Merge remote-tracking branch 'upstream/release-v1.60'

.git-blame-ignore-revs

@@ -6,3 +6,6 @@ aff1eb7c671b0a3813407321d2702ec46c71fa56
 
 # Update black to 20.8b1 (#9381).
 0a00b7ff14890987f09112a2ae696c61001e6cf1
+
+# Convert tests/rest/admin/test_room.py to unix file endings (#7953).
+c4268e3da64f1abb5b31deaeb5769adb6510c0a7

CHANGES.md

@@ -1,3 +1,115 @@
+Synapse 1.60.0rc1 (2022-05-24)
+==============================
+
+Features
+--------
+
+- Measure the time taken in spam-checking callbacks and expose those measurements as metrics. ([\#12513](https://github.com/matrix-org/synapse/issues/12513))
+- Add a `default_power_level_content_override` config option to set default room power levels per room preset. ([\#12618](https://github.com/matrix-org/synapse/issues/12618))
+- Add support for [MSC3787: Allowing knocks to restricted rooms](https://github.com/matrix-org/matrix-spec-proposals/pull/3787). ([\#12623](https://github.com/matrix-org/synapse/issues/12623))
+- Send `USER_IP` commands on a different Redis channel, in order to reduce traffic to workers that do not process these commands. ([\#12672](https://github.com/matrix-org/synapse/issues/12672), [\#12809](https://github.com/matrix-org/synapse/issues/12809))
+- Synapse will now reload [cache config](https://matrix-org.github.io/synapse/latest/usage/configuration/config_documentation.html#caching) when it receives a [SIGHUP](https://en.wikipedia.org/wiki/SIGHUP) signal. ([\#12673](https://github.com/matrix-org/synapse/issues/12673))
+- Add a config options to allow for auto-tuning of caches. ([\#12701](https://github.com/matrix-org/synapse/issues/12701))
+- Update [MSC2716](https://github.com/matrix-org/matrix-spec-proposals/pull/2716) implementation to process marker events from the current state to avoid markers being lost in timeline gaps for federated servers which would cause the imported history to be undiscovered. ([\#12718](https://github.com/matrix-org/synapse/issues/12718))
+- Add a `drop_federated_event` callback to `SpamChecker` to disregard inbound federated events before they take up much processing power, in an emergency. ([\#12744](https://github.com/matrix-org/synapse/issues/12744))
+- Implement [MSC3818: Copy room type on upgrade](https://github.com/matrix-org/matrix-spec-proposals/pull/3818). ([\#12786](https://github.com/matrix-org/synapse/issues/12786), [\#12792](https://github.com/matrix-org/synapse/issues/12792))
+- Update to `check_event_for_spam`. Deprecate the current callback signature, replace it with a new signature that is both less ambiguous (replacing booleans with explicit allow/block) and more powerful (ability to return explicit error codes). ([\#12808](https://github.com/matrix-org/synapse/issues/12808))
+
+
+Bugfixes
+--------
+
+- Fix a bug introduced in Synapse 1.7.0 that would prevent events from being sent to clients if there's a retention policy in the room when the support for retention policies is disabled. ([\#12611](https://github.com/matrix-org/synapse/issues/12611))
+- Fix a bug introduced in Synapse 1.57.0 where `/messages` would throw a 500 error when querying for a non-existent room. ([\#12683](https://github.com/matrix-org/synapse/issues/12683))
+- Add a unique index to `state_group_edges` to prevent duplicates being accidentally introduced and the consequential impact to performance. ([\#12687](https://github.com/matrix-org/synapse/issues/12687))
+- Fix a long-standing bug where an empty room would be created when a user with an insufficient power level tried to upgrade a room. ([\#12696](https://github.com/matrix-org/synapse/issues/12696))
+- Fix a bug introduced in Synapse 1.30.0 where empty rooms could be automatically created if a monthly active users limit is set. ([\#12713](https://github.com/matrix-org/synapse/issues/12713))
+- Fix push to dismiss notifications when read on another client. Contributed by @SpiritCroc @ Beeper. ([\#12721](https://github.com/matrix-org/synapse/issues/12721))
+- Fix poor database performance when reading the cache invalidation stream for large servers with lots of workers. ([\#12747](https://github.com/matrix-org/synapse/issues/12747))
+- Delete events from the `federation_inbound_events_staging` table when a room is purged through the admin API. ([\#12770](https://github.com/matrix-org/synapse/issues/12770))
+- Give a meaningful error message when a client tries to create a room with an invalid alias localpart. ([\#12779](https://github.com/matrix-org/synapse/issues/12779))
+- Fix a bug introduced in 1.43.0 where a file (`providers.json`) was never closed. Contributed by @arkamar. ([\#12794](https://github.com/matrix-org/synapse/issues/12794))
+- Fix a long-standing bug where finished log contexts would be re-started when failing to contact remote homeservers. ([\#12803](https://github.com/matrix-org/synapse/issues/12803))
+- Fix a bug, introduced in Synapse 1.21.0, that led to media thumbnails being unusable before the index has been added in the background. ([\#12823](https://github.com/matrix-org/synapse/issues/12823))
+
+
+Updates to the Docker image
+---------------------------
+
+- Fix the docker file after a dependency update. ([\#12853](https://github.com/matrix-org/synapse/issues/12853))
+
+
+Improved Documentation
+----------------------
+
+- Fix a typo in the Media Admin API documentation. ([\#12715](https://github.com/matrix-org/synapse/issues/12715))
+- Update the OpenID Connect example for Keycloak to be compatible with newer versions of Keycloak. Contributed by @nhh. ([\#12727](https://github.com/matrix-org/synapse/issues/12727))
+- Fix typo in server listener documentation. ([\#12742](https://github.com/matrix-org/synapse/issues/12742))
+- Link to the configuration manual from the welcome page of the documentation. ([\#12748](https://github.com/matrix-org/synapse/issues/12748))
+- Fix typo in 'run_background_tasks_on' option name in configuration manual documentation. ([\#12749](https://github.com/matrix-org/synapse/issues/12749))
+- Add information regarding the `rc_invites` ratelimiting option to the configuration docs. ([\#12759](https://github.com/matrix-org/synapse/issues/12759))
+- Add documentation for cancellation of request processing. ([\#12761](https://github.com/matrix-org/synapse/issues/12761))
+- Recommend using docker to run tests against postgres. ([\#12765](https://github.com/matrix-org/synapse/issues/12765))
+- Add missing user directory endpoint from the generic worker documentation. Contributed by @olmari. ([\#12773](https://github.com/matrix-org/synapse/issues/12773))
+- Add additional info to documentation of config option `cache_autotuning`. ([\#12776](https://github.com/matrix-org/synapse/issues/12776))
+- Update configuration manual documentation to document size-related suffixes. ([\#12777](https://github.com/matrix-org/synapse/issues/12777))
+- Fix invalid YAML syntax in the example documentation for the `url_preview_accept_language` config option. ([\#12785](https://github.com/matrix-org/synapse/issues/12785))
+
+
+Deprecations and Removals
+-------------------------
+
+- Require a body in POST requests to `/rooms/{roomId}/receipt/{receiptType}/{eventId}`, as required by the [Matrix specification](https://spec.matrix.org/v1.2/client-server-api/#post_matrixclientv3roomsroomidreceiptreceipttypeeventid). This breaks compatibility with Element Android 1.2.0 and earlier: users of those clients will be unable to send read receipts. ([\#12709](https://github.com/matrix-org/synapse/issues/12709))
+
+
+Internal Changes
+----------------
+
+- Improve event caching mechanism to avoid having multiple copies of an event in memory at a time. ([\#10533](https://github.com/matrix-org/synapse/issues/10533))
+- Add some type hints to datastore. ([\#12477](https://github.com/matrix-org/synapse/issues/12477), [\#12717](https://github.com/matrix-org/synapse/issues/12717), [\#12753](https://github.com/matrix-org/synapse/issues/12753))
+- Preparation for faster-room-join work: return subsets of room state which we already have, immediately. ([\#12498](https://github.com/matrix-org/synapse/issues/12498))
+- Replace string literal instances of stream key types with typed constants. ([\#12567](https://github.com/matrix-org/synapse/issues/12567))
+- Add `@cancellable` decorator, for use on endpoint methods that can be cancelled when clients disconnect. ([\#12586](https://github.com/matrix-org/synapse/issues/12586))
+- Add ability to cancel disconnected requests to `SynapseRequest`. ([\#12588](https://github.com/matrix-org/synapse/issues/12588))
+- Add a helper class for testing request cancellation. ([\#12630](https://github.com/matrix-org/synapse/issues/12630))
+- Improve documentation of the `synapse.push` module. ([\#12676](https://github.com/matrix-org/synapse/issues/12676))
+- Refactor functions to on `PushRuleEvaluatorForEvent`. ([\#12677](https://github.com/matrix-org/synapse/issues/12677))
+- Preparation for database schema simplifications: stop writing to `event_reference_hashes`. ([\#12679](https://github.com/matrix-org/synapse/issues/12679))
+- Remove code which updates unused database column `application_services_state.last_txn`. ([\#12680](https://github.com/matrix-org/synapse/issues/12680))
+- Refactor `EventContext` class. ([\#12689](https://github.com/matrix-org/synapse/issues/12689))
+- Remove an unneeded class in the push code. ([\#12691](https://github.com/matrix-org/synapse/issues/12691))
+- Consolidate parsing of relation information from events. ([\#12693](https://github.com/matrix-org/synapse/issues/12693))
+- Capture the `Deferred` for request cancellation in `_AsyncResource`. ([\#12694](https://github.com/matrix-org/synapse/issues/12694))
+- Fixes an incorrect type hint for `Filter._check_event_relations`. ([\#12695](https://github.com/matrix-org/synapse/issues/12695))
+- Respect the `@cancellable` flag for `DirectServe{Html,Json}Resource`s. ([\#12698](https://github.com/matrix-org/synapse/issues/12698))
+- Respect the `@cancellable` flag for `RestServlet`s and `BaseFederationServlet`s. ([\#12699](https://github.com/matrix-org/synapse/issues/12699))
+- Respect the `@cancellable` flag for `ReplicationEndpoint`s. ([\#12700](https://github.com/matrix-org/synapse/issues/12700))
+- Convert namespace class `Codes` into a string enum. ([\#12703](https://github.com/matrix-org/synapse/issues/12703))
+- Complain if a federation endpoint has the `@cancellable` flag, since some of the wrapper code may not handle cancellation correctly yet. ([\#12705](https://github.com/matrix-org/synapse/issues/12705))
+- Enable cancellation of `GET /rooms/$room_id/members`, `GET /rooms/$room_id/state` and `GET /rooms/$room_id/state/$event_type/*` requests. ([\#12708](https://github.com/matrix-org/synapse/issues/12708))
+- Optimize private read receipt filtering. ([\#12711](https://github.com/matrix-org/synapse/issues/12711))
+- Add type annotations to increase the number of modules passing `disallow-untyped-defs`. ([\#12716](https://github.com/matrix-org/synapse/issues/12716), [\#12726](https://github.com/matrix-org/synapse/issues/12726))
+- Drop the logging level of status messages for the URL preview cache expiry job from INFO to DEBUG. ([\#12720](https://github.com/matrix-org/synapse/issues/12720))
+- Downgrade some OIDC errors to warnings in the logs, to reduce the noise of Sentry reports. ([\#12723](https://github.com/matrix-org/synapse/issues/12723))
+- Update configs used by Complement to allow more invites/3PID validations during tests. ([\#12731](https://github.com/matrix-org/synapse/issues/12731))
+- Tidy up and type-hint the database engine modules. ([\#12734](https://github.com/matrix-org/synapse/issues/12734))
+- Fix a long-standing bug where the user directory background process would fail to make forward progress if a user included a null codepoint in their display name or avatar. ([\#12762](https://github.com/matrix-org/synapse/issues/12762))
+- Tweak the mypy plugin so that `@cached` can accept `on_invalidate=None`. ([\#12769](https://github.com/matrix-org/synapse/issues/12769))
+- Move methods that call `add_push_rule` to the `PushRuleStore` class. ([\#12772](https://github.com/matrix-org/synapse/issues/12772))
+- Make handling of federation Authorization header (more) compliant with RFC7230. ([\#12774](https://github.com/matrix-org/synapse/issues/12774))
+- Refactor `resolve_state_groups_for_events` to not pull out full state when no state resolution happens. ([\#12775](https://github.com/matrix-org/synapse/issues/12775))
+- Do not keep going if there are 5 back-to-back background update failures. ([\#12781](https://github.com/matrix-org/synapse/issues/12781))
+- Fix federation when using the demo scripts. ([\#12783](https://github.com/matrix-org/synapse/issues/12783))
+- The `hash_password` script now fails when it is called without specifying a config file. ([\#12789](https://github.com/matrix-org/synapse/issues/12789))
+- Simplify `disallow_untyped_defs` config in `mypy.ini`. ([\#12790](https://github.com/matrix-org/synapse/issues/12790))
+- Update EventContext `get_current_event_ids` and `get_prev_event_ids` to accept state filters and update calls where possible. ([\#12791](https://github.com/matrix-org/synapse/issues/12791))
+- Remove Caddy from the Synapse workers image used in Complement. ([\#12818](https://github.com/matrix-org/synapse/issues/12818))
+- Add Complement's shared registration secret to the Complement worker image. This fixes tests that depend on it. ([\#12819](https://github.com/matrix-org/synapse/issues/12819))
+- Support registering Application Services when running with workers under Complement. ([\#12826](https://github.com/matrix-org/synapse/issues/12826))
+- Add some type hints to test files. ([\#12833](https://github.com/matrix-org/synapse/issues/12833))
+- Disable 'faster room join' Complement tests when testing against Synapse with workers. ([\#12842](https://github.com/matrix-org/synapse/issues/12842))
+
+
 Synapse 1.59.1 (2022-05-18)
 ===========================
 

debian/changelog

@@ -1,3 +1,9 @@
+matrix-synapse-py3 (1.60.0~rc1) stable; urgency=medium
+
+  * New Synapse release 1.60.0rc1.
+
+ -- Synapse Packaging team <packages@matrix.org>  Tue, 24 May 2022 12:05:01 +0100
+
 matrix-synapse-py3 (1.59.1) stable; urgency=medium
 
   * New Synapse release 1.59.1.

demo/start.sh

@@ -12,6 +12,7 @@ export PYTHONPATH
 
 echo "$PYTHONPATH"
 
+# Create servers which listen on HTTP at 808x and HTTPS at 848x.
 for port in 8080 8081 8082; do
     echo "Starting server on port $port... "
 
@@ -19,10 +20,12 @@ for port in 8080 8081 8082; do
     mkdir -p demo/$port
     pushd demo/$port || exit
 
-    # Generate the configuration for the homeserver at localhost:848x.
+    # Generate the configuration for the homeserver at localhost:848x, note that
+    # the homeserver name needs to match the HTTPS listening port for federation
+    # to properly work..
     python3 -m synapse.app.homeserver \
         --generate-config \
-        --server-name "localhost:$port" \
+        --server-name "localhost:$https_port" \
         --config-path "$port.config" \
         --report-stats no
 

docker/Dockerfile

@@ -55,7 +55,7 @@ RUN \
 # NB: In poetry 1.2 `poetry export` will be moved into a plugin; we'll need to also
 # pip install poetry-plugin-export (https://github.com/python-poetry/poetry-plugin-export).
 RUN --mount=type=cache,target=/root/.cache/pip \
-  pip install --user git+https://github.com/python-poetry/poetry.git@fb13b3a676f476177f7937ffa480ee5cff9a90a5
+  pip install --user "poetry-core==1.1.0a7" "git+https://github.com/python-poetry/poetry.git@fb13b3a676f476177f7937ffa480ee5cff9a90a5"
 
 WORKDIR /synapse
 

docker/complement/SynapseWorkers.Dockerfile

@@ -6,12 +6,6 @@
 # https://github.com/matrix-org/synapse/blob/develop/docker/README-testing.md#testing-with-postgresql-and-single-or-multi-process-synapse
 FROM matrixdotorg/synapse-workers
 
-# Download a caddy server to stand in front of nginx and terminate TLS using Complement's
-# custom CA.
-# We include this near the top of the file in order to cache the result.
-RUN curl -OL "https://github.com/caddyserver/caddy/releases/download/v2.3.0/caddy_2.3.0_linux_amd64.tar.gz" && \
-  tar xzf caddy_2.3.0_linux_amd64.tar.gz && rm caddy_2.3.0_linux_amd64.tar.gz && mv caddy /root
-
 # Install postgresql
 RUN apt-get update && \
   DEBIAN_FRONTEND=noninteractive apt-get install --no-install-recommends -y postgresql-13
@@ -31,16 +25,12 @@ COPY conf-workers/workers-shared.yaml /conf/workers/shared.yaml
 
 WORKDIR /data
 
-# Copy the caddy config
-COPY conf-workers/caddy.complement.json /root/caddy.json
-
 COPY conf-workers/postgres.supervisord.conf /etc/supervisor/conf.d/postgres.conf
-COPY conf-workers/caddy.supervisord.conf /etc/supervisor/conf.d/caddy.conf
 
 # Copy the entrypoint
 COPY conf-workers/start-complement-synapse-workers.sh /
 
-# Expose caddy's listener ports
+# Expose nginx's listener ports
 EXPOSE 8008 8448
 
