github.com/keybase/client/go@v0.0.0-20241007131713-f10651d043c8/service/gregor.go

package service

import (
	"bytes"
	"errors"
	"fmt"
	"net/url"
	"sync"
	"time"

	"golang.org/x/net/context"

	"github.com/keybase/backoff"
	"github.com/keybase/client/go/badges"
	"github.com/keybase/client/go/chat"
	"github.com/keybase/client/go/chat/globals"
	chatstorage "github.com/keybase/client/go/chat/storage"
	"github.com/keybase/client/go/chat/utils"
	"github.com/keybase/client/go/engine"
	"github.com/keybase/client/go/gregor"
	grclient "github.com/keybase/client/go/gregor/client"
	"github.com/keybase/client/go/gregor/storage"
	grutils "github.com/keybase/client/go/gregor/utils"
	"github.com/keybase/client/go/libkb"
	"github.com/keybase/client/go/logger"
	"github.com/keybase/client/go/protocol/chat1"
	"github.com/keybase/client/go/protocol/gregor1"
	"github.com/keybase/client/go/protocol/keybase1"
	"github.com/keybase/clockwork"
	"github.com/keybase/go-framed-msgpack-rpc/rpc"
	jsonw "github.com/keybase/go-jsonw"
)

const GregorRequestTimeout time.Duration = 30 * time.Second
const GregorConnectionRetryInterval time.Duration = 2 * time.Second
const GregorGetClientTimeout time.Duration = 4 * time.Second
const slowConnSleepTime = 1 * time.Second

type IdentifyUIHandler struct {
	libkb.Contextified
	connID      libkb.ConnectionID
	alwaysAlive bool
}

var _ libkb.GregorInBandMessageHandler = (*IdentifyUIHandler)(nil)

func NewIdentifyUIHandler(g *libkb.GlobalContext, connID libkb.ConnectionID) IdentifyUIHandler {
	return IdentifyUIHandler{
		Contextified: libkb.NewContextified(g),
		connID:       connID,
		alwaysAlive:  false,
	}
}

func (h IdentifyUIHandler) IsAlive() bool {
	return (h.alwaysAlive || h.G().ConnectionManager.LookupConnection(h.connID) != nil)
}

func (h IdentifyUIHandler) Name() string {
	return "IdentifyUIHandler"
}

func (h *IdentifyUIHandler) toggleAlwaysAlive(alive bool) {
	h.alwaysAlive = alive
}

type oobmSystemSubscriptions map[string]bool

func newOOBMSystemSubscriptions(oobmSystems []string) oobmSystemSubscriptions {
	if oobmSystems == nil {
		return nil
	}
	ret := make(oobmSystemSubscriptions)
	for _, system := range oobmSystems {
		ret[system] = true
	}
	return ret
}

type gregorFirehoseHandler struct {
	libkb.Contextified
	connID     libkb.ConnectionID
	cli        keybase1.GregorUIClient
	oobmFilter oobmSystemSubscriptions
}

func newGregorFirehoseHandler(g *libkb.GlobalContext, connID libkb.ConnectionID, xp rpc.Transporter, oobmSystems []string) *gregorFirehoseHandler {
	return &gregorFirehoseHandler{
		Contextified: libkb.NewContextified(g),
		connID:       connID,
		cli:          keybase1.GregorUIClient{Cli: rpc.NewClient(xp, libkb.NewContextifiedErrorUnwrapper(g), nil)},
		oobmFilter:   newOOBMSystemSubscriptions(oobmSystems),
	}
}

func (h *gregorFirehoseHandler) IsAlive() bool {
	return h.G().ConnectionManager.LookupConnection(h.connID) != nil
}

func (h *gregorFirehoseHandler) PushState(s gregor1.State, r keybase1.PushReason) {
	defer h.G().Trace("gregorFirehoseHandler#PushState", nil)()
	err := h.cli.PushState(context.Background(), keybase1.PushStateArg{State: s, Reason: r})
	if err != nil {
		h.G().Log.Error(fmt.Sprintf("Error in firehose push state: %s", err))
	}
}

func (h *gregorFirehoseHandler) filterOOBMs(v []gregor1.OutOfBandMessage) []gregor1.OutOfBandMessage {
	// Filter OOBMs down to wanted systems if we have a filter installed
	if h.oobmFilter == nil {
		return v
	}

	var tmp []gregor1.OutOfBandMessage
	for _, m := range v {
		if h.oobmFilter[m.System().String()] {
			tmp = append(tmp, m)
		}
	}
	return tmp
}

func (h *gregorFirehoseHandler) PushOutOfBandMessages(v []gregor1.OutOfBandMessage) {
	defer h.G().Trace("gregorFirehoseHandler#PushOutOfBandMessages", nil)()
	nOrig := len(v)

	// Filter OOBMs down to wanted systems if we have a filter installed
	v = h.filterOOBMs(v)
	h.G().Log.Debug("gregorFirehoseHandler#PushOutOfBandMessages: %d message(s) (%d before filter)", len(v), nOrig)

	if len(v) == 0 {
		return
	}

	err := h.cli.PushOutOfBandMessages(context.Background(), v)
	if err != nil {
		h.G().Log.Error(fmt.Sprintf("Error in firehose push out-of-band messages: %s", err))
	}
}

type testingReplayRes struct {
	replayed []gregor.InBandMessage
	err      error
}

type testingEvents struct {
	broadcastSentCh chan error
	replayThreadCh  chan testingReplayRes
}

func newTestingEvents() *testingEvents {
	return &testingEvents{
		broadcastSentCh: make(chan error),
		replayThreadCh:  make(chan testingReplayRes, 10),
	}
}

type connectionAuthError struct {
	msg         string
	shouldRetry bool
}

func newConnectionAuthError(msg string, shouldRetry bool) connectionAuthError {
	return connectionAuthError{
		msg:         msg,
		shouldRetry: shouldRetry,
	}
}

func (c connectionAuthError) ShouldRetry() bool {
	return c.shouldRetry
}

func (c connectionAuthError) Error() string {
	return fmt.Sprintf("connection auth error: msg: %s shouldRetry: %v", c.msg, c.shouldRetry)
}

type replayThreadArg struct {
	cli gregor1.IncomingInterface
	t   time.Time
	ctx context.Context
}

type gregorHandler struct {
	globals.Contextified

	// This lock is to protect ibmHandlers and gregorCli and firehoseHandlers. Only public methods
	// should grab it.
	sync.Mutex
	ibmHandlers []libkb.GregorInBandMessageHandler

	// gregorCliMu just protecs the gregorCli pointer, since it can be swapped out
	// in one goroutine and accessed in another.
	gregorCliMu sync.Mutex
	gregorCli   *grclient.Client

	firehoseHandlers []libkb.GregorFirehoseHandler
	badger           *badges.Badger
	reachability     *reachability
	chatLog          utils.DebugLabeler

	// This mutex protects the con object
	connMutex sync.Mutex
	conn      *rpc.Connection
	uri       *rpc.FMPURI

	// connectHappened will be closed after gregor connection established
	connectHappened chan struct{}

	cli               rpc.GenericClient
	pingCli           rpc.GenericClient
	sessionID         gregor1.SessionID
	firstConnectMu    sync.Mutex
	firstConnect      bool
	forceSessionCheck bool

	// Function for determining if a new BroadcastMessage should trigger
	// a pushState call to firehose handlers
	pushStateFilter func(m gregor.Message) bool

	shutdownCh  chan struct{}
	broadcastCh chan gregor1.Message
	replayCh    chan replayThreadArg
	pushStateCh chan struct{}
	forcePingCh chan struct{}

	// Testing
	testingEvents       *testingEvents
	transportForTesting *connTransport
}

var _ libkb.GregorState = (*gregorHandler)(nil)
var _ libkb.GregorListener = (*gregorHandler)(nil)

func newGregorHandler(g *globals.Context) *gregorHandler {
	gh := &gregorHandler{
		Contextified:      globals.NewContextified(g),
		chatLog:           utils.NewDebugLabeler(g.ExternalG(), "PushHandler", false),
		firstConnect:      true,
		pushStateFilter:   func(m gregor.Message) bool { return true },
		badger:            nil,
		broadcastCh:       make(chan gregor1.Message, 10000),
		forceSessionCheck: false,
		connectHappened:   make(chan struct{}),
		replayCh:          make(chan replayThreadArg, 10),
		pushStateCh:       make(chan struct{}, 100),
		forcePingCh:       make(chan struct{}, 5),
	}
	return gh
}

