegedChangeFromLogin(msg)); err != nil {
 		return nil, err
 	}
@@ -695,7 +728,7 @@ func (s *Server) Login(callerCtx context.Context, msg *proto.LoginRequest) (*pro
 		}
 	}()
 
-	ctx, activeProf, err := s.authorizeAndPrepareLogin(callerCtx, msg, activeProf)
+	ctx, activeProf, switched, err := s.authorizeAndPrepareLogin(callerCtx, msg, activeProf)
 	if err != nil {
 		// The RPC boundary is where this gets recorded: nothing logs handler
 		// errors for us, and a caller that retries would otherwise leave no
@@ -725,6 +758,9 @@ func (s *Server) Login(callerCtx context.Context, msg *proto.LoginRequest) (*pro
 	}
 	s.mutex.Lock()
 	s.config = config
+	if switched {
+		s.jwtCache.clear()
+	}
 	s.mutex.Unlock()
 
 	s.localMetrics.Reconcile(config.LocalMetricsEnabled, config.LocalMetricsAddress)
@@ -747,7 +783,7 @@ func (s *Server) Login(callerCtx context.Context, msg *proto.LoginRequest) (*pro
 	}
 
 	if msg.SetupKey == "" {
-		return s.beginSSOLogin(ctx, config, msg)
+		return s.startSSOLogin(ctx, msg, config, state)
 	}
 
 	// Setup-key path: we are about to dial Management with the key, so the
@@ -763,23 +799,31 @@ func (s *Server) Login(callerCtx context.Context, msg *proto.LoginRequest) (*pro
 	return &proto.LoginResponse{}, nil
 }
 
-// beginSSOLogin starts the browser leg of a login that carries no setup key and
-// returns the response that parks the caller on it.
-func (s *Server) beginSSOLogin(ctx context.Context, config *profilemanager.Config, msg *proto.LoginRequest) (*proto.LoginResponse, error) {
-	state := internal.CtxGetState(s.rootCtx)
-
+// startSSOLogin opens the interactive leg of a login: it reuses the in-flight
+// OAuth flow when one is still valid for the same client, and otherwise
+// requests fresh auth info and parks the daemon on StatusNeedsLogin.
+func (s *Server) startSSOLogin(ctx context.Context, msg *proto.LoginRequest, config *profilemanager.Config, state statusSetter) (*proto.LoginResponse, error) {
 	hint := ""
 	if msg.Hint != nil {
 		hint = *msg.Hint
 	}
-	oAuthFlow, err := auth.NewOAuthFlow(ctx, config, msg.IsUnixDesktopClient, false, hint)
+	oAuthFlow, err := auth.NewOAuthFlow(ctx, config, msg.IsUnixDesktopClient, false, hint, false)
 	if err != nil {
 		state.Set(internal.StatusLoginFailed)
 		return nil, err
 	}
 
-	if resp := s.pendingOAuthFlowResponse(ctx, oAuthFlow); resp != nil {
-		state.Set(internal.StatusNeedsLogin)
+	s.mutex.Lock()
+	promptForAccount := s.forceAccountPrompt
+	s.forceAccountPrompt = false
+	s.mutex.Unlock()
+	if promptForAccount && auth.RetryFlowForAccount(oAuthFlow) == nil {
+		// The device flow cannot ask; run it as-is. accountPrompted still goes
+		// true below so a second mismatch is let through instead of looping.
+		log.Warnf("the previous login returned a different account, but this flow cannot ask the IdP to choose one")
+	}
+
+	if resp := s.reuseOAuthFlow(ctx, oAuthFlow, state, promptForAccount); resp != nil {
 		return resp, nil
 	}
 
@@ -789,11 +833,13 @@ func (s *Server) beginSSOLogin(ctx context.Context, config *profilemanager.Confi
 		return nil, err
 	}
 
-	s.mutex.Lock()
-	s.oauthAuthFlow.flow = oAuthFlow
-	s.oauthAuthFlow.info = authInfo
-	s.oauthAuthFlow.expiresAt = time.Now().Add(time.Duration(authInfo.ExpiresIn) * time.Second)
-	s.mutex.Unlock()
+	s.replaceOAuthFlow(oauthAuthFlow{
+		flow:            oAuthFlow,
+		info:            authInfo,
+		expiresAt:       time.Now().Add(time.Duration(authInfo.ExpiresIn) * time.Second),
+		hint:            hint,
+		accountPrompted: promptForAccount,
+	})
 
 	state.Set(internal.StatusNeedsLogin)
 
@@ -805,32 +851,74 @@ func (s *Server) beginSSOLogin(ctx context.Context, config *profilemanager.Confi
 	}, nil
 }
 
-// pendingOAuthFlowResponse returns the in-flight flow's response when it
-// targets the same IdP client and has enough time left for the user to finish
-// the browser leg, so a second login joins the pending flow instead of opening
-// a competing one. A flow too close to expiry has its waiter cancelled and nil
-// returned, leaving the caller to start a fresh flow.
-func (s *Server) pendingOAuthFlowResponse(ctx context.Context, oAuthFlow auth.OAuthFlow) *proto.LoginResponse {
-	if s.oauthAuthFlow.flow == nil || s.oauthAuthFlow.flow.GetClientID(ctx) != oAuthFlow.GetClientID(ctx) {
+// replaceOAuthFlow installs next as the shared OAuth flow record and takes over
+// the wait it displaces, so a WaitSSOLogin still parked on the old flow is not
+// left without an owner: nothing would preempt it, and it could go on to run
+// attemptLogin or mutate the record behind the new flow.
+//
+// The displaced cancel is read in the same critical section that replaces the
+// record, so two callers racing here cannot both take the same predecessor. The
+// cancel runs after the unlock — the displaced wait takes s.mutex as it unwinds.
+func (s *Server) replaceOAuthFlow(next oauthAuthFlow) {
+	s.mutex.Lock()
+	staleCancel := s.oauthAuthFlow.waitCancel
+	s.oauthAuthFlow = next
+	s.mutex.Unlock()
+
+	if staleCancel != nil {
+		staleCancel()
+	}
+}
+
+func (s *Server) expireOAuthFlow(flow auth.OAuthFlow) {
+	s.mutex.Lock()
+	if s.oauthAuthFlow.flow == flow {
+		s.oauthAuthFlow.expiresAt = time.Now()
+	}
+	s.mutex.Unlock()
+}
+
+// reuseOAuthFlow returns the cached auth info when the previous flow targets
+// the same client and still has enough life left, and otherwise cancels the
+// stale wait and returns nil so the caller requests a fresh flow.
+//
+// promptForAccount rules reuse out: the cached flow was built without the
+// account prompt, so handing its URL back would repeat the silent
+// authorization that returned the wrong account — and with the flag already
+// consumed, no later round would ask either. The predecessor's wait is still
+// cancelled on the way out, so it is not orphaned on its device-code window.
+//
+// The whole decision runs off one snapshot taken under s.mutex: a concurrent
+// WaitSSOLogin replaces waitCancel and expires the flow, so reading the fields
+// one at a time could cancel a wait that no longer belongs to the flow just
+// judged stale, or answer with auth info from a flow that was already replaced.
+// The cancel itself is called after unlocking — it runs arbitrary teardown, and
+// WaitSSOLogin takes s.mutex on the way out.
+func (s *Server) reuseOAuthFlow(ctx context.Context, oAuthFlow auth.OAuthFlow, state statusSetter, promptForAccount bool) *proto.LoginResponse {
+	s.mutex.Lock()
+	current := s.oauthAuthFlow
+	s.mutex.Unlock()
+
+	if current.flow == nil || current.flow.GetClientID(ctx) != oAuthFlow.GetClientID(ctx) {
 		return nil
 	}
 