 ENTRYPOINT ["/start-complement-synapse-workers.sh"]

docker/complement/conf-workers/caddy.complement.json

@@ -1,72 +0,0 @@
-{
-    "apps": {
-      "http": {
-        "servers": {
-          "srv0": {
-            "listen": [
-              ":8448"
-            ],
-            "routes": [
-              {
-                "match": [
-                  {
-                    "host": [
-                      "{{ server_name }}"
-                    ]
-                  }
-                ],
-                "handle": [
-                  {
-                    "handler": "subroute",
-                    "routes": [
-                      {
-                        "handle": [
-                          {
-                            "handler": "reverse_proxy",
-                            "upstreams": [
-                              {
-                                "dial": "localhost:8008"
-                              }
-                            ]
-                          }
-                        ]
-                      }
-                    ]
-                  }
-                ],
-                "terminal": true
-              }
-            ]
-          }
-        }
-      },
-      "tls": {
-        "automation": {
-          "policies": [
-            {
-              "subjects": [
-                "{{ server_name }}"
-              ],
-              "issuers": [
-                {
-                  "module": "internal"
-                }
-              ],
-              "on_demand": true
-            }
-          ]
-        }
-      },
-      "pki": {
-        "certificate_authorities": {
-          "local": {
-            "name": "Complement CA",
-            "root": {
-              "certificate": "/complement/ca/ca.crt",
-              "private_key": "/complement/ca/ca.key"
-            }
-          }
-        }
-      }
-    }
-  }

docker/complement/conf-workers/caddy.supervisord.conf

@@ -1,7 +0,0 @@
-[program:caddy]
-command=/usr/local/bin/prefix-log /root/caddy run --config /root/caddy.json
-autorestart=unexpected
-stdout_logfile=/dev/stdout
-stdout_logfile_maxbytes=0
-stderr_logfile=/dev/stderr
-stderr_logfile_maxbytes=0

docker/complement/conf-workers/start-complement-synapse-workers.sh

@@ -9,9 +9,6 @@ function log {
     echo "$d $@"
 }
 
-# Replace the server name in the caddy config
-sed -i "s/{{ server_name }}/${SERVER_NAME}/g" /root/caddy.json
-
 # Set the server name of the homeserver
 export SYNAPSE_SERVER_NAME=${SERVER_NAME}
 
@@ -39,6 +36,26 @@ export SYNAPSE_WORKER_TYPES="\
     appservice, \
     pusher"
 
+# Add Complement's appservice registration directory, if there is one
+# (It can be absent when there are no application services in this test!)
+if [ -d /complement/appservice ]; then
+    export SYNAPSE_AS_REGISTRATION_DIR=/complement/appservice
+fi
+
+# Generate a TLS key, then generate a certificate by having Complement's CA sign it
+# Note that both the key and certificate are in PEM format (not DER).
+openssl genrsa -out /conf/server.tls.key 2048
+
+openssl req -new -key /conf/server.tls.key -out /conf/server.tls.csr \
+  -subj "/CN=${SERVER_NAME}"
+
+openssl x509 -req -in /conf/server.tls.csr \
+  -CA /complement/ca/ca.crt -CAkey /complement/ca/ca.key -set_serial 1 \
+  -out /conf/server.tls.crt
+
+export SYNAPSE_TLS_CERT=/conf/server.tls.crt
+export SYNAPSE_TLS_KEY=/conf/server.tls.key
+
 # Run the script that writes the necessary config files and starts supervisord, which in turn
 # starts everything else
 exec /configure_workers_and_start.py

docker/complement/conf-workers/workers-shared.yaml

@@ -5,6 +5,12 @@ enable_registration: true
 enable_registration_without_verification: true
 bcrypt_rounds: 4
 
+## Registration ##
+
+# Needed by Complement to register admin users
+# DO NOT USE in a production configuration! This should be a random secret.
+registration_shared_secret: complement
+
 ## Federation ##
 
 # trust certs signed by Complement's CA
@@ -53,6 +59,18 @@ rc_joins:
     per_second: 9999
     burst_count: 9999
 
+rc_3pid_validation:
+  per_second: 1000
+  burst_count: 1000
+
+rc_invites:
+  per_room:
+    per_second: 1000
+    burst_count: 1000
+  per_user:
+    per_second: 1000
+    burst_count: 1000
+
 federation_rr_transactions_per_room_per_second: 9999
 
 ## Experimental Features ##

docker/complement/conf/homeserver.yaml

@@ -87,6 +87,18 @@ rc_joins:
     per_second: 9999
     burst_count: 9999
 
+rc_3pid_validation:
+  per_second: 1000
+  burst_count: 1000
+
+rc_invites:
+  per_room:
+    per_second: 1000
+    burst_count: 1000
+  per_user:
+    per_second: 1000
+    burst_count: 1000
+
 federation_rr_transactions_per_room_per_second: 9999
 
 ## API Configuration ##

docker/conf-workers/nginx.conf.j2

@@ -9,6 +9,22 @@ server {
     listen 8008;
     listen [::]:8008;
 
+    {% if tls_cert_path is not none and tls_key_path is not none %}
+        listen 8448 ssl;
+        listen [::]:8448 ssl;
+
+        ssl_certificate {{ tls_cert_path }};
+        ssl_certificate_key {{ tls_key_path }};
+
+        # Some directives from cipherlist.eu (fka cipherli.st):
+        ssl_protocols TLSv1 TLSv1.1 TLSv1.2 TLSv1.3;
+        ssl_prefer_server_ciphers on;
+        ssl_ciphers "EECDH+AESGCM:EDH+AESGCM:AES256+EECDH:AES256+EDH";
+        ssl_ecdh_curve secp384r1; # Requires nginx >= 1.1.0
+        ssl_session_cache shared:SSL:10m;
+        ssl_session_tickets off; # Requires nginx >= 1.5.9
+    {% endif %}
+
     server_name localhost;
 
     # Nginx by default only allows file uploads up to 1M in size

docker/conf-workers/shared.yaml.j2

@@ -6,4 +6,13 @@
 redis:
     enabled: true
 
-{{ shared_worker_config }}
+{% if appservice_registrations is not none %}
+## Application Services ##
+# A list of application service config files to use.
+app_service_config_files:
+{%- for path in appservice_registrations %}
+  - "{{ path }}"
+{%- endfor %}
+{%- endif %}
+
+{{ shared_worker_config }}

docker/configure_workers_and_start.py

@@ -21,6 +21,11 @@
 #   * SYNAPSE_REPORT_STATS: Whether to report stats.
 #   * SYNAPSE_WORKER_TYPES: A comma separated list of worker names as specified in WORKER_CONFIG
 #         below. Leave empty for no workers, or set to '*' for all possible workers.
+#   * SYNAPSE_AS_REGISTRATION_DIR: If specified, a directory in which .yaml and .yml files
+#         will be treated as Application Service registration files.
+#   * SYNAPSE_TLS_CERT: Path to a TLS certificate in PEM format.
+#   * SYNAPSE_TLS_KEY: Path to a TLS key. If this and SYNAPSE_TLS_CERT are specified,
+#         Nginx will be configured to serve TLS on port 8448.
 #
 # NOTE: According to Complement's ENTRYPOINT expectations for a homeserver image (as defined
 # in the project's README), this script may be run multiple times, and functionality should
@@ -29,6 +34,7 @@
 import os
 import subprocess
 import sys
+from pathlib import Path
 from typing import Any, Dict, List, Mapping, MutableMapping, NoReturn, Set
 
 import jinja2
@@ -488,11 +494,23 @@ def generate_worker_files(
     master_log_config = generate_worker_log_config(environ, "master", data_dir)
     shared_config["log_config"] = master_log_config
 
+    # Find application service registrations
+    appservice_registrations = None
+    appservice_registration_dir = os.environ.get("SYNAPSE_AS_REGISTRATION_DIR")
+    if appservice_registration_dir:
+        # Scan for all YAML files that should be application service registrations.
+        appservice_registrations = [
+            str(reg_path.resolve())
+            for reg_path in Path(appservice_registration_dir).iterdir()
+            if reg_path.suffix.lower() in (".yaml", ".yml")
+        ]
+
     # Shared homeserver config
     convert(
         "/conf/shared.yaml.j2",
         "/conf/workers/shared.yaml",
         shared_worker_config=yaml.dump(shared_config),
+        appservice_registrations=appservice_registrations,
     )
 
     # Nginx config
@@ -501,6 +519,8 @@ def generate_worker_files(
         "/etc/nginx/conf.d/matrix-synapse.conf",
         worker_locations=nginx_location_config,
         upstream_directives=nginx_upstream_config,
+        tls_cert_path=os.environ.get("SYNAPSE_TLS_CERT"),
+        tls_key_path=os.environ.get("SYNAPSE_TLS_KEY"),
     )
 
     # Supervisord config

docs/SUMMARY.md

@@ -89,6 +89,7 @@
   - [Database Schemas](development/database_schema.md)
   - [Experimental features](development/experimental_features.md)
   - [Synapse Architecture]()
+    - [Cancellation](development/synapse_architecture/cancellation.md)
     - [Log Contexts](log_contexts.md)
     - [Replication](replication.md)
     - [TCP Replication](tcp_replication.md)

docs/admin_api/media_admin_api.md

@@ -289,7 +289,7 @@ POST /_synapse/admin/v1/purge_media_cache?before_ts=<unix_timestamp_in_ms>
 
 URL Parameters
 
-* `unix_timestamp_in_ms`: string representing a positive integer - Unix timestamp in milliseconds.
+* `before_ts`: string representing a positive integer - Unix timestamp in milliseconds.
 All cached media that was last accessed before this timestamp will be removed.
 
 Response:

docs/development/contributing_guide.md

@@ -206,7 +206,32 @@ This means that we need to run our unit tests against PostgreSQL too. Our CI doe
 this automatically for pull requests and release candidates, but it's sometimes
 useful to reproduce this locally.
 
-To do so, [configure Postgres](../postgres.md) and run `trial` with the
+#### Using Docker
+
+The easiest way to do so is to run Postgres via a docker container. In one
+terminal:
+
+```shell
+docker run --rm -e POSTGRES_PASSWORD=mysecretpassword -e POSTGRES_USER=postgres -e POSTGRES_DB=postgress -p 5432:5432 postgres:14
+```
+
+If you see an error like
+
+```
+docker: Error response from daemon: driver failed programming external connectivity on endpoint nice_ride (b57bbe2e251b70015518d00c9981e8cb8346b5c785250341a6c53e3c899875f1): Error starting userland proxy: listen tcp4 0.0.0.0:5432: bind: address already in use.
+```
+
+then something is already bound to port 5432. You're probably already running postgres locally.
+
+Once you have a postgres server running, invoke `trial` in a second terminal:
+
+```shell
+SYNAPSE_POSTGRES=1 SYNAPSE_POSTGRES_HOST=127.0.0.1 SYNAPSE_POSTGRES_USER=postgres SYNAPSE_POSTGRES_PASSWORD=mysecretpassword poetry run trial tests
+````
+
+#### Using an existing Postgres installation
+
+If you have postgres already installed on your system, you can run `trial` with the
 following environment variables matching your configuration:
 
 - `SYNAPSE_POSTGRES` to anything nonempty
@@ -229,8 +254,8 @@ You don't need to specify the host, user, port or password if your Postgres
 server is set to authenticate you over the UNIX socket (i.e. if the `psql` command
 works without further arguments).
 
-Your Postgres account needs to be able to create databases.
+Your Postgres account needs to be able to create databases; see the postgres
-
+docs for [`ALTER ROLE`](https://www.postgresql.org/docs/current/sql-alterrole.html).
 