// Init starts all the background services for managing connection to Gregor
func (g *gregorHandler) Init() {
	// Start broadcast handler goroutine
	go g.broadcastMessageHandler()
	// Start the app state monitor thread
	go g.monitorAppState()
	// Start replay thread
	go g.syncReplayThread()
}

const (
	monitorConnect int = iota
	monitorDisconnect
	monitorNoop
)

func (g *gregorHandler) monitorAppState() {
	// Wait for state updates and react accordingly
	state := keybase1.MobileAppState_FOREGROUND
	suspended := false
	for {
		monitorAction := monitorNoop
		select {
		case state = <-g.G().MobileAppState.NextUpdate(&state):
			switch state {
			case keybase1.MobileAppState_FOREGROUND:
				g.forcePing(context.Background())
				monitorAction = monitorConnect
			case keybase1.MobileAppState_BACKGROUNDACTIVE:
				monitorAction = monitorConnect
			case keybase1.MobileAppState_BACKGROUND, keybase1.MobileAppState_INACTIVE:
				monitorAction = monitorDisconnect
			}
		case suspended = <-g.G().DesktopAppState.NextSuspendUpdate(&suspended):
			if !suspended {
				monitorAction = monitorConnect
				g.chatLog.Debug(context.Background(), "resumed, connecting")
			} else {
				g.chatLog.Debug(context.Background(), "suspended, disconnecting")
				monitorAction = monitorDisconnect
			}
		}
		switch monitorAction {
		case monitorConnect:
			// Make sure the URI is set before attempting this (possible it isn't in a race)
			if g.uri != nil {
				g.chatLog.Debug(context.Background(), "foregrounded, reconnecting")
				if err := g.Connect(g.uri); err != nil {
					g.chatLog.Debug(context.Background(), "error reconnecting: %s", err)
				}
			}
		case monitorDisconnect:
			g.chatLog.Debug(context.Background(), "backgrounded, shutting down connection")
			g.Shutdown()
		}
	}
}

func (g *gregorHandler) GetURI() *rpc.FMPURI {
	return g.uri
}

func (g *gregorHandler) GetIncomingClient() gregor1.IncomingInterface {
	cli := g.getRPCCli()
	if g.IsShutdown() || cli == nil {
		return gregor1.IncomingClient{Cli: chat.OfflineClient{}}
	}
	return gregor1.IncomingClient{Cli: cli}
}

func (g *gregorHandler) GetClient() chat1.RemoteInterface {
	cli := g.getRPCCli()
	if g.IsShutdown() || cli == nil {
		select {
		case <-g.connectHappened:
			cli = g.getRPCCli()
			if g.IsShutdown() || cli == nil {
				g.chatLog.Debug(context.Background(), "GetClient: connectHappened, but still shutdown, using OfflineClient for chat1.RemoteClient")
				return chat1.RemoteClient{Cli: chat.OfflineClient{}}

			}
			g.chatLog.Debug(context.Background(), "GetClient: successfully waited for connection")
			return chat1.RemoteClient{Cli: chat.NewRemoteClient(g.G(), cli)}
		case <-time.After(GregorGetClientTimeout):
			g.chatLog.Debug(context.Background(), "GetClient: shutdown, using OfflineClient for chat1.RemoteClient (waited %s for connectHappened)", GregorGetClientTimeout)
			return chat1.RemoteClient{Cli: chat.OfflineClient{}}
		}
	}
	g.chatLog.Debug(context.Background(), "GetClient: not shutdown, making new remote client")
	return chat1.RemoteClient{Cli: chat.NewRemoteClient(g.G(), cli)}
}

func (g *gregorHandler) isFirstConnect() bool {
	g.firstConnectMu.Lock()
	defer g.firstConnectMu.Unlock()
	return g.firstConnect
}

func (g *gregorHandler) setFirstConnect(val bool) {
	g.firstConnectMu.Lock()
	defer g.firstConnectMu.Unlock()
	g.firstConnect = val
}

func (g *gregorHandler) shutdownGregorClient(ctx context.Context) {
	g.gregorCliMu.Lock()
	gcliOld := g.gregorCli
	g.gregorCli = nil
	g.gregorCliMu.Unlock()
	if gcliOld != nil {
		gcliOld.Stop()
	}
}

func (g *gregorHandler) resetGregorClient(ctx context.Context, uid gregor1.UID, deviceID gregor1.DeviceID) (gcli *grclient.Client, err error) {
	defer g.G().Trace("gregorHandler#newGregorClient", &err)()
	// Create client object if we are logged in
	if uid != nil && deviceID != nil {
		gcli = grclient.NewClient(uid, deviceID, func() gregor.StateMachine {
			return storage.NewMemEngine(gregor1.ObjFactory{}, clockwork.NewRealClock(), g.G().Log)
		}, storage.NewLocalDB(g.G().ExternalG()), g.GetIncomingClient, g.G().Log, clockwork.NewRealClock())

		// Bring up local state
		g.Debug(ctx, "restoring state from leveldb")
		if err = gcli.Restore(ctx); err != nil {
			// If this fails, we'll keep trying since the server can bail us out
			g.Debug(ctx, "restore local state failed: %s", err)
		}
	}
	g.gregorCliMu.Lock()
	gcliOld := g.gregorCli
	g.gregorCli = gcli
	g.gregorCliMu.Unlock()
	if gcliOld != nil {
		gcliOld.Stop()
	}
	return gcli, nil
}

func (g *gregorHandler) getGregorCli() (*grclient.Client, error) {

	if g == nil {
		return nil, errors.New("gregorHandler client unset")
	}

	g.gregorCliMu.Lock()
	ret := g.gregorCli
	g.gregorCliMu.Unlock()

	if ret == nil {
		return nil, errors.New("client unset")
	}
	return ret, nil
}

func (g *gregorHandler) getRPCCli() rpc.GenericClient {
	g.connMutex.Lock()
	defer g.connMutex.Unlock()
	return g.cli
}

func (g *gregorHandler) Debug(ctx context.Context, s string, args ...interface{}) {
	g.G().Log.CloneWithAddedDepth(1).CDebugf(ctx, "PushHandler: "+s, args...)
}

func (g *gregorHandler) Warning(ctx context.Context, s string, args ...interface{}) {
	g.G().Log.CloneWithAddedDepth(1).CWarningf(ctx, "PushHandler: "+s, args...)
}

func (g *gregorHandler) Errorf(ctx context.Context, s string, args ...interface{}) {
	g.G().Log.CloneWithAddedDepth(1).CErrorf(ctx, "PushHandler: "+s, args...)
}

func (g *gregorHandler) SetPushStateFilter(f func(m gregor.Message) bool) {
	g.pushStateFilter = f
}

func (g *gregorHandler) setReachability(r *reachability) {
	g.reachability = r
}

func (g *gregorHandler) Connect(uri *rpc.FMPURI) (err error) {

	defer g.G().Trace("gregorHandler#Connect", &err)()

	g.connMutex.Lock()
	defer g.connMutex.Unlock()
	if g.conn != nil {
		g.chatLog.Debug(context.Background(), "skipping connect, conn is not nil")
		return nil
	}
	defer func() {
		close(g.connectHappened)
		g.connectHappened = make(chan struct{})
	}()

	// In case we need to interrupt auth'ing or the ping loop,
	// set up this channel.
	g.shutdownCh = make(chan struct{})
	g.uri = uri
	go g.pushStateNewDataDebouncer(g.shutdownCh)
	if uri.UseTLS() {
		err = g.connectTLS()
	} else {
		err = g.connectNoTLS()
	}

	return err
}

func (g *gregorHandler) HandlerName() string {
	return "gregor"
}

// PushHandler adds a new ibm handler to our list. This is usually triggered
// when an external entity (like Electron) connects to the service, and we can
// safely send Gregor information to it
func (g *gregorHandler) PushHandler(handler libkb.GregorInBandMessageHandler) {
	defer g.chatLog.Trace(context.Background(), nil, "PushHandler")()

	g.G().Log.Debug("pushing inband handler %s to position %d", handler.Name(), len(g.ibmHandlers))

	g.Lock()
	g.ibmHandlers = append(g.ibmHandlers, handler)
	g.Unlock()

	// Only try replaying if we are logged in, it's possible that a handler can
	// attach before that is true (like if we start the service logged out and
	// Electron connects)
	cli := g.getRPCCli()
	if g.IsConnected() && cli != nil {
		if _, err := g.replayInBandMessages(context.TODO(), gregor1.IncomingClient{Cli: cli},
			time.Time{}, handler); err != nil {
			g.Errorf(context.Background(), "replayInBandMessages on PushHandler failed: %s", err)
		}

		if g.badger != nil {
			s, err := g.getState(context.Background())
			if err != nil {
				g.Warning(context.Background(), "Cannot get state in PushHandler: %s", err)
				return
			}
			g.badger.PushState(context.Background(), s)
		}
	}
}

