bugfix: E2EE device keys could sometimes not be sent to remote servers (#2466)

* Fix flakey sytest 'Local device key changes get to remote servers'

* Debug logs

* Remove internal/test and use /test only

Remove a lot of ancient code too.

* Use FederationRoomserverAPI in more places

* Use more interfaces in federationapi; begin adding regression test

* Linting

* Add regression test

* Unbreak tests

* ALL THE LOGS

* Fix a race condition which could cause events to not be sent to servers

If a new room event which rewrites state arrives, we remove all joined hosts
then re-calculate them. This wasn't done in a transaction so for a brief period
we would have no joined hosts. During this interim, key change events which arrive
would not be sent to destination servers. This would sporadically fail on sytest.

* Unbreak new tests

* Linting
This commit is contained in:
kegsay 2022-05-17 13:23:35 +01:00 committed by GitHub
parent cd82460513
commit 6de29c1cd2
No known key found for this signature in database
GPG key ID: 4AEE18F83AFDEB23
48 changed files with 566 additions and 618 deletions

View file

@ -3,18 +3,250 @@ package federationapi_test
import (
"context"
"crypto/ed25519"
"encoding/json"
"fmt"
"strings"
"testing"
"time"
"github.com/matrix-org/dendrite/federationapi"
"github.com/matrix-org/dendrite/federationapi/api"
"github.com/matrix-org/dendrite/federationapi/internal"
"github.com/matrix-org/dendrite/internal/test"
keyapi "github.com/matrix-org/dendrite/keyserver/api"
rsapi "github.com/matrix-org/dendrite/roomserver/api"
"github.com/matrix-org/dendrite/setup/base"
"github.com/matrix-org/dendrite/setup/config"
"github.com/matrix-org/dendrite/setup/jetstream"
"github.com/matrix-org/dendrite/test"
"github.com/matrix-org/dendrite/test/testrig"
"github.com/matrix-org/gomatrix"
"github.com/matrix-org/gomatrixserverlib"
"github.com/nats-io/nats.go"
)
type fedRoomserverAPI struct {
rsapi.FederationRoomserverAPI
inputRoomEvents func(ctx context.Context, req *rsapi.InputRoomEventsRequest, res *rsapi.InputRoomEventsResponse)
queryRoomsForUser func(ctx context.Context, req *rsapi.QueryRoomsForUserRequest, res *rsapi.QueryRoomsForUserResponse) error
}
// PerformJoin will call this function
func (f *fedRoomserverAPI) InputRoomEvents(ctx context.Context, req *rsapi.InputRoomEventsRequest, res *rsapi.InputRoomEventsResponse) {
if f.inputRoomEvents == nil {
return
}
f.inputRoomEvents(ctx, req, res)
}
// keychange consumer calls this
func (f *fedRoomserverAPI) QueryRoomsForUser(ctx context.Context, req *rsapi.QueryRoomsForUserRequest, res *rsapi.QueryRoomsForUserResponse) error {
if f.queryRoomsForUser == nil {
return nil
}
return f.queryRoomsForUser(ctx, req, res)
}
// TODO: This struct isn't generic, only works for TestFederationAPIJoinThenKeyUpdate
type fedClient struct {
api.FederationClient
allowJoins []*test.Room
keys map[gomatrixserverlib.ServerName]struct {
key ed25519.PrivateKey
keyID gomatrixserverlib.KeyID
}
t *testing.T
sentTxn bool
}
func (f *fedClient) GetServerKeys(ctx context.Context, matrixServer gomatrixserverlib.ServerName) (gomatrixserverlib.ServerKeys, error) {
