Merge branches 'develop' and 't3chguy/fix_contains-url_filtering' of github.com:matrix-org/synapse into t3chguy/fix_contains-url_filtering

Initial impl of capping MAU

AUTHORS.rst

@@ -62,4 +62,7 @@ Christoph Witzany <christoph at web.crofting.com>
  * Add LDAP support for authentication
 
 Pierre Jaury <pierre at jaury.eu>
-* Docker packaging
+* Docker packaging
+
+Serban Constantin <serban.constantin at gmail dot com>
+ * Small bug fix

Dockerfile

@@ -1,16 +1,32 @@
 FROM docker.io/python:2-alpine3.7
 
-RUN apk add --no-cache --virtual .nacl_deps su-exec build-base libffi-dev zlib-dev libressl-dev libjpeg-turbo-dev linux-headers postgresql-dev libxslt-dev
+RUN apk add --no-cache --virtual .nacl_deps \
+        build-base \
+        libffi-dev \
+        libjpeg-turbo-dev \
+        libressl-dev \
+        libxslt-dev \
+        linux-headers \
+        postgresql-dev \
+        su-exec \
+        zlib-dev
 
 COPY . /synapse
 
 # A wheel cache may be provided in ./cache for faster build
 RUN cd /synapse \
- && pip install --upgrade pip setuptools psycopg2 lxml \
+ && pip install --upgrade \
+        lxml \
+        pip \
+        psycopg2 \
+        setuptools \
  && mkdir -p /synapse/cache \
  && pip install -f /synapse/cache --upgrade --process-dependency-links . \
  && mv /synapse/contrib/docker/start.py /synapse/contrib/docker/conf / \
- && rm -rf setup.py setup.cfg synapse
+ && rm -rf \
+        setup.cfg \
+        setup.py \
+        synapse
 
 VOLUME ["/data"]
 

MANIFEST.in

@@ -2,6 +2,7 @@ include synctl
 include LICENSE
 include VERSION
 include *.rst
+include *.md
 include demo/README
 include demo/demo.tls.dh
 include demo/*.py

README.rst

@@ -71,7 +71,7 @@ We'd like to invite you to join #matrix:matrix.org (via
 https://matrix.org/docs/projects/try-matrix-now.html), run a homeserver, take a look
 at the `Matrix spec <https://matrix.org/docs/spec>`_, and experiment with the
 `APIs <https://matrix.org/docs/api>`_ and `Client SDKs
-<http://matrix.org/docs/projects/try-matrix-now.html#client-sdks>`_.
+<https://matrix.org/docs/projects/try-matrix-now.html#client-sdks>`_.
 
 Thanks for using Matrix!
 
@@ -283,7 +283,7 @@ Connecting to Synapse from a client
 
 The easiest way to try out your new Synapse installation is by connecting to it
 from a web client. The easiest option is probably the one at
-http://riot.im/app. You will need to specify a "Custom server" when you log on
+https://riot.im/app. You will need to specify a "Custom server" when you log on
 or register: set this to ``https://domain.tld`` if you setup a reverse proxy
 following the recommended setup, or ``https://localhost:8448`` - remember to specify the
 port (``:8448``) if not ``:443`` unless you changed the configuration. (Leave the identity
@@ -329,7 +329,7 @@ Security Note
 =============
 
 Matrix serves raw user generated data in some APIs - specifically the `content
-repository endpoints <http://matrix.org/docs/spec/client_server/latest.html#get-matrix-media-r0-download-servername-mediaid>`_.
+repository endpoints <https://matrix.org/docs/spec/client_server/latest.html#get-matrix-media-r0-download-servername-mediaid>`_.
 
 Whilst we have tried to mitigate against possible XSS attacks (e.g.
 https://github.com/matrix-org/synapse/pull/1021) we recommend running
@@ -348,7 +348,7 @@ Platform-Specific Instructions
 Debian
 ------
 
-Matrix provides official Debian packages via apt from http://matrix.org/packages/debian/.
+Matrix provides official Debian packages via apt from https://matrix.org/packages/debian/.
 Note that these packages do not include a client - choose one from
 https://matrix.org/docs/projects/try-matrix-now.html (or build your own with one of our SDKs :)
 
@@ -362,6 +362,19 @@ Synapse is in the Fedora repositories as ``matrix-synapse``::
 Oleg Girko provides Fedora RPMs at
 https://obs.infoserver.lv/project/monitor/matrix-synapse
 
+OpenSUSE
+--------
+
+Synapse is in the OpenSUSE repositories as ``matrix-synapse``::
+
+    sudo zypper install matrix-synapse
+
+SUSE Linux Enterprise Server
+----------------------------
+
+Unofficial package are built for SLES 15 in the openSUSE:Backports:SLE-15 repository at
+https://download.opensuse.org/repositories/openSUSE:/Backports:/SLE-15/standard/
+
 ArchLinux
 ---------
 
@@ -524,7 +537,7 @@ Troubleshooting Running
 -----------------------
 
 If synapse fails with ``missing "sodium.h"`` crypto errors, you may need
-to manually upgrade PyNaCL, as synapse uses NaCl (http://nacl.cr.yp.to/) for
+to manually upgrade PyNaCL, as synapse uses NaCl (https://nacl.cr.yp.to/) for
 encryption and digital signatures.
 Unfortunately PyNACL currently has a few issues
 (https://github.com/pyca/pynacl/issues/53) and
@@ -672,8 +685,8 @@ useful just for development purposes. See `<demo/README>`_.
 Using PostgreSQL
 ================
 
-As of Synapse 0.9, `PostgreSQL <http://www.postgresql.org>`_ is supported as an
-alternative to the `SQLite <http://sqlite.org/>`_ database that Synapse has
+As of Synapse 0.9, `PostgreSQL <https://www.postgresql.org>`_ is supported as an
+alternative to the `SQLite <https://sqlite.org/>`_ database that Synapse has
 traditionally used for convenience and simplicity.
 
 The advantages of Postgres include:
@@ -697,7 +710,7 @@ Using a reverse proxy with Synapse
 It is recommended to put a reverse proxy such as
 `nginx <https://nginx.org/en/docs/http/ngx_http_proxy_module.html>`_,
 `Apache <https://httpd.apache.org/docs/current/mod/mod_proxy_http.html>`_ or
-`HAProxy <http://www.haproxy.org/>`_ in front of Synapse. One advantage of
+`HAProxy <https://www.haproxy.org/>`_ in front of Synapse. One advantage of
 doing so is that it means that you can expose the default https port (443) to
 Matrix clients without needing to run Synapse with root privileges.
 

changelog.d/2952.bugfix

@@ -0,0 +1 @@
+Make /directory/list API return 404 for room not found instead of 400

changelog.d/2970.feature

@@ -0,0 +1 @@
+add support for the lazy_loaded_members filter as per MSC1227

changelog.d/3331.feature

@@ -0,0 +1 @@
+add support for the include_redundant_members filter param as per MSC1227

changelog.d/3350.misc

@@ -0,0 +1 @@
+Remove redundant checks on who_forgot_in_room

changelog.d/3367.misc

@@ -0,0 +1 @@
+Remove unnecessary event re-signing hacks

changelog.d/3384.misc

@@ -0,0 +1 @@
+Rewrite cache list decorator

changelog.d/3391.bugfix

@@ -0,0 +1 @@
+Default inviter_display_name to mxid for email invites

changelog.d/3514.bugfix

@@ -0,0 +1 @@
+Don't generate TURN credentials if no TURN config options are set

changelog.d/3520.bugfix

@@ -0,0 +1 @@
+Correctly announce deleted devices over federation

changelog.d/3543.misc

@@ -0,0 +1 @@
+Improve Dockerfile and docker-compose instructions

changelog.d/3548.bugfix

@@ -0,0 +1 @@
+Catch failures saving metrics captured by Measure, and instead log the faulty metrics information for further analysis.

changelog.d/3552.misc

@@ -0,0 +1 @@
+Release notes are now in the Markdown format.

changelog.d/3553.feature

@@ -0,0 +1 @@
+Add metrics to track resource usage by background processes

changelog.d/3554.feature

@@ -0,0 +1 @@
+Add `code` label to `synapse_http_server_response_time_seconds` prometheus metric

changelog.d/3555.feature

@@ -0,0 +1 @@
+Add support for client_reader to handle more APIs

changelog.d/3556.feature

@@ -0,0 +1 @@
+Add metrics to track resource usage by background processes

changelog.d/3559.misc

@@ -0,0 +1 @@
+add config for pep8

changelog.d/3567.feature

@@ -0,0 +1 @@
+make the /context API filter & lazy-load aware as per MSC1227

changelog.d/3569.bugfix

@@ -0,0 +1 @@
+Unicode passwords are now normalised before hashing, preventing the instance where two different devices or browsers might send a different UTF-8 sequence for the password.

changelog.d/3570.bugfix

@@ -0,0 +1 @@
+Fix potential stack overflow and deadlock under heavy load

changelog.d/3571.misc

@@ -0,0 +1 @@
+Merge Linearizer and Limiter

changelog.d/3572.misc

@@ -0,0 +1 @@
+Merge Linearizer and Limiter

changelog.d/3579.misc

@@ -0,0 +1 @@
+Lazily load state on master process when using workers to reduce DB consumption

changelog.d/3581.misc

@@ -0,0 +1 @@
+Lazily load state on master process when using workers to reduce DB consumption

changelog.d/3582.misc

@@ -0,0 +1 @@
+Lazily load state on master process when using workers to reduce DB consumption

changelog.d/3584.misc

@@ -0,0 +1 @@
+Lazily load state on master process when using workers to reduce DB consumption

changelog.d/3586.misc

@@ -0,0 +1 @@
+Fixes and optimisations for resolve_state_groups

changelog.d/3587.misc

@@ -0,0 +1 @@
+Improve logging for exceptions when handling PDUs

changelog.d/3590.misc

@@ -0,0 +1 @@
+Add some measure blocks to persist_events

changelog.d/3591.misc

@@ -0,0 +1 @@
+Fix some random logcontext leaks.

changelog.d/3592.misc

@@ -0,0 +1 @@
+Speed up calculating state deltas in persist_event loop

changelog.d/3595.misc

@@ -0,0 +1 @@
+Attempt to reduce amount of state pulled out of DB during persist_events

changelog.d/3597.feature

@@ -0,0 +1 @@
+Add support for client_reader to handle more APIs

changelog.d/3601.bugfix

@@ -0,0 +1 @@
+Fix failure to persist events over federation under load

changelog.d/3604.feature

@@ -0,0 +1 @@
+Add metrics to track resource usage by background processes

changelog.d/3605.bugfix

@@ -0,0 +1 @@
+Fix updating of cached remote profiles

changelog.d/3606.misc

@@ -0,0 +1 @@
+Fix some random logcontext leaks.

changelog.d/3607.bugfix

@@ -0,0 +1 @@
+Fix 'tuple index out of range' error

changelog.d/3609.misc

@@ -0,0 +1 @@
+Fix a documentation typo in on_make_leave_request

changelog.d/3610.feature

@@ -0,0 +1 @@
+Add metrics to track resource usage by background processes

changelog.d/3612.misc

@@ -0,0 +1 @@
+Make EventStore inherit from EventFederationStore

changelog.d/3613.misc

@@ -0,0 +1 @@
+Remove some redundant joins on event_edges.room_id

changelog.d/3614.misc

@@ -0,0 +1 @@
+Stop populating events.content

changelog.d/3616.misc

@@ -0,0 +1 @@
+Update the /send_leave path registration to use event_id rather than a transaction ID.

changelog.d/3626.bugfix

@@ -0,0 +1 @@
+Only import secrets when available (fix for py < 3.6)

changelog.d/3628.misc

@@ -0,0 +1 @@
+Remove unused field "pdu_failures" from transactions.

changelog.d/3630.feature

@@ -0,0 +1 @@
+Add ability to limit number of monthly active users on the server

changelog.d/3634.misc

@@ -0,0 +1 @@
+rename replication_layer to federation_client

contrib/docker/README.md

@@ -9,13 +9,7 @@ use that server.
 
 ## Build
 
-Build the docker image with the `docker build` command from the root of the synapse repository.
-
-```
-docker build -t docker.io/matrixdotorg/synapse .
-```
-
-The `-t` option sets the image tag. Official images are tagged `matrixdotorg/synapse:<version>` where `<version>` is the same as the release tag in the synapse git repository.
+Build the docker image with the `docker-compose build` command.
 
 You may have a local Python wheel cache available, in which case copy the relevant packages in the ``cache/`` directory at the root of the project.
 

contrib/docker/docker-compose.yml

@@ -6,6 +6,7 @@ version: '3'
 services:
 
   synapse:
+    build: ../..
     image: docker.io/matrixdotorg/synapse:latest
     # Since snyapse does not retry to connect to the database, restart upon
     # failure

contrib/grafana/README.md

@@ -0,0 +1,6 @@
+# Using the Synapse Grafana dashboard
+
+0. Set up Prometheus and Grafana. Out of scope for this readme. Useful documentation about using Grafana with Prometheus: http://docs.grafana.org/features/datasources/prometheus/
+1. Have your Prometheus scrape your Synapse. https://github.com/matrix-org/synapse/blob/master/docs/metrics-howto.rst
+2. Import dashboard into Grafana. Download `synapse.json`. Import it to Grafana and select the correct Prometheus datasource. http://docs.grafana.org/reference/export_import/
+3. Set up additional recording rules

docs/admin_api/register_api.rst

@@ -0,0 +1,63 @@
+Shared-Secret Registration
+==========================
+
+This API allows for the creation of users in an administrative and
+non-interactive way. This is generally used for bootstrapping a Synapse
+instance with administrator accounts.
+
+To authenticate yourself to the server, you will need both the shared secret
+(``registration_shared_secret`` in the homeserver configuration), and a
+one-time nonce. If the registration shared secret is not configured, this API
+is not enabled.
+
+To fetch the nonce, you need to request one from the API::
+
+  > GET /_matrix/client/r0/admin/register
+
+  < {"nonce": "thisisanonce"}
+
+Once you have the nonce, you can make a ``POST`` to the same URL with a JSON
+body containing the nonce, username, password, whether they are an admin
+(optional, False by default), and a HMAC digest of the content.
+
+As an example::
+
+  > POST /_matrix/client/r0/admin/register
+  > {
+     "nonce": "thisisanonce",
+     "username": "pepper_roni",
+     "password": "pizza",
+     "admin": true,
+     "mac": "mac_digest_here"
+    }
+
+  < {
+     "access_token": "token_here",
+     "user_id": "@pepper_roni@test",
+     "home_server": "test",
+     "device_id": "device_id_here"
+    }
+
+The MAC is the hex digest output of the HMAC-SHA1 algorithm, with the key being
+the shared secret and the content being the nonce, user, password, and either
+the string "admin" or "notadmin", each separated by NULs. For an example of
+generation in Python::
+
+  import hmac, hashlib
+
+  def generate_mac(nonce, user, password, admin=False):
+
+      mac = hmac.new(
+        key=shared_secret,
+        digestmod=hashlib.sha1,
+      )
+
+      mac.update(nonce.encode('utf8'))
+      mac.update(b"\x00")
+      mac.update(user.encode('utf8'))
+      mac.update(b"\x00")
+      mac.update(password.encode('utf8'))
+      mac.update(b"\x00")
+      mac.update(b"admin" if admin else b"notadmin")
+
+      return mac.hexdigest()

docs/workers.rst

@@ -206,6 +206,10 @@ Handles client API endpoints. It can handle REST endpoints matching the
 following regular expressions::
 
     ^/_matrix/client/(api/v1|r0|unstable)/publicRooms$
+    ^/_matrix/client/(api/v1|r0|unstable)/rooms/.*/joined_members$
+    ^/_matrix/client/(api/v1|r0|unstable)/rooms/.*/context/.*$
+    ^/_matrix/client/(api/v1|r0|unstable)/rooms/.*/members$
+    ^/_matrix/client/(api/v1|r0|unstable)/rooms/.*/state$
 