// PushFirehoseHandler pushes a new firehose handler onto the list of currently
// active firehose handles. We can have several of these active at once. All
// get the "firehose" of gregor events. They're removed lazily as their underlying
// connections die.
func (g *gregorHandler) PushFirehoseHandler(handler libkb.GregorFirehoseHandler) {
	defer g.chatLog.Trace(context.Background(), nil, "PushFirehoseHandler")()
	g.Lock()
	g.firehoseHandlers = append(g.firehoseHandlers, handler)
	g.Unlock()

	s, err := g.getState(context.Background())
	if err != nil {
		g.Warning(context.Background(), "Cannot push state in firehose handler: %s", err)
		return
	}
	g.Debug(context.Background(), "PushFirehoseHandler: pushing state with %d items", len(s.Items_))
	handler.PushState(s, keybase1.PushReason_RECONNECTED)
}

// iterateOverFirehoseHandlers applies the function f to all live firehose handlers
// and then resets the list to only include the live ones.
func (g *gregorHandler) iterateOverFirehoseHandlers(f func(h libkb.GregorFirehoseHandler)) {
	var freshHandlers []libkb.GregorFirehoseHandler
	for _, h := range g.firehoseHandlers {
		if h.IsAlive() {
			f(h)
			freshHandlers = append(freshHandlers, h)
		}
	}
	g.firehoseHandlers = freshHandlers
}

func (g *gregorHandler) pushStateNewDataDebouncer(shutdownCh chan struct{}) {
	shouldSend := false
	var lastTime time.Time
	dur := time.Second
	trigger := func() {
		if shouldSend {
			go g.pushStateOnce(keybase1.PushReason_NEW_DATA)
			shouldSend = false
			lastTime = time.Now()
		}
	}
	for {
		select {
		case <-g.pushStateCh:
			shouldSend = true
			if time.Since(lastTime) > dur {
				trigger()
			}
		case <-time.After(dur):
			trigger()
		case <-shutdownCh:
			return
		}
	}
}

func (g *gregorHandler) pushStateOnce(r keybase1.PushReason) {
	s, err := g.getState(context.Background())
	if err != nil {
		g.Warning(context.Background(), "Cannot push state in firehose handler: %s", err)
		return
	}
	g.iterateOverFirehoseHandlers(func(h libkb.GregorFirehoseHandler) {
		g.Debug(context.Background(), "pushState: pushing state with %d items", len(s.Items_))
		h.PushState(s, r)
	})
	// Only send this state update on reception of new data, not a reconnect since we will
	// be sending that on a different code path altogether (see OnConnect).
	if g.badger != nil && r != keybase1.PushReason_RECONNECTED {
		g.badger.PushState(context.Background(), s)
	}
}

func (g *gregorHandler) pushState(r keybase1.PushReason) {
	switch r {
	case keybase1.PushReason_RECONNECTED, keybase1.PushReason_NONE:
		g.pushStateOnce(r)
	default:
		g.pushStateCh <- struct{}{}
	}
}

func (g *gregorHandler) pushOutOfBandMessages(m []gregor1.OutOfBandMessage) {
	g.iterateOverFirehoseHandlers(func(h libkb.GregorFirehoseHandler) { h.PushOutOfBandMessages(m) })
}

// replayInBandMessages will replay all the messages in the current state from
// the given time. If a handler is specified, it will only replay using it,
// otherwise it will try all of them. gregorHandler needs to be locked when calling
// this function.
func (g *gregorHandler) replayInBandMessages(ctx context.Context, cli gregor1.IncomingInterface,
	t time.Time, handler libkb.GregorInBandMessageHandler) ([]gregor.InBandMessage, error) {

	var msgs []gregor.InBandMessage
	var err error

	gcli, err := g.getGregorCli()
	if err != nil {
		return nil, err
	}

	if t.IsZero() {
		g.Debug(ctx, "replayInBandMessages: fresh replay: using state items")
		state, err := gcli.StateMachineState(ctx, nil, true)
		if err != nil {
			g.Debug(ctx, "replayInBandMessages: unable to fetch state for replay: %s", err)
			return nil, err
		}
		if msgs, err = gcli.InBandMessagesFromState(state); err != nil {
			g.Debug(ctx, "replayInBandMessages: unable to fetch messages from state for replay: %s", err)
			return nil, err
		}
	} else {
		g.Debug(ctx, "replayInBandMessages: incremental replay: using ibms since")
		if msgs, err = gcli.StateMachineInBandMessagesSince(ctx, t, true); err != nil {
			g.Debug(ctx, "replayInBandMessages: unable to fetch messages for replay: %s", err)
			return nil, err
		}
	}

	g.Debug(ctx, "replayInBandMessages: replaying %d messages", len(msgs))
	for _, msg := range msgs {
		g.Debug(ctx, "replayInBandMessages: replaying: %s", msg.Metadata().MsgID())
		// If we have a handler, just run it on that, otherwise run it against
		// all of the handlers we know about
		if handler == nil {
			err = g.handleInBandMessage(ctx, cli, msg)
		} else {
			_, err = g.handleInBandMessageWithHandler(ctx, cli, msg, handler)
		}

		// If an error happens when replaying, don't kill everything else that
		// follows, just make a warning.
		if err != nil {
			g.Debug(ctx, "replayInBandMessages: failure in message replay: %s", err.Error())
			err = nil
		}
	}

	return msgs, nil
}

func (g *gregorHandler) IsShutdown() bool {
	g.connMutex.Lock()
	defer g.connMutex.Unlock()
	return g.conn == nil
}

func (g *gregorHandler) IsConnected() bool {
	g.connMutex.Lock()
	defer g.connMutex.Unlock()
	return g.conn != nil && g.conn.IsConnected()
}

func (g *gregorHandler) syncReplayThread() {
	for rarg := range g.replayCh {
		var trr testingReplayRes
		now := time.Now()
		g.Debug(rarg.ctx, "serverSync: starting replay thread")
		replayedMsgs, err := g.replayInBandMessages(rarg.ctx, rarg.cli, rarg.t, nil)
		if err != nil {
			g.Debug(rarg.ctx, "serverSync: replayThread: replay messages failed: %s", err)
			trr.err = err
		} else {
			g.Debug(rarg.ctx, "serverSync: replayThread: replayed %d messages", len(replayedMsgs))
			trr.replayed = replayedMsgs
		}
		if g.testingEvents != nil {
			g.testingEvents.replayThreadCh <- trr
		}
		g.Debug(rarg.ctx, "serverSync: syncReplayThread complete: %v", time.Since(now))
	}
}

// serverSync is called from
// gregord. This can happen either on initial startup, or after a reconnect. Needs
// to be called with gregorHandler locked.
func (g *gregorHandler) serverSync(ctx context.Context,
	cli gregor1.IncomingInterface, gcli *grclient.Client, syncRes *chat1.SyncAllNotificationRes) (res []gregor.InBandMessage, err error) {
	defer g.chatLog.Trace(ctx, &err, "serverSync")()

	// Get time of the last message we synced (unless this is our first time syncing)
	var t time.Time
	if !g.isFirstConnect() {
		pt := gcli.StateMachineLatestCTime(ctx)
		if pt != nil {
			t = *pt
		}
		g.Debug(ctx, "serverSync: starting replay from: %s", t)
	} else {
		g.Debug(ctx, "serverSync: performing a fresh replay")
	}

	// Sync down everything from the server
	consumedMsgs, err := gcli.Sync(ctx, cli, syncRes)
	if err != nil {
		g.Debug(ctx, "serverSync: error syncing from the server, reason: %s", err)
		return nil, err
	}
	g.Debug(ctx, "serverSync: consumed %d messages", len(consumedMsgs))

	// Schedule replay of in-band messages
	g.replayCh <- replayThreadArg{
		cli: cli,
		t:   t,
		ctx: globals.BackgroundChatCtx(ctx, g.G()),
	}

	g.pushState(keybase1.PushReason_RECONNECTED)
	return consumedMsgs, nil
}

func (g *gregorHandler) makeReconnectOobm() gregor1.Message {
	return gregor1.Message{
		Oobm_: &gregor1.OutOfBandMessage{
			System_: "internal.reconnect",
		},
	}
}

func (g *gregorHandler) authParams(ctx context.Context) (uid gregor1.UID, deviceID gregor1.DeviceID,
	token gregor1.SessionToken, nist *libkb.NIST, err error) {
	var res loggedInRes
	var stoken string
	var kuid keybase1.UID
	var kdid keybase1.DeviceID
	if kuid, kdid, stoken, nist, res = g.loggedIn(ctx); res != loggedInYes {
		return uid, deviceID, token, nil,
			newConnectionAuthError("failed to check logged in status", res == loggedInMaybe)
	}
	deviceID = make([]byte, libkb.DeviceIDLen)
	if err := kdid.ToBytes(deviceID); err != nil {
		return uid, deviceID, token, nil, err
	}
	g.chatLog.Debug(ctx, "generated NIST for UID %s", kuid)
	return kuid.ToBytes(), deviceID, gregor1.SessionToken(stoken), nist, nil
}

func (g *gregorHandler) inboxParams(ctx context.Context, uid gregor1.UID) chat1.InboxVers {
	// Grab current on disk version
	ibox := chatstorage.NewInbox(g.G())
	vers, err := ibox.Version(ctx, uid)
	if err != nil {
		g.chatLog.Debug(ctx, "inboxParams: failed to get current inbox version (using 0): %s",
			err.Error())
		vers = chat1.InboxVers(0)
	}
	return vers
}

func (g *gregorHandler) notificationParams(ctx context.Context, gcli *grclient.Client) (t gregor1.Time) {
	pt := gcli.StateMachineLatestCTime(ctx)
	if pt != nil {
		t = gregor1.ToTime(*pt)
	}
	g.chatLog.Debug(ctx, "notificationParams: latest ctime: %v", t.Time())
	return t
}