-	if s.oauthAuthFlow.expiresAt.After(time.Now().Add(90 * time.Second)) {
-		log.Debugf("using previous oauth flow info")
-		return &proto.LoginResponse{
-			NeedsSSOLogin:           true,
-			VerificationURI:         s.oauthAuthFlow.info.VerificationURI,
-			VerificationURIComplete: s.oauthAuthFlow.info.VerificationURIComplete,
-			UserCode:                s.oauthAuthFlow.info.UserCode,
+	if promptForAccount || !current.expiresAt.After(time.Now().Add(90*time.Second)) {
+		log.Warnf("canceling previous waiting execution")
+		if current.waitCancel != nil {
+			current.waitCancel()
 		}
+		return nil
 	}
 
-	log.Warnf("canceling previous waiting execution")
-	if s.oauthAuthFlow.waitCancel != nil {
-		s.oauthAuthFlow.waitCancel()
+	log.Debugf("using previous oauth flow info")
+	state.Set(internal.StatusNeedsLogin)
+	return &proto.LoginResponse{
+		NeedsSSOLogin:           true,
+		VerificationURI:         current.info.VerificationURI,
+		VerificationURIComplete: current.info.VerificationURIComplete,
+		UserCode:                current.info.UserCode,
 	}
-
-	return nil
 }
 
 // WaitSSOLogin validates the supplied userCode against the in-flight OAuth
@@ -903,9 +991,16 @@ func (s *Server) WaitSSOLogin(callerCtx context.Context, msg *proto.WaitSSOLogin
 	}
 
 	s.actCancel = cancel
+	// One snapshot of the flow this wait belongs to. hint and accountPrompted
+	// are judged against the token that comes back below, and WaitToken blocks
+	// for the whole browser leg: a concurrent Login or RequestJWTAuth replaces
+	// s.oauthAuthFlow meanwhile, so re-reading them after the wait would judge
+	// this flow's token against another flow's account.
+	pending := s.oauthAuthFlow
+	flow := pending.flow
 	s.mutex.Unlock()
 
-	if s.oauthAuthFlow.flow == nil {
+	if flow == nil {
 		return nil, gstatus.Errorf(codes.Internal, "oauth flow is not initialized")
 	}
 
@@ -923,31 +1018,36 @@ func (s *Server) WaitSSOLogin(callerCtx context.Context, msg *proto.WaitSSOLogin
 	// the affordance instead of a Connecting that never resolves.
 	state.Set(internal.StatusNeedsLogin)
 
-	s.mutex.Lock()
-	flowInfo := s.oauthAuthFlow.info
-	s.mutex.Unlock()
+	flowInfo := pending.info
 
 	if flowInfo.UserCode != msg.UserCode {
 		state.Set(internal.StatusLoginFailed)
 		return nil, gstatus.Errorf(codes.InvalidArgument, "sso user code is invalid")
 	}
 
-	if s.oauthAuthFlow.waitCancel != nil {
-		s.oauthAuthFlow.waitCancel()
-	}
-
 	waitCTX, cancel := context.WithCancel(ctx)
 	defer cancel()
 
+	// Swap in this wait's cancel and take over the one it displaces in a single
+	// critical section, so two WaitSSOLogin calls racing here cannot both read
+	// the same predecessor and leave one wait uncancelled. Cancelling happens
+	// after the unlock: the displaced wait takes s.mutex as it unwinds.
 	s.mutex.Lock()
+	if s.oauthAuthFlow.flow != flow {
+		s.mutex.Unlock()
+		return nil, gstatus.Errorf(codes.Canceled, "sso login was replaced by a newer login")
+	}
+	staleCancel := s.oauthAuthFlow.waitCancel
 	s.oauthAuthFlow.waitCancel = cancel
 	s.mutex.Unlock()
 
-	tokenInfo, err := s.oauthAuthFlow.flow.WaitToken(waitCTX, flowInfo)
+	if staleCancel != nil {
+		staleCancel()
+	}
+
+	tokenInfo, err := flow.WaitToken(waitCTX, flowInfo)
 	if err != nil {
-		s.mutex.Lock()
-		s.oauthAuthFlow.expiresAt = time.Now()
-		s.mutex.Unlock()
+		s.expireOAuthFlow(flow)
 		switch {
 		case errors.Is(err, context.Canceled):
 			// External abort. If our caller cancelled (the client closed
@@ -960,7 +1060,9 @@ func (s *Server) WaitSSOLogin(callerCtx context.Context, msg *proto.WaitSSOLogin
 			// the new owner — don't clobber it.
 			if callerCtx.Err() != nil {
 				s.mutex.Lock()
-				s.oauthAuthFlow = oauthAuthFlow{}
+				if s.oauthAuthFlow.flow == flow {
+					s.oauthAuthFlow = oauthAuthFlow{}
+				}
 				s.mutex.Unlock()
 			}
 		case errors.Is(err, context.DeadlineExceeded):
@@ -976,9 +1078,32 @@ func (s *Server) WaitSSOLogin(callerCtx context.Context, msg *proto.WaitSSOLogin
 		return nil, err
 	}
 
-	s.mutex.Lock()
-	s.oauthAuthFlow.expiresAt = time.Now()
-	s.mutex.Unlock()
+	s.expireOAuthFlow(flow)
+
+	if !tokenInfo.MatchesAccount(pending.hint) {
+		if !pending.accountPrompted {
+			// The IdP answered from a session belonging to another account. The
+			// browser for this flow is gone, so a new URL cannot be handed out
+			// here — arm the prompt for the user's next connect and fail this
+			// round. Never log in with the token: on a registered peer the
+			// server would reject it, and on a fresh one it would silently
+			// register the peer under the wrong account.
+			log.Warnf("login returned an account other than the one this profile is bound to; the next connect will ask the IdP to choose")
+			s.mutex.Lock()
+			if s.oauthAuthFlow.flow == flow {
+				s.oauthAuthFlow = oauthAuthFlow{}
+				s.forceAccountPrompt = true
+			}
+			s.mutex.Unlock()
+			state.Set(internal.StatusNeedsLogin)
+			return nil, gstatus.Errorf(codes.FailedPrecondition, "the login used a different account than this profile; connect again to choose the account")
+		}
+		// Already asked once; the account may legitimately differ (a changed
+		// email address). Refusing again would lock the user out of the profile,
+		// and the management server still rejects a token that does not own the
+		// peer.
+		log.Warnf("login still returned a different account after the prompt, continuing with it")
+	}
 
 	if loginStatus, err := s.attemptLogin(ctx, "", tokenInfo.GetTokenToUse()); err != nil {
 		state.Set(loginStatus)
@@ -1076,11 +1201,15 @@ func (s *Server) Up(callerCtx context.Context, msg *proto.UpRequest) (*proto.UpR
 	}
 
 	if msg != nil && msg.ProfileName != nil {
-		if _, err := s.switchProfileIfNeeded(*msg.ProfileName, msg.Username, activeProf); err != nil {
+		switched, err := s.switchProfileIfNeeded(*msg.ProfileName, msg.Username, activeProf)
+		if err != nil {
 			s.mutex.Unlock()
 			log.Errorf("failed to switch profile: %v", err)
 			return nil, err
 		}
+		if switched {
+			s.dropPendingAuthFlows()
+		}
 	}
 