 ``synapse.app.user_dir``
 ~~~~~~~~~~~~~~~~~~~~~~~~

pyproject.toml

@@ -1,5 +1,30 @@
 [tool.towncrier]
     package = "synapse"
-    filename = "CHANGES.rst"
+    filename = "CHANGES.md"
     directory = "changelog.d"
-    issue_format = "`#{issue} <https://github.com/matrix-org/synapse/issues/{issue}>`_"
+    issue_format = "[\\#{issue}](https://github.com/matrix-org/synapse/issues/{issue}>)"
+
+    [[tool.towncrier.type]]
+        directory = "feature"
+        name = "Features"
+        showcontent = true
+
+    [[tool.towncrier.type]]
+        directory = "bugfix"
+        name = "Bugfixes"
+        showcontent = true
+
+    [[tool.towncrier.type]]
+        directory = "doc"
+        name = "Improved Documentation"
+        showcontent = true
+
+    [[tool.towncrier.type]]
+        directory = "removal"
+        name = "Deprecations and Removals"
+        showcontent = true
+
+    [[tool.towncrier.type]]
+        directory = "misc"
+        name = "Internal Changes"
+        showcontent = true

scripts/register_new_matrix_user

@@ -26,11 +26,37 @@ import yaml
 
 
 def request_registration(user, password, server_location, shared_secret, admin=False):
+    req = urllib2.Request(
+        "%s/_matrix/client/r0/admin/register" % (server_location,),
+        headers={'Content-Type': 'application/json'}
+    )
+
+    try:
+        if sys.version_info[:3] >= (2, 7, 9):
+            # As of version 2.7.9, urllib2 now checks SSL certs
+            import ssl
+            f = urllib2.urlopen(req, context=ssl.SSLContext(ssl.PROTOCOL_SSLv23))
+        else:
+            f = urllib2.urlopen(req)
+        body = f.read()
+        f.close()
+        nonce = json.loads(body)["nonce"]
+    except urllib2.HTTPError as e:
+        print "ERROR! Received %d %s" % (e.code, e.reason,)
+        if 400 <= e.code < 500:
+            if e.info().type == "application/json":
+                resp = json.load(e)
+                if "error" in resp:
+                    print resp["error"]
+        sys.exit(1)
+
     mac = hmac.new(
         key=shared_secret,
         digestmod=hashlib.sha1,
     )
 
+    mac.update(nonce)
+    mac.update("\x00")
     mac.update(user)
     mac.update("\x00")
     mac.update(password)
@@ -40,10 +66,10 @@ def request_registration(user, password, server_location, shared_secret, admin=F
     mac = mac.hexdigest()
 
     data = {
-        "user": user,
+        "nonce": nonce,
+        "username": user,
         "password": password,
         "mac": mac,
-        "type": "org.matrix.login.shared_secret",
         "admin": admin,
     }
 
@@ -52,7 +78,7 @@ def request_registration(user, password, server_location, shared_secret, admin=F
     print "Sending registration request..."
 
     req = urllib2.Request(
-        "%s/_matrix/client/api/v1/register" % (server_location,),
+        "%s/_matrix/client/r0/admin/register" % (server_location,),
         data=json.dumps(data),
         headers={'Content-Type': 'application/json'}
     )

setup.cfg

@@ -14,12 +14,17 @@ ignore =
     pylint.cfg
     tox.ini
 
-[flake8]
+[pep8]
 max-line-length = 90
-#  W503 requires that binary operators be at the end, not start, of lines. Erik doesn't like it.
-#  E203 is contrary to PEP8.
+#  W503 requires that binary operators be at the end, not start, of lines. Erik
+#  doesn't like it.  E203 is contrary to PEP8.
 ignore = W503,E203
 
+[flake8]
+# note that flake8 inherits the "ignore" settings from "pep8" (because it uses
+# pep8 to do those checks), but not the "max-line-length" setting
+max-line-length = 90
+
 [isort]
 line_length = 89
 not_skip = __init__.py
@@ -31,3 +36,4 @@ known_compat = mock,six
 known_twisted=twisted,OpenSSL
 multi_line_output=3
 include_trailing_comma=true
+combine_as_imports=true

synapse/api/auth.py

@@ -65,8 +65,9 @@ class Auth(object):
 
     @defer.inlineCallbacks
     def check_from_context(self, event, context, do_sig_check=True):
+        prev_state_ids = yield context.get_prev_state_ids(self.store)
         auth_events_ids = yield self.compute_auth_events(
-            event, context.prev_state_ids, for_verification=True,
+            event, prev_state_ids, for_verification=True,
         )
         auth_events = yield self.store.get_events(auth_events_ids)
         auth_events = {
@@ -251,10 +252,10 @@ class Auth(object):
             if ip_address not in app_service.ip_range_whitelist:
                 defer.returnValue((None, None))
 
-        if "user_id" not in request.args:
+        if b"user_id" not in request.args:
             defer.returnValue((app_service.sender, app_service))
 
-        user_id = request.args["user_id"][0]
+        user_id = request.args[b"user_id"][0].decode('utf8')
         if app_service.sender == user_id:
             defer.returnValue((app_service.sender, app_service))
 
@@ -544,7 +545,8 @@ class Auth(object):
 
     @defer.inlineCallbacks
     def add_auth_events(self, builder, context):
-        auth_ids = yield self.compute_auth_events(builder, context.prev_state_ids)
+        prev_state_ids = yield context.get_prev_state_ids(self.store)
+        auth_ids = yield self.compute_auth_events(builder, prev_state_ids)
 
         auth_events_entries = yield self.store.add_event_hashes(
             auth_ids
@@ -737,3 +739,37 @@ class Auth(object):
                 )
 
             return query_params[0]
+
+    @defer.inlineCallbacks
+    def check_in_room_or_world_readable(self, room_id, user_id):
+        """Checks that the user is or was in the room or the room is world
+        readable. If it isn't then an exception is raised.
+
+        Returns:
+            Deferred[tuple[str, str|None]]: Resolves to the current membership of
+            the user in the room and the membership event ID of the user. If
+            the user is not in the room and never has been, then
+            `(Membership.JOIN, None)` is returned.
+        """
+
+        try:
+            # check_user_was_in_room will return the most recent membership
+            # event for the user if:
+            #  * The user is a non-guest user, and was ever in the room
+            #  * The user is a guest user, and has joined the room
+            # else it will throw.
+            member_event = yield self.check_user_was_in_room(room_id, user_id)
+            defer.returnValue((member_event.membership, member_event.event_id))
+        except AuthError:
+            visibility = yield self.state.get_current_state(
+                room_id, EventTypes.RoomHistoryVisibility, ""
+            )
+            if (
+                visibility and
+                visibility.content["history_visibility"] == "world_readable"
+            ):
+                defer.returnValue((Membership.JOIN, None))
+                return
+            raise AuthError(
+                403, "Guest access not allowed", errcode=Codes.GUEST_ACCESS_FORBIDDEN
+            )

synapse/api/errors.py

@@ -55,6 +55,7 @@ class Codes(object):
     SERVER_NOT_TRUSTED = "M_SERVER_NOT_TRUSTED"
     CONSENT_NOT_GIVEN = "M_CONSENT_NOT_GIVEN"
     CANNOT_LEAVE_SERVER_NOTICE_ROOM = "M_CANNOT_LEAVE_SERVER_NOTICE_ROOM"
+    MAU_LIMIT_EXCEEDED = "M_MAU_LIMIT_EXCEEDED"
 
 
 class CodeMessageException(RuntimeError):

synapse/api/filtering.py

@@ -12,6 +12,9 @@
 # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 # See the License for the specific language governing permissions and
 # limitations under the License.
+
+import six
+import ujson as json
 import jsonschema
 from canonicaljson import json
 from jsonschema import FormatChecker
@@ -113,7 +116,13 @@ ROOM_EVENT_FILTER_SCHEMA = {
         },
         "contains_url": {
             "type": "boolean"
-        }
+        },
+        "lazy_load_members": {
+            "type": "boolean"
+        },
+        "include_redundant_members": {
+            "type": "boolean"
+        },
     }
 }
 
@@ -261,6 +270,12 @@ class FilterCollection(object):
     def ephemeral_limit(self):
         return self._room_ephemeral_filter.limit()
 
+    def lazy_load_members(self):
+        return self._room_state_filter.lazy_load_members()
+
+    def include_redundant_members(self):
+        return self._room_state_filter.include_redundant_members()
+
     def filter_presence(self, events):
         return self._presence_filter.filter(events)
 
@@ -352,7 +367,10 @@ class Filter(object):
 
             room_id = event.get("room_id", None)
             ev_type = event.get("type", None)
-            is_url = "url" in event.get("content", {})
+
+            content = event.get("content", {})
+            # check if there is a string url field in the content for filtering purposes
+            is_url = isinstance(content.get("url"), six.text_type)
 
         return self.check_fields(
             room_id,
@@ -417,6 +435,12 @@ class Filter(object):
     def limit(self):
         return self.filter_json.get("limit", 10)
 
+    def lazy_load_members(self):
+        return self.filter_json.get("lazy_load_members", False)
+
+    def include_redundant_members(self):
+        return self.filter_json.get("include_redundant_members", False)
+
 
 def _matches_wildcard(actual_value, filter_value):
     if filter_value.endswith("*"):

synapse/app/client_reader.py

@@ -31,6 +31,7 @@ from synapse.http.site import SynapseSite
 from synapse.metrics import RegistryProxy
 from synapse.metrics.resource import METRICS_PREFIX, MetricsResource
 from synapse.replication.slave.storage._base import BaseSlavedStore
+from synapse.replication.slave.storage.account_data import SlavedAccountDataStore
 from synapse.replication.slave.storage.appservice import SlavedApplicationServiceStore
 from synapse.replication.slave.storage.client_ips import SlavedClientIpStore
 from synapse.replication.slave.storage.directory import DirectoryStore
@@ -40,7 +41,13 @@ from synapse.replication.slave.storage.registration import SlavedRegistrationSto
 from synapse.replication.slave.storage.room import RoomStore
 from synapse.replication.slave.storage.transactions import TransactionStore
 from synapse.replication.tcp.client import ReplicationClientHandler
-from synapse.rest.client.v1.room import PublicRoomListRestServlet
+from synapse.rest.client.v1.room import (
+    JoinedRoomMemberListRestServlet,
+    PublicRoomListRestServlet,
+    RoomEventContextServlet,
+    RoomMemberListRestServlet,
+    RoomStateRestServlet,
+)
 from synapse.server import HomeServer
 from synapse.storage.engines import create_engine
 from synapse.util.httpresourcetree import create_resource_tree
@@ -52,6 +59,7 @@ logger = logging.getLogger("synapse.app.client_reader")
 
 
 class ClientReaderSlavedStore(
+    SlavedAccountDataStore,
     SlavedEventStore,
     SlavedKeyStore,
     RoomStore,
@@ -82,7 +90,13 @@ class ClientReaderServer(HomeServer):
                     resources[METRICS_PREFIX] = MetricsResource(RegistryProxy)
                 elif name == "client":
                     resource = JsonResource(self, canonical_json=False)
+
                     PublicRoomListRestServlet(self).register(resource)
+                    RoomMemberListRestServlet(self).register(resource)
+                    JoinedRoomMemberListRestServlet(self).register(resource)
+                    RoomStateRestServlet(self).register(resource)
+                    RoomEventContextServlet(self).register(resource)
+
                     resources.update({
                         "/_matrix/client/r0": resource,
                         "/_matrix/client/unstable": resource,

synapse/app/homeserver.py

@@ -18,6 +18,10 @@ import logging
 import os
 import sys
 
+from six import iteritems
+
+from prometheus_client import Gauge
+
 from twisted.application import service
 from twisted.internet import defer, reactor
 from twisted.web.resource import EncodingResourceWrapper, NoResource
@@ -47,6 +51,7 @@ from synapse.http.additional_resource import AdditionalResource
 from synapse.http.server import RootRedirect
 from synapse.http.site import SynapseSite
 from synapse.metrics import RegistryProxy
+from synapse.metrics.background_process_metrics import run_as_background_process
 from synapse.metrics.resource import METRICS_PREFIX, MetricsResource
 from synapse.module_api import ModuleApi
 from synapse.python_dependencies import CONDITIONAL_REQUIREMENTS, check_requirements
@@ -297,6 +302,11 @@ class SynapseHomeServer(HomeServer):
             quit_with_error(e.message)
 
 
+# Gauges to expose monthly active user control metrics
+current_mau_gauge = Gauge("synapse_admin_current_mau", "Current MAU")
+max_mau_value_gauge = Gauge("synapse_admin_max_mau_value", "MAU Limit")
+
+
 def setup(config_options):
     """
     Args:
@@ -425,6 +435,9 @@ def run(hs):
     # currently either 0 or 1
     stats_process = []
 
+    def start_phone_stats_home():
+        return run_as_background_process("phone_stats_home", phone_stats_home)
+
     @defer.inlineCallbacks
     def phone_stats_home():
         logger.info("Gathering stats for reporting")
@@ -442,7 +455,7 @@ def run(hs):
         stats["total_nonbridged_users"] = total_nonbridged_users
 
         daily_user_type_results = yield hs.get_datastore().count_daily_user_type()
-        for name, count in daily_user_type_results.iteritems():
+        for name, count in iteritems(daily_user_type_results):
             stats["daily_user_type_" + name] = count
 
         room_count = yield hs.get_datastore().get_room_count()
@@ -453,7 +466,7 @@ def run(hs):
         stats["daily_messages"] = yield hs.get_datastore().count_daily_messages()
 
         r30_results = yield hs.get_datastore().count_r30_users()
-        for name, count in r30_results.iteritems():
+        for name, count in iteritems(r30_results):
             stats["r30_users_" + name] = count
 
         daily_sent_messages = yield hs.get_datastore().count_daily_sent_messages()
@@ -496,16 +509,31 @@ def run(hs):
             )
 
     def generate_user_daily_visit_stats():
-        hs.get_datastore().generate_user_daily_visits()
+        return run_as_background_process(
+            "generate_user_daily_visits",
+            hs.get_datastore().generate_user_daily_visits,
+        )
 
     # Rather than update on per session basis, batch up the requests.
     # If you increase the loop period, the accuracy of user_daily_visits
     # table will decrease
     clock.looping_call(generate_user_daily_visit_stats, 5 * 60 * 1000)
 
+    @defer.inlineCallbacks
+    def generate_monthly_active_users():
+        count = 0
+        if hs.config.limit_usage_by_mau:
+            count = yield hs.get_datastore().count_monthly_users()
+        current_mau_gauge.set(float(count))
+        max_mau_value_gauge.set(float(hs.config.max_mau_value))
+
+    generate_monthly_active_users()
+    if hs.config.limit_usage_by_mau:
+        clock.looping_call(generate_monthly_active_users, 5 * 60 * 1000)
+
     if hs.config.report_stats:
         logger.info("Scheduling stats reporting for 3 hour intervals")
-        clock.looping_call(phone_stats_home, 3 * 60 * 60 * 1000)
+        clock.looping_call(start_phone_stats_home, 3 * 60 * 60 * 1000)
 
         # We need to defer this init for the cases that we daemonize
         # otherwise the process ID we get is that of the non-daemon process
@@ -513,7 +541,7 @@ def run(hs):
 
         # We wait 5 minutes to send the first set of stats as the server can
         # be quite busy the first few minutes
-        clock.call_later(5 * 60, phone_stats_home)
+        clock.call_later(5 * 60, start_phone_stats_home)
 
     if hs.config.daemonize and hs.config.print_pidfile:
         print (hs.config.pid_file)

synapse/app/synchrotron.py

@@ -55,7 +55,6 @@ from synapse.rest.client.v2_alpha import sync
 from synapse.server import HomeServer
 from synapse.storage.engines import create_engine
 from synapse.storage.presence import UserPresenceState
-from synapse.storage.roommember import RoomMemberStore
 from synapse.util.httpresourcetree import create_resource_tree
 from synapse.util.logcontext import LoggingContext, run_in_background
 from synapse.util.manhole import manhole
@@ -81,9 +80,7 @@ class SynchrotronSlavedStore(
     RoomStore,
     BaseSlavedStore,
 ):
-    did_forget = (
-        RoomMemberStore.__dict__["did_forget"]
-    )
+    pass
 
 
 UPDATE_SYNCING_USERS_MS = 10 * 1000

synapse/app/synctl.py

@@ -25,6 +25,8 @@ import subprocess
 import sys
 import time
 
+from six import iteritems
+
 import yaml
 
 SYNAPSE = [sys.executable, "-B", "-m", "synapse.app.homeserver"]
@@ -173,7 +175,7 @@ def main():
         os.environ["SYNAPSE_CACHE_FACTOR"] = str(cache_factor)
 
     cache_factors = config.get("synctl_cache_factors", {})
-    for cache_name, factor in cache_factors.iteritems():
+    for cache_name, factor in iteritems(cache_factors):
         os.environ["SYNAPSE_CACHE_FACTOR_" + cache_name.upper()] = str(factor)
 
     worker_configfiles = []

synapse/config/server.py

@@ -67,6 +67,14 @@ class ServerConfig(Config):
             "block_non_admin_invites", False,
         )
 
+        # Options to control access by tracking MAU
+        self.limit_usage_by_mau = config.get("limit_usage_by_mau", False)
+        if self.limit_usage_by_mau:
+            self.max_mau_value = config.get(
+                "max_mau_value", 0,
+            )
+        else:
+            self.max_mau_value = 0
         # FIXME: federation_domain_whitelist needs sytests
         self.federation_domain_whitelist = None
         federation_domain_whitelist = config.get(

synapse/config/voip.py

@@ -30,10 +30,10 @@ class VoipConfig(Config):
         ## Turn ##
 