// OnConnect is called by the rpc library to indicate we have connected to
// gregord
func (g *gregorHandler) OnConnect(ctx context.Context, conn *rpc.Connection,
	cli rpc.GenericClient, srv *rpc.Server) (err error) {

	ctx = libkb.WithLogTag(ctx, "GRGRONCONN")

	defer g.chatLog.Trace(ctx, &err, "OnConnect")()

	// If we get a random OnConnect on some other connection that is not g.conn, then
	// just reject it.
	g.connMutex.Lock()
	if conn != g.conn {
		g.connMutex.Unlock()
		g.chatLog.Debug(ctx, "aborting on dup connection")
		return chat.ErrDuplicateConnection
	}
	g.connMutex.Unlock()

	g.chatLog.Debug(ctx, "connected")
	timeoutCli := WrapGenericClientWithTimeout(cli, GregorRequestTimeout, chat.ErrChatServerTimeout)
	chatCli := chat1.RemoteClient{Cli: chat.NewRemoteClient(g.G(), cli)}
	if err := srv.Register(gregor1.OutgoingProtocol(g)); err != nil {
		return fmt.Errorf("error registering protocol: %s", err)
	}

	uid, deviceID, token, nist, err := g.authParams(ctx)
	if err != nil {
		return err
	}
	gcli, err := g.resetGregorClient(ctx, uid, deviceID)
	if err != nil {
		return fmt.Errorf("failed to get gregor client: %s", err)
	}
	iboxVers := g.inboxParams(ctx, uid)
	latestCtime := g.notificationParams(ctx, gcli)

	// Run SyncAll to both authenticate, and grab all the data we will need to run the
	// various resync procedures for chat and notifications
	var identBreaks []keybase1.TLFIdentifyFailure
	ctx = globals.ChatCtx(ctx, g.G(), keybase1.TLFIdentifyBehavior_CHAT_GUI, &identBreaks,
		chat.NewCachingIdentifyNotifier(g.G()))
	g.chatLog.Debug(ctx, "OnConnect begin")
	syncAllRes, err := chatCli.SyncAll(ctx, chat1.SyncAllArg{
		Uid:              uid,
		DeviceID:         deviceID,
		Session:          token,
		InboxVers:        iboxVers,
		Ctime:            latestCtime,
		Fresh:            g.isFirstConnect(),
		ProtVers:         chat1.SyncAllProtVers_V1,
		HostName:         g.GetURI().Host,
		SummarizeMaxMsgs: true,
		ParticipantsMode: chat1.InboxParticipantsMode_SKIP_TEAMS,
	})
	if err != nil {
		// This will cause us to try and refresh session on the next attempt
		if _, ok := err.(libkb.BadSessionError); ok {
			g.chatLog.Debug(ctx, "bad session from SyncAll(): forcing session check on next attempt")
			g.forceSessionCheck = true
			nist.MarkFailure()
		}
		return fmt.Errorf("error running SyncAll: %s", err)
	}

	// Use the client parameter instead of conn.GetClient(), since we can get stuck
	// in a recursive loop if we keep retrying on reconnect.
	if err := g.auth(ctx, timeoutCli, &syncAllRes.Auth); err != nil {
		return fmt.Errorf("error authenticating: %s", err)
	}

	// Update badging for chat.
	// This happens before Syncer.Connected for a reason.
	// If the new inbox version (e.g. 8) were committed to disk and then the
	// app lost connection and bailed out of OnConnect before applying the
	// badging update (7->8) then on reconnect an incomplete chat badge update (8->9)
	// could be received.
	// See: https://github.com/keybase/client/pull/12651
	if g.badger != nil {
		g.badger.PushChatFullUpdate(ctx, syncAllRes.Badge)
	}

	// Sync chat data using a Syncer object
	// This commits the new inbox version to persistent storage.
	if err := g.G().Syncer.Connected(ctx, chatCli, uid, &syncAllRes.Chat); err != nil {
		return fmt.Errorf("error running chat sync: %s", err)
	}

	// Sync down events since we have been dead
	if _, err := g.serverSync(ctx, gregor1.IncomingClient{Cli: timeoutCli}, gcli,
		&syncAllRes.Notification); err != nil {
		g.chatLog.Debug(ctx, "serverSync: failure: %s", err)
		return fmt.Errorf("error running state sync: %s", err)
	}

	// Update badging from gregor.
	if g.badger != nil {
		state, err := gcli.StateMachineState(ctx, nil, false)
		if err != nil {
			g.chatLog.Debug(ctx, "unable to get gregor state for badging: %v", err)
			g.badger.PushState(ctx, gregor1.State{})
		} else {
			g.badger.PushState(ctx, state)
		}
	}

	// Call out to reachability module if we have one
	if g.reachability != nil {
		g.chatLog.Debug(ctx, "setting reachability")
		g.reachability.setReachability(keybase1.Reachability{
			Reachable: keybase1.Reachable_YES,
		})
	}

	// Broadcast reconnect oobm. Spawn this off into a goroutine so that we don't delay
	// reconnection any longer than we have to.
	g.chatLog.Debug(ctx, "broadcasting reconnect oobm")
	go func(m gregor1.Message) {
		ctx := context.Background()
		err := g.BroadcastMessage(ctx, m)
		if err != nil {
			g.chatLog.Debug(ctx, "Gregor broadcast error: %+v", err)
		}
	}(g.makeReconnectOobm())

	// No longer first connect if we are now connected
	g.chatLog.Debug(ctx, "setting first connect to false")
	g.setFirstConnect(false)
	// On successful login we can reset this guy to not force a check
	g.forceSessionCheck = false
	g.chatLog.Debug(ctx, "OnConnect complete")

	return nil
}

func (g *gregorHandler) OnConnectError(err error, reconnectThrottleDuration time.Duration) {
	defer g.chatLog.Trace(context.Background(), nil, "OnConnectError")()
	g.chatLog.Debug(context.Background(), "OnConnectError: err: %s, reconnect throttle duration: %s", err,
		reconnectThrottleDuration)

	// Check reachability here to see the nature of our offline status
	go func() {
		if g.reachability != nil && !g.isReachable() {
			g.reachability.setReachability(keybase1.Reachability{
				Reachable: keybase1.Reachable_NO,
			})
		}
	}()
}

func (g *gregorHandler) OnDisconnected(ctx context.Context, status rpc.DisconnectStatus) {
	g.chatLog.Debug(context.Background(), "disconnected: %v", status)

	// Alert chat syncer that we are now disconnected
	g.G().Syncer.Disconnected(ctx)

	// Call out to reachability module if we have one (and we are currently connected)
	go func() {
		if g.reachability != nil && status != rpc.StartingFirstConnection && !g.isReachable() {
			g.reachability.setReachability(keybase1.Reachability{
				Reachable: keybase1.Reachable_NO,
			})
		}
	}()
}

func (g *gregorHandler) OnDoCommandError(err error, nextTime time.Duration) {
	g.chatLog.Debug(context.Background(), "do command error: %s, nextTime: %s", err, nextTime)
}

func (g *gregorHandler) ShouldRetry(name string, err error) bool {
	g.chatLog.Debug(context.Background(), "should retry: name %s, err %v (returning false)", name, err)
	return false
}

func (g *gregorHandler) ShouldRetryOnConnect(err error) bool {
	if err == nil {
		return false
	}

	ctx := context.Background()
	g.chatLog.Debug(ctx, "should retry on connect, err %v", err)
	if err == chat.ErrDuplicateConnection {
		g.chatLog.Debug(ctx, "duplicate connection error, not retrying")
		return false
	}
	if _, ok := err.(libkb.BadSessionError); ok {
		g.chatLog.Debug(ctx, "bad session error, not retrying")
		return false
	}
	if cerr, ok := err.(connectionAuthError); ok && !cerr.ShouldRetry() {
		g.chatLog.Debug(ctx, "should retry on connect, non-retry error, ending: %s", err.Error())
		return false
	}

	return true
}

func (g *gregorHandler) broadcastMessageOnce(ctx context.Context, m gregor1.Message) (err error) {
	defer g.chatLog.Trace(ctx, &err, "broadcastMessageOnce")()