 	activeProf, err = s.profileManager.GetActiveProfileState()
@@ -1181,12 +1310,16 @@ func (s *Server) storedLoginConfig(activeProf *profilemanager.ActiveProfileState
 // storedConfigAtPath reads a profile config file with the MDM policy applied on
 // top, yielding nil when the file does not exist yet.
 //
-// The overlay matters because the privilege gates are this function's only
-// consumers, and they have to reason about the config the engine will actually
-// run. An MDM policy that enables a remote-access server leaves the on-disk
-// value off, so without it the guards read a host publishing its console as one
-// running nothing, and let an unprivileged caller repoint the management
-// identity that decides who may connect to it.
+// The overlay matters because the privilege gates have to reason about the
+// config the engine will actually run. An MDM policy that enables a
+// remote-access server leaves the on-disk value off, so without it the guards
+// read a host publishing its console as one running nothing, and let an
+// unprivileged caller repoint the management identity that decides who may
+// connect to it.
+//
+// Reading it has no side effect: profilemanager.GetExistingConfig does not
+// write, so a request that the gates go on to refuse leaves the profile file as
+// it found it.
 func (s *Server) storedConfigAtPath(path string) (*profilemanager.Config, error) {
 	if _, err := os.Stat(path); err != nil {
 		if os.IsNotExist(err) {
@@ -1195,7 +1328,7 @@ func (s *Server) storedConfigAtPath(path string) (*profilemanager.Config, error)
 		return nil, fmt.Errorf("stat profile config: %w", err)
 	}
 
-	cfg, err := profilemanager.GetConfig(path)
+	cfg, err := profilemanager.GetExistingConfig(path)
 	if err != nil {
 		return nil, fmt.Errorf("read profile config: %w", err)
 	}
@@ -1222,12 +1355,12 @@ func (s *Server) resolveProfileHandle(handle, username string) (*profilemanager.
 }
 
 // switchProfileIfNeeded resolves the user-supplied handle, updates the
-// active profile state if it differs from the current one, and returns
-// the resolved profile so callers can include its ID in RPC responses.
-func (s *Server) switchProfileIfNeeded(handle string, userName *string, activeProf *profilemanager.ActiveProfileState) (*profilemanager.Profile, error) {
+// active profile state if it differs from the current one, and reports
+// whether the active profile changed.
+func (s *Server) switchProfileIfNeeded(handle string, userName *string, activeProf *profilemanager.ActiveProfileState) (bool, error) {
 	if handle != profilemanager.DefaultProfileName && (userName == nil || *userName == "") {
 		log.Errorf("profile name is set to %s, but username is not provided", handle)
-		return nil, fmt.Errorf("profile name is set to %s, but username is not provided", handle)
+		return false, fmt.Errorf("profile name is set to %s, but username is not provided", handle)
 	}
 
 	var username string
@@ -1237,26 +1370,48 @@ func (s *Server) switchProfileIfNeeded(handle string, userName *string, activePr
 
 	resolved, err := s.resolveProfileHandle(handle, username)
 	if err != nil {
-		return nil, err
+		return false, err
 	}
 
-	if resolved.ID != activeProf.ID || username != activeProf.Username {
-		if s.checkProfilesDisabled() {
-			log.Errorf("profiles are disabled, you cannot use this feature without profiles enabled")
-			return nil, gstatus.Errorf(codes.Unavailable, errProfilesDisabled)
-		}
-
-		log.Infof("switching to profile %s (%s) for user %s", resolved.Name, resolved.ID, username)
-		if err := s.profileManager.SetActiveProfileState(&profilemanager.ActiveProfileState{
-			ID:       resolved.ID,
-			Username: username,
-		}); err != nil {
-			log.Errorf("failed to set active profile state: %v", err)
-			return nil, fmt.Errorf("failed to set active profile state: %w", err)
-		}
+	if resolved.ID == activeProf.ID && username == activeProf.Username {
+		return false, nil
 	}
 
-	return resolved, nil
+	if s.checkProfilesDisabled() {
+		log.Errorf("profiles are disabled, you cannot use this feature without profiles enabled")
+		return false, gstatus.Errorf(codes.Unavailable, errProfilesDisabled)
+	}
+
+	log.Infof("switching to profile %s (%s) for user %s", resolved.Name, resolved.ID, username)
+	if err := s.profileManager.SetActiveProfileState(&profilemanager.ActiveProfileState{
+		ID:       resolved.ID,
+		Username: username,
+	}); err != nil {
+		log.Errorf("failed to set active profile state: %v", err)
+		return false, fmt.Errorf("failed to set active profile state: %w", err)
+	}
+
+	return true, nil
+}
+
+func (s *Server) dropPendingAuthFlows() {
+	// A pending login flow and the account-prompt flag describe the previous
+	// profile's login; carried across a switch they would judge the new
+	// profile's token against the old profile's account. CancelFunc is
+	// non-blocking, so calling it under the mutex is safe.
+	if cancel := s.oauthAuthFlow.waitCancel; cancel != nil {
+		cancel()
+	}
+	s.oauthAuthFlow = oauthAuthFlow{}
+	s.forceAccountPrompt = false
+
+	// A pending session extend belongs to the previous profile too: its device
+	// code was issued by that profile's IdP client, and WaitExtendAuthSession
+	// would submit the resulting token against the new profile's engine.
+	s.extendAuthSessionFlow.CancelWait()
+	s.extendAuthSessionFlow.Clear()
+
+	s.jwtCache.clear()
 }
 
 // SwitchProfile switches the active profile in the daemon.
@@ -1290,7 +1445,7 @@ func (s *Server) SwitchProfile(callerCtx context.Context, msg *proto.SwitchProfi
 	s.config = config
 	s.localMetrics.Reconcile(config.LocalMetricsEnabled, config.LocalMetricsAddress)
 
-	s.jwtCache.clear()
+	s.dropPendingAuthFlows()
 
 	if msg != nil && msg.ProfileName != nil {
 		s.publishProfileListChanged(*msg.ProfileName)
@@ -1499,8 +1654,16 @@ func (s *Server) handleActiveProfileLogout(ctx context.Context) (*proto.LogoutRe
 	return &proto.LogoutResponse{}, nil
 }
 
-// getConfig reads config file and returns Config and whether the config file already existed. Errors out if it does not exist
-func (s *Server) getConfig(activeProf *profilemanager.ActiveProfileState) (*profilemanager.Config, bool, error) {
+// provisionProfileIdentity resolves the active profile's config and puts the
+// keys that identify the peer on disk, reporting whether the config file
+// already existed.
+//
+// This is the daemon's provisioning point: the config resolved here is the one
+// the peer runs with, so it needs its identity, and that has to reach disk — a
+// key that stays in memory would come back different on the next start and
+// re-register the peer. Reads themselves are pure, so the write is here, in
+// the open, instead of hiding inside the reader.
+func provisionProfileIdentity(activeProf *profilemanager.ActiveProfileState) (*profilemanager.Config, bool, error) {
 	cfgPath, err := activeProf.FilePath()
 	if err != nil {
 		return nil, false, fmt.Errorf("failed to get active profile file path: %w", err)
@@ -1511,15 +1674,38 @@ func (s *Server) getConfig(activeProf *profilemanager.ActiveProfileState) (*prof
 
 	log.Infof("active profile config existed: %t, err %v", configExisted, err)
 