         # The public URIs of the TURN server to give to clients
-        turn_uris: []
+        #turn_uris: []
 
         # The shared secret used to compute passwords for the TURN server
-        turn_shared_secret: "YOUR_SHARED_SECRET"
+        #turn_shared_secret: "YOUR_SHARED_SECRET"
 
         # The Username and password if the TURN server needs them and
         # does not use a token

synapse/events/snapshot.py

@@ -13,22 +13,18 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 
+from six import iteritems
+
 from frozendict import frozendict
 
 from twisted.internet import defer
 
+from synapse.util.logcontext import make_deferred_yieldable, run_in_background
+
 
 class EventContext(object):
     """
     Attributes:
-        current_state_ids (dict[(str, str), str]):
-            The current state map including the current event.
-            (type, state_key) -> event_id
-
-        prev_state_ids (dict[(str, str), str]):
-            The current state map excluding the current event.
-            (type, state_key) -> event_id
-
         state_group (int|None): state group id, if the state has been stored
             as a state group. This is usually only None if e.g. the event is
             an outlier.
@@ -45,38 +41,77 @@ class EventContext(object):
 
         prev_state_events (?): XXX: is this ever set to anything other than
             the empty list?
+
+        _current_state_ids (dict[(str, str), str]|None):
+            The current state map including the current event. None if outlier
+            or we haven't fetched the state from DB yet.
+            (type, state_key) -> event_id
+
+        _prev_state_ids (dict[(str, str), str]|None):
+            The current state map excluding the current event. None if outlier
+            or we haven't fetched the state from DB yet.
+            (type, state_key) -> event_id
+
+        _fetching_state_deferred (Deferred|None): Resolves when *_state_ids have
+            been calculated. None if we haven't started calculating yet
+
+        _event_type (str): The type of the event the context is associated with.
+            Only set when state has not been fetched yet.
+
+        _event_state_key (str|None): The state_key of the event the context is
+            associated with. Only set when state has not been fetched yet.
+
+        _prev_state_id (str|None): If the event associated with the context is
+            a state event, then `_prev_state_id` is the event_id of the state
+            that was replaced.
+            Only set when state has not been fetched yet.
     """
 
     __slots__ = [
-        "current_state_ids",
-        "prev_state_ids",
         "state_group",
         "rejected",
         "prev_group",
         "delta_ids",
         "prev_state_events",
         "app_service",
+        "_current_state_ids",
+        "_prev_state_ids",
+        "_prev_state_id",
+        "_event_type",
+        "_event_state_key",
+        "_fetching_state_deferred",
     ]
 
     def __init__(self):
-        # The current state including the current event
-        self.current_state_ids = None
-        # The current state excluding the current event
-        self.prev_state_ids = None
-        self.state_group = None
-
+        self.prev_state_events = []
         self.rejected = False
+        self.app_service = None
+
+    @staticmethod
+    def with_state(state_group, current_state_ids, prev_state_ids,
+                   prev_group=None, delta_ids=None):
+        context = EventContext()
+
+        # The current state including the current event
+        context._current_state_ids = current_state_ids
+        # The current state excluding the current event
+        context._prev_state_ids = prev_state_ids
+        context.state_group = state_group
+
+        context._prev_state_id = None
+        context._event_type = None
+        context._event_state_key = None
+        context._fetching_state_deferred = defer.succeed(None)
 
         # A previously persisted state group and a delta between that
         # and this state.
-        self.prev_group = None
-        self.delta_ids = None
+        context.prev_group = prev_group
+        context.delta_ids = delta_ids
 
-        self.prev_state_events = None
+        return context
 
-        self.app_service = None
-
-    def serialize(self, event):
+    @defer.inlineCallbacks
+    def serialize(self, event, store):
         """Converts self to a type that can be serialized as JSON, and then
         deserialized by `deserialize`
 
@@ -92,11 +127,12 @@ class EventContext(object):
         # the prev_state_ids, so if we're a state event we include the event
         # id that we replaced in the state.
         if event.is_state():
-            prev_state_id = self.prev_state_ids.get((event.type, event.state_key))
+            prev_state_ids = yield self.get_prev_state_ids(store)
+            prev_state_id = prev_state_ids.get((event.type, event.state_key))
         else:
             prev_state_id = None
 
-        return {
+        defer.returnValue({
             "prev_state_id": prev_state_id,
             "event_type": event.type,
             "event_state_key": event.state_key if event.is_state() else None,
@@ -106,10 +142,9 @@ class EventContext(object):
             "delta_ids": _encode_state_dict(self.delta_ids),
             "prev_state_events": self.prev_state_events,
             "app_service_id": self.app_service.id if self.app_service else None
-        }
+        })
 
     @staticmethod
-    @defer.inlineCallbacks
     def deserialize(store, input):
         """Converts a dict that was produced by `serialize` back into a
         EventContext.
@@ -122,32 +157,115 @@ class EventContext(object):
             EventContext
         """
         context = EventContext()
-        context.state_group = input["state_group"]
-        context.rejected = input["rejected"]
-        context.prev_group = input["prev_group"]
-        context.delta_ids = _decode_state_dict(input["delta_ids"])
-        context.prev_state_events = input["prev_state_events"]
 
         # We use the state_group and prev_state_id stuff to pull the
         # current_state_ids out of the DB and construct prev_state_ids.
-        prev_state_id = input["prev_state_id"]
-        event_type = input["event_type"]
-        event_state_key = input["event_state_key"]
+        context._prev_state_id = input["prev_state_id"]
+        context._event_type = input["event_type"]
+        context._event_state_key = input["event_state_key"]
 
-        context.current_state_ids = yield store.get_state_ids_for_group(
-            context.state_group,
-        )
-        if prev_state_id and event_state_key:
-            context.prev_state_ids = dict(context.current_state_ids)
-            context.prev_state_ids[(event_type, event_state_key)] = prev_state_id
-        else:
-            context.prev_state_ids = context.current_state_ids
+        context._current_state_ids = None
+        context._prev_state_ids = None
+        context._fetching_state_deferred = None
+
+        context.state_group = input["state_group"]
+        context.prev_group = input["prev_group"]
+        context.delta_ids = _decode_state_dict(input["delta_ids"])
+
+        context.rejected = input["rejected"]
+        context.prev_state_events = input["prev_state_events"]
 
         app_service_id = input["app_service_id"]
         if app_service_id:
             context.app_service = store.get_app_service_by_id(app_service_id)
 
-        defer.returnValue(context)
+        return context
+
+    @defer.inlineCallbacks
+    def get_current_state_ids(self, store):
+        """Gets the current state IDs
+
+        Returns:
+            Deferred[dict[(str, str), str]|None]: Returns None if state_group
+            is None, which happens when the associated event is an outlier.
+        """
+
+        if not self._fetching_state_deferred:
+            self._fetching_state_deferred = run_in_background(
+                self._fill_out_state, store,
+            )
+
+        yield make_deferred_yieldable(self._fetching_state_deferred)
+
+        defer.returnValue(self._current_state_ids)
+
+    @defer.inlineCallbacks
+    def get_prev_state_ids(self, store):
+        """Gets the prev state IDs
+
+        Returns:
+            Deferred[dict[(str, str), str]|None]: Returns None if state_group
+            is None, which happens when the associated event is an outlier.
+        """
+
+        if not self._fetching_state_deferred:
+            self._fetching_state_deferred = run_in_background(
+                self._fill_out_state, store,
+            )
+
+        yield make_deferred_yieldable(self._fetching_state_deferred)
+
+        defer.returnValue(self._prev_state_ids)
+
+    def get_cached_current_state_ids(self):
+        """Gets the current state IDs if we have them already cached.
+
+        Returns:
+            dict[(str, str), str]|None: Returns None if we haven't cached the
+            state or if state_group is None, which happens when the associated
+            event is an outlier.
+        """
+
+        return self._current_state_ids
+
+    @defer.inlineCallbacks
+    def _fill_out_state(self, store):
+        """Called to populate the _current_state_ids and _prev_state_ids
+        attributes by loading from the database.
+        """
+        if self.state_group is None:
+            return
+
+        self._current_state_ids = yield store.get_state_ids_for_group(
+            self.state_group,
+        )
+        if self._prev_state_id and self._event_state_key is not None:
+            self._prev_state_ids = dict(self._current_state_ids)
+
+            key = (self._event_type, self._event_state_key)
+            self._prev_state_ids[key] = self._prev_state_id
+        else:
+            self._prev_state_ids = self._current_state_ids
+
+    @defer.inlineCallbacks
+    def update_state(self, state_group, prev_state_ids, current_state_ids,
+                     prev_group, delta_ids):
+        """Replace the state in the context
+        """
+
+        # We need to make sure we wait for any ongoing fetching of state
+        # to complete so that the updated state doesn't get clobbered
+        if self._fetching_state_deferred:
+            yield make_deferred_yieldable(self._fetching_state_deferred)
+
+        self.state_group = state_group
+        self._prev_state_ids = prev_state_ids
+        self.prev_group = prev_group
+        self._current_state_ids = current_state_ids
+        self.delta_ids = delta_ids
+
+        # We need to ensure that that we've marked as having fetched the state
+        self._fetching_state_deferred = defer.succeed(None)
 
 
 def _encode_state_dict(state_dict):
@@ -159,7 +277,7 @@ def _encode_state_dict(state_dict):
 
     return [
         (etype, state_key, v)
-        for (etype, state_key), v in state_dict.iteritems()
+        for (etype, state_key), v in iteritems(state_dict)
     ]
 
 

synapse/federation/federation_server.py

@@ -24,6 +24,7 @@ from prometheus_client import Counter
 
 from twisted.internet import defer
 from twisted.internet.abstract import isIPAddress
+from twisted.python import failure
 
 from synapse.api.constants import EventTypes
 from synapse.api.errors import AuthError, FederationError, NotFoundError, SynapseError
@@ -186,8 +187,12 @@ class FederationServer(FederationBase):
                     logger.warn("Error handling PDU %s: %s", event_id, e)
                     pdu_results[event_id] = {"error": str(e)}
                 except Exception as e:
+                    f = failure.Failure()
                     pdu_results[event_id] = {"error": str(e)}
-                    logger.exception("Failed to handle PDU %s", event_id)
+                    logger.error(
+                        "Failed to handle PDU %s: %s",
+                        event_id, f.getTraceback().rstrip(),
+                    )
 
         yield async.concurrently_execute(
             process_pdus_for_room, pdus_by_room.keys(),
@@ -202,10 +207,6 @@ class FederationServer(FederationBase):
                     edu.content
                 )
 
-        pdu_failures = getattr(transaction, "pdu_failures", [])
-        for failure in pdu_failures:
-            logger.info("Got failure %r", failure)
-
         response = {
             "pdus": pdu_results,
         }

synapse/federation/send_queue.py

@@ -62,8 +62,6 @@ class FederationRemoteSendQueue(object):
 
         self.edus = SortedDict()  # stream position -> Edu
 
-        self.failures = SortedDict()  # stream position -> (destination, Failure)
-
         self.device_messages = SortedDict()  # stream position -> destination
 
         self.pos = 1
@@ -79,7 +77,7 @@ class FederationRemoteSendQueue(object):
 
         for queue_name in [
             "presence_map", "presence_changed", "keyed_edu", "keyed_edu_changed",
-            "edus", "failures", "device_messages", "pos_time",
+            "edus", "device_messages", "pos_time",
         ]:
             register(queue_name, getattr(self, queue_name))
 
@@ -149,12 +147,6 @@ class FederationRemoteSendQueue(object):
             for key in keys[:i]:
                 del self.edus[key]
 
-            # Delete things out of failure map
-            keys = self.failures.keys()
-            i = self.failures.bisect_left(position_to_delete)
-            for key in keys[:i]:
-                del self.failures[key]
-
             # Delete things out of device map
             keys = self.device_messages.keys()
             i = self.device_messages.bisect_left(position_to_delete)
@@ -204,13 +196,6 @@ class FederationRemoteSendQueue(object):
 
         self.notifier.on_new_replication_data()
 
-    def send_failure(self, failure, destination):
-        """As per TransactionQueue"""
-        pos = self._next_pos()
-
-        self.failures[pos] = (destination, str(failure))
-        self.notifier.on_new_replication_data()
-
     def send_device_messages(self, destination):
         """As per TransactionQueue"""
         pos = self._next_pos()
@@ -285,17 +270,6 @@ class FederationRemoteSendQueue(object):
         for (pos, edu) in edus:
             rows.append((pos, EduRow(edu)))
 
-        # Fetch changed failures
-        i = self.failures.bisect_right(from_token)
-        j = self.failures.bisect_right(to_token) + 1
-        failures = self.failures.items()[i:j]
-
-        for (pos, (destination, failure)) in failures:
-            rows.append((pos, FailureRow(
-                destination=destination,
-                failure=failure,
-            )))
-
         # Fetch changed device messages
         i = self.device_messages.bisect_right(from_token)
         j = self.device_messages.bisect_right(to_token) + 1
@@ -417,34 +391,6 @@ class EduRow(BaseFederationRow, namedtuple("EduRow", (
         buff.edus.setdefault(self.edu.destination, []).append(self.edu)
 
 
-class FailureRow(BaseFederationRow, namedtuple("FailureRow", (
-    "destination",  # str
-    "failure",
-))):
-    """Streams failures to a remote server. Failures are issued when there was
-    something wrong with a transaction the remote sent us, e.g. it included
-    an event that was invalid.
-    """
-
-    TypeId = "f"
-
-    @staticmethod
-    def from_data(data):
-        return FailureRow(
-            destination=data["destination"],
-            failure=data["failure"],
-        )
-
-    def to_data(self):
-        return {
-            "destination": self.destination,
-            "failure": self.failure,
-        }
-
-    def add_to_buffer(self, buff):
-        buff.failures.setdefault(self.destination, []).append(self.failure)
-
-
 class DeviceRow(BaseFederationRow, namedtuple("DeviceRow", (
     "destination",  # str
 ))):
@@ -471,7 +417,6 @@ TypeToRow = {
         PresenceRow,
         KeyedEduRow,
         EduRow,
-        FailureRow,
         DeviceRow,
     )
 }
@@ -481,7 +426,6 @@ ParsedFederationStreamData = namedtuple("ParsedFederationStreamData", (
     "presence",  # list(UserPresenceState)
     "keyed_edus",  # dict of destination -> { key -> Edu }
     "edus",  # dict of destination -> [Edu]
-    "failures",  # dict of destination -> [failures]
     "device_destinations",  # set of destinations
 ))
 
@@ -503,7 +447,6 @@ def process_rows_for_federation(transaction_queue, rows):
         presence=[],
         keyed_edus={},
         edus={},
-        failures={},
         device_destinations=set(),
     )
 
@@ -532,9 +475,5 @@ def process_rows_for_federation(transaction_queue, rows):
                 edu.destination, edu.edu_type, edu.content, key=None,
             )
 
-    for destination, failure_list in iteritems(buff.failures):
-        for failure in failure_list:
-            transaction_queue.send_failure(destination, failure)
-
     for destination in buff.device_destinations:
         transaction_queue.send_device_messages(destination)

synapse/federation/transaction_queue.py

@@ -30,7 +30,8 @@ from synapse.metrics import (
     sent_edus_counter,
     sent_transactions_counter,
 )
-from synapse.util import PreserveLoggingContext, logcontext
+from synapse.metrics.background_process_metrics import run_as_background_process
+from synapse.util import logcontext
 from synapse.util.metrics import measure_func
 from synapse.util.retryutils import NotRetryingDestination, get_retry_limiter
 
@@ -115,9 +116,6 @@ class TransactionQueue(object):
             ),
         )
 
-        # destination -> list of tuple(failure, deferred)
-        self.pending_failures_by_dest = {}