	// Handle the message
	var obm gregor.OutOfBandMessage
	ibm := m.ToInBandMessage()
	if ibm != nil {
		gcli, err := g.getGregorCli()
		if err != nil {
			g.Debug(ctx, "BroadcastMessage: failed to get Gregor client: %s", err.Error())
			return err
		}
		// Check to see if this is already in our state
		msgID := ibm.Metadata().MsgID()
		state, err := gcli.StateMachineState(ctx, nil, false)
		if err != nil {
			g.Debug(ctx, "BroadcastMessage: no state machine available: %s", err.Error())
			return err
		}
		if _, ok := state.GetItem(msgID); ok {
			g.Debug(ctx, "BroadcastMessage: msgID: %s already in state, ignoring", msgID)
			return errors.New("ignored repeat message")
		}

		g.Debug(ctx, "broadcast: in-band message: msgID: %s Ctime: %s", msgID, ibm.Metadata().CTime())
		err = g.handleInBandMessage(ctx, g.GetIncomingClient(), ibm)

		// Send message to local state machine
		consumeErr := gcli.StateMachineConsumeMessage(ctx, m)
		if consumeErr != nil {
			g.Debug(ctx, "broadcast: error consuming message: %+v", consumeErr)
		}

		// Forward to electron or whichever UI is listening for the new gregor state
		if g.pushStateFilter(m) {
			g.pushState(keybase1.PushReason_NEW_DATA)
		}

		return err
	}

	obm = m.ToOutOfBandMessage()
	if obm != nil {
		g.Debug(ctx, "broadcast: out-of-band message: uid: %s",
			m.ToOutOfBandMessage().UID())
		if err := g.handleOutOfBandMessage(ctx, obm); err != nil {
			g.Debug(ctx, "BroadcastMessage: error handling oobm: %s", err.Error())
			return err
		}
		return nil
	}

	g.Debug(ctx, "BroadcastMessage: both in-band and out-of-band message nil")
	return errors.New("invalid message, no ibm or oobm")
}

func (g *gregorHandler) broadcastMessageHandler() {
	ctx := context.Background()
	for {
		m := <-g.broadcastCh
		if g.G().GetEnv().GetSlowGregorConn() {
			g.Debug(ctx, "[slow conn]: sleeping")
			time.Sleep(slowConnSleepTime)
			g.Debug(ctx, "[slow conn]: awake")
		}
		err := g.broadcastMessageOnce(ctx, m)
		if err != nil {
			g.Debug(context.Background(), "broadcast error: %v", err)
		}

		// Testing alerts
		if g.testingEvents != nil {
			g.testingEvents.broadcastSentCh <- err
		}
	}
}

// BroadcastMessage is called when we receive a new message from gregord. Grabs
// the lock protect the state machine and handleInBandMessage
func (g *gregorHandler) BroadcastMessage(ctx context.Context, m gregor1.Message) error {
	// Send the message on a channel so we can return to Gregor as fast as possible. Note
	// that this can block, but broadcastCh has a large buffer to try and mitigate
	g.broadcastCh <- m
	return nil
}

// handleInBandMessage runs a message on all the alive handlers. gregorHandler
// must be locked when calling this function.
func (g *gregorHandler) handleInBandMessage(ctx context.Context, cli gregor1.IncomingInterface,
	ibm gregor.InBandMessage) (err error) {

	defer g.G().Trace(fmt.Sprintf("gregorHandler#handleInBandMessage with %d handlers", len(g.ibmHandlers)), &err)()
	ctx = libkb.WithLogTag(ctx, "GRGIBM")

	var freshHandlers []libkb.GregorInBandMessageHandler

	// Loop over all handlers and run the messages against any that are alive
	// If the handler is not alive, we prune it from our list
	for i, handler := range g.ibmHandlers {
		g.Debug(ctx, "trying handler %s at position %d", handler.Name(), i)
		if handler.IsAlive() {
			if handled, err := g.handleInBandMessageWithHandler(ctx, cli, ibm, handler); err != nil {
				if handled {
					// Don't stop handling errors on a first failure.
					g.Errorf(ctx, "failed to run %s handler: %s", handler.Name(), err)
				} else {
					g.Debug(ctx, "handleInBandMessage() failed to run %s handler: %s", handler.Name(), err)
				}
			}
			freshHandlers = append(freshHandlers, handler)
		} else {
			g.Debug(ctx, "skipping handler as it's marked dead: %s", handler.Name())
		}
	}

	if len(g.ibmHandlers) != len(freshHandlers) {
		g.Debug(ctx, "Change # of live handlers from %d to %d", len(g.ibmHandlers), len(freshHandlers))
		g.ibmHandlers = freshHandlers
	}
	return nil
}

// handleInBandMessageWithHandler runs a message against the specified handler
func (g *gregorHandler) handleInBandMessageWithHandler(ctx context.Context, cli gregor1.IncomingInterface,
	ibm gregor.InBandMessage, handler libkb.GregorInBandMessageHandler) (bool, error) {
	g.Debug(ctx, "handleInBand: %+v", ibm)

	gcli, err := g.getGregorCli()
	if err != nil {
		return false, err
	}
	state, err := gcli.StateMachineState(ctx, nil, false)
	if err != nil {
		return false, err
	}

	sync := ibm.ToStateSyncMessage()
	if sync != nil {
		g.Debug(ctx, "state sync message")
		return false, nil
	}

	update := ibm.ToStateUpdateMessage()
	if update != nil {
		g.Debug(ctx, "state update message")

		item := update.Creation()
		if item != nil {
			id := item.Metadata().MsgID().String()
			g.Debug(ctx, "msg ID %s created ctime: %s", id,
				item.Metadata().CTime())

			category := ""
			if item.Category() != nil {
				category = item.Category().String()
				g.Debug(ctx, "item %s has category %s", id, category)
			}

			if handled, err := handler.Create(ctx, cli, category, item); err != nil {
				return handled, err
			}
		}

		dismissal := update.Dismissal()
		if dismissal != nil {
			g.Debug(ctx, "received dismissal")
			for _, id := range dismissal.MsgIDsToDismiss() {
				item, present := state.GetItem(id)
				if !present {
					g.Debug(ctx, "tried to dismiss item %s, not present", id.String())
					continue
				}
				g.Debug(ctx, "dismissing item %s", id.String())

				category := ""
				if item.Category() != nil {
					category = item.Category().String()
					g.Debug(ctx, "dismissal %s has category %s", id, category)
				}

				if handled, err := handler.Dismiss(ctx, cli, category, item); handled && err != nil {
					return handled, err
				}
			}
			if len(dismissal.RangesToDismiss()) > 0 {
				g.Debug(ctx, "message range dismissing not implemented")
			}
		}

		return true, nil
	}

	return false, nil
}

func (h IdentifyUIHandler) Create(ctx context.Context, cli gregor1.IncomingInterface, category string,
	item gregor.Item) (bool, error) {

	switch category {
	case "show_tracker_popup":
		return true, h.handleShowTrackerPopupCreate(ctx, cli, item)
	default:
		return false, nil
	}
}

func (h IdentifyUIHandler) Dismiss(ctx context.Context, cli gregor1.IncomingInterface, category string,
	item gregor.Item) (bool, error) {

	switch category {
	case "show_tracker_popup":
		return true, h.handleShowTrackerPopupDismiss(ctx, cli, item)
	default:
		return false, nil
	}
}

func (h IdentifyUIHandler) handleShowTrackerPopupCreate(ctx context.Context, cli gregor1.IncomingInterface,
	item gregor.Item) error {

	h.G().Log.Debug("handleShowTrackerPopupCreate: %+v", item)
	if item.Body() == nil {
		return errors.New("gregor handler for show_tracker_popup: nil message body")
	}
	body, err := jsonw.Unmarshal(item.Body().Bytes())
	if err != nil {
		h.G().Log.Debug("body failed to unmarshal", err)
		return err
	}
	uidString, err := body.AtPath("uid").GetString()
	if err != nil {
		h.G().Log.Debug("failed to extract uid", err)
		return err
	}
	uid, err := keybase1.UIDFromString(uidString)
	if err != nil {
		h.G().Log.Debug("failed to convert UID from string", err)
		return err
	}

	identifyUI, err := h.G().UIRouter.GetIdentifyUI()
	if err != nil {
		h.G().Log.Debug("failed to get IdentifyUI", err)
		return err
	}
	if identifyUI == nil {
		h.G().Log.Debug("got nil IdentifyUI")
		return errors.New("got nil IdentifyUI")
	}
	secretUI, err := h.G().UIRouter.GetSecretUI(0)
	if err != nil {
		h.G().Log.Debug("failed to get SecretUI", err)
		return err
	}
	if secretUI == nil {
		h.G().Log.Debug("got nil SecretUI")
		return errors.New("got nil SecretUI")
	}
	uis := libkb.UIs{
		IdentifyUI: identifyUI,
		SecretUI:   secretUI,
	}