fmt.Println("GetServerKeys:", matrixServer)
var keys gomatrixserverlib.ServerKeys
var keyID gomatrixserverlib.KeyID
var pkey ed25519.PrivateKey
for srv, data := range f.keys {
if srv == matrixServer {
pkey = data.key
keyID = data.keyID
break
}
}
if pkey == nil {
return keys, nil
}
keys.ServerName = matrixServer
keys.ValidUntilTS = gomatrixserverlib.AsTimestamp(time.Now().Add(10 * time.Hour))
publicKey := pkey.Public().(ed25519.PublicKey)
keys.VerifyKeys = map[gomatrixserverlib.KeyID]gomatrixserverlib.VerifyKey{
keyID: {
Key: gomatrixserverlib.Base64Bytes(publicKey),
},
}
toSign, err := json.Marshal(keys.ServerKeyFields)
if err != nil {
return keys, err
}
keys.Raw, err = gomatrixserverlib.SignJSON(
string(matrixServer), keyID, pkey, toSign,
)
if err != nil {
return keys, err
}
return keys, nil
}
func (f *fedClient) MakeJoin(ctx context.Context, s gomatrixserverlib.ServerName, roomID, userID string, roomVersions []gomatrixserverlib.RoomVersion) (res gomatrixserverlib.RespMakeJoin, err error) {
for _, r := range f.allowJoins {
if r.ID == roomID {
res.RoomVersion = r.Version
res.JoinEvent = gomatrixserverlib.EventBuilder{
Sender: userID,
RoomID: roomID,
Type: "m.room.member",
StateKey: &userID,
Content: gomatrixserverlib.RawJSON([]byte(`{"membership":"join"}`)),
PrevEvents: r.ForwardExtremities(),
}
var needed gomatrixserverlib.StateNeeded
needed, err = gomatrixserverlib.StateNeededForEventBuilder(&res.JoinEvent)
if err != nil {
f.t.Errorf("StateNeededForEventBuilder: %v", err)
return
}
res.JoinEvent.AuthEvents = r.MustGetAuthEventRefsForEvent(f.t, needed)
return
}
}
return
}
func (f *fedClient) SendJoin(ctx context.Context, s gomatrixserverlib.ServerName, event *gomatrixserverlib.Event) (res gomatrixserverlib.RespSendJoin, err error) {
for _, r := range f.allowJoins {
if r.ID == event.RoomID() {
r.InsertEvent(f.t, event.Headered(r.Version))
f.t.Logf("Join event: %v", event.EventID())
res.StateEvents = gomatrixserverlib.NewEventJSONsFromHeaderedEvents(r.CurrentState())
res.AuthEvents = gomatrixserverlib.NewEventJSONsFromHeaderedEvents(r.Events())
}
}
return
}
func (f *fedClient) SendTransaction(ctx context.Context, t gomatrixserverlib.Transaction) (res gomatrixserverlib.RespSend, err error) {
for _, edu := range t.EDUs {
if edu.Type == gomatrixserverlib.MDeviceListUpdate {
f.sentTxn = true
}
}
f.t.Logf("got /send")
return
}
// Regression test to make sure that /send_join is updating the destination hosts synchronously and
// isn't relying on the roomserver.
func TestFederationAPIJoinThenKeyUpdate(t *testing.T) {
test.WithAllDatabases(t, func(t *testing.T, dbType test.DBType) {
testFederationAPIJoinThenKeyUpdate(t, dbType)
})
}
func testFederationAPIJoinThenKeyUpdate(t *testing.T, dbType test.DBType) {
base, close := testrig.CreateBaseDendrite(t, dbType)
base.Cfg.FederationAPI.PreferDirectFetch = true
defer close()
jsctx, _ := base.NATS.Prepare(base.ProcessContext, &base.Cfg.Global.JetStream)
defer jetstream.DeleteAllStreams(jsctx, &base.Cfg.Global.JetStream)
serverA := gomatrixserverlib.ServerName("server.a")
serverAKeyID := gomatrixserverlib.KeyID("ed25519:servera")