-
         # destination -> stream_id of last successfully sent to-device message.
         # NB: may be a long or an int.
         self.last_device_stream_id_by_dest = {}
@@ -165,10 +163,11 @@ class TransactionQueue(object):
         if self._is_processing:
             return
 
-        # fire off a processing loop in the background. It's likely it will
-        # outlast the current request, so run it in the sentinel logcontext.
-        with PreserveLoggingContext():
-            self._process_event_queue_loop()
+        # fire off a processing loop in the background
+        run_as_background_process(
+            "process_event_queue_for_federation",
+            self._process_event_queue_loop,
+        )
 
     @defer.inlineCallbacks
     def _process_event_queue_loop(self):
@@ -380,19 +379,6 @@ class TransactionQueue(object):
 
         self._attempt_new_transaction(destination)
 
-    def send_failure(self, failure, destination):
-        if destination == self.server_name or destination == "localhost":
-            return
-
-        if not self.can_send_to(destination):
-            return
-
-        self.pending_failures_by_dest.setdefault(
-            destination, []
-        ).append(failure)
-
-        self._attempt_new_transaction(destination)
-
     def send_device_messages(self, destination):
         if destination == self.server_name or destination == "localhost":
             return
@@ -432,14 +418,11 @@ class TransactionQueue(object):
 
         logger.debug("TX [%s] Starting transaction loop", destination)
 
-        # Drop the logcontext before starting the transaction. It doesn't
-        # really make sense to log all the outbound transactions against
-        # whatever path led us to this point: that's pretty arbitrary really.
-        #
-        # (this also means we can fire off _perform_transaction without
-        # yielding)
-        with logcontext.PreserveLoggingContext():
-            self._transaction_transmission_loop(destination)
+        run_as_background_process(
+            "federation_transaction_transmission_loop",
+            self._transaction_transmission_loop,
+            destination,
+        )
 
     @defer.inlineCallbacks
     def _transaction_transmission_loop(self, destination):
@@ -470,7 +453,6 @@ class TransactionQueue(object):
                 pending_pdus = self.pending_pdus_by_dest.pop(destination, [])
                 pending_edus = self.pending_edus_by_dest.pop(destination, [])
                 pending_presence = self.pending_presence_by_dest.pop(destination, {})
-                pending_failures = self.pending_failures_by_dest.pop(destination, [])
 
                 pending_edus.extend(
                     self.pending_edus_keyed_by_dest.pop(destination, {}).values()
@@ -498,7 +480,7 @@ class TransactionQueue(object):
                     logger.debug("TX [%s] len(pending_pdus_by_dest[dest]) = %d",
                                  destination, len(pending_pdus))
 
-                if not pending_pdus and not pending_edus and not pending_failures:
+                if not pending_pdus and not pending_edus:
                     logger.debug("TX [%s] Nothing to send", destination)
                     self.last_device_stream_id_by_dest[destination] = (
                         device_stream_id
@@ -508,7 +490,7 @@ class TransactionQueue(object):
                 # END CRITICAL SECTION
 
                 success = yield self._send_new_transaction(
-                    destination, pending_pdus, pending_edus, pending_failures,
+                    destination, pending_pdus, pending_edus,
                 )
                 if success:
                     sent_transactions_counter.inc()
@@ -585,14 +567,12 @@ class TransactionQueue(object):
 
     @measure_func("_send_new_transaction")
     @defer.inlineCallbacks
-    def _send_new_transaction(self, destination, pending_pdus, pending_edus,
-                              pending_failures):
+    def _send_new_transaction(self, destination, pending_pdus, pending_edus):
 
         # Sort based on the order field
         pending_pdus.sort(key=lambda t: t[1])
         pdus = [x[0] for x in pending_pdus]
         edus = pending_edus
-        failures = [x.get_dict() for x in pending_failures]
 
         success = True
 
@@ -602,11 +582,10 @@ class TransactionQueue(object):
 
         logger.debug(
             "TX [%s] {%s} Attempting new transaction"
-            " (pdus: %d, edus: %d, failures: %d)",
+            " (pdus: %d, edus: %d)",
             destination, txn_id,
             len(pdus),
             len(edus),
-            len(failures)
         )
 
         logger.debug("TX [%s] Persisting transaction...", destination)
@@ -618,7 +597,6 @@ class TransactionQueue(object):
             destination=destination,
             pdus=pdus,
             edus=edus,
-            pdu_failures=failures,
         )
 
         self._next_txn_id += 1
@@ -628,12 +606,11 @@ class TransactionQueue(object):
         logger.debug("TX [%s] Persisted transaction", destination)
         logger.info(
             "TX [%s] {%s} Sending transaction [%s],"
-            " (PDUs: %d, EDUs: %d, failures: %d)",
+            " (PDUs: %d, EDUs: %d)",
             destination, txn_id,
             transaction.transaction_id,
             len(pdus),
             len(edus),
-            len(failures),
         )
 
         # Actually send the transaction

synapse/federation/transport/server.py

@@ -165,7 +165,7 @@ def _parse_auth_header(header_bytes):
         param_dict = dict(kv.split("=") for kv in params)
 
         def strip_quotes(value):
-            if value.startswith(b"\""):
+            if value.startswith("\""):
                 return value[1:-1]
             else:
                 return value
@@ -283,11 +283,10 @@ class FederationSendServlet(BaseFederationServlet):
             )
 
             logger.info(
-                "Received txn %s from %s. (PDUs: %d, EDUs: %d, failures: %d)",
+                "Received txn %s from %s. (PDUs: %d, EDUs: %d)",
                 transaction_id, origin,
                 len(transaction_data.get("pdus", [])),
                 len(transaction_data.get("edus", [])),
-                len(transaction_data.get("failures", [])),
             )
 
             # We should ideally be getting this from the security layer.
@@ -404,10 +403,10 @@ class FederationMakeLeaveServlet(BaseFederationServlet):
 
 
 class FederationSendLeaveServlet(BaseFederationServlet):
-    PATH = "/send_leave/(?P<room_id>[^/]*)/(?P<txid>[^/]*)"
+    PATH = "/send_leave/(?P<room_id>[^/]*)/(?P<event_id>[^/]*)"
 
     @defer.inlineCallbacks
-    def on_PUT(self, origin, content, query, room_id, txid):
+    def on_PUT(self, origin, content, query, room_id, event_id):
         content = yield self.handler.on_send_leave_request(origin, content)
         defer.returnValue((200, content))
 

synapse/federation/units.py

@@ -73,7 +73,6 @@ class Transaction(JsonEncodedObject):
         "previous_ids",
         "pdus",
         "edus",
-        "pdu_failures",
     ]
 
     internal_keys = [

synapse/groups/attestations.py

@@ -43,6 +43,7 @@ from signedjson.sign import sign_json
 from twisted.internet import defer
 
 from synapse.api.errors import SynapseError
+from synapse.metrics.background_process_metrics import run_as_background_process
 from synapse.types import get_domain_from_id
 from synapse.util.logcontext import run_in_background
 
@@ -129,7 +130,7 @@ class GroupAttestionRenewer(object):
         self.attestations = hs.get_groups_attestation_signing()
 
         self._renew_attestations_loop = self.clock.looping_call(
-            self._renew_attestations, 30 * 60 * 1000,
+            self._start_renew_attestations, 30 * 60 * 1000,
         )
 
     @defer.inlineCallbacks
@@ -151,6 +152,9 @@ class GroupAttestionRenewer(object):
 
         defer.returnValue({})
 
+    def _start_renew_attestations(self):
+        return run_as_background_process("renew_attestations", self._renew_attestations)
+
     @defer.inlineCallbacks
     def _renew_attestations(self):
         """Called periodically to check if we need to update any of our attestations

synapse/handlers/__init__.py

@@ -17,9 +17,7 @@ from .admin import AdminHandler
 from .directory import DirectoryHandler
 from .federation import FederationHandler
 from .identity import IdentityHandler
-from .message import MessageHandler
 from .register import RegistrationHandler
-from .room import RoomContextHandler
 from .search import SearchHandler
 
 
@@ -44,10 +42,8 @@ class Handlers(object):
 
     def __init__(self, hs):
         self.registration_handler = RegistrationHandler(hs)
-        self.message_handler = MessageHandler(hs)
         self.federation_handler = FederationHandler(hs)
         self.directory_handler = DirectoryHandler(hs)
         self.admin_handler = AdminHandler(hs)
         self.identity_handler = IdentityHandler(hs)
         self.search_handler = SearchHandler(hs)
-        self.room_context_handler = RoomContextHandler(hs)

synapse/handlers/_base.py

@@ -112,8 +112,9 @@ class BaseHandler(object):
             guest_access = event.content.get("guest_access", "forbidden")
             if guest_access != "can_join":
                 if context:
+                    current_state_ids = yield context.get_current_state_ids(self.store)
                     current_state = yield self.store.get_events(
-                        list(context.current_state_ids.values())
+                        list(current_state_ids.values())
                     )
                 else:
                     current_state = yield self.state_handler.get_current_state(

synapse/handlers/appservice.py

@@ -23,6 +23,7 @@ from twisted.internet import defer
 
 import synapse
 from synapse.api.constants import EventTypes
+from synapse.metrics.background_process_metrics import run_as_background_process
 from synapse.util.logcontext import make_deferred_yieldable, run_in_background
 from synapse.util.metrics import Measure
 
@@ -106,7 +107,9 @@ class ApplicationServicesHandler(object):
                             yield self._check_user_exists(event.state_key)
 
                         if not self.started_scheduler:
-                            self.scheduler.start().addErrback(log_failure)
+                            def start_scheduler():
+                                return self.scheduler.start().addErrback(log_failure)
+                            run_as_background_process("as_scheduler", start_scheduler)
                             self.started_scheduler = True
 
                         # Fork off pushes to these services

synapse/handlers/auth.py

@@ -15,6 +15,7 @@
 # limitations under the License.
 
 import logging
+import unicodedata
 
 import attr
 import bcrypt
@@ -519,6 +520,7 @@ class AuthHandler(BaseHandler):
         """
         logger.info("Logging in user %s on device %s", user_id, device_id)
         access_token = yield self.issue_access_token(user_id, device_id)
+        yield self._check_mau_limits()
 
         # the device *should* have been registered before we got here; however,
         # it's possible we raced against a DELETE operation. The thing we
@@ -626,6 +628,7 @@ class AuthHandler(BaseHandler):
         # special case to check for "password" for the check_password interface
         # for the auth providers
         password = login_submission.get("password")
+
         if login_type == LoginType.PASSWORD:
             if not self._password_enabled:
                 raise SynapseError(400, "Password login has been disabled.")
@@ -707,9 +710,10 @@ class AuthHandler(BaseHandler):
         multiple inexact matches.
 
         Args:
-            user_id (str): complete @user:id
+            user_id (unicode): complete @user:id
+            password (unicode): the provided password
         Returns:
-            (str) the canonical_user_id, or None if unknown user / bad password
+            (unicode) the canonical_user_id, or None if unknown user / bad password
         """
         lookupres = yield self._find_user_id_and_pwd_hash(user_id)
         if not lookupres:
@@ -728,15 +732,18 @@ class AuthHandler(BaseHandler):
                                                   device_id)
         defer.returnValue(access_token)
 
+    @defer.inlineCallbacks
     def validate_short_term_login_token_and_get_user_id(self, login_token):
+        yield self._check_mau_limits()
         auth_api = self.hs.get_auth()
+        user_id = None
         try:
             macaroon = pymacaroons.Macaroon.deserialize(login_token)
             user_id = auth_api.get_user_id_from_macaroon(macaroon)
             auth_api.validate_macaroon(macaroon, "login", True, user_id)
-            return user_id
         except Exception:
             raise AuthError(403, "Invalid token", errcode=Codes.FORBIDDEN)
+        defer.returnValue(user_id)
 
     @defer.inlineCallbacks
     def delete_access_token(self, access_token):
@@ -849,14 +856,19 @@ class AuthHandler(BaseHandler):
         """Computes a secure hash of password.
 
         Args:
-            password (str): Password to hash.
+            password (unicode): Password to hash.
 
         Returns:
-            Deferred(str): Hashed password.
+            Deferred(unicode): Hashed password.
         """
         def _do_hash():
-            return bcrypt.hashpw(password.encode('utf8') + self.hs.config.password_pepper,
-                                 bcrypt.gensalt(self.bcrypt_rounds))
+            # Normalise the Unicode in the password
+            pw = unicodedata.normalize("NFKC", password)
+
+            return bcrypt.hashpw(
+                pw.encode('utf8') + self.hs.config.password_pepper.encode("utf8"),
+                bcrypt.gensalt(self.bcrypt_rounds),
+            ).decode('ascii')
 
         return make_deferred_yieldable(
             threads.deferToThreadPool(
@@ -868,16 +880,19 @@ class AuthHandler(BaseHandler):
         """Validates that self.hash(password) == stored_hash.
 
         Args:
-            password (str): Password to hash.
-            stored_hash (str): Expected hash value.
+            password (unicode): Password to hash.
+            stored_hash (unicode): Expected hash value.
 
         Returns:
             Deferred(bool): Whether self.hash(password) == stored_hash.
         """
 
         def _do_validate_hash():
+            # Normalise the Unicode in the password
+            pw = unicodedata.normalize("NFKC", password)
+
             return bcrypt.checkpw(
-                password.encode('utf8') + self.hs.config.password_pepper,
+                pw.encode('utf8') + self.hs.config.password_pepper.encode("utf8"),
                 stored_hash.encode('utf8')
             )
 
@@ -892,6 +907,19 @@ class AuthHandler(BaseHandler):
         else:
             return defer.succeed(False)
 
+    @defer.inlineCallbacks
+    def _check_mau_limits(self):
+        """
+        Ensure that if mau blocking is enabled that invalid users cannot
+        log in.
+        """
+        if self.hs.config.limit_usage_by_mau is True:
+            current_mau = yield self.store.count_monthly_users()
+            if current_mau >= self.hs.config.max_mau_value:
+                raise AuthError(
+                    403, "MAU Limit Exceeded", errcode=Codes.MAU_LIMIT_EXCEEDED
+                )
+
 
 @attr.s
 class MacaroonGenerator(object):

synapse/handlers/federation.py

@@ -21,8 +21,8 @@ import logging
 import sys
 
 import six
-from six import iteritems
-from six.moves import http_client
+from six import iteritems, itervalues
+from six.moves import http_client, zip
 
 from signedjson.key import decode_verify_key_bytes
 from signedjson.sign import verify_signed_json
@@ -76,7 +76,7 @@ class FederationHandler(BaseHandler):
         self.hs = hs
 
         self.store = hs.get_datastore()
-        self.replication_layer = hs.get_federation_client()
+        self.federation_client = hs.get_federation_client()
         self.state_handler = hs.get_state_handler()
         self.server_name = hs.hostname
         self.keyring = hs.get_keyring()
@@ -255,7 +255,7 @@ class FederationHandler(BaseHandler):
                     # know about
                     for p in prevs - seen:
                         state, got_auth_chain = (
-                            yield self.replication_layer.get_state_for_room(
+                            yield self.federation_client.get_state_for_room(
                                 origin, pdu.room_id, p
                             )
                         )
@@ -338,7 +338,7 @@ class FederationHandler(BaseHandler):
         #
         # see https://github.com/matrix-org/synapse/pull/1744
 
-        missing_events = yield self.replication_layer.get_missing_events(
+        missing_events = yield self.federation_client.get_missing_events(
             origin,
             pdu.room_id,
             earliest_events_ids=list(latest),
@@ -486,7 +486,10 @@ class FederationHandler(BaseHandler):
                 # joined the room. Don't bother if the user is just
                 # changing their profile info.
                 newly_joined = True
-                prev_state_id = context.prev_state_ids.get(
+
+                prev_state_ids = yield context.get_prev_state_ids(self.store)
+
+                prev_state_id = prev_state_ids.get(
                     (event.type, event.state_key)
                 )
                 if prev_state_id:
@@ -519,7 +522,7 @@ class FederationHandler(BaseHandler):