	identifyReason := keybase1.IdentifyReason{
		Type: keybase1.IdentifyReasonType_TRACK,
		// TODO: text here?
	}
	identifyArg := keybase1.Identify2Arg{Uid: uid, Reason: identifyReason}
	m := libkb.NewMetaContext(ctx, h.G()).WithUIs(uis)
	identifyEng := engine.NewIdentify2WithUID(h.G(), &identifyArg)
	identifyEng.SetResponsibleGregorItem(item)
	return identifyEng.Run(m)
}

func (h IdentifyUIHandler) handleShowTrackerPopupDismiss(ctx context.Context, cli gregor1.IncomingInterface,
	item gregor.Item) error {
	mctx := libkb.NewMetaContext(ctx, h.G())

	mctx.Debug("handleShowTrackerPopupDismiss: %+v", item)
	if item.Body() == nil {
		return errors.New("gregor dismissal for show_tracker_popup: nil message body")
	}
	body, err := jsonw.Unmarshal(item.Body().Bytes())
	if err != nil {
		mctx.Debug("body failed to unmarshal", err)
		return err
	}
	uidString, err := body.AtPath("uid").GetString()
	if err != nil {
		mctx.Debug("failed to extract uid", err)
		return err
	}
	uid, err := keybase1.UIDFromString(uidString)
	if err != nil {
		mctx.Debug("failed to convert UID from string", err)
		return err
	}
	user, err := libkb.LoadUser(libkb.NewLoadUserByUIDArg(ctx, h.G(), uid))
	if err != nil {
		mctx.Debug("failed to load user from UID", err)
		return err
	}

	identifyUI, err := h.G().UIRouter.GetIdentifyUI()
	if err != nil {
		mctx.Debug("failed to get IdentifyUI", err)
		return err
	}
	if identifyUI == nil {
		mctx.Debug("got nil IdentifyUI")
		return errors.New("got nil IdentifyUI")
	}

	reason := keybase1.DismissReason{
		Type: keybase1.DismissReasonType_HANDLED_ELSEWHERE,
	}
	_ = identifyUI.Dismiss(mctx, user.GetName(), reason)

	return nil
}

func (g *gregorHandler) handleOutOfBandMessage(ctx context.Context, obm gregor.OutOfBandMessage) error {
	if obm.System() == nil {
		return errors.New("nil system in out of band message")
	}

	if tmp, ok := obm.(gregor1.OutOfBandMessage); ok {
		g.pushOutOfBandMessages([]gregor1.OutOfBandMessage{tmp})
	} else {
		g.G().Log.Warning("Got non-exportable out-of-band message")
	}

	// Send the oobm to the chat system so that it can potentially handle it
	if g.G().PushHandler != nil {
		handled, err := g.G().PushHandler.HandleOobm(ctx, obm)
		if err != nil {
			return err
		}
		if handled {
			return nil
		}
	}

	// Send the oobm to the wallet system so that it can potentially handle it
	if g.G().StellarPushHandler != nil {
		handled, err := g.G().StellarPushHandler.HandleOobm(ctx, obm)
		if err != nil {
			return err
		}
		if handled {
			return nil
		}
	}

	switch obm.System().String() {
	case "internal.reconnect":
		g.G().Log.Debug("reconnected to push server")
		return nil
	default:
		return fmt.Errorf("unhandled system: %s", obm.System())
	}
}

func (g *gregorHandler) Shutdown() {
	defer g.chatLog.Trace(context.Background(), nil, "Shutdown")()
	g.connMutex.Lock()
	defer g.connMutex.Unlock()

	if g.conn == nil {
		return
	}

	// Alert chat syncer that we are now disconnected
	g.G().Syncer.Disconnected(context.Background())

	close(g.shutdownCh)
	g.conn.Shutdown()
	g.conn = nil
	g.cli = nil
}

func (g *gregorHandler) Reset() error {
	g.Shutdown()
	g.setFirstConnect(true)
	g.shutdownGregorClient(context.TODO())
	return nil
}

type loggedInRes int

const (
	loggedInYes loggedInRes = iota
	loggedInNo
	loggedInMaybe
)

func (g *gregorHandler) loggedIn(ctx context.Context) (uid keybase1.UID, did keybase1.DeviceID, token string, nist *libkb.NIST, res loggedInRes) {

	// Check to see if we have been shut down,
	select {
	case <-g.shutdownCh:
		return uid, did, token, nil, loggedInMaybe
	default:
		// if we were going to block, then that means we are still alive
	}

	var err error

	nist, uid, did, err = g.G().ActiveDevice.NISTAndUIDDeviceID(ctx)
	if nist == nil {
		g.G().Log.CDebugf(ctx, "gregorHandler: no NIST for login; user isn't logged in")
		return uid, did, token, nil, loggedInNo
	}
	if err != nil {
		g.G().Log.CDebugf(ctx, "gregorHandler: error in generating NIST: %s", err.Error())
		return uid, did, token, nil, loggedInMaybe
	}

	return uid, did, nist.Token().String(), nist, loggedInYes
}

func (g *gregorHandler) auth(ctx context.Context, cli rpc.GenericClient, auth *gregor1.AuthResult) (err error) {
	var token string
	var res loggedInRes
	var uid keybase1.UID
	var nist *libkb.NIST

	if uid, _, token, nist, res = g.loggedIn(ctx); res != loggedInYes {
		return newConnectionAuthError("not logged in for auth", res == loggedInMaybe)
	}

	if auth == nil {
		g.chatLog.Debug(ctx, "logged in: authenticating")
		ac := gregor1.AuthClient{Cli: cli}
		auth = new(gregor1.AuthResult)
		*auth, err = ac.AuthenticateSessionToken(ctx, gregor1.SessionToken(token))
		if err != nil {
			g.chatLog.Debug(ctx, "auth error: %s", err)
			g.forceSessionCheck = true
			nist.DidFail()
			return err
		}
	} else {
		g.Debug(ctx, "using previously obtained auth result")
	}

	g.chatLog.Debug(ctx, "auth result: %+v", *auth)
	if !bytes.Equal(auth.Uid, uid.ToBytes()) {
		msg := fmt.Sprintf("auth result uid %x doesn't match session uid %q", auth.Uid, uid)
		return newConnectionAuthError(msg, false)
	}
	g.sessionID = auth.Sid

	return nil
}

func (g *gregorHandler) isReachable() bool {
	ctx := context.Background()
	timeout := g.G().Env.GetGregorPingTimeout()
	url, err := url.Parse(g.G().Env.GetGregorURI())
	if err != nil {
		g.chatLog.Debug(ctx, "isReachable: failed to parse server uri, exiting: %s", err.Error())
		return false
	}

	// If we currently think we are online, then make sure
	conn, err := libkb.ProxyDialTimeout(g.G().Env, "tcp", url.Host, timeout)
	if conn != nil {
		conn.Close()
		return true
	}
	if err != nil {
		g.chatLog.Debug(ctx, "isReachable: error: terminating connection: %s", err.Error())
		if _, err := g.Reconnect(ctx); err != nil {
			g.chatLog.Debug(ctx, "isReachable: error reconnecting: %s", err.Error())
		}
		return false
	}

	return true
}

func (g *gregorHandler) Reconnect(ctx context.Context) (didShutdown bool, err error) {
	if g.IsConnected() {
		didShutdown = true
		g.chatLog.Debug(ctx, "Reconnect: reconnecting to server")
		g.Shutdown()
		return didShutdown, g.Connect(g.uri)
	}

	didShutdown = false
	g.chatLog.Debug(ctx, "Reconnect: skipping reconnect, already disconnected")
	return didShutdown, nil
}

func (g *gregorHandler) forcePing(ctx context.Context) {
	select {
	case g.forcePingCh <- struct{}{}:
	default:
		g.Debug(ctx, "forcePing: failed to write to channel, its full")
	}
}

func (g *gregorHandler) pingOnce(ctx context.Context, id []byte, shutdownCancel context.CancelFunc) {
	var err error
	doneCh := make(chan error)
	timeout := g.G().Env.GetGregorPingTimeout()
	go func(ctx context.Context) {
		if g.IsConnected() {
			// If we are connected, subject the ping call to a fairly
			// aggressive timeout so our chat stuff can be responsive
			// to changes in connectivity
			var timeoutCancel context.CancelFunc
			var timeoutCtx context.Context
			timeoutCtx, timeoutCancel = context.WithTimeout(ctx, timeout)
			_, err = gregor1.IncomingClient{Cli: g.pingCli}.Ping(timeoutCtx)
			timeoutCancel()
		} else {
			// If we are not connected, we don't want to timeout anything
			// Just hook into the normal reconnect chan stuff in the RPC
			// library
			g.chatLog.Debug(ctx, "ping loop: id: %x normal ping, not connected", id)
			_, err = gregor1.IncomingClient{Cli: g.pingCli}.Ping(ctx)
			g.chatLog.Debug(ctx, "ping loop: id: %x normal ping success", id)
		}
		select {
		case <-ctx.Done():
			g.chatLog.Debug(ctx, "ping loop: id: %x context cancelled, so not sending err", id)
		default:
			doneCh <- err
		}
	}(ctx)