-	config, err := profilemanager.ReadConfig(cfgPath)
+	config, err := profilemanager.ReadConfigOrDefault(cfgPath)
 	if err != nil {
 		return nil, false, fmt.Errorf("failed to get config: %w", err)
 	}
 
-	// Apply the daemon-owned MDM policy on top of the just-resolved
-	// Config. profilemanager's apply() initialises the policy to
-	// empty — the Loader lives outside Config, so this overlay step
-	// is driven externally here.
+	generated, err := config.EnsureIdentity()
+	if err != nil {
+		return nil, false, fmt.Errorf("ensure profile identity: %w", err)
+	}
+
+	if generated || !configExisted {
+		if err := profilemanager.WriteOutConfig(cfgPath, config); err != nil {
+			return nil, false, fmt.Errorf("write out profile config: %w", err)
+		}
+	}
+
+	return config, configExisted, nil
+}
+
+// getConfig resolves the active profile's config, provisions its identity and
+// reports whether the config file already existed.
+func (s *Server) getConfig(activeProf *profilemanager.ActiveProfileState) (*profilemanager.Config, bool, error) {
+	config, configExisted, err := provisionProfileIdentity(activeProf)
+	if err != nil {
+		return nil, false, err
+	}
+
+	// Apply the daemon-owned MDM policy on top of the just-resolved Config.
+	// profilemanager's apply() initialises the policy to empty — the Loader
+	// lives outside Config, so this overlay step is driven externally here.
+	// After the write above, on purpose: the overlay is runtime-only and
+	// re-derived on every load, so the file keeps the profile's own values.
 	config.ApplyMDMPolicy(s.mdmLoader.Load())
 
 	return config, configExisted, nil
@@ -1574,7 +1760,7 @@ func (s *Server) logoutFromProfile(ctx context.Context, profile *profilemanager.
 		cfgPath = profilemanager.DefaultConfigPath
 	}
 
-	config, err := profilemanager.GetConfig(cfgPath)
+	config, err := profilemanager.GetExistingConfig(cfgPath)
 	if err != nil {
 		return fmt.Errorf("profile '%s' not found", profile.ID)
 	}
@@ -1593,6 +1779,19 @@ func (s *Server) sendLogoutRequestWithConfig(ctx context.Context, config *profil
 	// Privilege gate: deregistering frees this machine's key to be registered
 	// against another management server, which is only restricted while the SSH
 	// server makes that a privilege handover.
+	// Ahead of the privilege gate on purpose. A profile with no identity was
+	// never registered — a logout clears the keys in place, so logging the same
+	// profile out twice lands here — so there is nothing to deregister and
+	// nothing for the gate to protect: what it guards against is handing this
+	// machine's registered key to another management server. Behind the gate,
+	// an unprivileged caller would be refused instead, and for a profile whose
+	// ServerSSHAllowed is unset that is every caller, since an absent value
+	// counts as SSH enabled.
+	if config.PrivateKey == "" {
+		log.Infof("profile carries no identity, nothing to deregister")
+		return nil
+	}
+
 	if err := requirePrivilegeForDeregistration(ctx, config); err != nil {
 		return err
 	}
@@ -1924,7 +2123,7 @@ func (s *Server) RequestJWTAuth(
 	}
 
 	// the daemon has no graphical session of its own, only the caller can answer this
-	oAuthFlow, err := auth.NewOAuthFlow(ctx, config, msg.GetHasGraphicalSession(), false, hint)
+	oAuthFlow, err := auth.NewOAuthFlow(ctx, config, msg.GetHasGraphicalSession(), false, hint, false)
 	if err != nil {
 		return nil, gstatus.Errorf(codes.Internal, "failed to create OAuth flow: %v", err)
 	}
@@ -1934,12 +2133,15 @@ func (s *Server) RequestJWTAuth(
 		return nil, gstatus.Errorf(codes.Internal, "failed to request auth info: %v", err)
 	}
 
-	s.mutex.Lock()
-	s.oauthAuthFlow.flow = oAuthFlow
-	s.oauthAuthFlow.info = authInfo
-	s.oauthAuthFlow.expiresAt = time.Now().Add(time.Duration(authInfo.ExpiresIn) * time.Second)
-	s.oauthAuthFlow.cacheGeneration = cacheGeneration
-	s.mutex.Unlock()
+	// This flow carries no profile hint: leaving the previous login's hint and
+	// accountPrompted in place would have WaitSSOLogin judge a later token
+	// against them.
+	s.replaceOAuthFlow(oauthAuthFlow{
+		flow:            oAuthFlow,
+		info:            authInfo,
+		expiresAt:       time.Now().Add(time.Duration(authInfo.ExpiresIn) * time.Second),
+		cacheGeneration: cacheGeneration,
+	})
 
 	return &proto.RequestJWTAuthResponse{
 		VerificationURI:         authInfo.VerificationURI,
@@ -2039,7 +2241,7 @@ func (s *Server) RequestExtendAuthSession(
 	}
 
 	// the daemon has no graphical session of its own, only the caller can answer this
-	oAuthFlow, err := auth.NewOAuthFlow(ctx, config, msg.GetHasGraphicalSession(), false, hint)
+	oAuthFlow, err := auth.NewOAuthFlow(ctx, config, msg.GetHasGraphicalSession(), false, hint, true)
 	if err != nil {
 		return nil, gstatus.Errorf(codes.Internal, "failed to create OAuth flow: %v", err)
 	}
@@ -2284,7 +2486,7 @@ func (s *Server) GetConfig(ctx context.Context, req *proto.GetConfigRequest) (*p
 		cfgPath = profilemanager.DefaultConfigPath
 	}
 
-	cfg, err := profilemanager.GetConfig(cfgPath)
+	cfg, err := profilemanager.GetExistingConfig(cfgPath)
 	if err != nil {
 		log.Errorf("failed to get active profile config: %v", err)
 		return nil, fmt.Errorf("failed to get active profile config: %w", err)
@@ -2605,9 +2807,23 @@ func (s *Server) checkDisableAdvancedView() *bool {
 	return nil
 }
 
+// profileOwnerOption passes the OS account of the active profile to the connect client,
+// which reads that account's certificate store for user certificate posture checks.
+func (s *Server) profileOwnerOption() []internal.ConnectClientOption {
+	if s.profileManager == nil {
+		return nil
+	}
+	activeProf, err := s.profileManager.GetActiveProfileState()
+	if err != nil {
+		log.Warnf("failed to read the active profile owner, no user certificate store is used for certificate posture: %v", err)
+		return []internal.ConnectClientOption{internal.WithUnknownProfileOwner()}
+	}
+	return []internal.ConnectClientOption{internal.WithProfileOwner(activeProf.Username)}
+}
+
 func (s *Server) connect(ctx context.Context, config *profilemanager.Config, statusRecorder *peer.Status, runningChan chan struct{}) error {
 	log.Tracef("running client connection")
-	client := internal.NewConnectClient(ctx, config, statusRecorder)
+	client := internal.NewConnectClient(ctx, config, statusRecorder, s.profileOwnerOption()...)
 	client.SetUpdateManager(s.updateManager)
 	client.SetSyncResponsePersistence(s.persistSyncResponse)
 
@@ -2749,8 +2965,6 @@ func sendTerminalNotification() error {
 	return wallCmd.Wait()
 }
 