         if dest == self.server_name:
             raise SynapseError(400, "Can't backfill from self.")
 
-        events = yield self.replication_layer.backfill(
+        events = yield self.federation_client.backfill(
             dest,
             room_id,
             limit=limit,
@@ -567,7 +570,7 @@ class FederationHandler(BaseHandler):
         state_events = {}
         events_to_state = {}
         for e_id in edges:
-            state, auth = yield self.replication_layer.get_state_for_room(
+            state, auth = yield self.federation_client.get_state_for_room(
                 destination=dest,
                 room_id=room_id,
                 event_id=e_id
@@ -609,7 +612,7 @@ class FederationHandler(BaseHandler):
                 results = yield logcontext.make_deferred_yieldable(defer.gatherResults(
                     [
                         logcontext.run_in_background(
-                            self.replication_layer.get_pdu,
+                            self.federation_client.get_pdu,
                             [dest],
                             event_id,
                             outlier=True,
@@ -731,7 +734,7 @@ class FederationHandler(BaseHandler):
             """
             joined_users = [
                 (state_key, int(event.depth))
-                for (e_type, state_key), event in state.iteritems()
+                for (e_type, state_key), event in iteritems(state)
                 if e_type == EventTypes.Member
                 and event.membership == Membership.JOIN
             ]
@@ -748,7 +751,7 @@ class FederationHandler(BaseHandler):
                 except Exception:
                     pass
 
-            return sorted(joined_domains.iteritems(), key=lambda d: d[1])
+            return sorted(joined_domains.items(), key=lambda d: d[1])
 
         curr_domains = get_domains_from_state(curr_state)
 
@@ -811,7 +814,7 @@ class FederationHandler(BaseHandler):
         tried_domains = set(likely_domains)
         tried_domains.add(self.server_name)
 
-        event_ids = list(extremities.iterkeys())
+        event_ids = list(extremities.keys())
 
         logger.debug("calling resolve_state_groups in _maybe_backfill")
         resolve = logcontext.preserve_fn(
@@ -827,15 +830,15 @@ class FederationHandler(BaseHandler):
         states = dict(zip(event_ids, [s.state for s in states]))
 
         state_map = yield self.store.get_events(
-            [e_id for ids in states.itervalues() for e_id in ids.itervalues()],
+            [e_id for ids in itervalues(states) for e_id in itervalues(ids)],
             get_prev_content=False
         )
         states = {
             key: {
                 k: state_map[e_id]
-                for k, e_id in state_dict.iteritems()
+                for k, e_id in iteritems(state_dict)
                 if e_id in state_map
-            } for key, state_dict in states.iteritems()
+            } for key, state_dict in iteritems(states)
         }
 
         for e_id, _ in sorted_extremeties_tuple:
@@ -890,7 +893,7 @@ class FederationHandler(BaseHandler):
 
         Invites must be signed by the invitee's server before distribution.
         """
-        pdu = yield self.replication_layer.send_invite(
+        pdu = yield self.federation_client.send_invite(
             destination=target_host,
             room_id=event.room_id,
             event_id=event.event_id,
@@ -906,16 +909,6 @@ class FederationHandler(BaseHandler):
             [auth_id for auth_id, _ in event.auth_events],
             include_given=True
         )
-
-        for event in auth:
-            event.signatures.update(
-                compute_event_signature(
-                    event,
-                    self.hs.hostname,
-                    self.hs.config.signing_key[0]
-                )
-            )
-
         defer.returnValue([e for e in auth])
 
     @log_function
@@ -962,7 +955,7 @@ class FederationHandler(BaseHandler):
                 target_hosts.insert(0, origin)
             except ValueError:
                 pass
-            ret = yield self.replication_layer.send_join(target_hosts, event)
+            ret = yield self.federation_client.send_join(target_hosts, event)
 
             origin = ret["origin"]
             state = ret["state"]
@@ -1116,10 +1109,12 @@ class FederationHandler(BaseHandler):
                 user = UserID.from_string(event.state_key)
                 yield user_joined_room(self.distributor, user, event.room_id)
 
-        state_ids = list(context.prev_state_ids.values())
+        prev_state_ids = yield context.get_prev_state_ids(self.store)
+
+        state_ids = list(prev_state_ids.values())
         auth_chain = yield self.store.get_auth_chain(state_ids)
 
-        state = yield self.store.get_events(list(context.prev_state_ids.values()))
+        state = yield self.store.get_events(list(prev_state_ids.values()))
 
         defer.returnValue({
             "state": list(state.values()),
@@ -1216,7 +1211,7 @@ class FederationHandler(BaseHandler):
         except ValueError:
             pass
 
-        yield self.replication_layer.send_leave(
+        yield self.federation_client.send_leave(
             target_hosts,
             event
         )
@@ -1239,7 +1234,7 @@ class FederationHandler(BaseHandler):
     @defer.inlineCallbacks
     def _make_and_verify_event(self, target_hosts, room_id, user_id, membership,
                                content={},):
-        origin, pdu = yield self.replication_layer.make_membership_event(
+        origin, pdu = yield self.federation_client.make_membership_event(
             target_hosts,
             room_id,
             user_id,
@@ -1284,7 +1279,7 @@ class FederationHandler(BaseHandler):
     @log_function
     def on_make_leave_request(self, room_id, user_id):
         """ We've received a /make_leave/ request, so we create a partial
-        join event for the room and return that. We do *not* persist or
+        leave event for the room and return that. We do *not* persist or
         process it until the other server has signed it and sent it back.
         """
         builder = self.event_builder_factory.new({
@@ -1371,18 +1366,6 @@ class FederationHandler(BaseHandler):
                     del results[(event.type, event.state_key)]
 
             res = list(results.values())
-            for event in res:
-                # We sign these again because there was a bug where we
-                # incorrectly signed things the first time round
-                if self.is_mine_id(event.event_id):
-                    event.signatures.update(
-                        compute_event_signature(
-                            event,
-                            self.hs.hostname,
-                            self.hs.config.signing_key[0]
-                        )
-                    )
-
             defer.returnValue(res)
         else:
             defer.returnValue([])
@@ -1454,18 +1437,6 @@ class FederationHandler(BaseHandler):
         )
 
         if event:
-            if self.is_mine_id(event.event_id):
-                # FIXME: This is a temporary work around where we occasionally
-                # return events slightly differently than when they were
-                # originally signed
-                event.signatures.update(
-                    compute_event_signature(
-                        event,
-                        self.hs.hostname,
-                        self.hs.config.signing_key[0]
-                    )
-                )
-
             in_room = yield self.auth.check_host_in_room(
                 event.room_id,
                 origin
@@ -1549,7 +1520,7 @@ class FederationHandler(BaseHandler):
         yield self.store.persist_events(
             [
                 (ev_info["event"], context)
-                for ev_info, context in itertools.izip(event_infos, contexts)
+                for ev_info, context in zip(event_infos, contexts)
             ],
             backfilled=backfilled,
         )
@@ -1596,7 +1567,7 @@ class FederationHandler(BaseHandler):
                     missing_auth_events.add(e_id)
 
         for e_id in missing_auth_events:
-            m_ev = yield self.replication_layer.get_pdu(
+            m_ev = yield self.federation_client.get_pdu(
                 [origin],
                 e_id,
                 outlier=True,
@@ -1669,8 +1640,9 @@ class FederationHandler(BaseHandler):
         )
 
         if not auth_events:
+            prev_state_ids = yield context.get_prev_state_ids(self.store)
             auth_events_ids = yield self.auth.compute_auth_events(
-                event, context.prev_state_ids, for_verification=True,
+                event, prev_state_ids, for_verification=True,
             )
             auth_events = yield self.store.get_events(auth_events_ids)
             auth_events = {
@@ -1730,15 +1702,6 @@ class FederationHandler(BaseHandler):
             local_auth_chain, remote_auth_chain
         )
 
-        for event in ret["auth_chain"]:
-            event.signatures.update(
-                compute_event_signature(
-                    event,
-                    self.hs.hostname,
-                    self.hs.config.signing_key[0]
-                )
-            )
-
         logger.debug("on_query_auth returning: %s", ret)
 
         defer.returnValue(ret)
@@ -1814,7 +1777,7 @@ class FederationHandler(BaseHandler):
             logger.info("Missing auth: %s", missing_auth)
             # If we don't have all the auth events, we need to get them.
             try:
-                remote_auth_chain = yield self.replication_layer.get_event_auth(
+                remote_auth_chain = yield self.federation_client.get_event_auth(
                     origin, event.room_id, event.event_id
                 )
 
@@ -1919,9 +1882,10 @@ class FederationHandler(BaseHandler):
                         break
 
             if do_resolution:
+                prev_state_ids = yield context.get_prev_state_ids(self.store)
                 # 1. Get what we think is the auth chain.
                 auth_ids = yield self.auth.compute_auth_events(
-                    event, context.prev_state_ids
+                    event, prev_state_ids
                 )
                 local_auth_chain = yield self.store.get_auth_chain(
                     auth_ids, include_given=True
@@ -1929,7 +1893,7 @@ class FederationHandler(BaseHandler):
 
                 try:
                     # 2. Get remote difference.
-                    result = yield self.replication_layer.query_auth(
+                    result = yield self.federation_client.query_auth(
                         origin,
                         event.room_id,
                         event.event_id,
@@ -2011,21 +1975,34 @@ class FederationHandler(BaseHandler):
             k: a.event_id for k, a in iteritems(auth_events)
             if k != event_key
         }
-        context.current_state_ids = dict(context.current_state_ids)
-        context.current_state_ids.update(state_updates)
-        if context.delta_ids is not None:
-            context.delta_ids = dict(context.delta_ids)
-            context.delta_ids.update(state_updates)
-        context.prev_state_ids = dict(context.prev_state_ids)
-        context.prev_state_ids.update({
+        current_state_ids = yield context.get_current_state_ids(self.store)
+        current_state_ids = dict(current_state_ids)
+
+        current_state_ids.update(state_updates)
+
+        prev_state_ids = yield context.get_prev_state_ids(self.store)
+        prev_state_ids = dict(prev_state_ids)
+
+        prev_state_ids.update({
             k: a.event_id for k, a in iteritems(auth_events)
         })
-        context.state_group = yield self.store.store_state_group(
+
+        # create a new state group as a delta from the existing one.
+        prev_group = context.state_group
+        state_group = yield self.store.store_state_group(
             event.event_id,
             event.room_id,
-            prev_group=context.prev_group,
-            delta_ids=context.delta_ids,
-            current_state_ids=context.current_state_ids,
+            prev_group=prev_group,
+            delta_ids=state_updates,
+            current_state_ids=current_state_ids,
+        )
+
+        yield context.update_state(
+            state_group=state_group,
+            current_state_ids=current_state_ids,
+            prev_state_ids=prev_state_ids,
+            prev_group=prev_group,
+            delta_ids=state_updates,
         )
 
     @defer.inlineCallbacks
@@ -2215,7 +2192,7 @@ class FederationHandler(BaseHandler):
             yield member_handler.send_membership_event(None, event, context)
         else:
             destinations = set(x.split(":", 1)[-1] for x in (sender_user_id, room_id))
-            yield self.replication_layer.forward_third_party_invite(
+            yield self.federation_client.forward_third_party_invite(
                 destinations,
                 room_id,
                 event_dict,
@@ -2265,7 +2242,8 @@ class FederationHandler(BaseHandler):
             event.content["third_party_invite"]["signed"]["token"]
         )
         original_invite = None
-        original_invite_id = context.prev_state_ids.get(key)
+        prev_state_ids = yield context.get_prev_state_ids(self.store)
+        original_invite_id = prev_state_ids.get(key)
         if original_invite_id:
             original_invite = yield self.store.get_event(
                 original_invite_id, allow_none=True
@@ -2307,7 +2285,8 @@ class FederationHandler(BaseHandler):
         signed = event.content["third_party_invite"]["signed"]
         token = signed["token"]
 
-        invite_event_id = context.prev_state_ids.get(
+        prev_state_ids = yield context.get_prev_state_ids(self.store)
+        invite_event_id = prev_state_ids.get(
             (EventTypes.ThirdPartyInvite, token,)
         )
 

synapse/handlers/initial_sync.py

@@ -148,13 +148,15 @@ class InitialSyncHandler(BaseHandler):
             try:
                 if event.membership == Membership.JOIN:
                     room_end_token = now_token.room_key
-                    deferred_room_state = self.state_handler.get_current_state(
-                        event.room_id
+                    deferred_room_state = run_in_background(
+                        self.state_handler.get_current_state,
+                        event.room_id,
                     )
                 elif event.membership == Membership.LEAVE:
                     room_end_token = "s%d" % (event.stream_ordering,)
-                    deferred_room_state = self.store.get_state_for_events(
-                        [event.event_id], None
+                    deferred_room_state = run_in_background(
+                        self.store.get_state_for_events,
+                        [event.event_id], None,
                     )
                     deferred_room_state.addCallback(
                         lambda states: states[event.event_id]
@@ -387,19 +389,21 @@ class InitialSyncHandler(BaseHandler):
                 receipts = []
             defer.returnValue(receipts)
 
-        presence, receipts, (messages, token) = yield defer.gatherResults(
-            [
-                run_in_background(get_presence),
-                run_in_background(get_receipts),
-                run_in_background(
-                    self.store.get_recent_events_for_room,
-                    room_id,
-                    limit=limit,
-                    end_token=now_token.room_key,
-                )
-            ],
-            consumeErrors=True,
-        ).addErrback(unwrapFirstError)
+        presence, receipts, (messages, token) = yield make_deferred_yieldable(
+            defer.gatherResults(
+                [
+                    run_in_background(get_presence),
+                    run_in_background(get_receipts),
+                    run_in_background(
+                        self.store.get_recent_events_for_room,
+                        room_id,
+                        limit=limit,
+                        end_token=now_token.room_key,
+                    )
+                ],
+                consumeErrors=True,
+            ).addErrback(unwrapFirstError),
+        )
 
         messages = yield filter_events_for_client(
             self.store, user_id, messages, is_peeking=is_peeking,

synapse/handlers/message.py

@@ -23,7 +23,6 @@ from canonicaljson import encode_canonical_json, json
 
 from twisted.internet import defer
 from twisted.internet.defer import succeed
-from twisted.python.failure import Failure
 
 from synapse.api.constants import MAX_DEPTH, EventTypes, Membership
 from synapse.api.errors import AuthError, Codes, ConsentNotGivenError, SynapseError
@@ -32,247 +31,26 @@ from synapse.crypto.event_signing import add_hashes_and_signatures
 from synapse.events.utils import serialize_event
 from synapse.events.validator import EventValidator
 from synapse.replication.http.send_event import send_event_to_master
-from synapse.types import RoomAlias, RoomStreamToken, UserID
-from synapse.util.async import Limiter, ReadWriteLock
+from synapse.types import RoomAlias, UserID
+from synapse.util.async import Linearizer
 from synapse.util.frozenutils import frozendict_json_encoder
 from synapse.util.logcontext import run_in_background
 from synapse.util.metrics import measure_func
-from synapse.util.stringutils import random_string
-from synapse.visibility import filter_events_for_client
 
 from ._base import BaseHandler
 