 ## Run the integration tests ([Sytest](https://github.com/matrix-org/sytest)).
 
@@ -397,8 +422,8 @@ same lightweight approach that the Linux Kernel
 [submitting patches process](
 https://www.kernel.org/doc/html/latest/process/submitting-patches.html#sign-your-work-the-developer-s-certificate-of-origin>),
 [Docker](https://github.com/docker/docker/blob/master/CONTRIBUTING.md), and many other
-projects use: the DCO (Developer Certificate of Origin:
+projects use: the DCO ([Developer Certificate of Origin](http://developercertificate.org/)).
-http://developercertificate.org/). This is a simple declaration that you wrote
+This is a simple declaration that you wrote
 the contribution or otherwise have the right to contribute it to Matrix:
 
 ```

docs/development/demo.md

@@ -5,7 +5,7 @@
 Requires you to have a [Synapse development environment setup](https://matrix-org.github.io/synapse/develop/development/contributing_guide.html#4-install-the-dependencies).
 
 The demo setup allows running three federation Synapse servers, with server
-names `localhost:8080`, `localhost:8081`, and `localhost:8082`.
+names `localhost:8480`, `localhost:8481`, and `localhost:8482`.
 
 You can access them via any Matrix client over HTTP at `localhost:8080`,
 `localhost:8081`, and `localhost:8082` or over HTTPS at `localhost:8480`,
@@ -20,9 +20,10 @@ and the servers are configured in a highly insecure way, including:
 The servers are configured to store their data under `demo/8080`, `demo/8081`, and
 `demo/8082`. This includes configuration, logs, SQLite databases, and media.
 
-Note that when joining a public room on a different HS via "#foo:bar.net", then
+Note that when joining a public room on a different homeserver via "#foo:bar.net",
-you are (in the current impl) joining a room with room_id "foo". This means that
+then you are (in the current implementation) joining a room with room_id "foo".
-it won't work if your HS already has a room with that name.
+This means that it won't work if your homeserver already has a room with that
+name.
 
 ## Using the demo scripts
 

docs/development/synapse_architecture/cancellation.md

@@ -0,0 +1,392 @@
+# Cancellation
+Sometimes, requests take a long time to service and clients disconnect
+before Synapse produces a response. To avoid wasting resources, Synapse
+can cancel request processing for select endpoints marked with the
+`@cancellable` decorator.
+
+Synapse makes use of Twisted's `Deferred.cancel()` feature to make
+cancellation work. The `@cancellable` decorator does nothing by itself
+and merely acts as a flag, signalling to developers and other code alike
+that a method can be cancelled.
+
+## Enabling cancellation for an endpoint
+1. Check that the endpoint method, and any `async` functions in its call
+   tree handle cancellation correctly. See
+   [Handling cancellation correctly](#handling-cancellation-correctly)
+   for a list of things to look out for.
+2. Add the `@cancellable` decorator to the `on_GET/POST/PUT/DELETE`
+   method. It's not recommended to make non-`GET` methods cancellable,
+   since cancellation midway through some database updates is less
+   likely to be handled correctly.
+
+## Mechanics
+There are two stages to cancellation: downward propagation of a
+`cancel()` call, followed by upwards propagation of a `CancelledError`
+out of a blocked `await`.
+Both Twisted and asyncio have a cancellation mechanism.
+
+|               | Method              | Exception                               | Exception inherits from |
+|---------------|---------------------|-----------------------------------------|-------------------------|
+| Twisted       | `Deferred.cancel()` | `twisted.internet.defer.CancelledError` | `Exception` (!)         |
+| asyncio       | `Task.cancel()`     | `asyncio.CancelledError`                | `BaseException`         |
+
+### Deferred.cancel()
+When Synapse starts handling a request, it runs the async method
+responsible for handling it using `defer.ensureDeferred`, which returns
+a `Deferred`. For example:
+
+```python
+def do_something() -> Deferred[None]:
+    ...
+
+@cancellable
+async def on_GET() -> Tuple[int, JsonDict]:
+    d = make_deferred_yieldable(do_something())
+    await d
+    return 200, {}
+
+request = defer.ensureDeferred(on_GET())
+```
+
+When a client disconnects early, Synapse checks for the presence of the
+`@cancellable` decorator on `on_GET`. Since `on_GET` is cancellable,
+`Deferred.cancel()` is called on the `Deferred` from
+`defer.ensureDeferred`, ie. `request`. Twisted knows which `Deferred`
+`request` is waiting on and passes the `cancel()` call on to `d`.
+
+The `Deferred` being waited on, `d`, may have its own handling for
+`cancel()` and pass the call on to other `Deferred`s.
+
+Eventually, a `Deferred` handles the `cancel()` call by resolving itself
+with a `CancelledError`.
+
+### CancelledError
+The `CancelledError` gets raised out of the `await` and bubbles up, as
+per normal Python exception handling.
+
+## Handling cancellation correctly
+In general, when writing code that might be subject to cancellation, two
+things must be considered:
+ * The effect of `CancelledError`s raised out of `await`s.
+ * The effect of `Deferred`s being `cancel()`ed.
+
+Examples of code that handles cancellation incorrectly include:
+ * `try-except` blocks which swallow `CancelledError`s.
+ * Code that shares the same `Deferred`, which may be cancelled, between
+   multiple requests.
+ * Code that starts some processing that's exempt from cancellation, but
+   uses a logging context from cancellable code. The logging context
+   will be finished upon cancellation, while the uncancelled processing
+   is still using it.
+
+Some common patterns are listed below in more detail.
+
+### `async` function calls
+Most functions in Synapse are relatively straightforward from a
+cancellation standpoint: they don't do anything with `Deferred`s and
+purely call and `await` other `async` functions.
+
+An `async` function handles cancellation correctly if its own code
+handles cancellation correctly and all the async function it calls
+handle cancellation correctly. For example:
+```python
+async def do_two_things() -> None:
+    check_something()
+    await do_something()
+    await do_something_else()
+```
+`do_two_things` handles cancellation correctly if `do_something` and
+`do_something_else` handle cancellation correctly.
+
+That is, when checking whether a function handles cancellation
+correctly, its implementation and all its `async` function calls need to
+be checked, recursively.
+
+As `check_something` is not `async`, it does not need to be checked.
+
+### CancelledErrors
+Because Twisted's `CancelledError`s are `Exception`s, it's easy to
+accidentally catch and suppress them. Care must be taken to ensure that
+`CancelledError`s are allowed to propagate upwards.
+
+<table width="100%">
+<tr>
+<td width="50%" valign="top">
+
+**Bad**:
+```python
+try:
+    await do_something()
+except Exception:
+    # `CancelledError` gets swallowed here.
+    logger.info(...)
+```
+</td>
+<td width="50%" valign="top">
+
+**Good**:
+```python
+try:
+    await do_something()
+except CancelledError:
+    raise
+except Exception:
+    logger.info(...)
+```
+</td>
+</tr>
+<tr>
+<td width="50%" valign="top">
+
+**OK**:
+```python
+try:
+    check_something()
+    # A `CancelledError` won't ever be raised here.
+except Exception:
+    logger.info(...)
+```
+</td>
+<td width="50%" valign="top">
+
+**Good**:
+```python
+try:
+    await do_something()
+except ValueError:
+    logger.info(...)
+```
+</td>
+</tr>
+</table>
+
+#### defer.gatherResults
+`defer.gatherResults` produces a `Deferred` which:
+ * broadcasts `cancel()` calls to every `Deferred` being waited on.
+ * wraps the first exception it sees in a `FirstError`.
+
+Together, this means that `CancelledError`s will be wrapped in
+a `FirstError` unless unwrapped. Such `FirstError`s are liable to be
+swallowed, so they must be unwrapped.
+
+<table width="100%">
+<tr>
+<td width="50%" valign="top">
+
+**Bad**:
+```python
+async def do_something() -> None:
+    await make_deferred_yieldable(
+        defer.gatherResults([...], consumeErrors=True)
+    )
+
+try:
+    await do_something()
+except CancelledError:
+    raise
+except Exception:
+    # `FirstError(CancelledError)` gets swallowed here.
+    logger.info(...)
+```
+
+</td>
+<td width="50%" valign="top">
+
+**Good**:
+```python
+async def do_something() -> None:
+    await make_deferred_yieldable(
+        defer.gatherResults([...], consumeErrors=True)
+    ).addErrback(unwrapFirstError)
+
+try:
+    await do_something()
+except CancelledError:
+    raise
+except Exception:
+    logger.info(...)
+```
+</td>
+</tr>
+</table>
+
+### Creation of `Deferred`s
+If a function creates a `Deferred`, the effect of cancelling it must be considered. `Deferred`s that get shared are likely to have unintended behaviour when cancelled.
+
+<table width="100%">
+<tr>
+<td width="50%" valign="top">
+
+**Bad**:
+```python
+cache: Dict[str, Deferred[None]] = {}
+
+def wait_for_room(room_id: str) -> Deferred[None]:
+    deferred = cache.get(room_id)
+    if deferred is None:
+        deferred = Deferred()
+        cache[room_id] = deferred
+    # `deferred` can have multiple waiters.
+    # All of them will observe a `CancelledError`
+    # if any one of them is cancelled.
+    return make_deferred_yieldable(deferred)
+
+# Request 1
+await wait_for_room("!aAAaaAaaaAAAaAaAA:matrix.org")
+# Request 2
+await wait_for_room("!aAAaaAaaaAAAaAaAA:matrix.org")
+```
+</td>
+<td width="50%" valign="top">
+
+**Good**:
+```python
+cache: Dict[str, Deferred[None]] = {}
+
+def wait_for_room(room_id: str) -> Deferred[None]:
+    deferred = cache.get(room_id)
+    if deferred is None:
+        deferred = Deferred()
+        cache[room_id] = deferred
+    # `deferred` will never be cancelled now.
+    # A `CancelledError` will still come out of
+    # the `await`.
+    # `delay_cancellation` may also be used.
+    return make_deferred_yieldable(stop_cancellation(deferred))
+
+# Request 1
+await wait_for_room("!aAAaaAaaaAAAaAaAA:matrix.org")
+# Request 2
+await wait_for_room("!aAAaaAaaaAAAaAaAA:matrix.org")
+```
+</td>
+</tr>
+<tr>
+<td width="50%" valign="top">
+</td>
+<td width="50%" valign="top">
+
+**Good**:
+```python
+cache: Dict[str, List[Deferred[None]]] = {}
+
+def wait_for_room(room_id: str) -> Deferred[None]:
+    if room_id not in cache:
+        cache[room_id] = []
+    # Each request gets its own `Deferred` to wait on.
+    deferred = Deferred()
+    cache[room_id]].append(deferred)
+    return make_deferred_yieldable(deferred)
+
+# Request 1
+await wait_for_room("!aAAaaAaaaAAAaAaAA:matrix.org")
+# Request 2
+await wait_for_room("!aAAaaAaaaAAAaAaAA:matrix.org")
+```
+</td>
+</table>
+
+### Uncancelled processing
+Some `async` functions may kick off some `async` processing which is
+intentionally protected from cancellation, by `stop_cancellation` or
+other means. If the `async` processing inherits the logcontext of the
+request which initiated it, care must be taken to ensure that the
+logcontext is not finished before the `async` processing completes.
+
+<table width="100%">
+<tr>
+<td width="50%" valign="top">
+
+**Bad**:
+```python
+cache: Optional[ObservableDeferred[None]] = None
+
+async def do_something_else(
+    to_resolve: Deferred[None]
+) -> None:
+    await ...
+    logger.info("done!")
+    to_resolve.callback(None)
+
+async def do_something() -> None:
+    if not cache:
+        to_resolve = Deferred()
+        cache = ObservableDeferred(to_resolve)
+        # `do_something_else` will never be cancelled and
+        # can outlive the `request-1` logging context.
+        run_in_background(do_something_else, to_resolve)
+
+    await make_deferred_yieldable(cache.observe())
+
+with LoggingContext("request-1"):
+    await do_something()
+```
+</td>
+<td width="50%" valign="top">
+
+**Good**:
+```python
+cache: Optional[ObservableDeferred[None]] = None
+
+async def do_something_else(
+    to_resolve: Deferred[None]
+) -> None:
+    await ...
+    logger.info("done!")
+    to_resolve.callback(None)
+
+async def do_something() -> None:
+    if not cache:
+        to_resolve = Deferred()
+        cache = ObservableDeferred(to_resolve)
+        run_in_background(do_something_else, to_resolve)
+        # We'll wait until `do_something_else` is
+        # done before raising a `CancelledError`.
+        await make_deferred_yieldable(
+            delay_cancellation(cache.observe())
+        )
+    else:
+        await make_deferred_yieldable(cache.observe())
+
+with LoggingContext("request-1"):
+    await do_something()
+```
+</td>
+</tr>
+<tr>
+<td width="50%">
+
+**OK**:
+```python
+cache: Optional[ObservableDeferred[None]] = None
+
+async def do_something_else(
+    to_resolve: Deferred[None]
+) -> None:
+    await ...
+    logger.info("done!")
+    to_resolve.callback(None)
+
+async def do_something() -> None:
+    if not cache:
+        to_resolve = Deferred()
+        cache = ObservableDeferred(to_resolve)
+        # `do_something_else` will get its own independent
+        # logging context. `request-1` will not count any
+        # metrics from `do_something_else`.
+        run_as_background_process(
+            "do_something_else",
+            do_something_else,
+            to_resolve,
+        )
+
+    await make_deferred_yieldable(cache.observe())
+
+with LoggingContext("request-1"):
+    await do_something()
+```
+</td>
+<td width="50%">
+</td>
+</tr>
+</table>

docs/modules/spam_checker_callbacks.md

@@ -11,22 +11,29 @@ The available spam checker callbacks are:
 ### `check_event_for_spam`
 
 _First introduced in Synapse v1.37.0_
+_Signature extended to support Allow and Code in Synapse v1.60.0_
+_Boolean and string return value types deprecated in Synapse v1.60.0_
 
 ```python
-async def check_event_for_spam(event: "synapse.events.EventBase") -> Union[bool, str]
+async def check_event_for_spam(event: "synapse.module_api.EventBase") -> Union["synapse.module_api.ALLOW", "synapse.module_api.error.Codes", str, bool]
 ```
 
-Called when receiving an event from a client or via federation. The callback must return
+Called when receiving an event from a client or via federation. The callback must return either:
-either:
+  - `synapse.module_api.ALLOW`, to allow the operation. Other callbacks
-- an error message string, to indicate the event must be rejected because of spam and 
+    may still decide to reject it.
-  give a rejection reason to forward to clients;
+  - `synapse.api.Codes` to reject the operation with an error code. In case
-- the boolean `True`, to indicate that the event is spammy, but not provide further details; or
+    of doubt, `synapse.api.error.Codes.FORBIDDEN` is a good error code.
-- the booelan `False`, to indicate that the event is not considered spammy.
+  - (deprecated) a `str` to reject the operation and specify an error message. Note that clients
+    typically will not localize the error message to the user's preferred locale.
+  - (deprecated) on `False`, behave as `ALLOW`. Deprecated as confusing, as some
+    callbacks in expect `True` to allow and others `True` to reject.
+  - (deprecated) on `True`, behave as `synapse.api.error.Codes.FORBIDDEN`. Deprecated as confusing, as
+    some callbacks in expect `True` to allow and others `True` to reject.
 
 If multiple modules implement this callback, they will be considered in order. If a
-callback returns `False`, Synapse falls through to the next one. The value of the first
+callback returns `synapse.module_api.ALLOW`, Synapse falls through to the next one. The value of the
-callback that does not return `False` will be used. If this happens, Synapse will not call
+first callback that does not return `synapse.module_api.ALLOW` will be used. If this happens, Synapse
-any of the subsequent implementations of this callback.
+will not call any of the subsequent implementations of this callback.
 
 ### `user_may_join_room`
 
@@ -249,6 +256,24 @@ callback returns `False`, Synapse falls through to the next one. The value of th
 callback that does not return `False` will be used. If this happens, Synapse will not call
 any of the subsequent implementations of this callback.
 
+### `should_drop_federated_event`
+
+_First introduced in Synapse v1.60.0_
+
+```python
+async def should_drop_federated_event(event: "synapse.events.EventBase") -> bool