-// persistLoginOverrides writes management URL and pre-shared key from a LoginRequest to the
-// active profile config so that subsequent reads pick them up. Empty/nil values are ignored.
 // afterLoginPreCheck is a seam for tests to run a concurrent config change
 // between Login's first privilege check and the authoritative one.
 var afterLoginPreCheck func()
@@ -2765,7 +2979,7 @@ var afterLoginPreCheck func()
 // of this is reached; this one exists because that check is not synchronized
 // against a concurrent privileged request that enables the SSH server, and a
 // caller refused here must not have cancelled or switched anything either.
-func (s *Server) authorizeAndPrepareLogin(callerCtx context.Context, msg *proto.LoginRequest, activeProf *profilemanager.ActiveProfileState) (context.Context, *profilemanager.ActiveProfileState, error) {
+func (s *Server) authorizeAndPrepareLogin(callerCtx context.Context, msg *proto.LoginRequest, activeProf *profilemanager.ActiveProfileState) (context.Context, *profilemanager.ActiveProfileState, bool, error) {
 	if afterLoginPreCheck != nil {
 		afterLoginPreCheck()
 	}
@@ -2775,10 +2989,19 @@ func (s *Server) authorizeAndPrepareLogin(callerCtx context.Context, msg *proto.
 
 	stored, err := s.storedLoginConfig(activeProf, msg)
 	if err != nil {
-		return nil, nil, err
+		return nil, nil, false, err
 	}
 	if err := requirePrivilegeForConfigChange(callerCtx, stored, privilegedChangeFromLogin(msg)); err != nil {
-		return nil, nil, err
+		return nil, nil, false, err
+	}
+
+	// The update-settings decision is re-taken here for the same reason as the
+	// privilege one: Login's earlier check ran outside this lock, so the stored
+	// config it compared against could have moved since. This one is the
+	// authoritative check, and it is the last read before persistLoginOverrides
+	// writes.
+	if s.checkUpdateSettingsDisabled() && configChangeRequested(stored, loginOverridesInput(msg)) {
+		return nil, nil, false, gstatus.Errorf(codes.FailedPrecondition, errUpdateSettingsDisabled)
 	}
 
 	s.mutex.Lock()
@@ -2796,29 +3019,46 @@ func (s *Server) authorizeAndPrepareLogin(callerCtx context.Context, msg *proto.
 		log.Warnf(errRestoreResidualState, err)
 	}
 
+	switched := false
 	if msg.ProfileName != nil {
-		if _, err := s.switchProfileIfNeeded(*msg.ProfileName, msg.Username, activeProf); err != nil {
-			return nil, nil, fmt.Errorf("switch profile: %w", err)
+		switched, err = s.switchProfileIfNeeded(*msg.ProfileName, msg.Username, activeProf)
+		if err != nil {
+			return nil, nil, false, fmt.Errorf("switch profile: %w", err)
+		}
+		if switched {
+			s.mutex.Lock()
+			s.dropPendingAuthFlows()
+			s.mutex.Unlock()
 		}
 	}
 
 	activeProf, err = s.profileManager.GetActiveProfileState()
 	if err != nil {
-		return nil, nil, fmt.Errorf("active profile state: %w", err)
+		return nil, nil, false, fmt.Errorf("active profile state: %w", err)
 	}
 
-	if err := persistLoginOverrides(activeProf, msg.ManagementUrl, msg.OptionalPreSharedKey); err != nil {
-		return nil, nil, fmt.Errorf("persist login overrides: %w", err)
+	if err := persistLoginOverrides(activeProf, msg); err != nil {
+		return nil, nil, false, fmt.Errorf("persist login overrides: %w", err)
 	}
 
-	return ctx, activeProf, nil
+	// Provisioning under the same lock as the decision above, and next to the
+	// write it guards. getConfig would otherwise mint the identity and persist
+	// it once this returns: between its read and its write, a SetConfig that
+	// had already answered its caller would be overwritten by the config this
+	// login read before it landed.
+	if _, _, err := provisionProfileIdentity(activeProf); err != nil {
+		return nil, nil, false, err
+	}
+
+	return ctx, activeProf, switched, nil
 }
 
-func persistLoginOverrides(activeProf *profilemanager.ActiveProfileState, managementURL string, preSharedKey *string) error {
-	if preSharedKey != nil && *preSharedKey == "" {
-		preSharedKey = nil
-	}
-	if managementURL == "" && preSharedKey == nil {
+// persistLoginOverrides writes the config fields a login request is allowed to
+// carry into the active profile. It shares its input builder with the
+// update-settings gate, so the gate judges exactly the fields this writes.
+func persistLoginOverrides(activeProf *profilemanager.ActiveProfileState, msg *proto.LoginRequest) error {
+	input := loginOverridesInput(msg)
+	if input.ManagementURL == "" && input.PreSharedKey == nil {
 		return nil
 	}
 
@@ -2827,11 +3067,7 @@ func persistLoginOverrides(activeProf *profilemanager.ActiveProfileState, manage
 		return fmt.Errorf("active profile file path: %w", err)
 	}
 
-	input := profilemanager.ConfigInput{
-		ConfigPath:    cfgPath,
-		ManagementURL: managementURL,
-		PreSharedKey:  preSharedKey,
-	}
+	input.ConfigPath = cfgPath
 	if _, err := profilemanager.UpdateOrCreateConfig(input); err != nil {
 		return fmt.Errorf("update config: %w", err)
 	}

client/server/server_connect_test.go

@@ -10,6 +10,7 @@ import (
 	"github.com/stretchr/testify/require"
 
 	"github.com/netbirdio/netbird/client/internal"
+	"github.com/netbirdio/netbird/client/internal/auth"
 	"github.com/netbirdio/netbird/client/internal/peer"
 	"github.com/netbirdio/netbird/client/proto"
 )
@@ -21,7 +22,8 @@ func newTestServer() *Server {
 		// New always populates the SSH JWT cache and the logout and
 		// profile-switch paths call into it unconditionally, so a Server
 		// assembled field by field has to populate it too.
-		jwtCache: newJWTCache(),
+		jwtCache:              newJWTCache(),
+		extendAuthSessionFlow: auth.NewPendingFlow(),
 	}
 }
 

client/server/server_privileged_test.go

@@ -10,9 +10,9 @@ import (
 	"testing"
 	"time"
 
-	"go.uber.org/mock/gomock"
 	"github.com/stretchr/testify/require"
 	"go.opentelemetry.io/otel"
+	"go.uber.org/mock/gomock"
 
 	"github.com/netbirdio/netbird/management/server/integrations/integrated_validator/validator"
 