	select {
	case err = <-doneCh:
	case <-g.shutdownCh:
		g.chatLog.Debug(ctx, "ping loop: id: %x shutdown received", id)
		shutdownCancel()
		return
	}
	if err != nil {
		g.Debug(ctx, "ping loop: id: %x error: %s", id, err)
		if err == context.DeadlineExceeded {
			g.chatLog.Debug(ctx, "ping loop: timeout: terminating connection")
			var didShutdown bool
			var err error
			if didShutdown, err = g.Reconnect(ctx); err != nil {
				g.chatLog.Debug(ctx, "ping loop: id: %x error reconnecting: %s", id, err)
			}
			// It is possible that we have already reconnected by the time we call Reconnect
			// above. If that is the case, we don't want to terminate the ping loop. Only
			// if Reconnect has actually reset the connection do we stop this ping loop.
			if didShutdown {
				shutdownCancel()
				return
			}
		}
	}
}

func (g *gregorHandler) pingLoop() {
	ctx := context.Background()
	id, _ := libkb.RandBytes(4)
	duration := g.G().Env.GetGregorPingInterval()
	timeout := g.G().Env.GetGregorPingTimeout()
	url, err := url.Parse(g.G().Env.GetGregorURI())
	if err != nil {
		g.chatLog.Debug(ctx, "ping loop: failed to parse server uri, exiting: %s", err.Error())
		return
	}
	g.chatLog.Debug(ctx, "ping loop: starting up: id: %x duration: %v timeout: %v url: %s",
		id, duration, timeout, url.Host)
	defer g.chatLog.Debug(ctx, "ping loop: id: %x terminating", id)
	ticker := time.NewTicker(duration)
	for {
		ctx, shutdownCancel := context.WithCancel(context.Background())
		select {
		case <-g.forcePingCh:
			g.chatLog.Debug(ctx, "ping loop: forced attempt")
			g.pingOnce(ctx, id, shutdownCancel)
		case <-ticker.C:
			g.pingOnce(ctx, id, shutdownCancel)
		case <-g.shutdownCh:
			g.chatLog.Debug(ctx, "ping loop: id: %x shutdown received", id)
			shutdownCancel()
			return
		}
		shutdownCancel()
	}
}

// connMutex must be locked before calling this
func (g *gregorHandler) connectTLS() error {
	ctx := context.Background()
	if g.conn != nil {
		g.chatLog.Debug(ctx, "skipping connect, conn is not nil")
		return nil
	}

	uri := g.uri
	g.chatLog.Debug(ctx, "connecting to gregord via TLS at %s", uri)
	rawCA := g.G().Env.GetBundledCA(uri.Host)
	if len(rawCA) == 0 {
		return fmt.Errorf("No bundled CA for %s", uri.Host)
	}
	g.chatLog.Debug(ctx, "Using CA for gregor: %s", libkb.ShortCA(rawCA))
	// Let people know we are trying to sync
	g.G().NotifyRouter.HandleChatInboxSyncStarted(ctx, g.G().Env.GetUID())

	opts := rpc.ConnectionOpts{
		TagsFunc:      logger.LogTagsFromContextRPC,
		WrapErrorFunc: libkb.MakeWrapError(g.G().ExternalG()),
		ReconnectBackoff: func() backoff.BackOff {
			return backoff.NewConstantBackOff(GregorConnectionRetryInterval)
		},
		DialerTimeout:    10 * time.Second,
		HandshakeTimeout: 10 * time.Second,
		// We deliberately avoid ForceInitialBackoff here, becuase we don't
		// want to penalize mobile, which tears down its connection frequently.
	}
	g.conn = rpc.NewTLSConnectionWithDialable(rpc.NewFixedRemote(uri.HostPort),
		[]byte(rawCA), libkb.NewContextifiedErrorUnwrapper(g.G().ExternalG()),
		g, libkb.NewRPCLogFactory(g.G().ExternalG()),
		g.G().ExternalG().RemoteNetworkInstrumenterStorage,
		logger.LogOutputWithDepthAdder{Logger: g.G().Log},
		rpc.DefaultMaxFrameLength, opts,
		libkb.NewProxyDialable(g.G().Env))

	// The client we get here will reconnect to gregord on disconnect if necessary.
	// We should grab it here instead of in OnConnect, since the connection is not
	// fully established in OnConnect. Anything that wants to make calls outside
	// of OnConnect should use g.cli, everything else should the client that is
	// a parameter to OnConnect
	g.cli = WrapGenericClientWithTimeout(g.conn.GetClient(), GregorRequestTimeout,
		chat.ErrChatServerTimeout)
	g.pingCli = g.conn.GetClient() // Don't want this to have a timeout from here

	// Start up ping loop to keep the connection to gregord alive, and to kick
	// off the reconnect logic in the RPC library
	go g.pingLoop()

	return nil
}

// connMutex must be locked before calling this
func (g *gregorHandler) connectNoTLS() error {
	ctx := context.Background()
	if g.conn != nil {
		g.chatLog.Debug(ctx, "skipping connect, conn is not nil")
		return nil
	}
	uri := g.uri
	g.chatLog.Debug(ctx, "connecting to gregord without TLS at %s", uri)
	t := newConnTransport(g.G().ExternalG(), uri.HostPort)
	g.transportForTesting = t

	opts := rpc.ConnectionOpts{
		TagsFunc:      logger.LogTagsFromContextRPC,
		WrapErrorFunc: libkb.MakeWrapError(g.G().ExternalG()),
		ReconnectBackoff: func() backoff.BackOff {
			return backoff.NewConstantBackOff(GregorConnectionRetryInterval)
		},
	}
	g.conn = rpc.NewConnectionWithTransport(g, t,
		libkb.NewContextifiedErrorUnwrapper(g.G().ExternalG()),
		logger.LogOutputWithDepthAdder{Logger: g.G().Log}, opts)

	g.cli = WrapGenericClientWithTimeout(g.conn.GetClient(), GregorRequestTimeout,
		chat.ErrChatServerTimeout)
	g.pingCli = g.conn.GetClient()

	// Start up ping loop to keep the connection to gregord alive, and to kick
	// off the reconnect logic in the RPC library
	go g.pingLoop()

	return nil
}

func (g *gregorHandler) currentUID() gregor1.UID {
	return gregor1.UID(g.G().ActiveDevice.UID().ToBytes())
}

// `cli` is the interface used to talk to gregor.
// If nil then the global cli will be used.
// Be sure to pass a cli when called from within OnConnect, as the global cli would deadlock.
func (g *gregorHandler) DismissItem(ctx context.Context, cli gregor1.IncomingInterface, id gregor.MsgID) error {
	if id == nil {
		return nil
	}
	var err error
	defer g.G().CTrace(ctx, fmt.Sprintf("gregorHandler.dismissItem(%s)", id.String()),
		&err,
	)()
	defer g.pushState(keybase1.PushReason_NEW_DATA)
	dismissal, err := grutils.FormMessageForDismissItem(ctx, g.currentUID(), id)
	if err != nil {
		return err
	}
	gcli, err := g.getGregorCli()
	if err != nil {
		return err
	}
	return gcli.ConsumeMessage(ctx, dismissal)
}

func (g *gregorHandler) LocalDismissItem(ctx context.Context, id gregor.MsgID) (err error) {
	if id == nil {
		return nil
	}
	defer g.G().CTrace(ctx, fmt.Sprintf("gregorHandler.localDismissItem(%s)", id.String()),
		&err,
	)()
	defer g.pushState(keybase1.PushReason_NEW_DATA)

	cli, err := g.getGregorCli()
	if err != nil {
		return err
	}
	return cli.StateMachineConsumeLocalDismissal(ctx, id)
}

func (g *gregorHandler) DismissCategory(ctx context.Context, category gregor1.Category) error {
	var err error
	defer g.G().CTrace(ctx, fmt.Sprintf("gregorHandler.DismissCategory(%s)", category.String()),
		&err,
	)()
	defer g.pushState(keybase1.PushReason_NEW_DATA)

	dismissal, err := grutils.FormMessageForDismissCategory(ctx, g.currentUID(), category)
	if err != nil {
		return err
	}

	gcli, err := g.getGregorCli()
	if err != nil {
		return err
	}
	return gcli.ConsumeMessage(ctx, dismissal)
}