+```
+
+Called when checking whether a remote server can federate an event with us. **Returning
+`True` from this function will silently drop a federated event and split-brain our view
+of a room's DAG, and thus you shouldn't use this callback unless you know what you are
+doing.**
+
+If multiple modules implement this callback, they will be considered in order. If a
+callback returns `False`, Synapse falls through to the next one. The value of the first
+callback that does not return `False` will be used. If this happens, Synapse will not call
+any of the subsequent implementations of this callback.
+
 ## Example
 
 The example below is a module that implements the spam checker callback

docs/openid.md

@@ -159,7 +159,7 @@ Follow the [Getting Started Guide](https://www.keycloak.org/getting-started) to
 oidc_providers:
   - idp_id: keycloak
     idp_name: "My KeyCloak server"
-    issuer: "https://127.0.0.1:8443/auth/realms/{realm_name}"
+    issuer: "https://127.0.0.1:8443/realms/{realm_name}"
     client_id: "synapse"
     client_secret: "copy secret generated from above"
     scopes: ["openid", "profile"]
@@ -293,7 +293,7 @@ can be used to retrieve information on the authenticated user. As the Synapse
 login mechanism needs an attribute to uniquely identify users, and that endpoint
 does not return a `sub` property, an alternative `subject_claim` has to be set.
 
-1. Create a new OAuth application: https://github.com/settings/applications/new.
+1. Create a new OAuth application: [https://github.com/settings/applications/new](https://github.com/settings/applications/new).
 2. Set the callback URL to `[synapse public baseurl]/_synapse/client/oidc/callback`.
 
 Synapse config:
@@ -322,10 +322,10 @@ oidc_providers:
 
 [Google][google-idp] is an OpenID certified authentication and authorisation provider.
 
-1. Set up a project in the Google API Console (see
+1. Set up a project in the Google API Console (see 
-   https://developers.google.com/identity/protocols/oauth2/openid-connect#appsetup).
+   [documentation](https://developers.google.com/identity/protocols/oauth2/openid-connect#appsetup)).
-2. Add an "OAuth Client ID" for a Web Application under "Credentials".
+3. Add an "OAuth Client ID" for a Web Application under "Credentials".
-3. Copy the Client ID and Client Secret, and add the following to your synapse config:
+4. Copy the Client ID and Client Secret, and add the following to your synapse config:
    ```yaml
    oidc_providers:
      - idp_id: google
@@ -501,8 +501,8 @@ As well as the private key file, you will need:
  * Team ID: a 10-character ID associated with your developer account.
  * Key ID: the 10-character identifier for the key.
 
-https://help.apple.com/developer-account/?lang=en#/dev77c875b7e has more
+[Apple's developer documentation](https://help.apple.com/developer-account/?lang=en#/dev77c875b7e)
-documentation on setting up SiWA.
+has more information on setting up SiWA.
 
 The synapse config will look like this:
 
@@ -535,8 +535,8 @@ needed to add OAuth2 capabilities to your Django projects. It supports
 
 Configuration on Django's side:
 
-1. Add an application: https://example.com/admin/oauth2_provider/application/add/ and choose parameters like this:
+1. Add an application: `https://example.com/admin/oauth2_provider/application/add/` and choose parameters like this:
-* `Redirect uris`: https://synapse.example.com/_synapse/client/oidc/callback
+* `Redirect uris`: `https://synapse.example.com/_synapse/client/oidc/callback`
 * `Client type`: `Confidential`
 * `Authorization grant type`: `Authorization code`
 * `Algorithm`: `HMAC with SHA-2 256`

docs/sample_config.yaml

@@ -289,7 +289,7 @@ presence:
 #   federation: the server-server API (/_matrix/federation). Also implies
 #       'media', 'keys', 'openid'
 #
-#   keys: the key discovery API (/_matrix/keys).
+#   keys: the key discovery API (/_matrix/key).
 #
 #   media: the media API (/_matrix/media).
 #
@@ -730,6 +730,12 @@ retention:
 # A cache 'factor' is a multiplier that can be applied to each of
 # Synapse's caches in order to increase or decrease the maximum
 # number of entries that can be stored.
+#
+# The configuration for cache factors (caches.global_factor and
+# caches.per_cache_factors) can be reloaded while the application is running,
+# by sending a SIGHUP signal to the Synapse process. Changes to other parts of
+# the caching config will NOT be applied after a SIGHUP is received; a restart
+# is necessary.
 
 # The number of events to cache in memory. Not affected by
 # caches.global_factor.
@@ -778,6 +784,24 @@ caches:
   #
   #cache_entry_ttl: 30m
 
+  # This flag enables cache autotuning, and is further specified by the sub-options `max_cache_memory_usage`,
+  # `target_cache_memory_usage`, `min_cache_ttl`. These flags work in conjunction with each other to maintain
+  # a balance between cache memory usage and cache entry availability. You must be using jemalloc to utilize
+  # this option, and all three of the options must be specified for this feature to work.
+  #cache_autotuning:
+    # This flag sets a ceiling on much memory the cache can use before caches begin to be continuously evicted.
+    # They will continue to be evicted until the memory usage drops below the `target_memory_usage`, set in
+    # the flag below, or until the `min_cache_ttl` is hit.
+    #max_cache_memory_usage: 1024M
+
+    # This flag sets a rough target for the desired memory usage of the caches.
+    #target_cache_memory_usage: 758M
+
+    # 'min_cache_ttl` sets a limit under which newer cache entries are not evicted and is only applied when
+    # caches are actively being evicted/`max_cache_memory_usage` has been exceeded. This is to protect hot caches
+    # from being emptied while Synapse is evicting due to memory.
+    #min_cache_ttl: 5m
+
   # Controls how long the results of a /sync request are cached for after
   # a successful response is returned. A higher duration can help clients with
   # intermittent connections, at the cost of higher memory usage.
@@ -2462,6 +2486,40 @@ push:
 #
 #encryption_enabled_by_default_for_room_type: invite
 
+# Override the default power levels for rooms created on this server, per
+# room creation preset.
+#
+# The appropriate dictionary for the room preset will be applied on top
+# of the existing power levels content.
+#
+# Useful if you know that your users need special permissions in rooms
+# that they create (e.g. to send particular types of state events without
+# needing an elevated power level).  This takes the same shape as the
+# `power_level_content_override` parameter in the /createRoom API, but
+# is applied before that parameter.
+#
+# Valid keys are some or all of `private_chat`, `trusted_private_chat`
+# and `public_chat`. Inside each of those should be any of the
+# properties allowed in `power_level_content_override` in the
+# /createRoom API. If any property is missing, its default value will
+# continue to be used. If any property is present, it will overwrite
+# the existing default completely (so if the `events` property exists,
+# the default event power levels will be ignored).
+#
+#default_power_level_content_override:
+#    private_chat:
+#        "events":
+#            "com.example.myeventtype" : 0
+#            "m.room.avatar": 50
+#            "m.room.canonical_alias": 50
+#            "m.room.encryption": 100
+#            "m.room.history_visibility": 100
+#            "m.room.name": 50
+#            "m.room.power_levels": 100
+#            "m.room.server_acl": 100
+#            "m.room.tombstone": 100
+#        "events_default": 1
+
 
 # Uncomment to allow non-server-admin users to create groups on this server
 #

docs/upgrade.md

@@ -89,6 +89,125 @@ process, for example:
     dpkg -i matrix-synapse-py3_1.3.0+stretch1_amd64.deb
     ```
 
+# Upgrading to v1.60.0
+
+## Adding a new unique index to `state_group_edges` could fail if your database is corrupted
+
+This release of Synapse will add a unique index to the `state_group_edges` table, in order
+to prevent accidentally introducing duplicate information (for example, because a database
+backup was restored multiple times).
+
+Duplicate rows being present in this table could cause drastic performance problems; see
+[issue 11779](https://github.com/matrix-org/synapse/issues/11779) for more details.
+
+If your Synapse database already has had duplicate rows introduced into this table,
+this could fail, with either of these errors:
+
+
+**On Postgres:**
+```
+synapse.storage.background_updates - 623 - INFO - background_updates-0 - Adding index state_group_edges_unique_idx to state_group_edges
+synapse.storage.background_updates - 282 - ERROR - background_updates-0 - Error doing update
+...
+psycopg2.errors.UniqueViolation: could not create unique index "state_group_edges_unique_idx"
+DETAIL:  Key (state_group, prev_state_group)=(2, 1) is duplicated.
+```
+(The numbers may be different.)
+
+**On SQLite:**
+```
+synapse.storage.background_updates - 623 - INFO - background_updates-0 - Adding index state_group_edges_unique_idx to state_group_edges
+synapse.storage.background_updates - 282 - ERROR - background_updates-0 - Error doing update
+...
+sqlite3.IntegrityError: UNIQUE constraint failed: state_group_edges.state_group, state_group_edges.prev_state_group
+```
+
+
+<details>
+<summary><b>Expand this section for steps to resolve this problem</b></summary>
+
+### On Postgres
+
+Connect to your database with `psql`.
+
+```sql
+BEGIN;
+DELETE FROM state_group_edges WHERE (ctid, state_group, prev_state_group) IN (
+  SELECT row_id, state_group, prev_state_group
+  FROM (
+    SELECT
+      ctid AS row_id,
+      MIN(ctid) OVER (PARTITION BY state_group, prev_state_group) AS min_row_id,
+      state_group,
+      prev_state_group
+    FROM state_group_edges
+  ) AS t1
+  WHERE row_id <> min_row_id
+);
+COMMIT;
+```
+
+
+### On SQLite
+
+At the command-line, use `sqlite3 path/to/your-homeserver-database.db`:
+
+```sql
+BEGIN;
+DELETE FROM state_group_edges WHERE (rowid, state_group, prev_state_group) IN (
+  SELECT row_id, state_group, prev_state_group
+  FROM (
+    SELECT
+      rowid AS row_id,
+      MIN(rowid) OVER (PARTITION BY state_group, prev_state_group) AS min_row_id,
+      state_group,
+      prev_state_group
+    FROM state_group_edges
+  )
+  WHERE row_id <> min_row_id
+);
+COMMIT;
+```
+
+
+### For more details
+
+[This comment on issue 11779](https://github.com/matrix-org/synapse/issues/11779#issuecomment-1131545970)
+has queries that can be used to check a database for this problem in advance.
+
+</details>
+
+## SpamChecker API's `check_event_for_spam` has a new signature.
+
+The previous signature has been deprecated.
+
+Whereas `check_event_for_spam` callbacks used to return `Union[str, bool]`, they should now return `Union["synapse.module_api.Allow", "synapse.module_api.errors.Codes"]`.
+
+This is part of an ongoing refactoring of the SpamChecker API to make it less ambiguous and more powerful.
+
+If your module implements `check_event_for_spam` as follows:
+
+```python
+async def check_event_for_spam(event):
+    if ...:
+        # Event is spam
+        return True
+    # Event is not spam
+    return False
+```
+
+you should rewrite it as follows:
+
+```python
+async def check_event_for_spam(event):
+    if ...:
+        # Event is spam, mark it as forbidden (you may use some more precise error
+        # code if it is useful).
+        return synapse.module_api.errors.Codes.FORBIDDEN
+    # Event is not spam, mark it as `ALLOW`.
+    return synapse.module_api.ALLOW
+```
+
 # Upgrading to v1.59.0
 
 ## Device name lookup over federation has been disabled by default

docs/usage/configuration/config_documentation.md

@@ -23,6 +23,14 @@ followed by a letter. Letters have the following meanings:
 For example, setting `redaction_retention_period: 5m` would remove redacted
 messages from the database after 5 minutes, rather than 5 months.
 
+In addition, configuration options referring to size use the following suffixes:
+
+* `M` = MiB, or 1,048,576 bytes
+* `K` = KiB, or 1024 bytes 
+
+For example, setting `max_avatar_size: 10M` means that Synapse will not accept files larger than 10,485,760 bytes
+for a user avatar. 
+
 ### YAML 
 The configuration file is a [YAML](https://yaml.org/) file, which means that certain syntax rules
 apply if you want your config file to be read properly. A few helpful things to know:
@@ -467,13 +475,13 @@ Sub-options for each listener include:
 
 Valid resource names are:
 
-* `client`: the client-server API (/_matrix/client), and the synapse admin API (/_synapse/admin). Also implies 'media' and 'static'.
+* `client`: the client-server API (/_matrix/client), and the synapse admin API (/_synapse/admin). Also implies `media` and `static`.
 
 * `consent`: user consent forms (/_matrix/consent). See [here](../../consent_tracking.md) for more.
 
 * `federation`: the server-server API (/_matrix/federation). Also implies `media`, `keys`, `openid`
 
-* `keys`: the key discovery API (/_matrix/keys).
+* `keys`: the key discovery API (/_matrix/key).
 
 * `media`: the media API (/_matrix/media).
 
@@ -1119,7 +1127,22 @@ Caching can be configured through the following sub-options:
   with intermittent connections, at the cost of higher memory usage.
   By default, this is zero, which means that sync responses are not cached
   at all.
-
+* `cache_autotuning` and its sub-options `max_cache_memory_usage`, `target_cache_memory_usage`, and
+   `min_cache_ttl` work in conjunction with each other to maintain a balance between cache memory 
+   usage and cache entry availability. You must be using [jemalloc](https://github.com/matrix-org/synapse#help-synapse-is-slow-and-eats-all-my-ramcpu) 
+   to utilize this option, and all three of the options must be specified for this feature to work. This option
+   defaults to off, enable it by providing values for the sub-options listed below. Please note that the feature will not work
+   and may cause unstable behavior (such as excessive emptying of caches or exceptions) if all of the values are not provided.
+   Please see the [Config Conventions](#config-conventions) for information on how to specify memory size and cache expiry
+   durations.
+     * `max_cache_memory_usage` sets a ceiling on how much memory the cache can use before caches begin to be continuously evicted.
+        They will continue to be evicted until the memory usage drops below the `target_memory_usage`, set in
+        the setting below, or until the `min_cache_ttl` is hit. There is no default value for this option.
+     * `target_memory_usage` sets a rough target for the desired memory usage of the caches. There is no default value
+        for this option.
+     * `min_cache_ttl` sets a limit under which newer cache entries are not evicted and is only applied when
+        caches are actively being evicted/`max_cache_memory_usage` has been exceeded. This is to protect hot caches
+        from being emptied while Synapse is evicting due to memory. There is no default value for this option. 
 
 Example configuration:
 ```yaml
@@ -1127,9 +1150,29 @@ caches:
   global_factor: 1.0
   per_cache_factors:
     get_users_who_share_room_with_user: 2.0
-  expire_caches: false
   sync_response_cache_duration: 2m
+  cache_autotuning:
+    max_cache_memory_usage: 1024M
+    target_cache_memory_usage: 758M
+    min_cache_ttl: 5m
 ```
+
+### Reloading cache factors
+
+The cache factors (i.e. `caches.global_factor` and `caches.per_cache_factors`)  may be reloaded at any time by sending a
+[`SIGHUP`](https://en.wikipedia.org/wiki/SIGHUP) signal to Synapse using e.g.
+
+```commandline
+kill -HUP [PID_OF_SYNAPSE_PROCESS]
+```
+
+If you are running multiple workers, you must individually update the worker 
+config file and send this signal to each worker process.
+
+If you're using the [example systemd service](https://github.com/matrix-org/synapse/blob/develop/contrib/systemd/matrix-synapse.service)
+file in Synapse's `contrib` directory, you can send a `SIGHUP` signal by using
+`systemctl reload matrix-synapse`.
+
 ---
 ## Database ##
 Config options related to database settings.
@@ -1164,7 +1207,7 @@ For more information on using Synapse with Postgres,
 see [here](../../postgres.md).
 
 Example SQLite configuration:
-```
+```yaml
 database:
   name: sqlite3
   args:
@@ -1172,7 +1215,7 @@ database:
 ```
 