@@ -36,7 +36,6 @@ import (
 	"github.com/netbirdio/netbird/management/server"
 	"github.com/netbirdio/netbird/management/server/activity"
 	nbcache "github.com/netbirdio/netbird/management/server/cache"
-	"github.com/netbirdio/netbird/management/server/integrations/port_forwarding"
 	"github.com/netbirdio/netbird/management/server/permissions"
 	"github.com/netbirdio/netbird/management/server/settings"
 	"github.com/netbirdio/netbird/management/server/store"
@@ -200,8 +199,8 @@ func startManagement(t *testing.T, signalAddr string, counter *int) (*grpc.Serve
 
 	requestBuffer := server.NewAccountRequestBuffer(context.Background(), store)
 	peersUpdateManager := update_channel.NewPeersUpdateManager(metrics)
-	networkMapController := controller.NewController(context.Background(), store, metrics, peersUpdateManager, requestBuffer, server.MockIntegratedValidator{}, settingsMockManager, "netbird.selfhosted", port_forwarding.NewControllerMock(), manager.NewEphemeralManager(store, peersManager), config, nil)
-	accountManager, err := server.BuildManager(context.Background(), config, store, networkMapController, jobManager, nil, "", eventStore, nil, false, ia, metrics, port_forwarding.NewControllerMock(), settingsMockManager, permissionsManagerMock, false, cacheStore)
+	networkMapController := controller.NewController(context.Background(), store, metrics, peersUpdateManager, requestBuffer, server.MockIntegratedValidator{}, settingsMockManager, "netbird.selfhosted", manager.NewEphemeralManager(store, peersManager), config, nil)
+	accountManager, err := server.BuildManager(context.Background(), config, store, networkMapController, jobManager, nil, "", eventStore, nil, false, ia, metrics, settingsMockManager, permissionsManagerMock, false, cacheStore)
 	if err != nil {
 		return nil, "", err
 	}

client/server/setconfig_mdm_test.go

@@ -313,7 +313,7 @@ func TestSetConfig_MDMReject_AllOrNothing(t *testing.T) {
 
 	// Confirm RosenpassEnabled was NOT applied even though it was not
 	// in the conflict list: the request was rejected as a whole.
-	reloaded, err := profilemanager.GetConfig(cfgPath)
+	reloaded, err := profilemanager.GetExistingConfig(cfgPath)
 	require.NoError(t, err)
 	assert.False(t, reloaded.RosenpassEnabled, "non-conflicting field must not be applied when request is rejected")
 }

client/server/setconfig_test.go

@@ -129,7 +129,7 @@ func TestSetConfig_AllFieldsSaved(t *testing.T) {
 	cfgPath, err := profState.FilePath()
 	require.NoError(t, err)
 
-	cfg, err := profilemanager.GetConfig(cfgPath)
+	cfg, err := profilemanager.GetExistingConfig(cfgPath)
 	require.NoError(t, err)
 
 	require.Equal(t, "https://new-api.netbird.io:443", cfg.ManagementURL.String())

client/server/ssh_gate.go

@@ -394,21 +394,5 @@ func sameManagementURL(stored *url.URL, requested string) bool {
 		return false
 	}
 
-	return stored.Scheme == parsed.Scheme &&
-		stored.Hostname() == parsed.Hostname() &&
-		effectivePort(stored) == effectivePort(parsed)
-}
-
-func effectivePort(u *url.URL) string {
-	if port := u.Port(); port != "" {
-		return port
-	}
-	switch u.Scheme {
-	case "https":
-		return "443"
-	case "http":
-		return "80"
-	default:
-		return ""
-	}
+	return profilemanager.SameServiceURL(stored, parsed)
 }

client/server/update_settings_gate.go

@@ -0,0 +1,55 @@
+package server
+
+import (
+	log "github.com/sirupsen/logrus"
+
+	"github.com/netbirdio/netbird/client/internal/profilemanager"
+	"github.com/netbirdio/netbird/client/proto"
+)
+
+// configChangeRequested reports whether applying input would move the target
+// profile away from the configuration it already persists. It is the decision
+// procedure of the update-settings kill switch (--disable-update-settings /
+// NB_DISABLE_UPDATE_SETTINGS / the MDM DisableUpdateSettings key): that switch
+// forbids *changing* settings, so a request that restates the stored values is
+// not a change and must not be refused.
+//
+// This has to be judged on values, not on field presence. `netbird up` rebuilds
+// the whole config surface of SetConfigRequest and LoginRequest from its flags
+// and environment on every invocation, so a service or container configured by
+// environment restates its own configuration on every start. A presence-based
+// gate refused those requests, and because Login carries the same fields it
+// refused the login too — leaving such a client unable to come up at all.
+//
+// A dry run that cannot be evaluated fails closed: the request counts as a
+// change, so a malformed field can never open the gate. The error itself is
+// reported to the caller by the real update path.
+func configChangeRequested(stored *profilemanager.Config, input profilemanager.ConfigInput) bool {
+	changed, err := stored.WouldChange(input)
+	if err != nil {
+		log.Warnf("cannot evaluate the requested config change, treating it as a change: %v", err)
+		return true
+	}
+	return changed
+}
+
+// loginOverridesInput builds the ConfigInput a login request persists. The
+// management URL and the pre-shared key are the only config fields the daemon
+// applies from a LoginRequest; everything else on that message is either pure
+// auth or ignored. An empty pre-shared key is dropped rather than written, so
+// a login cannot clear the stored key by omission.
+//
+// Both the write (persistLoginOverrides) and the update-settings gate go
+// through this builder, so the gate can neither refuse a field the write
+// ignores nor miss one it applies.
+func loginOverridesInput(msg *proto.LoginRequest) profilemanager.ConfigInput {
+	preSharedKey := msg.OptionalPreSharedKey
+	if preSharedKey != nil && *preSharedKey == "" {
+		preSharedKey = nil
+	}
+
+	return profilemanager.ConfigInput{
+		ManagementURL: msg.ManagementUrl,
+		PreSharedKey:  preSharedKey,
+	}
+}