 logger = logging.getLogger(__name__)
 
 
-class PurgeStatus(object):
-    """Object tracking the status of a purge request
-
-    This class contains information on the progress of a purge request, for
-    return by get_purge_status.
-
-    Attributes:
-        status (int): Tracks whether this request has completed. One of
-            STATUS_{ACTIVE,COMPLETE,FAILED}
+class MessageHandler(object):
+    """Contains some read only APIs to get state about a room
     """
 
-    STATUS_ACTIVE = 0
-    STATUS_COMPLETE = 1
-    STATUS_FAILED = 2
-
-    STATUS_TEXT = {
-        STATUS_ACTIVE: "active",
-        STATUS_COMPLETE: "complete",
-        STATUS_FAILED: "failed",
-    }
-
-    def __init__(self):
-        self.status = PurgeStatus.STATUS_ACTIVE
-
-    def asdict(self):
-        return {
-            "status": PurgeStatus.STATUS_TEXT[self.status]
-        }
-
-
-class MessageHandler(BaseHandler):
-
     def __init__(self, hs):
-        super(MessageHandler, self).__init__(hs)
-        self.hs = hs
-        self.state = hs.get_state_handler()
+        self.auth = hs.get_auth()
         self.clock = hs.get_clock()
-
-        self.pagination_lock = ReadWriteLock()
-        self._purges_in_progress_by_room = set()
-        # map from purge id to PurgeStatus
-        self._purges_by_id = {}
-
-    def start_purge_history(self, room_id, token,
-                            delete_local_events=False):
-        """Start off a history purge on a room.
-
-        Args:
-            room_id (str): The room to purge from
-
-            token (str): topological token to delete events before
-            delete_local_events (bool): True to delete local events as well as
-                remote ones
-
-        Returns:
-            str: unique ID for this purge transaction.
-        """
-        if room_id in self._purges_in_progress_by_room:
-            raise SynapseError(
-                400,
-                "History purge already in progress for %s" % (room_id, ),
-            )
-
-        purge_id = random_string(16)
-
-        # we log the purge_id here so that it can be tied back to the
-        # request id in the log lines.
-        logger.info("[purge] starting purge_id %s", purge_id)
-
-        self._purges_by_id[purge_id] = PurgeStatus()
-        run_in_background(
-            self._purge_history,
-            purge_id, room_id, token, delete_local_events,
-        )
-        return purge_id
-
-    @defer.inlineCallbacks
-    def _purge_history(self, purge_id, room_id, token,
-                       delete_local_events):
-        """Carry out a history purge on a room.
-
-        Args:
-            purge_id (str): The id for this purge
-            room_id (str): The room to purge from
-            token (str): topological token to delete events before
-            delete_local_events (bool): True to delete local events as well as
-                remote ones
-
-        Returns:
-            Deferred
-        """
-        self._purges_in_progress_by_room.add(room_id)
-        try:
-            with (yield self.pagination_lock.write(room_id)):
-                yield self.store.purge_history(
-                    room_id, token, delete_local_events,
-                )
-            logger.info("[purge] complete")
-            self._purges_by_id[purge_id].status = PurgeStatus.STATUS_COMPLETE
-        except Exception:
-            logger.error("[purge] failed: %s", Failure().getTraceback().rstrip())
-            self._purges_by_id[purge_id].status = PurgeStatus.STATUS_FAILED
-        finally:
-            self._purges_in_progress_by_room.discard(room_id)
-
-            # remove the purge from the list 24 hours after it completes
-            def clear_purge():
-                del self._purges_by_id[purge_id]
-            self.hs.get_reactor().callLater(24 * 3600, clear_purge)
-
-    def get_purge_status(self, purge_id):
-        """Get the current status of an active purge
-
-        Args:
-            purge_id (str): purge_id returned by start_purge_history
-
-        Returns:
-            PurgeStatus|None
-        """
-        return self._purges_by_id.get(purge_id)
-
-    @defer.inlineCallbacks
-    def get_messages(self, requester, room_id=None, pagin_config=None,
-                     as_client_event=True, event_filter=None):
-        """Get messages in a room.
-
-        Args:
-            requester (Requester): The user requesting messages.
-            room_id (str): The room they want messages from.
-            pagin_config (synapse.api.streams.PaginationConfig): The pagination
-                config rules to apply, if any.
-            as_client_event (bool): True to get events in client-server format.
-            event_filter (Filter): Filter to apply to results or None
-        Returns:
-            dict: Pagination API results
-        """
-        user_id = requester.user.to_string()
-
-        if pagin_config.from_token:
-            room_token = pagin_config.from_token.room_key
-        else:
-            pagin_config.from_token = (
-                yield self.hs.get_event_sources().get_current_token_for_room(
-                    room_id=room_id
-                )
-            )
-            room_token = pagin_config.from_token.room_key
-
-        room_token = RoomStreamToken.parse(room_token)
-
-        pagin_config.from_token = pagin_config.from_token.copy_and_replace(
-            "room_key", str(room_token)
-        )
-
-        source_config = pagin_config.get_source_config("room")
-
-        with (yield self.pagination_lock.read(room_id)):
-            membership, member_event_id = yield self._check_in_room_or_world_readable(
-                room_id, user_id
-            )
-
-            if source_config.direction == 'b':
-                # if we're going backwards, we might need to backfill. This
-                # requires that we have a topo token.
-                if room_token.topological:
-                    max_topo = room_token.topological
-                else:
-                    max_topo = yield self.store.get_max_topological_token(
-                        room_id, room_token.stream
-                    )
-
-                if membership == Membership.LEAVE:
-                    # If they have left the room then clamp the token to be before
-                    # they left the room, to save the effort of loading from the
-                    # database.
-                    leave_token = yield self.store.get_topological_token_for_event(
-                        member_event_id
-                    )
-                    leave_token = RoomStreamToken.parse(leave_token)
-                    if leave_token.topological < max_topo:
-                        source_config.from_key = str(leave_token)
-
-                yield self.hs.get_handlers().federation_handler.maybe_backfill(
-                    room_id, max_topo
-                )
-
-            events, next_key = yield self.store.paginate_room_events(
-                room_id=room_id,
-                from_key=source_config.from_key,
-                to_key=source_config.to_key,
-                direction=source_config.direction,
-                limit=source_config.limit,
-                event_filter=event_filter,
-            )
-
-            next_token = pagin_config.from_token.copy_and_replace(
-                "room_key", next_key
-            )
-
-        if not events:
-            defer.returnValue({
-                "chunk": [],
-                "start": pagin_config.from_token.to_string(),
-                "end": next_token.to_string(),
-            })
-
-        if event_filter:
-            events = event_filter.filter(events)
-
-        events = yield filter_events_for_client(
-            self.store,
-            user_id,
-            events,
-            is_peeking=(member_event_id is None),
-        )
-
-        time_now = self.clock.time_msec()
-
-        chunk = {
-            "chunk": [
-                serialize_event(e, time_now, as_client_event)
-                for e in events
-            ],
-            "start": pagin_config.from_token.to_string(),
-            "end": next_token.to_string(),
-        }
-
-        defer.returnValue(chunk)
+        self.state = hs.get_state_handler()
+        self.store = hs.get_datastore()
 
     @defer.inlineCallbacks
     def get_room_data(self, user_id=None, room_id=None,
@@ -286,12 +64,12 @@ class MessageHandler(BaseHandler):
         Raises:
             SynapseError if something went wrong.
         """
-        membership, membership_event_id = yield self._check_in_room_or_world_readable(
+        membership, membership_event_id = yield self.auth.check_in_room_or_world_readable(
             room_id, user_id
         )
 
         if membership == Membership.JOIN:
-            data = yield self.state_handler.get_current_state(
+            data = yield self.state.get_current_state(
                 room_id, event_type, state_key
             )
         elif membership == Membership.LEAVE:
@@ -303,31 +81,6 @@ class MessageHandler(BaseHandler):
 
         defer.returnValue(data)
 
-    @defer.inlineCallbacks
-    def _check_in_room_or_world_readable(self, room_id, user_id):
-        try:
-            # check_user_was_in_room will return the most recent membership
-            # event for the user if:
-            #  * The user is a non-guest user, and was ever in the room
-            #  * The user is a guest user, and has joined the room
-            # else it will throw.
-            member_event = yield self.auth.check_user_was_in_room(room_id, user_id)
-            defer.returnValue((member_event.membership, member_event.event_id))
-            return
-        except AuthError:
-            visibility = yield self.state_handler.get_current_state(
-                room_id, EventTypes.RoomHistoryVisibility, ""
-            )
-            if (
-                visibility and
-                visibility.content["history_visibility"] == "world_readable"
-            ):
-                defer.returnValue((Membership.JOIN, None))
-                return
-            raise AuthError(
-                403, "Guest access not allowed", errcode=Codes.GUEST_ACCESS_FORBIDDEN
-            )
-
     @defer.inlineCallbacks
     def get_state_events(self, user_id, room_id, is_guest=False):
         """Retrieve all state events for a given room. If the user is
@@ -340,12 +93,12 @@ class MessageHandler(BaseHandler):
         Returns:
             A list of dicts representing state events. [{}, {}, {}]
         """
-        membership, membership_event_id = yield self._check_in_room_or_world_readable(
+        membership, membership_event_id = yield self.auth.check_in_room_or_world_readable(
             room_id, user_id
         )
 
         if membership == Membership.JOIN:
-            room_state = yield self.state_handler.get_current_state(room_id)
+            room_state = yield self.state.get_current_state(room_id)
         elif membership == Membership.LEAVE:
             room_state = yield self.store.get_state_for_events(
                 [membership_event_id], None
@@ -373,7 +126,7 @@ class MessageHandler(BaseHandler):
         if not requester.app_service:
             # We check AS auth after fetching the room membership, as it
             # requires us to pull out all joined members anyway.
-            membership, _ = yield self._check_in_room_or_world_readable(
+            membership, _ = yield self.auth.check_in_room_or_world_readable(
                 room_id, user_id
             )
             if membership != Membership.JOIN:
@@ -427,7 +180,7 @@ class EventCreationHandler(object):
 
         # We arbitrarily limit concurrent event creation for a room to 5.
         # This is to stop us from diverging history *too* much.
-        self.limiter = Limiter(max_count=5)
+        self.limiter = Linearizer(max_count=5, name="room_event_creation_limit")
 
         self.action_generator = hs.get_action_generator()
 
@@ -630,7 +383,8 @@ class EventCreationHandler(object):
         If so, returns the version of the event in context.
         Otherwise, returns None.
         """
-        prev_event_id = context.prev_state_ids.get((event.type, event.state_key))
+        prev_state_ids = yield context.get_prev_state_ids(self.store)
+        prev_event_id = prev_state_ids.get((event.type, event.state_key))
         prev_event = yield self.store.get_event(prev_event_id, allow_none=True)
         if not prev_event:
             return
@@ -752,8 +506,8 @@ class EventCreationHandler(object):
         event = builder.build()
 
         logger.debug(
-            "Created event %s with state: %s",
-            event.event_id, context.prev_state_ids,
+            "Created event %s",
+            event.event_id,
         )
 
         defer.returnValue(
@@ -806,8 +560,9 @@ class EventCreationHandler(object):
             # If we're a worker we need to hit out to the master.
             if self.config.worker_app:
                 yield send_event_to_master(
-                    self.hs.get_clock(),
-                    self.http_client,
+                    clock=self.hs.get_clock(),
+                    store=self.store,
+                    client=self.http_client,
                     host=self.config.worker_replication_host,
                     port=self.config.worker_replication_http_port,
                     requester=requester,
@@ -884,9 +639,11 @@ class EventCreationHandler(object):
                         e.sender == event.sender
                     )
 
+                current_state_ids = yield context.get_current_state_ids(self.store)
+
                 state_to_include_ids = [
                     e_id
-                    for k, e_id in iteritems(context.current_state_ids)
+                    for k, e_id in iteritems(current_state_ids)
                     if k[0] in self.hs.config.room_invite_state_types
                     or k == (EventTypes.Member, event.sender)
                 ]
@@ -922,8 +679,9 @@ class EventCreationHandler(object):
                     )
 
         if event.type == EventTypes.Redaction:
+            prev_state_ids = yield context.get_prev_state_ids(self.store)
             auth_events_ids = yield self.auth.compute_auth_events(
-                event, context.prev_state_ids, for_verification=True,
+                event, prev_state_ids, for_verification=True,
             )
             auth_events = yield self.store.get_events(auth_events_ids)
             auth_events = {
@@ -943,11 +701,13 @@ class EventCreationHandler(object):
                         "You don't have permission to redact events"
                     )
 
-        if event.type == EventTypes.Create and context.prev_state_ids:
-            raise AuthError(
-                403,
-                "Changing the room create event is forbidden",
-            )
+        if event.type == EventTypes.Create:
+            prev_state_ids = yield context.get_prev_state_ids(self.store)
+            if prev_state_ids:
+                raise AuthError(
+                    403,
+                    "Changing the room create event is forbidden",
+                )
 
         (event_stream_id, max_stream_id) = yield self.store.persist_event(
             event, context=context

synapse/handlers/pagination.py

@@ -0,0 +1,265 @@
+# -*- coding: utf-8 -*-
+# Copyright 2014 - 2016 OpenMarket Ltd
+# Copyright 2017 - 2018 New Vector Ltd
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+import logging
+
+from twisted.internet import defer
+from twisted.python.failure import Failure
+
+from synapse.api.constants import Membership
+from synapse.api.errors import SynapseError
+from synapse.events.utils import serialize_event
+from synapse.types import RoomStreamToken
+from synapse.util.async import ReadWriteLock
+from synapse.util.logcontext import run_in_background
+from synapse.util.stringutils import random_string
+from synapse.visibility import filter_events_for_client
+
+logger = logging.getLogger(__name__)
+
+
+class PurgeStatus(object):
+    """Object tracking the status of a purge request
+
+    This class contains information on the progress of a purge request, for
+    return by get_purge_status.
+
+    Attributes:
+        status (int): Tracks whether this request has completed. One of
+            STATUS_{ACTIVE,COMPLETE,FAILED}
+    """
+
+    STATUS_ACTIVE = 0
+    STATUS_COMPLETE = 1
+    STATUS_FAILED = 2
+
+    STATUS_TEXT = {
+        STATUS_ACTIVE: "active",
+        STATUS_COMPLETE: "complete",
+        STATUS_FAILED: "failed",
+    }
+
+    def __init__(self):
+        self.status = PurgeStatus.STATUS_ACTIVE
+
+    def asdict(self):
+        return {
+            "status": PurgeStatus.STATUS_TEXT[self.status]
+        }
+
+
+class PaginationHandler(object):
+    """Handles pagination and purge history requests.
+
+    These are in the same handler due to the fact we need to block clients
+    paginating during a purge.
+    """
+
+    def __init__(self, hs):
+        self.hs = hs
+        self.auth = hs.get_auth()
+        self.store = hs.get_datastore()
+        self.clock = hs.get_clock()
+
+        self.pagination_lock = ReadWriteLock()
+        self._purges_in_progress_by_room = set()
+        # map from purge id to PurgeStatus
+        self._purges_by_id = {}
+
+    def start_purge_history(self, room_id, token,
+                            delete_local_events=False):
+        """Start off a history purge on a room.
+
+        Args:
+            room_id (str): The room to purge from
+