 Example Postgres configuration:
-```
+```yaml
 database:
   name: psycopg2
   txn_limit: 10000
@@ -1327,6 +1370,20 @@ This option sets ratelimiting how often invites can be sent in a room or to a
 specific user. `per_room` defaults to `per_second: 0.3`, `burst_count: 10` and
 `per_user` defaults to `per_second: 0.003`, `burst_count: 5`. 
 
+Client requests that invite user(s) when [creating a
+room](https://spec.matrix.org/v1.2/client-server-api/#post_matrixclientv3createroom)
+will count against the `rc_invites.per_room` limit, whereas
+client requests to [invite a single user to a
+room](https://spec.matrix.org/v1.2/client-server-api/#post_matrixclientv3roomsroomidinvite)
+will count against both the `rc_invites.per_user` and `rc_invites.per_room` limits.
+
+Federation requests to invite a user will count against the `rc_invites.per_user`
+limit only, as Synapse presumes ratelimiting by room will be done by the sending server.
+
+The `rc_invites.per_user` limit applies to the *receiver* of the invite, rather than the
+sender, meaning that a `rc_invite.per_user.burst_count` of 5 mandates that a single user
+cannot *receive* more than a burst of 5 invites at a time.
+
 Example configuration:
 ```yaml
 rc_invites:
@@ -1635,10 +1692,10 @@ Defaults to "en".
 Example configuration:
 ```yaml
  url_preview_accept_language:
-   - en-UK
+   - 'en-UK'
-   - en-US;q=0.9
+   - 'en-US;q=0.9'
-   - fr;q=0.8
+   - 'fr;q=0.8'
-   - *;q=0.7
+   - '*;q=0.7'
 ```
 ----
 Config option: `oembed`
@@ -3298,6 +3355,32 @@ room_list_publication_rules:
     room_id: "*"
     action: allow
 ```
+
+---
+Config option: `default_power_level_content_override`
+
+The `default_power_level_content_override` option controls the default power
+levels for rooms.
+
+Useful if you know that your users need special permissions in rooms
+that they create (e.g. to send particular types of state events without
+needing an elevated power level).  This takes the same shape as the
+`power_level_content_override` parameter in the /createRoom API, but
+is applied before that parameter.
+
+Note that each key provided inside a preset (for example `events` in the example
+below) will overwrite all existing defaults inside that key. So in the example
+below, newly-created private_chat rooms will have no rules for any event types
+except `com.example.foo`.
+
+Example configuration:
+```yaml
+default_power_level_content_override:
+   private_chat: { "events": { "com.example.foo" : 0 } }
+   trusted_private_chat: null
+   public_chat: null
+```
+
 ---
 ## Opentracing ##
 Configuration options related to Opentracing support.
@@ -3398,7 +3481,7 @@ stream_writers:
   typing: worker1
 ```
 ---
-Config option: `run_background_task_on`
+Config option: `run_background_tasks_on`
 
 The worker that is used to run background tasks (e.g. cleaning up expired
 data). If not provided this defaults to the main process.

docs/welcome_and_overview.md

@@ -7,10 +7,10 @@ team.
 ## Installing and using Synapse
 
 This documentation covers topics for **installation**, **configuration** and
-**maintainence** of your Synapse process:
+**maintenance** of your Synapse process:
 
 * Learn how to [install](setup/installation.md) and
-  [configure](usage/configuration/index.html) your own instance, perhaps with [Single
+  [configure](usage/configuration/config_documentation.md) your own instance, perhaps with [Single
   Sign-On](usage/configuration/user_authentication/index.html).
 
 * See how to [upgrade](upgrade.md) between Synapse versions.
@@ -65,7 +65,7 @@ following documentation:
 
 Want to help keep Synapse going but don't know how to code? Synapse is a
 [Matrix.org Foundation](https://matrix.org) project. Consider becoming a
-supportor on [Liberapay](https://liberapay.com/matrixdotorg),
+supporter on [Liberapay](https://liberapay.com/matrixdotorg),
 [Patreon](https://patreon.com/matrixdotorg) or through
 [PayPal](https://paypal.me/matrixdotorg) via a one-time donation.
 

docs/workers.md

@@ -251,6 +251,8 @@ information.
     # Presence requests
     ^/_matrix/client/(api/v1|r0|v3|unstable)/presence/
 
+    # User directory search requests
+    ^/_matrix/client/(r0|v3|unstable)/user_directory/search$
 
 Additionally, the following REST endpoints can be handled for GET requests:
 
@@ -448,6 +450,14 @@ update_user_directory_from_worker: worker_name
 This work cannot be load-balanced; please ensure the main process is restarted
 after setting this option in the shared configuration!
 
+User directory updates allow REST endpoints matching the following regular
+expressions to work:
+
+    ^/_matrix/client/(r0|v3|unstable)/user_directory/search$
+
+The above endpoints can be routed to any worker, though you may choose to route
+it to the chosen user directory worker.
+
 This style of configuration supersedes the legacy `synapse.app.user_dir`
 worker application type.
 

mypy.ini

@@ -10,6 +10,7 @@ warn_unreachable = True
 warn_unused_ignores = True
 local_partial_types = True
 no_implicit_optional = True
+disallow_untyped_defs = True
 
 files =
   docker/,
@@ -27,9 +28,6 @@ exclude = (?x)
    |synapse/storage/databases/__init__.py
    |synapse/storage/databases/main/cache.py
    |synapse/storage/databases/main/devices.py
-   |synapse/storage/databases/main/event_federation.py
-   |synapse/storage/databases/main/push_rule.py
-   |synapse/storage/databases/main/roommember.py
    |synapse/storage/schema/
 
    |tests/api/test_auth.py
@@ -43,16 +41,11 @@ exclude = (?x)
    |tests/events/test_utils.py
    |tests/federation/test_federation_catch_up.py
    |tests/federation/test_federation_sender.py
-   |tests/federation/test_federation_server.py
    |tests/federation/transport/test_knocking.py
-   |tests/federation/transport/test_server.py
    |tests/handlers/test_typing.py
    |tests/http/federation/test_matrix_federation_agent.py
    |tests/http/federation/test_srv_resolver.py
-   |tests/http/test_fedclient.py
    |tests/http/test_proxyagent.py
-   |tests/http/test_servlet.py
-   |tests/http/test_site.py
    |tests/logging/__init__.py
    |tests/logging/test_terse_json.py
    |tests/module_api/test_api.py
@@ -61,12 +54,9 @@ exclude = (?x)
    |tests/push/test_push_rule_evaluator.py
    |tests/rest/client/test_transactions.py
    |tests/rest/media/v1/test_media_storage.py
-   |tests/scripts/test_new_matrix_user.py
    |tests/server.py
    |tests/server_notices/test_resource_limits_server_notices.py
    |tests/state/test_v2.py
-   |tests/storage/test_base.py
-   |tests/storage/test_roommember.py
    |tests/test_metrics.py
    |tests/test_server.py
    |tests/test_state.py
@@ -89,131 +79,39 @@ exclude = (?x)
    |tests/utils.py
    )$
 
-[mypy-synapse._scripts.*]
-disallow_untyped_defs = True
-
-[mypy-synapse.api.*]
-disallow_untyped_defs = True
-
-[mypy-synapse.app.*]
-disallow_untyped_defs = True
-
-[mypy-synapse.appservice.*]
-disallow_untyped_defs = True
-
-[mypy-synapse.config.*]
-disallow_untyped_defs = True
-
-[mypy-synapse.crypto.*]
-disallow_untyped_defs = True
-
-[mypy-synapse.event_auth]
-disallow_untyped_defs = True
-
-[mypy-synapse.events.*]
-disallow_untyped_defs = True
-
-[mypy-synapse.federation.*]
-disallow_untyped_defs = True
-
 [mypy-synapse.federation.transport.client]
 disallow_untyped_defs = False
 
-[mypy-synapse.handlers.*]
+[mypy-synapse.http.client]
-disallow_untyped_defs = True
+disallow_untyped_defs = False
 
-[mypy-synapse.http.server]
+[mypy-synapse.http.matrixfederationclient]
-disallow_untyped_defs = True
+disallow_untyped_defs = False
 
-[mypy-synapse.logging.context]
+[mypy-synapse.logging.opentracing]
-disallow_untyped_defs = True
+disallow_untyped_defs = False
 
-[mypy-synapse.metrics.*]
+[mypy-synapse.logging.scopecontextmanager]
-disallow_untyped_defs = True
+disallow_untyped_defs = False
 
 [mypy-synapse.metrics._reactor_metrics]
+disallow_untyped_defs = False
 # This module imports select.epoll. That exists on Linux, but doesn't on macOS.
 # See https://github.com/matrix-org/synapse/pull/11771.
 warn_unused_ignores = False
 
-[mypy-synapse.module_api.*]
-disallow_untyped_defs = True
-
-[mypy-synapse.notifier]
-disallow_untyped_defs = True
-
-[mypy-synapse.push.*]
-disallow_untyped_defs = True
-
-[mypy-synapse.replication.*]
-disallow_untyped_defs = True
-
-[mypy-synapse.rest.*]
-disallow_untyped_defs = True
-
-[mypy-synapse.server_notices.*]
-disallow_untyped_defs = True
-
-[mypy-synapse.state.*]
-disallow_untyped_defs = True
-
-[mypy-synapse.storage.databases.main.account_data]
-disallow_untyped_defs = True
-
-[mypy-synapse.storage.databases.main.client_ips]
-disallow_untyped_defs = True
-
-[mypy-synapse.storage.databases.main.directory]
-disallow_untyped_defs = True
-
-[mypy-synapse.storage.databases.main.e2e_room_keys]
-disallow_untyped_defs = True
-
-[mypy-synapse.storage.databases.main.end_to_end_keys]
-disallow_untyped_defs = True
-
-[mypy-synapse.storage.databases.main.event_push_actions]
-disallow_untyped_defs = True
-
-[mypy-synapse.storage.databases.main.events_bg_updates]
-disallow_untyped_defs = True
-
-[mypy-synapse.storage.databases.main.events_worker]
-disallow_untyped_defs = True
-
-[mypy-synapse.storage.databases.main.room]
-disallow_untyped_defs = True
-
-[mypy-synapse.storage.databases.main.room_batch]
-disallow_untyped_defs = True
-
-[mypy-synapse.storage.databases.main.profile]
-disallow_untyped_defs = True
-
-[mypy-synapse.storage.databases.main.stats]
-disallow_untyped_defs = True
-
-[mypy-synapse.storage.databases.main.state_deltas]
-disallow_untyped_defs = True
-
-[mypy-synapse.storage.databases.main.transactions]
-disallow_untyped_defs = True
-
-[mypy-synapse.storage.databases.main.user_erasure_store]
-disallow_untyped_defs = True
-
-[mypy-synapse.storage.util.*]
-disallow_untyped_defs = True
-
-[mypy-synapse.streams.*]
-disallow_untyped_defs = True
-
-[mypy-synapse.util.*]
-disallow_untyped_defs = True
-
 [mypy-synapse.util.caches.treecache]
 disallow_untyped_defs = False
 
+[mypy-synapse.server]
+disallow_untyped_defs = False
+
+[mypy-synapse.storage.database]
+disallow_untyped_defs = False
+
+[mypy-tests.*]
+disallow_untyped_defs = False
+
 [mypy-tests.handlers.test_user_directory]
 disallow_untyped_defs = True
 

pyproject.toml

@@ -54,7 +54,7 @@ skip_gitignore = true
 
 [tool.poetry]
 name = "matrix-synapse"
-version = "1.59.1"
+version = "1.60.0rc1"
 description = "Homeserver for the Matrix decentralised comms protocol"
 authors = ["Matrix.org Team and Contributors <packages@matrix.org>"]
 license = "Apache-2.0"

scripts-dev/complement.sh

@@ -45,6 +45,8 @@ docker build -t matrixdotorg/synapse -f "docker/Dockerfile" .
 
 extra_test_args=()
 
+test_tags="synapse_blacklist,msc2716,msc3030"
+
 # If we're using workers, modify the docker files slightly.
 if [[ -n "$WORKERS" ]]; then
   # Build the workers docker image (from the base Synapse image).
@@ -65,6 +67,10 @@ if [[ -n "$WORKERS" ]]; then
 else
   export COMPLEMENT_BASE_IMAGE=complement-synapse
   COMPLEMENT_DOCKERFILE=Dockerfile
+
+  # We only test faster room joins on monoliths, because they are purposefully
+  # being developed without worker support to start with.
+  test_tags="$test_tags,faster_joins"
 fi
 
 # Build the Complement image from the Synapse image we just built.
@@ -73,4 +79,5 @@ docker build -t $COMPLEMENT_BASE_IMAGE -f "docker/complement/$COMPLEMENT_DOCKERF
 # Run the tests!
 echo "Images built; running complement"
 cd "$COMPLEMENT_DIR"
-go test -v -tags synapse_blacklist,msc2716,msc3030,faster_joins -count=1 "${extra_test_args[@]}" "$@" ./tests/...
+
+go test -v -tags $test_tags -count=1 "${extra_test_args[@]}" "$@" ./tests/...

scripts-dev/mypy_synapse_plugin.py

@@ -21,7 +21,7 @@ from typing import Callable, Optional, Type
 from mypy.nodes import ARG_NAMED_OPT
 from mypy.plugin import MethodSigContext, Plugin
 from mypy.typeops import bind_self
-from mypy.types import CallableType, NoneType
+from mypy.types import CallableType, NoneType, UnionType
 
 
 class SynapsePlugin(Plugin):
@@ -72,13 +72,20 @@ def cached_function_method_signature(ctx: MethodSigContext) -> CallableType:
 
     # Third, we add an optional "on_invalidate" argument.
     #
-    # This is a callable which accepts no input and returns nothing.
+    # This is a either
-    calltyp = CallableType(
+    # - a callable which accepts no input and returns nothing, or
-        arg_types=[],
+    # - None.
-        arg_kinds=[],
+    calltyp = UnionType(
-        arg_names=[],
+        [
-        ret_type=NoneType(),
+            NoneType(),
-        fallback=ctx.api.named_generic_type("builtins.function", []),
+            CallableType(
+                arg_types=[],
+                arg_kinds=[],
+                arg_names=[],
+                ret_type=NoneType(),
+                fallback=ctx.api.named_generic_type("builtins.function", []),
+            ),
+        ]
     )
 
     arg_types.append(calltyp)
@@ -95,7 +102,7 @@ def cached_function_method_signature(ctx: MethodSigContext) -> CallableType:
 
 
 def plugin(version: str) -> Type[SynapsePlugin]:
-    # This is the entry point of the plugin, and let's us deal with the fact
+    # This is the entry point of the plugin, and lets us deal with the fact
     # that the mypy plugin interface is *not* stable by looking at the version
     # string.
     #

synapse/_scripts/hash_password.py

@@ -46,14 +46,14 @@ def main() -> None:
             "Path to server config file. "
             "Used to read in bcrypt_rounds and password_pepper."
         ),
+        required=True,
     )
 
     args = parser.parse_args()
-    if "config" in args and args.config:
+    config = yaml.safe_load(args.config)
-        config = yaml.safe_load(args.config)
+    bcrypt_rounds = config.get("bcrypt_rounds", bcrypt_rounds)
-        bcrypt_rounds = config.get("bcrypt_rounds", bcrypt_rounds)
+    password_config = config.get("password_config", None) or {}
-        password_config = config.get("password_config", None) or {}
+    password_pepper = password_config.get("pepper", password_pepper)
-        password_pepper = password_config.get("pepper", password_pepper)
     password = args.password
 
     if not password:

synapse/api/constants.py

@@ -65,6 +65,8 @@ class JoinRules:
     PRIVATE: Final = "private"
     # As defined for MSC3083.
     RESTRICTED: Final = "restricted"
+    # As defined for MSC3787.
+    KNOCK_RESTRICTED: Final = "knock_restricted"
 
 
 class RestrictedJoinRuleTypes:

synapse/api/errors.py

@@ -17,6 +17,7 @@
 
 import logging
 import typing
+from enum import Enum
 from http import HTTPStatus
 from typing import Any, Dict, List, Optional, Union
 
@@ -30,7 +31,11 @@ if typing.TYPE_CHECKING:
 logger = logging.getLogger(__name__)
 
 
-class Codes:
+class Codes(str, Enum):
+    """
+    All known error codes, as an enum of strings.
+    """
+
     UNRECOGNIZED = "M_UNRECOGNIZED"
     UNAUTHORIZED = "M_UNAUTHORIZED"
     FORBIDDEN = "M_FORBIDDEN"

synapse/api/filtering.py

@@ -19,6 +19,7 @@ from typing import (
     TYPE_CHECKING,
     Awaitable,
     Callable,
+    Collection,
     Dict,
     Iterable,
     List,
@@ -444,9 +445,9 @@ class Filter:
         return room_ids
 
     async def _check_event_relations(
-        self, events: Iterable[FilterEvent]
+        self, events: Collection[FilterEvent]
     ) -> List[FilterEvent]:
-        # The event IDs to check, mypy doesn't understand the ifinstance check.
+        # The event IDs to check, mypy doesn't understand the isinstance check.
         event_ids = [event.event_id for event in events if isinstance(event, EventBase)]  # type: ignore[attr-defined]
         event_ids_to_keep = set(
             await self._store.events_have_relations(

synapse/api/room_versions.py

@@ -81,6 +81,9 @@ class RoomVersion:
     msc2716_historical: bool
     # MSC2716: Adds support for redacting "insertion", "chunk", and "marker" events
     msc2716_redactions: bool
+    # MSC3787: Adds support for a `knock_restricted` join rule, mixing concepts of
+    # knocks and restricted join rules into the same join condition.
+    msc3787_knock_restricted_join_rule: bool
 
 
 class RoomVersions:
@@ -99,6 +102,7 @@ class RoomVersions:
         msc2403_knocking=False,
         msc2716_historical=False,
         msc2716_redactions=False,
+        msc3787_knock_restricted_join_rule=False,
     )
     V2 = RoomVersion(
         "2",
@@ -115,6 +119,7 @@ class RoomVersions:
         msc2403_knocking=False,
         msc2716_historical=False,
         msc2716_redactions=False,
+        msc3787_knock_restricted_join_rule=False,
     )
     V3 = RoomVersion(
         "3",
@@ -131,6 +136,7 @@ class RoomVersions:
         msc2403_knocking=False,
         msc2716_historical=False,
         msc2716_redactions=False,
+        msc3787_knock_restricted_join_rule=False,
     )
     V4 = RoomVersion(
         "4",
@@ -147,6 +153,7 @@ class RoomVersions:
         msc2403_knocking=False,
         msc2716_historical=False,
         msc2716_redactions=False,
+        msc3787_knock_restricted_join_rule=False,
     )
     V5 = RoomVersion(
         "5",
@@ -163,6 +170,7 @@ class RoomVersions:
         msc2403_knocking=False,
         msc2716_historical=False,
         msc2716_redactions=False,
+        msc3787_knock_restricted_join_rule=False,
     )
     V6 = RoomVersion(
         "6",
@@ -179,6 +187,7 @@ class RoomVersions:
         msc2403_knocking=False,
         msc2716_historical=False,
         msc2716_redactions=False,
+        msc3787_knock_restricted_join_rule=False,
     )
     MSC2176 = RoomVersion(
         "org.matrix.msc2176",
@@ -195,6 +204,7 @@ class RoomVersions:
         msc2403_knocking=False,
         msc2716_historical=False,
         msc2716_redactions=False,
+        msc3787_knock_restricted_join_rule=False,
     )
     V7 = RoomVersion(
         "7",
@@ -211,6 +221,7 @@ class RoomVersions:
         msc2403_knocking=True,
         msc2716_historical=False,
         msc2716_redactions=False,
+        msc3787_knock_restricted_join_rule=False,
     )
     V8 = RoomVersion(
         "8",
@@ -227,6 +238,7 @@ class RoomVersions:
         msc2403_knocking=True,
         msc2716_historical=False,
         msc2716_redactions=False,
+        msc3787_knock_restricted_join_rule=False,
     )
     V9 = RoomVersion(
         "9",
@@ -243,6 +255,7 @@ class RoomVersions:
         msc2403_knocking=True,
         msc2716_historical=False,
         msc2716_redactions=False,
+        msc3787_knock_restricted_join_rule=False,
     )
     MSC2716v3 = RoomVersion(
         "org.matrix.msc2716v3",
@@ -259,6 +272,24 @@ class RoomVersions:
         msc2403_knocking=True,
         msc2716_historical=True,
         msc2716_redactions=True,
+        msc3787_knock_restricted_join_rule=False,
+    )
+    MSC3787 = RoomVersion(
+        "org.matrix.msc3787",
+        RoomDisposition.UNSTABLE,
+        EventFormatVersions.V3,
+        StateResolutionVersions.V2,
+        enforce_key_validity=True,
+        special_case_aliases_auth=False,
+        strict_canonicaljson=True,
+        limit_notifications_power_levels=True,
+        msc2176_redaction_rules=False,
+        msc3083_join_rules=True,
+        msc3375_redaction_rules=True,
+        msc2403_knocking=True,
+        msc2716_historical=False,
+        msc2716_redactions=False,
+        msc3787_knock_restricted_join_rule=True,
     )
 
 
@@ -276,6 +307,7 @@ KNOWN_ROOM_VERSIONS: Dict[str, RoomVersion] = {
         RoomVersions.V8,
         RoomVersions.V9,
         RoomVersions.MSC2716v3,
+        RoomVersions.MSC3787,
     )
 }
 

synapse/app/_base.py

@@ -49,9 +49,12 @@ from twisted.logger import LoggingFile, LogLevel
 from twisted.protocols.tls import TLSMemoryBIOFactory
 from twisted.python.threadpool import ThreadPool
 
+import synapse.util.caches
 from synapse.api.constants import MAX_PDU_SIZE
 from synapse.app import check_bind_error
 from synapse.app.phone_stats_home import start_phone_stats_home
+from synapse.config import ConfigError
+from synapse.config._base import format_config_error
 from synapse.config.homeserver import HomeServerConfig
 from synapse.config.server import ManholeConfig
 from synapse.crypto import context_factory
@@ -432,6 +435,10 @@ async def start(hs: "HomeServer") -> None:
         signal.signal(signal.SIGHUP, run_sighup)
 
         register_sighup(refresh_certificate, hs)
+        register_sighup(reload_cache_config, hs.config)
+
+    # Apply the cache config.
+    hs.config.caches.resize_all_caches()
 
     # Load the certificate from disk.
     refresh_certificate(hs)
@@ -486,6 +493,43 @@ async def start(hs: "HomeServer") -> None:
         atexit.register(gc.freeze)
 
 
+def reload_cache_config(config: HomeServerConfig) -> None:
+    """Reload cache config from disk and immediately apply it.resize caches accordingly.
+
+    If the config is invalid, a `ConfigError` is logged and no changes are made.
+
+    Otherwise, this:
+        - replaces the `caches` section on the given `config` object,
+        - resizes all caches according to the new cache factors, and
+
+    Note that the following cache config keys are read, but not applied:
+        - event_cache_size: used to set a max_size and _original_max_size on
+              EventsWorkerStore._get_event_cache when it is created. We'd have to update
+              the _original_max_size (and maybe
+        - sync_response_cache_duration: would have to update the timeout_sec attribute on
+              HomeServer ->  SyncHandler -> ResponseCache.
+        - track_memory_usage. This affects synapse.util.caches.TRACK_MEMORY_USAGE which
+              influences Synapse's self-reported metrics.
+
+    Also, the HTTPConnectionPool in SimpleHTTPClient sets its maxPersistentPerHost