client/server/update_settings_gate_test.go

@@ -0,0 +1,390 @@
+package server
+
+import (
+	"context"
+	"os"
+	"path/filepath"
+	"testing"
+
+	"github.com/stretchr/testify/require"
+	"google.golang.org/grpc/codes"
+	gstatus "google.golang.org/grpc/status"
+
+	"github.com/netbirdio/netbird/client/internal"
+	"github.com/netbirdio/netbird/client/internal/profilemanager"
+	"github.com/netbirdio/netbird/client/mdm"
+	"github.com/netbirdio/netbird/client/proto"
+)
+
+// The seeded profile of setupServerWithProfile is created with this management
+// URL, so a request carrying it restates what the profile already holds.
+const storedManagementURL = "https://api.netbird.io:443"
+
+// A client configured by environment re-sends its whole configuration on every
+// `netbird up`: the CLI fills the request from its flags and env regardless of
+// what changed. With the update-settings kill switch on, such a request must
+// pass — nothing about the configuration moves.
+func TestSetConfig_RestatingTheStoredConfigPassesTheGate(t *testing.T) {
+	s, ctx, profName, username, _ := setupServerWithProfile(t)
+	s.updateSettingsDisabled = true
+
+	_, err := s.SetConfig(ctx, &proto.SetConfigRequest{
+		ProfileName:   profName,
+		Username:      username,
+		ManagementUrl: storedManagementURL,
+	})
+	require.NoError(t, err, "restating the stored management URL is not a settings change")
+}
+
+// The same endpoint written without its default port is the same endpoint. A
+// gate that compared raw strings refused NB_MANAGEMENT_URL=https://host, which
+// is how the URL is normally spelled.
+func TestSetConfig_EquivalentManagementURLPassesTheGate(t *testing.T) {
+	s, ctx, profName, username, _ := setupServerWithProfile(t)
+	s.updateSettingsDisabled = true
+
+	_, err := s.SetConfig(ctx, &proto.SetConfigRequest{
+		ProfileName:   profName,
+		Username:      username,
+		ManagementUrl: "https://api.netbird.io",
+	})
+	require.NoError(t, err, "an implicit :443 is the same management URL")
+}
+
+// The kill switch still has to do its job: a request that moves a setting is
+// refused, and the profile keeps the value it had.
+func TestSetConfig_ChangingASettingIsRefused(t *testing.T) {
+	s, ctx, profName, username, cfgPath := setupServerWithProfile(t)
+	s.updateSettingsDisabled = true
+
+	_, err := s.SetConfig(ctx, &proto.SetConfigRequest{
+		ProfileName:   profName,
+		Username:      username,
+		ManagementUrl: "https://mgmt.elsewhere.example:443",
+	})
+	require.Error(t, err, "moving the management URL is a settings change")
+	require.Equal(t, codes.FailedPrecondition, gstatus.Code(err), "want the update-settings refusal, got %v", err)
+
+	cfg, err := profilemanager.GetExistingConfig(cfgPath)
+	require.NoError(t, err)
+	require.Equal(t, storedManagementURL, cfg.ManagementURL.String(), "the refused request changed the config anyway")
+}
+
+// A field whose requested value differs from the stored one is a change even
+// when the rest of the request restates the configuration.
+func TestSetConfig_SingleDivergingFieldIsRefused(t *testing.T) {
+	s, ctx, profName, username, _ := setupServerWithProfile(t)
+	s.updateSettingsDisabled = true
+
+	rosenpass := true
+	_, err := s.SetConfig(ctx, &proto.SetConfigRequest{
+		ProfileName:      profName,
+		Username:         username,
+		ManagementUrl:    storedManagementURL,
+		RosenpassEnabled: &rosenpass,
+	})
+	require.Error(t, err, "enabling Rosenpass is a settings change")
+	require.Equal(t, codes.FailedPrecondition, gstatus.Code(err), "want the update-settings refusal, got %v", err)
+}
+
+// With the switch off, the same diverging request goes through: the gate must
+// not leak into a daemon that never enabled it.
+func TestSetConfig_ChangeAllowedWhenTheSwitchIsOff(t *testing.T) {
+	s, ctx, profName, username, cfgPath := setupServerWithProfile(t)
+
+	_, err := s.SetConfig(ctx, &proto.SetConfigRequest{
+		ProfileName:   profName,
+		Username:      username,
+		ManagementUrl: "https://mgmt.elsewhere.example:443",
+	})
+	require.NoError(t, err)
+
+	cfg, err := profilemanager.GetExistingConfig(cfgPath)
+	require.NoError(t, err)
+	require.Equal(t, "https://mgmt.elsewhere.example:443", cfg.ManagementURL.String())
+}
+
+// Login carries the same config surface as SetConfig, so it is gated the same
+// way: a login that would move a protected setting is refused before it can
+// touch daemon state.
+func TestLogin_ChangingTheManagementURLIsRefused(t *testing.T) {
+	s, _, profName, username, cfgPath := setupServerWithProfile(t)
+	s.updateSettingsDisabled = true
+	s.rootCtx = internal.CtxInitState(context.Background())
+
+	cancelled := false
+	s.actCancel = func() { cancelled = true }
+
+	_, err := s.Login(userCtx(), &proto.LoginRequest{
+		Username:      &username,
+		ManagementUrl: "https://mgmt.elsewhere.example:443",
+	})
+	require.Error(t, err, "moving the management URL through Login is a settings change")
+	require.Equal(t, codes.FailedPrecondition, gstatus.Code(err), "want the update-settings refusal, got %v", err)
+
+	// "Refused before it can touch daemon state" is the contract, so check the
+	// state as well as the error.
+	cfg, err := profilemanager.GetExistingConfig(cfgPath)
+	require.NoError(t, err)
+	require.Equal(t, storedManagementURL, cfg.ManagementURL.String(), "the refused login moved the management URL")
+	require.False(t, cancelled, "the refused login cancelled the login already in progress")
+
+	active, err := s.profileManager.GetActiveProfileState()
+	require.NoError(t, err)
+	require.Equal(t, profilemanager.ID(profName), active.ID, "the refused login switched the active profile")
+}
+
+// seedProfileConfig writes a profile config carrying the given management URL
+// and pre-shared key into a temp dir, and returns its path.
+func seedProfileConfig(t *testing.T, managementURL, preSharedKey string) string {
+	t.Helper()
+
+	path := filepath.Join(t.TempDir(), "seeded.json")
+	_, err := profilemanager.UpdateOrCreateConfig(profilemanager.ConfigInput{
+		ConfigPath:    path,
+		ManagementURL: managementURL,
+		PreSharedKey:  &preSharedKey,
+	})
+	require.NoError(t, err, "seed profile config")
+	return path
+}
+
+// The decision procedure itself, over the fields a login actually persists.
+// A login that restates the stored values must not be refused: that is what
+// keeps a re-login, or a container restart carrying NB_MANAGEMENT_URL, working
+// with the kill switch on.
+func TestLoginGateDecision(t *testing.T) {
+	stored, err := profilemanager.GetExistingConfig(seedProfileConfig(t, storedManagementURL, "stored-key"))
+	require.NoError(t, err)
+
+	redacted := mdm.PreSharedKeyRedactedSentinel
+	empty := ""
+	sameKey := "stored-key"
+	otherKey := "other-key"
+
+	tests := []struct {
+		name        string
+		msg         *proto.LoginRequest
+		wantChanged bool
+	}{
+		{
+			name:        "pure auth carries no config",
+			msg:         &proto.LoginRequest{SetupKey: "ABC"},
+			wantChanged: false,
+		},
+		{
+			name:        "stored management URL restated",
+			msg:         &proto.LoginRequest{ManagementUrl: storedManagementURL},
+			wantChanged: false,
+		},
+		{
+			name:        "stored management URL without its default port",
+			msg:         &proto.LoginRequest{ManagementUrl: "https://api.netbird.io"},
+			wantChanged: false,
+		},
+		{
+			name:        "different management URL",
+			msg:         &proto.LoginRequest{ManagementUrl: "https://mgmt.elsewhere.example:443"},
+			wantChanged: true,
+		},
+		{
+			name:        "stored pre-shared key restated",
+			msg:         &proto.LoginRequest{OptionalPreSharedKey: &sameKey},
+			wantChanged: false,
+		},
+		{
+			name:        "redacted pre-shared key echoed back",
+			msg:         &proto.LoginRequest{OptionalPreSharedKey: &redacted},
+			wantChanged: false,
+		},
+		{
+			name:        "empty pre-shared key is not a request to clear it",
+			msg:         &proto.LoginRequest{OptionalPreSharedKey: &empty},
+			wantChanged: false,
+		},
+		{
+			name:        "different pre-shared key",
+			msg:         &proto.LoginRequest{OptionalPreSharedKey: &otherKey},
+			wantChanged: true,