+            token (str): topological token to delete events before
+            delete_local_events (bool): True to delete local events as well as
+                remote ones
+
+        Returns:
+            str: unique ID for this purge transaction.
+        """
+        if room_id in self._purges_in_progress_by_room:
+            raise SynapseError(
+                400,
+                "History purge already in progress for %s" % (room_id, ),
+            )
+
+        purge_id = random_string(16)
+
+        # we log the purge_id here so that it can be tied back to the
+        # request id in the log lines.
+        logger.info("[purge] starting purge_id %s", purge_id)
+
+        self._purges_by_id[purge_id] = PurgeStatus()
+        run_in_background(
+            self._purge_history,
+            purge_id, room_id, token, delete_local_events,
+        )
+        return purge_id
+
+    @defer.inlineCallbacks
+    def _purge_history(self, purge_id, room_id, token,
+                       delete_local_events):
+        """Carry out a history purge on a room.
+
+        Args:
+            purge_id (str): The id for this purge
+            room_id (str): The room to purge from
+            token (str): topological token to delete events before
+            delete_local_events (bool): True to delete local events as well as
+                remote ones
+
+        Returns:
+            Deferred
+        """
+        self._purges_in_progress_by_room.add(room_id)
+        try:
+            with (yield self.pagination_lock.write(room_id)):
+                yield self.store.purge_history(
+                    room_id, token, delete_local_events,
+                )
+            logger.info("[purge] complete")
+            self._purges_by_id[purge_id].status = PurgeStatus.STATUS_COMPLETE
+        except Exception:
+            logger.error("[purge] failed: %s", Failure().getTraceback().rstrip())
+            self._purges_by_id[purge_id].status = PurgeStatus.STATUS_FAILED
+        finally:
+            self._purges_in_progress_by_room.discard(room_id)
+
+            # remove the purge from the list 24 hours after it completes
+            def clear_purge():
+                del self._purges_by_id[purge_id]
+            self.hs.get_reactor().callLater(24 * 3600, clear_purge)
+
+    def get_purge_status(self, purge_id):
+        """Get the current status of an active purge
+
+        Args:
+            purge_id (str): purge_id returned by start_purge_history
+
+        Returns:
+            PurgeStatus|None
+        """
+        return self._purges_by_id.get(purge_id)
+
+    @defer.inlineCallbacks
+    def get_messages(self, requester, room_id=None, pagin_config=None,
+                     as_client_event=True, event_filter=None):
+        """Get messages in a room.
+
+        Args:
+            requester (Requester): The user requesting messages.
+            room_id (str): The room they want messages from.
+            pagin_config (synapse.api.streams.PaginationConfig): The pagination
+                config rules to apply, if any.
+            as_client_event (bool): True to get events in client-server format.
+            event_filter (Filter): Filter to apply to results or None
+        Returns:
+            dict: Pagination API results
+        """
+        user_id = requester.user.to_string()
+
+        if pagin_config.from_token:
+            room_token = pagin_config.from_token.room_key
+        else:
+            pagin_config.from_token = (
+                yield self.hs.get_event_sources().get_current_token_for_room(
+                    room_id=room_id
+                )
+            )
+            room_token = pagin_config.from_token.room_key
+
+        room_token = RoomStreamToken.parse(room_token)
+
+        pagin_config.from_token = pagin_config.from_token.copy_and_replace(
+            "room_key", str(room_token)
+        )
+
+        source_config = pagin_config.get_source_config("room")
+
+        with (yield self.pagination_lock.read(room_id)):
+            membership, member_event_id = yield self.auth.check_in_room_or_world_readable(
+                room_id, user_id
+            )
+
+            if source_config.direction == 'b':
+                # if we're going backwards, we might need to backfill. This
+                # requires that we have a topo token.
+                if room_token.topological:
+                    max_topo = room_token.topological
+                else:
+                    max_topo = yield self.store.get_max_topological_token(
+                        room_id, room_token.stream
+                    )
+
+                if membership == Membership.LEAVE:
+                    # If they have left the room then clamp the token to be before
+                    # they left the room, to save the effort of loading from the
+                    # database.
+                    leave_token = yield self.store.get_topological_token_for_event(
+                        member_event_id
+                    )
+                    leave_token = RoomStreamToken.parse(leave_token)
+                    if leave_token.topological < max_topo:
+                        source_config.from_key = str(leave_token)
+
+                yield self.hs.get_handlers().federation_handler.maybe_backfill(
+                    room_id, max_topo
+                )
+
+            events, next_key = yield self.store.paginate_room_events(
+                room_id=room_id,
+                from_key=source_config.from_key,
+                to_key=source_config.to_key,
+                direction=source_config.direction,
+                limit=source_config.limit,
+                event_filter=event_filter,
+            )
+
+            next_token = pagin_config.from_token.copy_and_replace(
+                "room_key", next_key
+            )
+
+        if not events:
+            defer.returnValue({
+                "chunk": [],
+                "start": pagin_config.from_token.to_string(),
+                "end": next_token.to_string(),
+            })
+
+        if event_filter:
+            events = event_filter.filter(events)
+
+        events = yield filter_events_for_client(
+            self.store,
+            user_id,
+            events,
+            is_peeking=(member_event_id is None),
+        )
+
+        time_now = self.clock.time_msec()
+
+        chunk = {
+            "chunk": [
+                serialize_event(e, time_now, as_client_event)
+                for e in events
+            ],
+            "start": pagin_config.from_token.to_string(),
+            "end": next_token.to_string(),
+        }
+
+        defer.returnValue(chunk)

synapse/handlers/profile.py

@@ -18,6 +18,7 @@ import logging
 from twisted.internet import defer
 
 from synapse.api.errors import AuthError, CodeMessageException, SynapseError
+from synapse.metrics.background_process_metrics import run_as_background_process
 from synapse.types import UserID, get_domain_from_id
 
 from ._base import BaseHandler
@@ -41,7 +42,7 @@ class ProfileHandler(BaseHandler):
 
         if hs.config.worker_app is None:
             self.clock.looping_call(
-                self._update_remote_profile_cache, self.PROFILE_UPDATE_MS,
+                self._start_update_remote_profile_cache, self.PROFILE_UPDATE_MS,
             )
 
     @defer.inlineCallbacks
@@ -254,6 +255,12 @@ class ProfileHandler(BaseHandler):
                     room_id, str(e.message)
                 )
 
+    def _start_update_remote_profile_cache(self):
+        return run_as_background_process(
+            "Update remote profile", self._update_remote_profile_cache,
+        )
+
+    @defer.inlineCallbacks
     def _update_remote_profile_cache(self):
         """Called periodically to check profiles of remote users we haven't
         checked in a while.

synapse/handlers/register.py

@@ -45,7 +45,7 @@ class RegistrationHandler(BaseHandler):
             hs (synapse.server.HomeServer):
         """
         super(RegistrationHandler, self).__init__(hs)
-
+        self.hs = hs
         self.auth = hs.get_auth()
         self._auth_handler = hs.get_auth_handler()
         self.profile_handler = hs.get_profile_handler()
@@ -131,7 +131,7 @@ class RegistrationHandler(BaseHandler):
         Args:
             localpart : The local part of the user ID to register. If None,
               one will be generated.
-            password (str) : The password to assign to this user so they can
+            password (unicode) : The password to assign to this user so they can
               login again. This can be None which means they cannot login again
               via a password (e.g. the user is an application service user).
             generate_token (bool): Whether a new access token should be
@@ -144,6 +144,7 @@ class RegistrationHandler(BaseHandler):
         Raises:
             RegistrationError if there was a problem registering.
         """
+        yield self._check_mau_limits()
         password_hash = None
         if password:
             password_hash = yield self.auth_handler().hash(password)
@@ -288,6 +289,7 @@ class RegistrationHandler(BaseHandler):
                 400,
                 "User ID can only contain characters a-z, 0-9, or '=_-./'",
             )
+        yield self._check_mau_limits()
         user = UserID(localpart, self.hs.hostname)
         user_id = user.to_string()
 
@@ -437,7 +439,7 @@ class RegistrationHandler(BaseHandler):
         """
         if localpart is None:
             raise SynapseError(400, "Request must include user id")
-
+        yield self._check_mau_limits()
         need_register = True
 
         try:
@@ -531,3 +533,16 @@ class RegistrationHandler(BaseHandler):
             remote_room_hosts=remote_room_hosts,
             action="join",
         )
+
+    @defer.inlineCallbacks
+    def _check_mau_limits(self):
+        """
+        Do not accept registrations if monthly active user limits exceeded
+         and limiting is enabled
+        """
+        if self.hs.config.limit_usage_by_mau is True:
+            current_mau = yield self.store.count_monthly_users()
+            if current_mau >= self.hs.config.max_mau_value:
+                raise RegistrationError(
+                    403, "MAU Limit Exceeded", Codes.MAU_LIMIT_EXCEEDED
+                )

synapse/handlers/room.py

@@ -15,6 +15,7 @@
 # limitations under the License.
 
 """Contains functions for performing events on rooms."""
+import itertools
 import logging
 import math
 import string
@@ -24,7 +25,7 @@ from twisted.internet import defer
 
 from synapse.api.constants import EventTypes, JoinRules, RoomCreationPreset
 from synapse.api.errors import AuthError, Codes, StoreError, SynapseError
-from synapse.types import RoomAlias, RoomID, RoomStreamToken, UserID
+from synapse.types import RoomAlias, RoomID, RoomStreamToken, StreamToken, UserID
 from synapse.util import stringutils
 from synapse.visibility import filter_events_for_client
 
@@ -395,9 +396,13 @@ class RoomCreationHandler(BaseHandler):
             )
 
 
-class RoomContextHandler(BaseHandler):
+class RoomContextHandler(object):
+    def __init__(self, hs):
+        self.hs = hs
+        self.store = hs.get_datastore()
+
     @defer.inlineCallbacks
-    def get_event_context(self, user, room_id, event_id, limit):
+    def get_event_context(self, user, room_id, event_id, limit, event_filter):
         """Retrieves events, pagination tokens and state around a given event
         in a room.
 
@@ -407,6 +412,8 @@ class RoomContextHandler(BaseHandler):
             event_id (str)
             limit (int): The maximum number of events to return in total
                 (excluding state).
+            event_filter (Filter|None): the filter to apply to the events returned
+                (excluding the target event_id)
 
         Returns:
             dict, or None if the event isn't found
@@ -414,8 +421,6 @@ class RoomContextHandler(BaseHandler):
         before_limit = math.floor(limit / 2.)
         after_limit = limit - before_limit
 
-        now_token = yield self.hs.get_event_sources().get_current_token()
-
         users = yield self.store.get_users_in_room(room_id)
         is_peeking = user.to_string() not in users
 
@@ -441,7 +446,7 @@ class RoomContextHandler(BaseHandler):
             )
 
         results = yield self.store.get_events_around(
-            room_id, event_id, before_limit, after_limit
+            room_id, event_id, before_limit, after_limit, event_filter
         )
 
         results["events_before"] = yield filter_evts(results["events_before"])
@@ -453,16 +458,35 @@ class RoomContextHandler(BaseHandler):
         else:
             last_event_id = event_id
 
+        types = None
+        filtered_types = None
+        if event_filter and event_filter.lazy_load_members():
+            members = set(ev.sender for ev in itertools.chain(
+                results["events_before"],
+                (results["event"],),
+                results["events_after"],
+            ))
+            filtered_types = [EventTypes.Member]
+            types = [(EventTypes.Member, member) for member in members]
+
+        # XXX: why do we return the state as of the last event rather than the
+        # first? Shouldn't we be consistent with /sync?
+        # https://github.com/matrix-org/matrix-doc/issues/687
+
         state = yield self.store.get_state_for_events(
-            [last_event_id], None
+            [last_event_id], types, filtered_types=filtered_types,
         )
         results["state"] = list(state[last_event_id].values())
 
-        results["start"] = now_token.copy_and_replace(
+        # We use a dummy token here as we only care about the room portion of
+        # the token, which we replace.
+        token = StreamToken.START
+
+        results["start"] = token.copy_and_replace(
             "room_key", results["start"]
         ).to_string()
 
-        results["end"] = now_token.copy_and_replace(
+        results["end"] = token.copy_and_replace(
             "room_key", results["end"]
         ).to_string()
 

synapse/handlers/room_member.py

@@ -201,7 +201,9 @@ class RoomMemberHandler(object):
             ratelimit=ratelimit,
         )
 
-        prev_member_event_id = context.prev_state_ids.get(
+        prev_state_ids = yield context.get_prev_state_ids(self.store)
+
+        prev_member_event_id = prev_state_ids.get(
             (EventTypes.Member, target.to_string()),
             None
         )
@@ -496,9 +498,10 @@ class RoomMemberHandler(object):
         if prev_event is not None:
             return
 
+        prev_state_ids = yield context.get_prev_state_ids(self.store)
         if event.membership == Membership.JOIN:
             if requester.is_guest:
-                guest_can_join = yield self._can_guest_join(context.prev_state_ids)
+                guest_can_join = yield self._can_guest_join(prev_state_ids)
                 if not guest_can_join:
                     # This should be an auth check, but guests are a local concept,
                     # so don't really fit into the general auth process.
@@ -517,7 +520,7 @@ class RoomMemberHandler(object):
             ratelimit=ratelimit,
         )
 
-        prev_member_event_id = context.prev_state_ids.get(
+        prev_member_event_id = prev_state_ids.get(
             (EventTypes.Member, event.state_key),
             None
         )
@@ -705,6 +708,10 @@ class RoomMemberHandler(object):
             inviter_display_name = member_event.content.get("displayname", "")
             inviter_avatar_url = member_event.content.get("avatar_url", "")
 
+        # if user has no display name, default to their MXID
+        if not inviter_display_name:
+            inviter_display_name = user.to_string()
+
         canonical_room_alias = ""
         canonical_alias_event = room_state.get((EventTypes.CanonicalAlias, ""))
         if canonical_alias_event:

synapse/handlers/search.py

@@ -287,7 +287,7 @@ class SearchHandler(BaseHandler):
             contexts = {}
             for event in allowed_events:
                 res = yield self.store.get_events_around(
-                    event.room_id, event.event_id, before_limit, after_limit
+                    event.room_id, event.event_id, before_limit, after_limit,
                 )
 
                 logger.info(

synapse/handlers/sync.py

@@ -1,5 +1,6 @@
 # -*- coding: utf-8 -*-
-# Copyright 2015 - 2016 OpenMarket Ltd
+# Copyright 2015, 2016 OpenMarket Ltd
+# Copyright 2018 New Vector Ltd
 #
 # Licensed under the Apache License, Version 2.0 (the "License");
 # you may not use this file except in compliance with the License.
@@ -25,6 +26,8 @@ from synapse.api.constants import EventTypes, Membership
 from synapse.push.clientformat import format_push_rules_for_user
 from synapse.types import RoomStreamToken
 from synapse.util.async import concurrently_execute
+from synapse.util.caches.expiringcache import ExpiringCache
+from synapse.util.caches.lrucache import LruCache
 from synapse.util.caches.response_cache import ResponseCache
 from synapse.util.logcontext import LoggingContext
 from synapse.util.metrics import Measure, measure_func
@@ -32,6 +35,14 @@ from synapse.visibility import filter_events_for_client
 
 logger = logging.getLogger(__name__)
 
+# Store the cache that tracks which lazy-loaded members have been sent to a given
+# client for no more than 30 minutes.
+LAZY_LOADED_MEMBERS_CACHE_MAX_AGE = 30 * 60 * 1000
+
+# Remember the last 100 members we sent to a client for the purposes of