+    parameter based on the global_factor. This won't be applied on a config reload.
+    """
+    try:
+        previous_cache_config = config.reload_config_section("caches")
+    except ConfigError as e:
+        logger.warning("Failed to reload cache config")
+        for f in format_config_error(e):
+            logger.warning(f)
+    else:
+        logger.debug(
+            "New cache config. Was:\n %s\nNow:\n",
+            previous_cache_config.__dict__,
+            config.caches.__dict__,
+        )
+        synapse.util.caches.TRACK_MEMORY_USAGE = config.caches.track_memory_usage
+        config.caches.resize_all_caches()
+
+
 def setup_sentry(hs: "HomeServer") -> None:
     """Enable sentry integration, if enabled in configuration"""
 

synapse/app/homeserver.py

@@ -16,7 +16,7 @@
 import logging
 import os
 import sys
-from typing import Dict, Iterable, Iterator, List
+from typing import Dict, Iterable, List
 
 from matrix_common.versionstring import get_distribution_version_string
 
@@ -45,7 +45,7 @@ from synapse.app._base import (
     redirect_stdio_to_logs,
     register_start,
 )
-from synapse.config._base import ConfigError
+from synapse.config._base import ConfigError, format_config_error
 from synapse.config.emailconfig import ThreepidBehaviour
 from synapse.config.homeserver import HomeServerConfig
 from synapse.config.server import ListenerConfig
@@ -399,38 +399,6 @@ def setup(config_options: List[str]) -> SynapseHomeServer:
     return hs
 
 
-def format_config_error(e: ConfigError) -> Iterator[str]:
-    """
-    Formats a config error neatly
-
-    The idea is to format the immediate error, plus the "causes" of those errors,
-    hopefully in a way that makes sense to the user. For example:
-
-        Error in configuration at 'oidc_config.user_mapping_provider.config.display_name_template':
-          Failed to parse config for module 'JinjaOidcMappingProvider':
-            invalid jinja template:
-              unexpected end of template, expected 'end of print statement'.
-
-    Args:
-        e: the error to be formatted
-
-    Returns: An iterator which yields string fragments to be formatted
-    """
-    yield "Error in configuration"
-
-    if e.path:
-        yield " at '%s'" % (".".join(e.path),)
-
-    yield ":\n  %s" % (e.msg,)
-
-    parent_e = e.__cause__
-    indent = 1
-    while parent_e:
-        indent += 1
-        yield ":\n%s%s" % ("  " * indent, str(parent_e))
-        parent_e = parent_e.__cause__
-
-
 def run(hs: HomeServer) -> None:
     _base.start_reactor(
         "synapse-homeserver",

synapse/config/_base.py

@@ -16,14 +16,18 @@
 
 import argparse
 import errno
+import logging
 import os
 from collections import OrderedDict
 from hashlib import sha256
 from textwrap import dedent
 from typing import (
     Any,
+    ClassVar,
+    Collection,
     Dict,
     Iterable,
+    Iterator,
     List,
     MutableMapping,
     Optional,
@@ -40,6 +44,8 @@ import yaml
 
 from synapse.util.templates import _create_mxc_to_http_filter, _format_ts_filter
 
+logger = logging.getLogger(__name__)
+
 
 class ConfigError(Exception):
     """Represents a problem parsing the configuration
@@ -55,6 +61,38 @@ class ConfigError(Exception):
         self.path = path
 
 
+def format_config_error(e: ConfigError) -> Iterator[str]:
+    """
+    Formats a config error neatly
+
+    The idea is to format the immediate error, plus the "causes" of those errors,
+    hopefully in a way that makes sense to the user. For example:
+
+        Error in configuration at 'oidc_config.user_mapping_provider.config.display_name_template':
+          Failed to parse config for module 'JinjaOidcMappingProvider':
+            invalid jinja template:
+              unexpected end of template, expected 'end of print statement'.
+
+    Args:
+        e: the error to be formatted
+
+    Returns: An iterator which yields string fragments to be formatted
+    """
+    yield "Error in configuration"
+
+    if e.path:
+        yield " at '%s'" % (".".join(e.path),)
+
+    yield ":\n  %s" % (e.msg,)
+
+    parent_e = e.__cause__
+    indent = 1
+    while parent_e:
+        indent += 1
+        yield ":\n%s%s" % ("  " * indent, str(parent_e))
+        parent_e = parent_e.__cause__
+
+
 # We split these messages out to allow packages to override with package
 # specific instructions.
 MISSING_REPORT_STATS_CONFIG_INSTRUCTIONS = """\
@@ -119,7 +157,7 @@ class Config:
             defined in subclasses.
     """
 
-    section: str
+    section: ClassVar[str]
 
     def __init__(self, root_config: "RootConfig" = None):
         self.root = root_config
@@ -309,9 +347,12 @@ class RootConfig:
     class, lower-cased and with "Config" removed.
     """
 
-    config_classes = []
+    config_classes: List[Type[Config]] = []
+
+    def __init__(self, config_files: Collection[str] = ()):
+        # Capture absolute paths here, so we can reload config after we daemonize.
+        self.config_files = [os.path.abspath(path) for path in config_files]
 
-    def __init__(self):
         for config_class in self.config_classes:
             if config_class.section is None:
                 raise ValueError("%r requires a section name" % (config_class,))
@@ -512,12 +553,10 @@ class RootConfig:
             object from parser.parse_args(..)`
         """
 
-        obj = cls()
-
         config_args = parser.parse_args(argv)
 
         config_files = find_config_files(search_paths=config_args.config_path)
-
+        obj = cls(config_files)
         if not config_files:
             parser.error("Must supply a config file.")
 
@@ -627,7 +666,7 @@ class RootConfig:
 
         generate_missing_configs = config_args.generate_missing_configs
 
-        obj = cls()
+        obj = cls(config_files)
 
         if config_args.generate_config:
             if config_args.report_stats is None:
@@ -727,6 +766,34 @@ class RootConfig:
     ) -> None:
         self.invoke_all("generate_files", config_dict, config_dir_path)
 
+    def reload_config_section(self, section_name: str) -> Config:
+        """Reconstruct the given config section, leaving all others unchanged.
+
+        This works in three steps:
+
+        1. Create a new instance of the relevant `Config` subclass.
+        2. Call `read_config` on that instance to parse the new config.
+        3. Replace the existing config instance with the new one.
+
+        :raises ValueError: if the given `section` does not exist.
+        :raises ConfigError: for any other problems reloading config.
+
+        :returns: the previous config object, which no longer has a reference to this
+            RootConfig.
+        """
+        existing_config: Optional[Config] = getattr(self, section_name, None)
+        if existing_config is None:
+            raise ValueError(f"Unknown config section '{section_name}'")
+        logger.info("Reloading config section '%s'", section_name)
+
+        new_config_data = read_config_files(self.config_files)
+        new_config = type(existing_config)(self)
+        new_config.read_config(new_config_data)
+        setattr(self, section_name, new_config)
+
+        existing_config.root = None
+        return existing_config
+
 
 def read_config_files(config_files: Iterable[str]) -> Dict[str, Any]:
     """Read the config files into a dict

synapse/config/_base.pyi

@@ -1,15 +1,19 @@
 import argparse
 from typing import (
     Any,
+    Collection,
     Dict,
     Iterable,
+    Iterator,
     List,
+    Literal,
     MutableMapping,
     Optional,
     Tuple,
     Type,
     TypeVar,
     Union,
+    overload,
 )
 
 import jinja2
@@ -65,6 +69,8 @@ class ConfigError(Exception):
         self.msg = msg
         self.path = path
 
+def format_config_error(e: ConfigError) -> Iterator[str]: ...
+
 MISSING_REPORT_STATS_CONFIG_INSTRUCTIONS: str
 MISSING_REPORT_STATS_SPIEL: str
 MISSING_SERVER_NAME: str
@@ -119,7 +125,8 @@ class RootConfig:
     background_updates: background_updates.BackgroundUpdateConfig
 
     config_classes: List[Type["Config"]] = ...
-    def __init__(self) -> None: ...
+    config_files: List[str]
+    def __init__(self, config_files: Collection[str] = ...) -> None: ...
     def invoke_all(
         self, func_name: str, *args: Any, **kwargs: Any
     ) -> MutableMapping[str, Any]: ...
@@ -159,6 +166,12 @@ class RootConfig:
     def generate_missing_files(
         self, config_dict: dict, config_dir_path: str
     ) -> None: ...
+    @overload
+    def reload_config_section(
+        self, section_name: Literal["caches"]
+    ) -> cache.CacheConfig: ...
+    @overload
+    def reload_config_section(self, section_name: str) -> Config: ...
 
 class Config:
     root: RootConfig

synapse/config/cache.py

@@ -69,11 +69,11 @@ def _canonicalise_cache_name(cache_name: str) -> str:
 def add_resizable_cache(
     cache_name: str, cache_resize_callback: Callable[[float], None]
 ) -> None:
-    """Register a cache that's size can dynamically change
+    """Register a cache whose size can dynamically change
 
     Args:
         cache_name: A reference to the cache
-        cache_resize_callback: A callback function that will be ran whenever
+        cache_resize_callback: A callback function that will run whenever
             the cache needs to be resized
     """
     # Some caches have '*' in them which we strip out.
@@ -96,6 +96,13 @@ class CacheConfig(Config):
     section = "caches"
     _environ = os.environ
 
+    event_cache_size: int
+    cache_factors: Dict[str, float]
+    global_factor: float
+    track_memory_usage: bool
+    expiry_time_msec: Optional[int]
+    sync_response_cache_duration: int
+
     @staticmethod
     def reset() -> None:
         """Resets the caches to their defaults. Used for tests."""
@@ -115,6 +122,12 @@ class CacheConfig(Config):
         # A cache 'factor' is a multiplier that can be applied to each of
         # Synapse's caches in order to increase or decrease the maximum
         # number of entries that can be stored.
+        #
+        # The configuration for cache factors (caches.global_factor and
+        # caches.per_cache_factors) can be reloaded while the application is running,
+        # by sending a SIGHUP signal to the Synapse process. Changes to other parts of
+        # the caching config will NOT be applied after a SIGHUP is received; a restart
+        # is necessary.
 
         # The number of events to cache in memory. Not affected by
         # caches.global_factor.
@@ -163,6 +176,24 @@ class CacheConfig(Config):
           #
           #cache_entry_ttl: 30m
 
+          # This flag enables cache autotuning, and is further specified by the sub-options `max_cache_memory_usage`,
+          # `target_cache_memory_usage`, `min_cache_ttl`. These flags work in conjunction with each other to maintain
+          # a balance between cache memory usage and cache entry availability. You must be using jemalloc to utilize
+          # this option, and all three of the options must be specified for this feature to work.
+          #cache_autotuning:
+            # This flag sets a ceiling on much memory the cache can use before caches begin to be continuously evicted.
+            # They will continue to be evicted until the memory usage drops below the `target_memory_usage`, set in
+            # the flag below, or until the `min_cache_ttl` is hit.
+            #max_cache_memory_usage: 1024M
+
+            # This flag sets a rough target for the desired memory usage of the caches.
+            #target_cache_memory_usage: 758M
+
+            # 'min_cache_ttl` sets a limit under which newer cache entries are not evicted and is only applied when
+            # caches are actively being evicted/`max_cache_memory_usage` has been exceeded. This is to protect hot caches
+            # from being emptied while Synapse is evicting due to memory.
+            #min_cache_ttl: 5m
+
           # Controls how long the results of a /sync request are cached for after
           # a successful response is returned. A higher duration can help clients with
           # intermittent connections, at the cost of higher memory usage.
@@ -174,21 +205,21 @@ class CacheConfig(Config):
         """
 
     def read_config(self, config: JsonDict, **kwargs: Any) -> None:
+        """Populate this config object with values from `config`.
+
+        This method does NOT resize existing or future caches: use `resize_all_caches`.
+        We use two separate methods so that we can reject bad config before applying it.
+        """
         self.event_cache_size = self.parse_size(
             config.get("event_cache_size", _DEFAULT_EVENT_CACHE_SIZE)
         )
-        self.cache_factors: Dict[str, float] = {}
+        self.cache_factors = {}
 
         cache_config = config.get("caches") or {}
-        self.global_factor = cache_config.get(
+        self.global_factor = cache_config.get("global_factor", _DEFAULT_FACTOR_SIZE)
-            "global_factor", properties.default_factor_size
-        )
         if not isinstance(self.global_factor, (int, float)):
             raise ConfigError("caches.global_factor must be a number.")
 
-        # Set the global one so that it's reflected in new caches
-        properties.default_factor_size = self.global_factor
-
         # Load cache factors from the config
         individual_factors = cache_config.get("per_cache_factors") or {}
         if not isinstance(individual_factors, dict):
@@ -230,7 +261,7 @@ class CacheConfig(Config):
         cache_entry_ttl = cache_config.get("cache_entry_ttl", "30m")
 
         if expire_caches:
-            self.expiry_time_msec: Optional[int] = self.parse_duration(cache_entry_ttl)
+            self.expiry_time_msec = self.parse_duration(cache_entry_ttl)
         else:
             self.expiry_time_msec = None
 
@@ -250,23 +281,38 @@ class CacheConfig(Config):
             )
             self.expiry_time_msec = self.parse_duration(expiry_time)
 
+        self.cache_autotuning = cache_config.get("cache_autotuning")
+        if self.cache_autotuning:
+            max_memory_usage = self.cache_autotuning.get("max_cache_memory_usage")
+            self.cache_autotuning["max_cache_memory_usage"] = self.parse_size(
+                max_memory_usage
+            )
+
+            target_mem_size = self.cache_autotuning.get("target_cache_memory_usage")
+            self.cache_autotuning["target_cache_memory_usage"] = self.parse_size(
+                target_mem_size
+            )
+
+            min_cache_ttl = self.cache_autotuning.get("min_cache_ttl")
+            self.cache_autotuning["min_cache_ttl"] = self.parse_duration(min_cache_ttl)
+
         self.sync_response_cache_duration = self.parse_duration(
             cache_config.get("sync_response_cache_duration", 0)
         )
 
-        # Resize all caches (if necessary) with the new factors we've loaded
-        self.resize_all_caches()
-
-        # Store this function so that it can be called from other classes without
-        # needing an instance of Config
-        properties.resize_all_caches_func = self.resize_all_caches
-
     def resize_all_caches(self) -> None:
-        """Ensure all cache sizes are up to date
+        """Ensure all cache sizes are up-to-date.
 
         For each cache, run the mapped callback function with either
         a specific cache factor or the default, global one.
         """
+        # Set the global factor size, so that new caches are appropriately sized.
+        properties.default_factor_size = self.global_factor
+
+        # Store this function so that it can be called from other classes without
+        # needing an instance of CacheConfig
+        properties.resize_all_caches_func = self.resize_all_caches
+
         # block other threads from modifying _CACHES while we iterate it.
         with _CACHES_LOCK:
             for cache_name, callback in _CACHES.items():

synapse/config/oembed.py

@@ -57,9 +57,9 @@ class OembedConfig(Config):
         """
         # Whether to use the packaged providers.json file.
         if not oembed_config.get("disable_default_providers") or False:
-            providers = json.load(
+            with pkg_resources.resource_stream("synapse", "res/providers.json") as s:
-                pkg_resources.resource_stream("synapse", "res/providers.json")
+                providers = json.load(s)
-            )
+
             yield from self._parse_and_validate_provider(
                 providers, config_path=("oembed",)
             )

synapse/config/room.py

@@ -63,6 +63,19 @@ class RoomConfig(Config):
                 "Invalid value for encryption_enabled_by_default_for_room_type"
             )
 
+        self.default_power_level_content_override = config.get(
+            "default_power_level_content_override",
+            None,
+        )
+        if self.default_power_level_content_override is not None:
+            for preset in self.default_power_level_content_override:
+                if preset not in vars(RoomCreationPreset).values():
+                    raise ConfigError(
+                        "Unrecognised room preset %s in default_power_level_content_override"
+                        % preset
+                    )
+                # We validate the actual overrides when we try to apply them.
+
     def generate_config_section(self, **kwargs: Any) -> str:
         return """\
         ## Rooms ##
@@ -83,4 +96,38 @@ class RoomConfig(Config):
         # will also not affect rooms created by other servers.
         #
         #encryption_enabled_by_default_for_room_type: invite
+
+        # Override the default power levels for rooms created on this server, per
+        # room creation preset.
+        #
+        # The appropriate dictionary for the room preset will be applied on top
+        # of the existing power levels content.
+        #
+        # Useful if you know that your users need special permissions in rooms
+        # that they create (e.g. to send particular types of state events without
+        # needing an elevated power level).  This takes the same shape as the
+        # `power_level_content_override` parameter in the /createRoom API, but
+        # is applied before that parameter.
+        #
+        # Valid keys are some or all of `private_chat`, `trusted_private_chat`
+        # and `public_chat`. Inside each of those should be any of the
+        # properties allowed in `power_level_content_override` in the
+        # /createRoom API. If any property is missing, its default value will
+        # continue to be used. If any property is present, it will overwrite
+        # the existing default completely (so if the `events` property exists,
+        # the default event power levels will be ignored).
+        #
+        #default_power_level_content_override:
+        #    private_chat:
+        #        "events":
+        #            "com.example.myeventtype" : 0
+        #            "m.room.avatar": 50
+        #            "m.room.canonical_alias": 50
+        #            "m.room.encryption": 100
+        #            "m.room.history_visibility": 100
+        #            "m.room.name": 50
+        #            "m.room.power_levels": 100
+        #            "m.room.server_acl": 100
+        #            "m.room.tombstone": 100
+        #        "events_default": 1
         """

synapse/config/server.py

@@ -996,7 +996,7 @@ class ServerConfig(Config):
         #   federation: the server-server API (/_matrix/federation). Also implies
         #       'media', 'keys', 'openid'
         #
-        #   keys: the key discovery API (/_matrix/keys).
+        #   keys: the key discovery API (/_matrix/key).
         #
         #   media: the media API (/_matrix/media).
         #

synapse/event_auth.py

@@ -414,7 +414,12 @@ def _is_membership_change_allowed(
             raise AuthError(403, "You are banned from this room")
         elif join_rule == JoinRules.PUBLIC:
             pass
-        elif room_version.msc3083_join_rules and join_rule == JoinRules.RESTRICTED:
+        elif (
+            room_version.msc3083_join_rules and join_rule == JoinRules.RESTRICTED
+        ) or (
+            room_version.msc3787_knock_restricted_join_rule
+            and join_rule == JoinRules.KNOCK_RESTRICTED
+        ):
             # This is the same as public, but the event must contain a reference
             # to the server who authorised the join. If the event does not contain
             # the proper content it is rejected.
@@ -440,8 +445,13 @@ def _is_membership_change_allowed(
                 if authorising_user_level < invite_level:
                     raise AuthError(403, "Join event authorised by invalid server.")
 
-        elif join_rule == JoinRules.INVITE or (
+        elif (
-            room_version.msc2403_knocking and join_rule == JoinRules.KNOCK
+            join_rule == JoinRules.INVITE
+            or (room_version.msc2403_knocking and join_rule == JoinRules.KNOCK)
+            or (
+                room_version.msc3787_knock_restricted_join_rule
+                and join_rule == JoinRules.KNOCK_RESTRICTED
+            )
         ):
             if not caller_in_room and not caller_invited:
                 raise AuthError(403, "You are not invited to this room.")
@@ -462,7 +472,10 @@ def _is_membership_change_allowed(
         if user_level < ban_level or user_level <= target_level:
             raise AuthError(403, "You don't have permission to ban")
     elif room_version.msc2403_knocking and Membership.KNOCK == membership:
-        if join_rule != JoinRules.KNOCK:
+        if join_rule != JoinRules.KNOCK and (
+            not room_version.msc3787_knock_restricted_join_rule