serverAPrivKey := test.PrivateKeyA
creator := test.NewUser(t, test.WithSigningServer(serverA, serverAKeyID, serverAPrivKey))
myServer := base.Cfg.Global.ServerName
myServerKeyID := base.Cfg.Global.KeyID
myServerPrivKey := base.Cfg.Global.PrivateKey
joiningUser := test.NewUser(t, test.WithSigningServer(myServer, myServerKeyID, myServerPrivKey))
fmt.Printf("creator: %v joining user: %v\n", creator.ID, joiningUser.ID)
room := test.NewRoom(t, creator)
rsapi := &fedRoomserverAPI{
inputRoomEvents: func(ctx context.Context, req *rsapi.InputRoomEventsRequest, res *rsapi.InputRoomEventsResponse) {
if req.Asynchronous {
t.Errorf("InputRoomEvents from PerformJoin MUST be synchronous")
}
},
queryRoomsForUser: func(ctx context.Context, req *rsapi.QueryRoomsForUserRequest, res *rsapi.QueryRoomsForUserResponse) error {
if req.UserID == joiningUser.ID && req.WantMembership == "join" {
res.RoomIDs = []string{room.ID}
return nil
}
return fmt.Errorf("unexpected queryRoomsForUser: %+v", *req)
},
}
fc := &fedClient{
allowJoins: []*test.Room{room},
t: t,
keys: map[gomatrixserverlib.ServerName]struct {
key ed25519.PrivateKey
keyID gomatrixserverlib.KeyID
}{
serverA: {
key: serverAPrivKey,
keyID: serverAKeyID,
},
myServer: {
key: myServerPrivKey,
keyID: myServerKeyID,
},
},
}
fsapi := federationapi.NewInternalAPI(base, fc, rsapi, base.Caches, nil, false)
var resp api.PerformJoinResponse
fsapi.PerformJoin(context.Background(), &api.PerformJoinRequest{
RoomID: room.ID,
UserID: joiningUser.ID,
ServerNames: []gomatrixserverlib.ServerName{serverA},
}, &resp)
if resp.JoinedVia != serverA {
t.Errorf("PerformJoin: joined via %v want %v", resp.JoinedVia, serverA)
}
if resp.LastError != nil {
t.Fatalf("PerformJoin: returned error: %+v", *resp.LastError)
}
// Inject a keyserver key change event and ensure we try to send it out. If we don't, then the
// federationapi is incorrectly waiting for an output room event to arrive to update the joined
// hosts table.
key := keyapi.DeviceMessage{
Type: keyapi.TypeDeviceKeyUpdate,
DeviceKeys: &keyapi.DeviceKeys{
UserID: joiningUser.ID,
DeviceID: "MY_DEVICE",
DisplayName: "BLARGLE",
KeyJSON: []byte(`{}`),
},
}
b, err := json.Marshal(key)
if err != nil {
t.Fatalf("Failed to marshal device message: %s", err)
}
msg := &nats.Msg{
Subject: base.Cfg.Global.JetStream.Prefixed(jetstream.OutputKeyChangeEvent),
Header: nats.Header{},
Data: b,
}
msg.Header.Set(jetstream.UserID, key.UserID)
testrig.MustPublishMsgs(t, jsctx, msg)
time.Sleep(500 * time.Millisecond)
if !fc.sentTxn {
t.Fatalf("did not send device list update")
}
}
// Tests that event IDs with '/' in them (escaped as %2F) are correctly passed to the right handler and don't 404.
// Relevant for v3 rooms and a cause of flakey sytests as the IDs are randomly generated.
func TestRoomsV3URLEscapeDoNot404(t *testing.T) {
@ -86,7 +318,7 @@ func TestRoomsV3URLEscapeDoNot404(t *testing.T) {
}
gerr, ok := err.(gomatrix.HTTPError)
if !ok {
t.Errorf("failed to cast response error as gomatrix.HTTPError")
t.Errorf("failed to cast response error as gomatrix.HTTPError: %s", err)
continue
}
t.Logf("Error: %+v", gerr)