+# avoiding redundantly sending the same lazy-loaded members to the client
+LAZY_LOADED_MEMBERS_CACHE_MAX_SIZE = 100
+
 
 SyncConfig = collections.namedtuple("SyncConfig", [
     "user",
@@ -181,6 +192,12 @@ class SyncHandler(object):
         self.response_cache = ResponseCache(hs, "sync")
         self.state = hs.get_state_handler()
 
+        # ExpiringCache((User, Device)) -> LruCache(state_key => event_id)
+        self.lazy_loaded_members_cache = ExpiringCache(
+            "lazy_loaded_members_cache", self.clock,
+            max_len=0, expiry_ms=LAZY_LOADED_MEMBERS_CACHE_MAX_AGE,
+        )
+
     def wait_for_sync_for_user(self, sync_config, since_token=None, timeout=0,
                                full_state=False):
         """Get the sync for a client if we have new data for it now. Otherwise
@@ -416,29 +433,44 @@ class SyncHandler(object):
         ))
 
     @defer.inlineCallbacks
-    def get_state_after_event(self, event):
+    def get_state_after_event(self, event, types=None, filtered_types=None):
         """
         Get the room state after the given event
 
         Args:
             event(synapse.events.EventBase): event of interest
+            types(list[(str, str|None)]|None): List of (type, state_key) tuples
+                which are used to filter the state fetched. If `state_key` is None,
+                all events are returned of the given type.
+                May be None, which matches any key.
+            filtered_types(list[str]|None): Only apply filtering via `types` to this
+                list of event types.  Other types of events are returned unfiltered.
+                If None, `types` filtering is applied to all events.
 
         Returns:
             A Deferred map from ((type, state_key)->Event)
         """
-        state_ids = yield self.store.get_state_ids_for_event(event.event_id)
+        state_ids = yield self.store.get_state_ids_for_event(
+            event.event_id, types, filtered_types=filtered_types,
+        )
         if event.is_state():
             state_ids = state_ids.copy()
             state_ids[(event.type, event.state_key)] = event.event_id
         defer.returnValue(state_ids)
 
     @defer.inlineCallbacks
-    def get_state_at(self, room_id, stream_position):
+    def get_state_at(self, room_id, stream_position, types=None, filtered_types=None):
         """ Get the room state at a particular stream position
 
         Args:
             room_id(str): room for which to get state
             stream_position(StreamToken): point at which to get state
+            types(list[(str, str|None)]|None): List of (type, state_key) tuples
+                which are used to filter the state fetched. If `state_key` is None,
+                all events are returned of the given type.
+            filtered_types(list[str]|None): Only apply filtering via `types` to this
+                list of event types.  Other types of events are returned unfiltered.
+                If None, `types` filtering is applied to all events.
 
         Returns:
             A Deferred map from ((type, state_key)->Event)
@@ -453,7 +485,9 @@ class SyncHandler(object):
 
         if last_events:
             last_event = last_events[-1]
-            state = yield self.get_state_after_event(last_event)
+            state = yield self.get_state_after_event(
+                last_event, types, filtered_types=filtered_types,
+            )
 
         else:
             # no events in this room - so presumably no state
@@ -485,59 +519,129 @@ class SyncHandler(object):
         # TODO(mjark) Check for new redactions in the state events.
 
         with Measure(self.clock, "compute_state_delta"):
+
+            types = None
+            filtered_types = None
+
+            lazy_load_members = sync_config.filter_collection.lazy_load_members()
+            include_redundant_members = (
+                sync_config.filter_collection.include_redundant_members()
+            )
+
+            if lazy_load_members:
+                # We only request state for the members needed to display the
+                # timeline:
+
+                types = [
+                    (EventTypes.Member, state_key)
+                    for state_key in set(
+                        event.sender  # FIXME: we also care about invite targets etc.
+                        for event in batch.events
+                    )
+                ]
+
+                # only apply the filtering to room members
+                filtered_types = [EventTypes.Member]
+
+            timeline_state = {
+                (event.type, event.state_key): event.event_id
+                for event in batch.events if event.is_state()
+            }
+
             if full_state:
                 if batch:
                     current_state_ids = yield self.store.get_state_ids_for_event(
-                        batch.events[-1].event_id
+                        batch.events[-1].event_id, types=types,
+                        filtered_types=filtered_types,
                     )
 
                     state_ids = yield self.store.get_state_ids_for_event(
-                        batch.events[0].event_id
+                        batch.events[0].event_id, types=types,
+                        filtered_types=filtered_types,
                     )
+
                 else:
                     current_state_ids = yield self.get_state_at(
-                        room_id, stream_position=now_token
+                        room_id, stream_position=now_token, types=types,
+                        filtered_types=filtered_types,
                     )
 
                     state_ids = current_state_ids
 
-                timeline_state = {
-                    (event.type, event.state_key): event.event_id
-                    for event in batch.events if event.is_state()
-                }
-
                 state_ids = _calculate_state(
                     timeline_contains=timeline_state,
                     timeline_start=state_ids,
                     previous={},
                     current=current_state_ids,
+                    lazy_load_members=lazy_load_members,
                 )
             elif batch.limited:
                 state_at_previous_sync = yield self.get_state_at(
-                    room_id, stream_position=since_token
+                    room_id, stream_position=since_token, types=types,
+                    filtered_types=filtered_types,
                 )
 
                 current_state_ids = yield self.store.get_state_ids_for_event(
-                    batch.events[-1].event_id
+                    batch.events[-1].event_id, types=types,
+                    filtered_types=filtered_types,
                 )
 
                 state_at_timeline_start = yield self.store.get_state_ids_for_event(
-                    batch.events[0].event_id
+                    batch.events[0].event_id, types=types,
+                    filtered_types=filtered_types,
                 )
 
-                timeline_state = {
-                    (event.type, event.state_key): event.event_id
-                    for event in batch.events if event.is_state()
-                }
-
                 state_ids = _calculate_state(
                     timeline_contains=timeline_state,
                     timeline_start=state_at_timeline_start,
                     previous=state_at_previous_sync,
                     current=current_state_ids,
+                    lazy_load_members=lazy_load_members,
                 )
             else:
                 state_ids = {}
+                if lazy_load_members:
+                    if types:
+                        state_ids = yield self.store.get_state_ids_for_event(
+                            batch.events[0].event_id, types=types,
+                            filtered_types=filtered_types,
+                        )
+
+            if lazy_load_members and not include_redundant_members:
+                cache_key = (sync_config.user.to_string(), sync_config.device_id)
+                cache = self.lazy_loaded_members_cache.get(cache_key)
+                if cache is None:
+                    logger.debug("creating LruCache for %r", cache_key)
+                    cache = LruCache(LAZY_LOADED_MEMBERS_CACHE_MAX_SIZE)
+                    self.lazy_loaded_members_cache[cache_key] = cache
+                else:
+                    logger.debug("found LruCache for %r", cache_key)
+
+                # if it's a new sync sequence, then assume the client has had
+                # amnesia and doesn't want any recent lazy-loaded members
+                # de-duplicated.
+                if since_token is None:
+                    logger.debug("clearing LruCache for %r", cache_key)
+                    cache.clear()
+                else:
+                    # only send members which aren't in our LruCache (either
+                    # because they're new to this client or have been pushed out
+                    # of the cache)
+                    logger.debug("filtering state from %r...", state_ids)
+                    state_ids = {
+                        t: event_id
+                        for t, event_id in state_ids.iteritems()
+                        if cache.get(t[1]) != event_id
+                    }
+                    logger.debug("...to %r", state_ids)
+
+                # add any member IDs we are about to send into our LruCache
+                for t, event_id in itertools.chain(
+                    state_ids.items(),
+                    timeline_state.items(),
+                ):
+                    if t[0] == EventTypes.Member:
+                        cache.set(t[1], event_id)
 
         state = {}
         if state_ids:
@@ -1448,7 +1552,9 @@ def _action_has_highlight(actions):
     return False
 
 
-def _calculate_state(timeline_contains, timeline_start, previous, current):
+def _calculate_state(
+    timeline_contains, timeline_start, previous, current, lazy_load_members,
+):
     """Works out what state to include in a sync response.
 
     Args:
@@ -1457,6 +1563,9 @@ def _calculate_state(timeline_contains, timeline_start, previous, current):
         previous (dict): state at the end of the previous sync (or empty dict
             if this is an initial sync)
         current (dict): state at the end of the timeline
+        lazy_load_members (bool): whether to return members from timeline_start
+            or not.  assumes that timeline_start has already been filtered to
+            include only the members the client needs to know about.
 
     Returns:
         dict
@@ -1472,9 +1581,25 @@ def _calculate_state(timeline_contains, timeline_start, previous, current):
     }
 
     c_ids = set(e for e in current.values())
-    tc_ids = set(e for e in timeline_contains.values())
-    p_ids = set(e for e in previous.values())
     ts_ids = set(e for e in timeline_start.values())
+    p_ids = set(e for e in previous.values())
+    tc_ids = set(e for e in timeline_contains.values())
+
+    # If we are lazyloading room members, we explicitly add the membership events
+    # for the senders in the timeline into the state block returned by /sync,
+    # as we may not have sent them to the client before.  We find these membership
+    # events by filtering them out of timeline_start, which has already been filtered
+    # to only include membership events for the senders in the timeline.
+    # In practice, we can do this by removing them from the p_ids list,
+    # which is the list of relevant state we know we have already sent to the client.
+    # see https://github.com/matrix-org/synapse/pull/2970
+    #            /files/efcdacad7d1b7f52f879179701c7e0d9b763511f#r204732809
+
+    if lazy_load_members:
+        p_ids.difference_update(
+            e for t, e in timeline_start.iteritems()
+            if t[0] == EventTypes.Member
+        )
 
     state_ids = ((c_ids | ts_ids) - p_ids) - tc_ids
 

synapse/http/client.py

@@ -26,9 +26,11 @@ from OpenSSL.SSL import VERIFY_NONE
 from twisted.internet import defer, protocol, reactor, ssl, task
 from twisted.internet.endpoints import HostnameEndpoint, wrapClientTLS
 from twisted.web._newclient import ResponseDone
-from twisted.web.client import Agent, BrowserLikeRedirectAgent, ContentDecoderAgent
-from twisted.web.client import FileBodyProducer as TwistedFileBodyProducer
 from twisted.web.client import (
+    Agent,
+    BrowserLikeRedirectAgent,
+    ContentDecoderAgent,
+    FileBodyProducer as TwistedFileBodyProducer,
     GzipDecoder,
     HTTPConnectionPool,
     PartialDownloadError,

synapse/http/request_metrics.py

@@ -38,7 +38,8 @@ outgoing_responses_counter = Counter(
 )
 
 response_timer = Histogram(
-    "synapse_http_server_response_time_seconds", "sec", ["method", "servlet", "tag"]
+    "synapse_http_server_response_time_seconds", "sec",
+    ["method", "servlet", "tag", "code"],
 )
 
 response_ru_utime = Counter(
@@ -171,11 +172,13 @@ class RequestMetrics(object):
                 )
                 return
 
-        outgoing_responses_counter.labels(request.method, str(request.code)).inc()
+        response_code = str(request.code)
+
+        outgoing_responses_counter.labels(request.method, response_code).inc()
 
         response_count.labels(request.method, self.name, tag).inc()
 
-        response_timer.labels(request.method, self.name, tag).observe(
+        response_timer.labels(request.method, self.name, tag, response_code).observe(
             time_sec - self.start
         )
 

synapse/http/server.py

@@ -13,12 +13,13 @@
 # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 # See the License for the specific language governing permissions and
 # limitations under the License.
+
 import cgi
 import collections
 import logging
-import urllib
 
-from six.moves import http_client
+from six import PY3
+from six.moves import http_client, urllib
 
 from canonicaljson import encode_canonical_json, encode_pretty_printed_json, json
 
@@ -264,6 +265,7 @@ class JsonResource(HttpServer, resource.Resource):
         self.hs = hs
 
     def register_paths(self, method, path_patterns, callback):
+        method = method.encode("utf-8")  # method is bytes on py3
         for path_pattern in path_patterns:
             logger.debug("Registering for %s %s", method, path_pattern.pattern)
             self.path_regexs.setdefault(method, []).append(
@@ -296,8 +298,19 @@ class JsonResource(HttpServer, resource.Resource):
         # here. If it throws an exception, that is handled by the wrapper
         # installed by @request_handler.
 
+        def _unquote(s):
+            if PY3:
+                # On Python 3, unquote is unicode -> unicode
+                return urllib.parse.unquote(s)
+            else:
+                # On Python 2, unquote is bytes -> bytes We need to encode the
+                # URL again (as it was decoded by _get_handler_for request), as
+                # ASCII because it's a URL, and then decode it to get the UTF-8
+                # characters that were quoted.
+                return urllib.parse.unquote(s.encode('ascii')).decode('utf8')
+
         kwargs = intern_dict({
-            name: urllib.unquote(value).decode("UTF-8") if value else value
+            name: _unquote(value) if value else value
             for name, value in group_dict.items()
         })
 
@@ -313,9 +326,9 @@ class JsonResource(HttpServer, resource.Resource):
             request (twisted.web.http.Request):
 
         Returns:
-            Tuple[Callable, dict[str, str]]: callback method, and the dict
-                mapping keys to path components as specified in the handler's
-                path match regexp.
+            Tuple[Callable, dict[unicode, unicode]]: callback method, and the
+                dict mapping keys to path components as specified in the
+                handler's path match regexp.
 
                 The callback will normally be a method registered via
                 register_paths, so will return (possibly via Deferred) either
@@ -327,7 +340,7 @@ class JsonResource(HttpServer, resource.Resource):
         # Loop through all the registered callbacks to check if the method
         # and path regex match
         for path_entry in self.path_regexs.get(request.method, []):
-            m = path_entry.pattern.match(request.path)
+            m = path_entry.pattern.match(request.path.decode('ascii'))
             if m:
                 # We found a match!
                 return path_entry.callback, m.groupdict()
@@ -383,7 +396,7 @@ class RootRedirect(resource.Resource):
         self.url = path
 
     def render_GET(self, request):
-        return redirectTo(self.url, request)
+        return redirectTo(self.url.encode('ascii'), request)
 
     def getChild(self, name, request):
         if len(name) == 0:
@@ -404,12 +417,14 @@ def respond_with_json(request, code, json_object, send_cors=False,
         return
 
     if pretty_print:
-        json_bytes = encode_pretty_printed_json(json_object) + "\n"
+        json_bytes = (encode_pretty_printed_json(json_object) + "\n"
+                      ).encode("utf-8")
     else:
         if canonical_json or synapse.events.USE_FROZEN_DICTS:
+            # canonicaljson already encodes to bytes
             json_bytes = encode_canonical_json(json_object)
         else:
-            json_bytes = json.dumps(json_object)
+            json_bytes = json.dumps(json_object).encode("utf-8")
 
     return respond_with_json_bytes(
         request, code, json_bytes,