+		},
+	}
+
+	for _, tt := range tests {
+		t.Run(tt.name, func(t *testing.T) {
+			require.Equal(t, tt.wantChanged, configChangeRequested(stored, loginOverridesInput(tt.msg)))
+		})
+	}
+}
+
+// A profile with no config on disk yet is judged against the config the daemon
+// would create for it, so a first login that asks for the defaults is not a
+// change while one that asks for a different management URL is.
+func TestGateDecisionWithoutStoredConfig(t *testing.T) {
+	require.False(t, configChangeRequested(nil, profilemanager.ConfigInput{}),
+		"a request carrying nothing cannot change anything")
+	require.False(t, configChangeRequested(nil, profilemanager.ConfigInput{ManagementURL: profilemanager.DefaultManagementURL}),
+		"asking for the default management URL is what the daemon would write anyway")
+	require.True(t, configChangeRequested(nil, profilemanager.ConfigInput{ManagementURL: "https://mgmt.elsewhere.example:443"}),
+		"asking for a non-default management URL is a change")
+}
+
+// A dry run that cannot be evaluated must fail closed, or a malformed field
+// would open the gate.
+func TestGateDecisionFailsClosedOnAnInvalidRequest(t *testing.T) {
+	require.True(t, configChangeRequested(nil, profilemanager.ConfigInput{ManagementURL: "not-a-url"}),
+		"an unevaluable request must count as a change")
+}
+
+// The gate reads the stored config to decide, and reading it must not write it:
+// a refused request has to leave the profile file byte-for-byte as it was.
+// A config file missing a field the config layer fills in (MTU, here) is what
+// makes the normalization write fire.
+func TestSetConfig_RefusedRequestLeavesTheConfigFileUntouched(t *testing.T) {
+	s, ctx, profName, username, cfgPath := setupServerWithProfile(t)
+	s.updateSettingsDisabled = true
+
+	require.NoError(t, os.WriteFile(cfgPath, []byte(`{"WgIface":"wt0"}`), 0o600))
+	before, err := os.ReadFile(cfgPath)
+	require.NoError(t, err)
+
+	_, err = s.SetConfig(ctx, &proto.SetConfigRequest{
+		ProfileName:   profName,
+		Username:      username,
+		ManagementUrl: "https://mgmt.elsewhere.example:443",
+	})
+	require.Error(t, err)
+	require.Equal(t, codes.FailedPrecondition, gstatus.Code(err), "want the update-settings refusal, got %v", err)
+
+	after, err := os.ReadFile(cfgPath)
+	require.NoError(t, err)
+	require.Equal(t, string(before), string(after), "the refused request rewrote the profile config")
+}
+
+// The container case that the string comparison still broke: the management URL
+// supplied through the environment is the stored one, written with a trailing
+// slash.
+func TestSetConfig_ManagementURLSpellingsPassTheGate(t *testing.T) {
+	for _, spelling := range []string{
+		"https://api.netbird.io",
+		"https://api.netbird.io/",
+		"https://api.netbird.io:443/",
+		"https://API.netbird.io:443",
+	} {
+		t.Run(spelling, func(t *testing.T) {
+			s, ctx, profName, username, _ := setupServerWithProfile(t)
+			s.updateSettingsDisabled = true
+
+			_, err := s.SetConfig(ctx, &proto.SetConfigRequest{
+				ProfileName:   profName,
+				Username:      username,
+				ManagementUrl: spelling,
+			})
+			require.NoError(t, err, "%q is the stored management URL written differently", spelling)
+		})
+	}
+}
+
+// The RPC the whole fix hangs on. Login is retried by the CLI in a backoff
+// loop, so a login that restates the stored configuration — which is what a
+// container configured by environment sends on every start — must get past the
+// gate, or the client never comes up at all.
+//
+// Past the gate the handler goes on to do real work this test does not stand
+// up, so the assertion is only that the refusal did not happen.
+func TestLogin_RestatingTheStoredConfigPassesTheGate(t *testing.T) {
+	s, _, _, username, _ := setupServerWithProfile(t)
+	s.updateSettingsDisabled = true
+	s.rootCtx = internal.CtxInitState(context.Background())
+
+	// Stand in for the management round trip the handler makes once the gate
+	// lets it through, so this test exercises the gate and not the network:
+	// without it the profile's management URL is dialed for real.
+	s.isLoginRequiredFn = func(context.Context) (bool, error) { return false, nil }
+
+	_, err := s.Login(userCtx(), &proto.LoginRequest{
+		Username:      &username,
+		ManagementUrl: storedManagementURL,
+	})
+	if err != nil {
+		require.NotEqual(t, codes.FailedPrecondition, gstatus.Code(err),
+			"the gate refused a login that changes nothing: %v", err)
+		require.NotContains(t, err.Error(), "update settings are disabled",
+			"the gate refused a login that changes nothing: %v", err)
+	}
+}
+
+// The value-aware decision has the same synchronization problem as the
+// privileged-change one: Login's first check runs outside guardedConfigMu, so
+// the stored config it compared against can move before the write. A login that
+// was a no-op when it was checked must not be written once it has become a
+// change.
+func TestLogin_ChangeThatAppearsMidRequestIsRefused(t *testing.T) {
+	s, _, _, username, _ := setupServerWithProfile(t)
+	s.updateSettingsDisabled = true
+	s.rootCtx = internal.CtxInitState(context.Background())
+
+	target := "moved-under-us"
+	targetPath := filepath.Join(profilemanager.DefaultConfigPathDir, target+".json")
+	_, err := profilemanager.UpdateOrCreateConfig(profilemanager.ConfigInput{
+		ConfigPath:    targetPath,
+		ManagementURL: storedManagementURL,
+	})
+	require.NoError(t, err)
+
+	cancelled := false
+	s.actCancel = func() { cancelled = true }
+
+	// Stand in for a concurrent writer that repoints the profile between the two
+	// checks, which is the interleaving the lock has to make safe. The login
+	// restates the URL the profile held when it was checked, so the first check
+	// sees a no-op and lets it through.
+	afterLoginPreCheck = func() {
+		_, err := profilemanager.UpdateOrCreateConfig(profilemanager.ConfigInput{
+			ConfigPath:    targetPath,
+			ManagementURL: "https://mgmt.elsewhere.example:443",
+		})
+		require.NoError(t, err)
+	}
+	t.Cleanup(func() { afterLoginPreCheck = nil })
+
+	_, err = s.Login(userCtx(), &proto.LoginRequest{
+		ProfileName:   &target,
+		Username:      &username,
+		ManagementUrl: storedManagementURL,
+	})
+	require.Error(t, err, "the login became a settings change before it was written")
+	require.Equal(t, codes.FailedPrecondition, gstatus.Code(err), "want the update-settings refusal, got %v", err)
+	require.False(t, cancelled, "the refused login cancelled the login already in progress")
+
+	stored, err := profilemanager.GetExistingConfig(targetPath)
+	require.NoError(t, err)
+	require.Equal(t, "https://mgmt.elsewhere.example:443", stored.ManagementURL.String(),
+		"the refused login wrote the management URL it was asked for")
+}
+
+// Logging out a profile that was already logged out must not fail: the logout
+// clears the keys in place, so the second attempt finds a profile with no
+// identity, which was never registered and has nothing to deregister.
+func TestLogout_ProfileWithoutAnIdentityIsANoOp(t *testing.T) {
+	s, _, _, _, cfgPath := setupServerWithProfile(t)
+
+	loggedOut, err := profilemanager.GetExistingConfig(cfgPath)
+	require.NoError(t, err)
+	loggedOut.PrivateKey = ""
+	loggedOut.SSHKey = ""
+	require.NoError(t, profilemanager.WriteOutConfig(cfgPath, loggedOut))
+
+	stored, err := profilemanager.GetExistingConfig(cfgPath)
+	require.NoError(t, err)
+	require.NoError(t, s.sendLogoutRequestWithConfig(privilegedTestCtx(), stored),
+		"logging out an identity-less profile must not fail")
+
+	// And for an unprivileged caller too: the deregistration privilege gate
+	// guards the handover of a registered key, so with no key there is nothing
+	// to guard. An unset SSH setting is what arms that gate — sshServerEnabled
+	// reads an absent value as enabled — so this stands in for every legacy
+	// profile, where behind the gate the caller would be refused.
+	stored.ServerSSHAllowed = nil
+	require.NoError(t, s.sendLogoutRequestWithConfig(userCtx(), stored),
+		"an unprivileged caller could not log out a profile with nothing to deregister")
+}