func (g *gregorHandler) InjectItem(ctx context.Context, cat string, body []byte, dtime gregor1.TimeOrOffset) (gregor1.MsgID, error) {
	var err error
	defer g.G().CTrace(ctx, fmt.Sprintf("gregorHandler.InjectItem(%s)", cat),
		&err,
	)()
	defer g.pushState(keybase1.PushReason_NEW_DATA)

	creation, err := grutils.FormMessageForInjectItem(ctx, g.currentUID(), cat, body, dtime)
	if err != nil {
		return nil, err
	}

	gcli, err := g.getGregorCli()
	if err != nil {
		return nil, err
	}
	retMsgID := gregor1.MsgID(creation.ToInBandMessage().Metadata().MsgID().Bytes())
	return retMsgID, gcli.ConsumeMessage(ctx, creation)
}

func (g *gregorHandler) UpdateItem(ctx context.Context, msgID gregor1.MsgID, cat string, body []byte, dtime gregor1.TimeOrOffset) (gregor1.MsgID, error) {
	var err error
	defer g.G().CTrace(ctx, fmt.Sprintf("gregorHandler.UpdateItem(%s,%s)", msgID.String(), cat),
		&err,
	)()
	defer g.pushState(keybase1.PushReason_NEW_DATA)

	msg, err := grutils.TemplateMessage(g.currentUID())
	if err != nil {
		return nil, err
	}
	msg.Ibm_.StateUpdate_.Creation_ = &gregor1.Item{
		Category_: gregor1.Category(cat),
		Body_:     gregor1.Body(body),
		Dtime_:    dtime,
	}
	msg.Ibm_.StateUpdate_.Dismissal_ = &gregor1.Dismissal{
		MsgIDs_: []gregor1.MsgID{msgID},
	}

	gcli, err := g.getGregorCli()
	if err != nil {
		return nil, err
	}
	return msg.Ibm_.StateUpdate_.Md_.MsgID_, gcli.ConsumeMessage(ctx, msg)
}

func (g *gregorHandler) UpdateCategory(ctx context.Context, cat string, body []byte,
	dtime gregor1.TimeOrOffset) (res gregor1.MsgID, err error) {
	defer g.G().CTrace(ctx, fmt.Sprintf("gregorHandler.UpdateCategory(%s)", cat),
		&err,
	)()
	defer g.pushState(keybase1.PushReason_NEW_DATA)

	msg, err := grutils.TemplateMessage(g.currentUID())
	if err != nil {
		return nil, err
	}
	msgID := msg.Ibm_.StateUpdate_.Md_.MsgID_
	msg.Ibm_.StateUpdate_.Creation_ = &gregor1.Item{
		Category_: gregor1.Category(cat),
		Body_:     gregor1.Body(body),
		Dtime_:    dtime,
	}
	msg.Ibm_.StateUpdate_.Dismissal_ = &gregor1.Dismissal{
		Ranges_: []gregor1.MsgRange{
			{
				Category_:   gregor1.Category(cat),
				SkipMsgIDs_: []gregor1.MsgID{msgID},
			}},
	}

	gcli, err := g.getGregorCli()
	if err != nil {
		return nil, err
	}
	return msgID, gcli.ConsumeMessage(ctx, msg)
}

func (g *gregorHandler) InjectOutOfBandMessage(ctx context.Context, system string, body []byte) error {
	var err error
	defer g.G().CTrace(ctx, fmt.Sprintf("gregorHandler.InjectOutOfBandMessage(%s)", system),
		&err,
	)()

	uid := g.G().Env.GetUID()
	if uid.IsNil() {
		return libkb.LoggedInError{}
	}
	gregorUID := gregor1.UID(uid.ToBytes())

	msg := gregor1.Message{
		Oobm_: &gregor1.OutOfBandMessage{
			Uid_:    gregorUID,
			System_: gregor1.System(system),
			Body_:   gregor1.Body(body),
		},
	}

	gcli, err := g.getGregorCli()
	if err != nil {
		return err
	}
	return gcli.ConsumeMessage(ctx, msg)
}

func (g *gregorHandler) simulateCrashForTesting() {
	g.transportForTesting.Reset()
	_, _ = gregor1.IncomingClient{Cli: g.cli}.Ping(context.Background())
}

type gregorRPCHandler struct {
	libkb.Contextified
	xp rpc.Transporter
	gh *gregorHandler
}

func newGregorRPCHandler(xp rpc.Transporter, g *libkb.GlobalContext, gh *gregorHandler) *gregorRPCHandler {
	return &gregorRPCHandler{
		Contextified: libkb.NewContextified(g),
		xp:           xp,
		gh:           gh,
	}
}

func (g *gregorHandler) getState(ctx context.Context) (res gregor1.State, err error) {
	var s gregor.State

	gcli, err := g.getGregorCli()
	if err != nil {
		return res, err
	}

	s, err = gcli.StateMachineState(ctx, nil, true)
	if err != nil {
		return res, err
	}

	ps, err := s.Export()
	if err != nil {
		return res, err
	}

	var ok bool
	if res, ok = ps.(gregor1.State); !ok {
		return res, errors.New("failed to convert state to exportable format")
	}

	return res, nil
}

func (g *gregorHandler) State(ctx context.Context) (res gregor.State, err error) {
	defer g.G().CTrace(ctx, "gregorHandler#State", &err)()
	gcli, err := g.getGregorCli()
	if err != nil {
		return res, err
	}
	return gcli.StateMachineState(ctx, nil, true)
}

func (g *gregorRPCHandler) GetState(ctx context.Context) (res gregor1.State, err error) {
	defer g.G().CTrace(ctx, "gregorRPCHandler#GetState", &err)()
	if res, err = g.gh.getState(ctx); err != nil {
		return res, err
	}
	g.G().Log.CDebugf(ctx, "GetState: returning %d items", len(res.Items_))
	return res, nil
}

func (g *gregorRPCHandler) InjectItem(ctx context.Context, arg keybase1.InjectItemArg) (res gregor1.MsgID, err error) {
	defer g.G().CTrace(ctx, "gregorRPCHandler#InjectItem", &err)()
	return g.gh.InjectItem(ctx, arg.Cat, []byte(arg.Body), arg.Dtime)
}

func (g *gregorRPCHandler) UpdateItem(ctx context.Context, arg keybase1.UpdateItemArg) (res gregor1.MsgID, err error) {
	defer g.G().CTrace(ctx, "gregorRPCHandler#UpdateItem", &err)()
	return g.gh.UpdateItem(ctx, arg.MsgID, arg.Cat, []byte(arg.Body), arg.Dtime)
}

func (g *gregorRPCHandler) UpdateCategory(ctx context.Context, arg keybase1.UpdateCategoryArg) (res gregor1.MsgID, err error) {
	defer g.G().CTrace(ctx, "gregorRPCHandler#UpdateCategory", &err)()
	return g.gh.UpdateCategory(ctx, arg.Category, []byte(arg.Body), arg.Dtime)
}

func (g *gregorRPCHandler) DismissCategory(ctx context.Context, category gregor1.Category) (err error) {
	defer g.G().CTrace(ctx, "gregorRPCHandler#DismissCategory", &err)()
	return g.gh.DismissCategory(ctx, category)
}

func (g *gregorRPCHandler) DismissItem(ctx context.Context, id gregor1.MsgID) (err error) {
	defer g.G().CTrace(ctx, "gregorRPCHandler#DismissItem", &err)()
	return g.gh.DismissItem(ctx, nil, id)
}

func WrapGenericClientWithTimeout(client rpc.GenericClient, timeout time.Duration, timeoutErr error) rpc.GenericClient {
	return &timeoutClient{client, timeout, timeoutErr}
}

type timeoutClient struct {
	inner      rpc.GenericClient
	timeout    time.Duration
	timeoutErr error
}

var _ rpc.GenericClient = (*timeoutClient)(nil)

func (t *timeoutClient) Call(ctx context.Context, method string, arg interface{},
	res interface{}, timeout time.Duration) error {
	if timeout == 0 {
		timeout = t.timeout
	}
	err := t.inner.Call(ctx, method, arg, res, timeout)
	if err == context.DeadlineExceeded {
		return t.timeoutErr
	}
	return err
}

func (t *timeoutClient) CallCompressed(ctx context.Context, method string, arg interface{},
	res interface{}, ctype rpc.CompressionType, timeout time.Duration) error {
	if timeout == 0 {
		timeout = t.timeout
	}
	err := t.inner.CallCompressed(ctx, method, arg, res, ctype, timeout)
	if err == context.DeadlineExceeded {
		return t.timeoutErr
	}
	return err
}

func (t *timeoutClient) Notify(ctx context.Context, method string, arg interface{}, timeout time.Duration) error {
	if timeout == 0 {
		timeout = t.timeout
	}
	err := t.inner.Notify(ctx, method, arg, timeout)
	if err == context.DeadlineExceeded {
		return t.timeoutErr
	}
	return err
}