+            or join_rule != JoinRules.KNOCK_RESTRICTED
+        ):
             raise AuthError(403, "You don't have permission to knock")
         elif target_user_id != event.user_id:
             raise AuthError(403, "You cannot knock for other users")

synapse/events/__init__.py

@@ -15,6 +15,7 @@
 # limitations under the License.
 
 import abc
+import collections.abc
 import os
 from typing import (
     TYPE_CHECKING,
@@ -32,9 +33,11 @@ from typing import (
     overload,
 )
 
+import attr
 from typing_extensions import Literal
 from unpaddedbase64 import encode_base64
 
+from synapse.api.constants import RelationTypes
 from synapse.api.room_versions import EventFormatVersions, RoomVersion, RoomVersions
 from synapse.types import JsonDict, RoomStreamToken
 from synapse.util.caches import intern_dict
@@ -615,3 +618,45 @@ def make_event_from_dict(
     return event_type(
         event_dict, room_version, internal_metadata_dict or {}, rejected_reason
     )
+
+
+@attr.s(slots=True, frozen=True, auto_attribs=True)
+class _EventRelation:
+    # The target event of the relation.
+    parent_id: str
+    # The relation type.
+    rel_type: str
+    # The aggregation key. Will be None if the rel_type is not m.annotation or is
+    # not a string.
+    aggregation_key: Optional[str]
+
+
+def relation_from_event(event: EventBase) -> Optional[_EventRelation]:
+    """
+    Attempt to parse relation information an event.
+
+    Returns:
+        The event relation information, if it is valid. None, otherwise.
+    """
+    relation = event.content.get("m.relates_to")
+    if not relation or not isinstance(relation, collections.abc.Mapping):
+        # No relation information.
+        return None
+
+    # Relations must have a type and parent event ID.
+    rel_type = relation.get("rel_type")
+    if not isinstance(rel_type, str):
+        return None
+
+    parent_id = relation.get("event_id")
+    if not isinstance(parent_id, str):
+        return None
+
+    # Annotations have a key field.
+    aggregation_key = None
+    if rel_type == RelationTypes.ANNOTATION:
+        aggregation_key = relation.get("key")
+        if not isinstance(aggregation_key, str):
+            aggregation_key = None
+
+    return _EventRelation(parent_id, rel_type, aggregation_key)

synapse/events/snapshot.py

@@ -15,17 +15,16 @@ from typing import TYPE_CHECKING, List, Optional, Tuple, Union
 
 import attr
 from frozendict import frozendict
-
+from typing_extensions import Literal
-from twisted.internet.defer import Deferred
 
 from synapse.appservice import ApplicationService
 from synapse.events import EventBase
-from synapse.logging.context import make_deferred_yieldable, run_in_background
 from synapse.types import JsonDict, StateMap
 
 if TYPE_CHECKING:
     from synapse.storage import Storage
     from synapse.storage.databases.main import DataStore
+    from synapse.storage.state import StateFilter
 
 
 @attr.s(slots=True, auto_attribs=True)
@@ -60,6 +59,9 @@ class EventContext:
             If ``state_group`` is None (ie, the event is an outlier),
             ``state_group_before_event`` will always also be ``None``.
 
+        state_delta_due_to_event: If `state_group` and `state_group_before_event` are not None
+            then this is the delta of the state between the two groups.
+
         prev_group: If it is known, ``state_group``'s prev_group. Note that this being
             None does not necessarily mean that ``state_group`` does not have
             a prev_group!
@@ -78,73 +80,47 @@ class EventContext:
         app_service: If this event is being sent by a (local) application service, that
             app service.
 
-        _current_state_ids: The room state map, including this event - ie, the state
-            in ``state_group``.
-
-            (type, state_key) -> event_id
-
-            For an outlier, this is {}
-
-            Note that this is a private attribute: it should be accessed via
-            ``get_current_state_ids``. _AsyncEventContext impl calculates this
-            on-demand: it will be None until that happens.
-
-        _prev_state_ids: The room state map, excluding this event - ie, the state
-            in ``state_group_before_event``. For a non-state
-            event, this will be the same as _current_state_events.
-
-            Note that it is a completely different thing to prev_group!
-
-            (type, state_key) -> event_id
-
-            For an outlier, this is {}
-
-            As with _current_state_ids, this is a private attribute. It should be
-            accessed via get_prev_state_ids.
-
         partial_state: if True, we may be storing this event with a temporary,
             incomplete state.
     """
 
-    rejected: Union[bool, str] = False
+    _storage: "Storage"
+    rejected: Union[Literal[False], str] = False
     _state_group: Optional[int] = None
     state_group_before_event: Optional[int] = None
+    _state_delta_due_to_event: Optional[StateMap[str]] = None
     prev_group: Optional[int] = None
     delta_ids: Optional[StateMap[str]] = None
     app_service: Optional[ApplicationService] = None
 
-    _current_state_ids: Optional[StateMap[str]] = None
-    _prev_state_ids: Optional[StateMap[str]] = None
-
     partial_state: bool = False
 
     @staticmethod
     def with_state(
+        storage: "Storage",
         state_group: Optional[int],
         state_group_before_event: Optional[int],
-        current_state_ids: Optional[StateMap[str]],
+        state_delta_due_to_event: Optional[StateMap[str]],
-        prev_state_ids: Optional[StateMap[str]],
         partial_state: bool,
         prev_group: Optional[int] = None,
         delta_ids: Optional[StateMap[str]] = None,
     ) -> "EventContext":
         return EventContext(
-            current_state_ids=current_state_ids,
+            storage=storage,
-            prev_state_ids=prev_state_ids,
             state_group=state_group,
             state_group_before_event=state_group_before_event,
+            state_delta_due_to_event=state_delta_due_to_event,
             prev_group=prev_group,
             delta_ids=delta_ids,
             partial_state=partial_state,
         )
 
     @staticmethod
-    def for_outlier() -> "EventContext":
+    def for_outlier(
+        storage: "Storage",
+    ) -> "EventContext":
         """Return an EventContext instance suitable for persisting an outlier event"""
-        return EventContext(
+        return EventContext(storage=storage)
-            current_state_ids={},
-            prev_state_ids={},
-        )
 
     async def serialize(self, event: EventBase, store: "DataStore") -> JsonDict:
         """Converts self to a type that can be serialized as JSON, and then
@@ -157,24 +133,14 @@ class EventContext:
             The serialized event.
         """
 
-        # We don't serialize the full state dicts, instead they get pulled out
-        # of the DB on the other side. However, the other side can't figure out
-        # 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_ids = await self.get_prev_state_ids()
-            prev_state_id = prev_state_ids.get((event.type, event.state_key))
-        else:
-            prev_state_id = None
-
         return {
-            "prev_state_id": prev_state_id,
-            "event_type": event.type,
-            "event_state_key": event.get_state_key(),
             "state_group": self._state_group,
             "state_group_before_event": self.state_group_before_event,
             "rejected": self.rejected,
             "prev_group": self.prev_group,
+            "state_delta_due_to_event": _encode_state_dict(
+                self._state_delta_due_to_event
+            ),
             "delta_ids": _encode_state_dict(self.delta_ids),
             "app_service_id": self.app_service.id if self.app_service else None,
             "partial_state": self.partial_state,
@@ -192,16 +158,16 @@ class EventContext:
         Returns:
             The event context.
         """
-        context = _AsyncEventContextImpl(
+        context = EventContext(
             # 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.
             storage=storage,
-            prev_state_id=input["prev_state_id"],
-            event_type=input["event_type"],
-            event_state_key=input["event_state_key"],
             state_group=input["state_group"],
             state_group_before_event=input["state_group_before_event"],
             prev_group=input["prev_group"],
+            state_delta_due_to_event=_decode_state_dict(
+                input["state_delta_due_to_event"]
+            ),
             delta_ids=_decode_state_dict(input["delta_ids"]),
             rejected=input["rejected"],
             partial_state=input.get("partial_state", False),
@@ -231,7 +197,9 @@ class EventContext:
 
         return self._state_group
 
-    async def get_current_state_ids(self) -> Optional[StateMap[str]]:
+    async def get_current_state_ids(
+        self, state_filter: Optional["StateFilter"] = None
+    ) -> Optional[StateMap[str]]:
         """
         Gets the room state map, including this event - ie, the state in ``state_group``
 
@@ -239,6 +207,9 @@ class EventContext:
         not make it into the room state. This method will raise an exception if
         ``rejected`` is set.
 
+        Arg:
+           state_filter: specifies the type of state event to fetch from DB, example: EventTypes.JoinRules
+
         Returns:
             Returns None if state_group is None, which happens when the associated
             event is an outlier.
@@ -249,15 +220,27 @@ class EventContext:
         if self.rejected:
             raise RuntimeError("Attempt to access state_ids of rejected event")
 
-        await self._ensure_fetched()
+        assert self._state_delta_due_to_event is not None
-        return self._current_state_ids
 
-    async def get_prev_state_ids(self) -> StateMap[str]:
+        prev_state_ids = await self.get_prev_state_ids(state_filter)
+
+        if self._state_delta_due_to_event:
+            prev_state_ids = dict(prev_state_ids)
+            prev_state_ids.update(self._state_delta_due_to_event)
+
+        return prev_state_ids
+
+    async def get_prev_state_ids(
+        self, state_filter: Optional["StateFilter"] = None
+    ) -> StateMap[str]:
         """
         Gets the room state map, excluding this event.
 
         For a non-state event, this will be the same as get_current_state_ids().
 
+        Args:
+            state_filter: specifies the type of state event to fetch from DB, example: EventTypes.JoinRules
+
         Returns:
             Returns {} if state_group is None, which happens when the associated
             event is an outlier.
@@ -265,94 +248,10 @@ class EventContext:
             Maps a (type, state_key) to the event ID of the state event matching
             this tuple.
         """
-        await self._ensure_fetched()
+        assert self.state_group_before_event is not None
-        # There *should* be previous state IDs now.
+        return await self._storage.state.get_state_ids_for_group(
-        assert self._prev_state_ids is not None
+            self.state_group_before_event, state_filter
-        return self._prev_state_ids
-
-    def get_cached_current_state_ids(self) -> Optional[StateMap[str]]:
-        """Gets the current state IDs if we have them already cached.
-
-        It is an error to access this for a rejected event, since rejected state should
-        not make it into the room state. This method will raise an exception if
-        ``rejected`` is set.
-
-        Returns:
-            Returns None if we haven't cached the state or if state_group is None
-            (which happens when the associated event is an outlier).
-
-            Otherwise, returns the the current state IDs.
-        """
-        if self.rejected:
-            raise RuntimeError("Attempt to access state_ids of rejected event")
-
-        return self._current_state_ids
-
-    async def _ensure_fetched(self) -> None:
-        return None
-
-
-@attr.s(slots=True)
-class _AsyncEventContextImpl(EventContext):
-    """
-    An implementation of EventContext which fetches _current_state_ids and
-    _prev_state_ids from the database on demand.
-
-    Attributes:
-
-        _storage
-
-        _fetching_state_deferred: Resolves when *_state_ids have been calculated.
-            None if we haven't started calculating yet
-
-        _event_type: The type of the event the context is associated with.
-
-        _event_state_key: The state_key of the event the context is associated with.
-
-        _prev_state_id: 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.
-    """
-
-    # This needs to have a default as we're inheriting
-    _storage: "Storage" = attr.ib(default=None)
-    _prev_state_id: Optional[str] = attr.ib(default=None)
-    _event_type: str = attr.ib(default=None)
-    _event_state_key: Optional[str] = attr.ib(default=None)
-    _fetching_state_deferred: Optional["Deferred[None]"] = attr.ib(default=None)
-
-    async def _ensure_fetched(self) -> None:
-        if not self._fetching_state_deferred:
-            self._fetching_state_deferred = run_in_background(self._fill_out_state)
-
-        await make_deferred_yieldable(self._fetching_state_deferred)
-
-    async def _fill_out_state(self) -> None:
-        """Called to populate the _current_state_ids and _prev_state_ids
-        attributes by loading from the database.
-        """
-        if self.state_group is None:
-            # No state group means the event is an outlier. Usually the state_ids dicts are also
-            # pre-set to empty dicts, but they get reset when the context is serialized, so set
-            # them to empty dicts again here.
-            self._current_state_ids = {}
-            self._prev_state_ids = {}
-            return
-
-        current_state_ids = await self._storage.state.get_state_ids_for_group(
-            self.state_group
         )
-        # Set this separately so mypy knows current_state_ids is not None.
-        self._current_state_ids = current_state_ids
-        if self._event_state_key is not None:
-            self._prev_state_ids = dict(current_state_ids)
-
-            key = (self._event_type, self._event_state_key)
-            if self._prev_state_id:
-                self._prev_state_ids[key] = self._prev_state_id
-            else:
-                self._prev_state_ids.pop(key, None)
-        else:
-            self._prev_state_ids = current_state_ids
 
 
 def _encode_state_dict(

synapse/events/spamcheck.py

@@ -27,11 +27,13 @@ from typing import (
     Union,
 )
 
+from synapse.api.errors import Codes
 from synapse.rest.media.v1._base import FileInfo
 from synapse.rest.media.v1.media_storage import ReadableFileWrapper
-from synapse.spam_checker_api import RegistrationBehaviour
+from synapse.spam_checker_api import Allow, Decision, RegistrationBehaviour
 from synapse.types import RoomAlias, UserProfile
 from synapse.util.async_helpers import delay_cancellation, maybe_awaitable
+from synapse.util.metrics import Measure
 
 if TYPE_CHECKING:
     import synapse.events
@@ -39,7 +41,21 @@ if TYPE_CHECKING:
 
 logger = logging.getLogger(__name__)
 
+
 CHECK_EVENT_FOR_SPAM_CALLBACK = Callable[
+    ["synapse.events.EventBase"],
+    Awaitable[
+        Union[
+            Allow,
+            Codes,
+            # Deprecated
+            bool,
+            # Deprecated
+            str,
+        ]
+    ],
+]
+SHOULD_DROP_FEDERATED_EVENT_CALLBACK = Callable[
     ["synapse.events.EventBase"],
     Awaitable[Union[bool, str]],
 ]
@@ -162,8 +178,14 @@ def load_legacy_spam_checkers(hs: "synapse.server.HomeServer") -> None:
 
 
 class SpamChecker:
-    def __init__(self) -> None:
+    def __init__(self, hs: "synapse.server.HomeServer") -> None:
+        self.hs = hs
+        self.clock = hs.get_clock()
+
         self._check_event_for_spam_callbacks: List[CHECK_EVENT_FOR_SPAM_CALLBACK] = []
+        self._should_drop_federated_event_callbacks: List[
+            SHOULD_DROP_FEDERATED_EVENT_CALLBACK
+        ] = []
         self._user_may_join_room_callbacks: List[USER_MAY_JOIN_ROOM_CALLBACK] = []
         self._user_may_invite_callbacks: List[USER_MAY_INVITE_CALLBACK] = []
         self._user_may_send_3pid_invite_callbacks: List[
@@ -187,6 +209,9 @@ class SpamChecker:
     def register_callbacks(
         self,
         check_event_for_spam: Optional[CHECK_EVENT_FOR_SPAM_CALLBACK] = None,
+        should_drop_federated_event: Optional[
+            SHOULD_DROP_FEDERATED_EVENT_CALLBACK
+        ] = None,
         user_may_join_room: Optional[USER_MAY_JOIN_ROOM_CALLBACK] = None,
         user_may_invite: Optional[USER_MAY_INVITE_CALLBACK] = None,
         user_may_send_3pid_invite: Optional[USER_MAY_SEND_3PID_INVITE_CALLBACK] = None,
@@ -205,6 +230,11 @@ class SpamChecker:
         if check_event_for_spam is not None:
             self._check_event_for_spam_callbacks.append(check_event_for_spam)
 
+        if should_drop_federated_event is not None:
+            self._should_drop_federated_event_callbacks.append(
+                should_drop_federated_event
+            )
+
         if user_may_join_room is not None:
             self._user_may_join_room_callbacks.append(user_may_join_room)
 
@@ -240,7 +270,7 @@ class SpamChecker:
 
     async def check_event_for_spam(
         self, event: "synapse.events.EventBase"
-    ) -> Union[bool, str]:
+    ) -> Union[Decision, str]:
         """Checks if a given event is considered "spammy" by this server.
 
         If the server considers an event spammy, then it will be rejected if
@@ -251,11 +281,57 @@ class SpamChecker:
             event: the event to be checked
 
         Returns:
-            True or a string if the event is spammy. If a string is returned it
+            - on `ALLOW`, the event is considered good (non-spammy) and should
-            will be used as the error message returned to the user.
+                be let through. Other spamcheck filters may still reject it.
+            - on `Code`, the event is considered spammy and is rejected with a specific
+                error message/code.
+            - on `str`, the event is considered spammy and the string is used as error
+                message. This usage is generally discouraged as it doesn't support
+                internationalization.
         """
         for callback in self._check_event_for_spam_callbacks:
-            res: Union[bool, str] = await delay_cancellation(callback(event))
+            with Measure(
+                self.clock, "{}.{}".format(callback.__module__, callback.__qualname__)
+            ):
+                res: Union[Decision, str, bool] = await delay_cancellation(
+                    callback(event)
+                )
+                if res is False or res is Allow.ALLOW:
+                    # This spam-checker accepts the event.
+                    # Other spam-checkers may reject it, though.
+                    continue
+                elif res is True:
+                    # This spam-checker rejects the event with deprecated
+                    # return value `True`
+                    return Codes.FORBIDDEN
+                else:
+                    # This spam-checker rejects the event either with a `str`
+                    # or with a `Codes`. In either case, we stop here.
+                    return res
+
+        # No spam-checker has rejected the event, let it pass.
+        return Allow.ALLOW
+
+    async def should_drop_federated_event(
+        self, event: "synapse.events.EventBase"
+    ) -> Union[bool, str]:
+        """Checks if a given federated event is considered "spammy" by this
+        server.
+
+        If the server considers an event spammy, it will be silently dropped,
+        and in doing so will split-brain our view of the room's DAG.
+
+        Args:
+            event: the event to be checked
+
+        Returns:
+            True if the event should be silently dropped
+        """
+        for callback in self._should_drop_federated_event_callbacks:
+            with Measure(
+                self.clock, "{}.{}".format(callback.__module__, callback.__qualname__)
+            ):
+                res: Union[bool, str] = await delay_cancellation(callback(event))
             if res:
                 return res
 
@@ -276,9 +352,12 @@ class SpamChecker:
             Whether the user may join the room
         """
         for callback in self._user_may_join_room_callbacks:
-            may_join_room = await delay_cancellation(
+            with Measure(
-                callback(user_id, room_id, is_invited)
+                self.clock, "{}.{}".format(callback.__module__, callback.__qualname__)
-            )
+            ):
+                may_join_room = await delay_cancellation(
+                    callback(user_id, room_id, is_invited)
+                )
             if may_join_room is False:
                 return False
 
@@ -300,9 +379,12 @@ class SpamChecker:
             True if the user may send an invite, otherwise False
         """
         for callback in self._user_may_invite_callbacks:
-            may_invite = await delay_cancellation(
+            with Measure(
-                callback(inviter_userid, invitee_userid, room_id)
+                self.clock, "{}.{}".format(callback.__module__, callback.__qualname__)
-            )
+            ):
+                may_invite = await delay_cancellation(
+                    callback(inviter_userid, invitee_userid, room_id)
+                )
             if may_invite is False:
                 return False
 
@@ -328,9 +410,12 @@ class SpamChecker:
             True if the user may send the invite, otherwise False
         """
         for callback in self._user_may_send_3pid_invite_callbacks:
-            may_send_3pid_invite = await delay_cancellation(
+            with Measure(
-                callback(inviter_userid, medium, address, room_id)
+                self.clock, "{}.{}".format(callback.__module__, callback.__qualname__)
-            )
+            ):
+                may_send_3pid_invite = await delay_cancellation(
+                    callback(inviter_userid, medium, address, room_id)
+                )
             if may_send_3pid_invite is False:
                 return False
 
@@ -348,7 +433,10 @@ class SpamChecker:
             True if the user may create a room, otherwise False
         """
         for callback in self._user_may_create_room_callbacks:
-            may_create_room = await delay_cancellation(callback(userid))
+            with Measure(
+                self.clock, "{}.{}".format(callback.__module__, callback.__qualname__)
+            ):
+                may_create_room = await delay_cancellation(callback(userid))
             if may_create_room is False:
                 return False
 
@@ -369,9 +457,12 @@ class SpamChecker:
             True if the user may create a room alias, otherwise False
         """
         for callback in self._user_may_create_room_alias_callbacks:
-            may_create_room_alias = await delay_cancellation(
+            with Measure(
-                callback(userid, room_alias)
+                self.clock, "{}.{}".format(callback.__module__, callback.__qualname__)
-            )
+            ):
+                may_create_room_alias = await delay_cancellation(
+                    callback(userid, room_alias)
+                )
             if may_create_room_alias is False:
                 return False
 
@@ -390,7 +481,10 @@ class SpamChecker:
             True if the user may publish the room, otherwise False
         """
         for callback in self._user_may_publish_room_callbacks:
-            may_publish_room = await delay_cancellation(callback(userid, room_id))
+            with Measure(
+                self.clock, "{}.{}".format(callback.__module__, callback.__qualname__)
+            ):
+                may_publish_room = await delay_cancellation(callback(userid, room_id))
             if may_publish_room is False:
                 return False
 
@@ -412,9 +506,13 @@ class SpamChecker:
             True if the user is spammy.
         """
         for callback in self._check_username_for_spam_callbacks:
-            # Make a copy of the user profile object to ensure the spam checker cannot
+            with Measure(
-            # modify it.
+                self.clock, "{}.{}".format(callback.__module__, callback.__qualname__)
-            if await delay_cancellation(callback(user_profile.copy())):
+            ):
+                # Make a copy of the user profile object to ensure the spam checker cannot
+                # modify it.
+                res = await delay_cancellation(callback(user_profile.copy()))
+            if res:
                 return True
 
         return False
@@ -442,9 +540,12 @@ class SpamChecker:
         """
 
         for callback in self._check_registration_for_spam_callbacks:
-            behaviour = await delay_cancellation(
+            with Measure(
-                callback(email_threepid, username, request_info, auth_provider_id)
+                self.clock, "{}.{}".format(callback.__module__, callback.__qualname__)
-            )
+            ):
+                behaviour = await delay_cancellation(
+                    callback(email_threepid, username, request_info, auth_provider_id)
+                )
             assert isinstance(behaviour, RegistrationBehaviour)
             if behaviour != RegistrationBehaviour.ALLOW:
                 return behaviour
@@ -486,7 +587,10 @@ class SpamChecker:
         """
 
         for callback in self._check_media_file_for_spam_callbacks:
-            spam = await delay_cancellation(callback(file_wrapper, file_info))
+            with Measure(
+                self.clock, "{}.{}".format(callback.__module__, callback.__qualname__)
+            ):
+                spam = await delay_cancellation(callback(file_wrapper, file_info))
             if spam:
                 return True
 

synapse/federation/federation_base.py

@@ -15,6 +15,7 @@
 import logging
 from typing import TYPE_CHECKING
 
+import synapse
 from synapse.api.constants import MAX_DEPTH, EventContentFields, EventTypes, Membership
 from synapse.api.errors import Codes, SynapseError
 from synapse.api.room_versions import EventFormatVersions, RoomVersion
@@ -98,9 +99,9 @@ class FederationBase:
                 )
             return redacted_event
 
-        result = await self.spam_checker.check_event_for_spam(pdu)
+        spam_check = await self.spam_checker.check_event_for_spam(pdu)
 
-        if result:
+        if spam_check is not synapse.spam_checker_api.Allow.ALLOW:
             logger.warning("Event contains spam, soft-failing %s", pdu.event_id)
             # we redact (to save disk space) as well as soft-failing (to stop
             # using the event in prev_events).

synapse/federation/federation_server.py

@@ -110,6 +110,7 @@ class FederationServer(FederationBase):
 
         self.handler = hs.get_federation_handler()
         self.storage = hs.get_storage()
+        self._spam_checker = hs.get_spam_checker()
         self._federation_event_handler = hs.get_federation_event_handler()
         self.state = hs.get_state_handler()
         self._event_auth_handler = hs.get_event_auth_handler()
@@ -1019,6 +1020,12 @@ class FederationServer(FederationBase):
         except SynapseError as e:
             raise FederationError("ERROR", e.code, e.msg, affected=pdu.event_id)
 
+        if await self._spam_checker.should_drop_federated_event(pdu):
+            logger.warning(
+                "Unstaged federated event contains spam, dropping %s", pdu.event_id
+            )
+            return
+
         # Add the event to our staging area
         await self.store.insert_received_event_to_staging(origin, pdu)
 
@@ -1032,6 +1039,41 @@ class FederationServer(FederationBase):
                 pdu.room_id, room_version, lock, origin, pdu
             )
 
+    async def _get_next_nonspam_staged_event_for_room(
+        self, room_id: str, room_version: RoomVersion
+    ) -> Optional[Tuple[str, EventBase]]:
+        """Fetch the first non-spam event from staging queue.
+
+        Args:
+            room_id: the room to fetch the first non-spam event in.
+            room_version: the version of the room.
+
+        Returns:
+            The first non-spam event in that room.
+        """
+
+        while True:
+            # We need to do this check outside the lock to avoid a race between
+            # a new event being inserted by another instance and it attempting
+            # to acquire the lock.
+            next = await self.store.get_next_staged_event_for_room(
+                room_id, room_version
+            )
+
+            if next is None:
+                return None
+
+            origin, event = next
+
+            if await self._spam_checker.should_drop_federated_event(event):
+                logger.warning(
+                    "Staged federated event contains spam, dropping %s",
+                    event.event_id,
+                )
+                continue
+
+            return next
+
     @wrap_as_background_process("_process_incoming_pdus_in_room_inner")
     async def _process_incoming_pdus_in_room_inner(
         self,
@@ -1109,12 +1151,10 @@ class FederationServer(FederationBase):
                         (self._clock.time_msec() - received_ts) / 1000
                     )
 
-            # We need to do this check outside the lock to avoid a race between
+            next = await self._get_next_nonspam_staged_event_for_room(
-            # a new event being inserted by another instance and it attempting
-            # to acquire the lock.
-            next = await self.store.get_next_staged_event_for_room(
                 room_id, room_version
             )
+
             if not next:
                 break
 

synapse/federation/sender/__init__.py

@@ -15,7 +15,17 @@
 import abc
 import logging
 from collections import OrderedDict
-from typing import TYPE_CHECKING, Dict, Hashable, Iterable, List, Optional, Set, Tuple
+from typing import (
+    TYPE_CHECKING,
+    Collection,
+    Dict,
+    Hashable,
+    Iterable,
+    List,
+    Optional,
+    Set,
+    Tuple,
+)
 
 import attr
 from prometheus_client import Counter
@@ -409,7 +419,7 @@ class FederationSender(AbstractFederationSender):
                         )
                         return
 
-                    destinations: Optional[Set[str]] = None
+                    destinations: Optional[Collection[str]] = None
                     if not event.prev_event_ids():
                         # If there are no prev event IDs then the state is empty
                         # and so no remote servers in the room
@@ -444,7 +454,7 @@ class FederationSender(AbstractFederationSender):
                             )
                             return
 
-                    destinations = {
+                    sharded_destinations = {
                         d
                         for d in destinations
                         if self._federation_shard_config.should_handle(
@@ -456,12 +466,12 @@ class FederationSender(AbstractFederationSender):
                         # If we are sending the event on behalf of another server
                         # then it already has the event and there is no reason to
                         # send the event to it.
-                        destinations.discard(send_on_behalf_of)
+                        sharded_destinations.discard(send_on_behalf_of)
 
-                    logger.debug("Sending %s to %r", event, destinations)
+                    logger.debug("Sending %s to %r", event, sharded_destinations)
 
-                    if destinations:
+                    if sharded_destinations:
-                        await self._send_pdu(event, destinations)
+                        await self._send_pdu(event, sharded_destinations)
 
                         now = self.clock.time_msec()
                         ts = await self.store.get_received_ts(event.event_id)

synapse/federation/transport/server/_base.py

@@ -21,7 +21,7 @@ from typing import TYPE_CHECKING, Any, Awaitable, Callable, Dict, Optional, Tupl
 
 from synapse.api.errors import Codes, FederationDeniedError, SynapseError
 from synapse.api.urls import FEDERATION_V1_PREFIX
-from synapse.http.server import HttpServer, ServletCallback
+from synapse.http.server import HttpServer, ServletCallback, is_method_cancellable
 from synapse.http.servlet import parse_json_object_from_request
 from synapse.http.site import SynapseRequest
 from synapse.logging.context import run_in_background
@@ -169,14 +169,16 @@ def _parse_auth_header(header_bytes: bytes) -> Tuple[str, str, str, Optional[str
     """
     try:
         header_str = header_bytes.decode("utf-8")
-        params = header_str.split(" ")[1].split(",")
+        params = re.split(" +", header_str)[1].split(",")
         param_dict: Dict[str, str] = {
-            k: v for k, v in [param.split("=", maxsplit=1) for param in params]
+            k.lower(): v for k, v in [param.split("=", maxsplit=1) for param in params]
         }
 
         def strip_quotes(value: str) -> str:
             if value.startswith('"'):
-                return value[1:-1]
+                return re.sub(
+                    "\\\\(.)", lambda matchobj: matchobj.group(1), value[1:-1]
+                )
             else:
                 return value
 
@@ -373,6 +375,17 @@ class BaseFederationServlet:
             if code is None:
                 continue
 
+            if is_method_cancellable(code):
+                # The wrapper added by `self._wrap` will inherit the cancellable flag,
+                # but the wrapper itself does not support cancellation yet.
+                # Once resolved, the cancellation tests in
+                # `tests/federation/transport/server/test__base.py` can be re-enabled.
+                raise Exception(
+                    f"{self.__class__.__name__}.on_{method} has been marked as "
+                    "cancellable, but federation servlets do not support cancellation "
+                    "yet."
+                )
+
             server.register_paths(
                 method,
                 (pattern,),

synapse/groups/groups_server.py

@@ -934,7 +934,7 @@ class GroupsServerHandler(GroupsServerWorkerHandler):
         # Before deleting the group lets kick everyone out of it
         users = await self.store.get_users_in_group(group_id, include_private=True)
 
-        async def _kick_user_from_group(user_id):
+        async def _kick_user_from_group(user_id: str) -> None:
             if self.hs.is_mine_id(user_id):
                 groups_local = self.hs.get_groups_local_handler()
                 assert isinstance(

synapse/handlers/account_data.py

@@ -23,7 +23,7 @@ from synapse.replication.http.account_data import (
     ReplicationUserAccountDataRestServlet,
 )
 from synapse.streams import EventSource
-from synapse.types import JsonDict, UserID
+from synapse.types import JsonDict, StreamKeyType, UserID
 
 if TYPE_CHECKING:
     from synapse.server import HomeServer
@@ -105,7 +105,7 @@ class AccountDataHandler:
             )
 
             self._notifier.on_new_event(
-                "account_data_key", max_stream_id, users=[user_id]
+                StreamKeyType.ACCOUNT_DATA, max_stream_id, users=[user_id]
             )
 
             await self._notify_modules(user_id, room_id, account_data_type, content)
@@ -141,7 +141,7 @@ class AccountDataHandler:
             )
 
             self._notifier.on_new_event(
-                "account_data_key", max_stream_id, users=[user_id]
+                StreamKeyType.ACCOUNT_DATA, max_stream_id, users=[user_id]
             )
 
             await self._notify_modules(user_id, None, account_data_type, content)
@@ -176,7 +176,7 @@ class AccountDataHandler:
             )
 
             self._notifier.on_new_event(
-                "account_data_key", max_stream_id, users=[user_id]
+                StreamKeyType.ACCOUNT_DATA, max_stream_id, users=[user_id]
             )
             return max_stream_id
         else:
@@ -201,7 +201,7 @@ class AccountDataHandler:
             )
 
             self._notifier.on_new_event(
-                "account_data_key", max_stream_id, users=[user_id]
+                StreamKeyType.ACCOUNT_DATA, max_stream_id, users=[user_id]
             )
             return max_stream_id
         else:

synapse/handlers/appservice.py

@@ -38,6 +38,7 @@ from synapse.types import (
     JsonDict,
     RoomAlias,
     RoomStreamToken,
+    StreamKeyType,
     UserID,
 )
 from synapse.util.async_helpers import Linearizer
@@ -213,8 +214,8 @@ class ApplicationServicesHandler:
         Args:
             stream_key: The stream the event came from.
 
-                `stream_key` can be "typing_key", "receipt_key", "presence_key",
+                `stream_key` can be StreamKeyType.TYPING, StreamKeyType.RECEIPT, StreamKeyType.PRESENCE,
-                "to_device_key" or "device_list_key". Any other value for `stream_key`
+                StreamKeyType.TO_DEVICE or StreamKeyType.DEVICE_LIST. Any other value for `stream_key`
                 will cause this function to return early.
 
                 Ephemeral events will only be pushed to appservices that have opted into
@@ -235,11 +236,11 @@ class ApplicationServicesHandler:
         # Only the following streams are currently supported.
         # FIXME: We should use constants for these values.
         if stream_key not in (
-            "typing_key",
+            StreamKeyType.TYPING,
-            "receipt_key",
+            StreamKeyType.RECEIPT,
-            "presence_key",
+            StreamKeyType.PRESENCE,
-            "to_device_key",
+            StreamKeyType.TO_DEVICE,
-            "device_list_key",
+            StreamKeyType.DEVICE_LIST,
         ):
             return
 
@@ -258,14 +259,14 @@ class ApplicationServicesHandler:
 
         # Ignore to-device messages if the feature flag is not enabled
         if (
-            stream_key == "to_device_key"
+            stream_key == StreamKeyType.TO_DEVICE
             and not self._msc2409_to_device_messages_enabled
         ):
             return
 
         # Ignore device lists if the feature flag is not enabled
         if (
-            stream_key == "device_list_key"
+            stream_key == StreamKeyType.DEVICE_LIST
             and not self._msc3202_transaction_extensions_enabled
         ):
             return
@@ -283,15 +284,15 @@ class ApplicationServicesHandler:
             if (
                 stream_key
                 in (
-                    "typing_key",
+                    StreamKeyType.TYPING,
-                    "receipt_key",
+                    StreamKeyType.RECEIPT,
-                    "presence_key",
+                    StreamKeyType.PRESENCE,
-                    "to_device_key",
+                    StreamKeyType.TO_DEVICE,
                 )
                 and service.supports_ephemeral
             )
             or (
-                stream_key == "device_list_key"
+                stream_key == StreamKeyType.DEVICE_LIST
                 and service.msc3202_transaction_extensions
             )
         ]
@@ -317,7 +318,7 @@ class ApplicationServicesHandler:
         logger.debug("Checking interested services for %s", stream_key)
         with Measure(self.clock, "notify_interested_services_ephemeral"):
             for service in services:
-                if stream_key == "typing_key":
+                if stream_key == StreamKeyType.TYPING:
                     # Note that we don't persist the token (via set_appservice_stream_type_pos)
                     # for typing_key due to performance reasons and due to their highly
                     # ephemeral nature.
@@ -333,7 +334,7 @@ class ApplicationServicesHandler:
                 async with self._ephemeral_events_linearizer.queue(
                     (service.id, stream_key)
                 ):
-                    if stream_key == "receipt_key":
+                    if stream_key == StreamKeyType.RECEIPT:
                         events = await self._handle_receipts(service, new_token)
                         self.scheduler.enqueue_for_appservice(service, ephemeral=events)
 
@@ -342,7 +343,7 @@ class ApplicationServicesHandler:
                             service, "read_receipt", new_token
                         )
 
-                    elif stream_key == "presence_key":
+                    elif stream_key == StreamKeyType.PRESENCE:
                         events = await self._handle_presence(service, users, new_token)
                         self.scheduler.enqueue_for_appservice(service, ephemeral=events)
 
@@ -351,7 +352,7 @@ class ApplicationServicesHandler:
                             service, "presence", new_token
                         )
 
-                    elif stream_key == "to_device_key":
+                    elif stream_key == StreamKeyType.TO_DEVICE:
                         # Retrieve a list of to-device message events, as well as the
                         # maximum stream token of the messages we were able to retrieve.
                         to_device_messages = await self._get_to_device_messages(
@@ -366,7 +367,7 @@ class ApplicationServicesHandler:
                             service, "to_device", new_token
                         )
 
-                    elif stream_key == "device_list_key":
+                    elif stream_key == StreamKeyType.DEVICE_LIST:
                         device_list_summary = await self._get_device_list_summary(
                             service, new_token
                         )

synapse/handlers/device.py

@@ -43,6 +43,7 @@ from synapse.metrics.background_process_metrics import (
 )
 from synapse.types import (
     JsonDict,
+    StreamKeyType,
     StreamToken,
     UserID,
     get_domain_from_id,
@@ -502,7 +503,7 @@ class DeviceHandler(DeviceWorkerHandler):
         # specify the user ID too since the user should always get their own device list
         # updates, even if they aren't in any rooms.
         self.notifier.on_new_event(
-            "device_list_key", position, users={user_id}, rooms=room_ids
+            StreamKeyType.DEVICE_LIST, position, users={user_id}, rooms=room_ids
         )
 
         # We may need to do some processing asynchronously for local user IDs.
@@ -523,7 +524,9 @@ class DeviceHandler(DeviceWorkerHandler):
             from_user_id, user_ids
         )
 
-        self.notifier.on_new_event("device_list_key", position, users=[from_user_id])
+        self.notifier.on_new_event(
+            StreamKeyType.DEVICE_LIST, position, users=[from_user_id]
+        )
 
     async def user_left_room(self, user: UserID, room_id: str) -> None:
         user_id = user.to_string()

synapse/handlers/devicemessage.py

@@ -26,7 +26,7 @@ from synapse.logging.opentracing import (
     set_tag,
 )
 from synapse.replication.http.devices import ReplicationUserDevicesResyncRestServlet
-from synapse.types import JsonDict, Requester, UserID, get_domain_from_id
+from synapse.types import JsonDict, Requester, StreamKeyType, UserID, get_domain_from_id
 from synapse.util import json_encoder
 from synapse.util.stringutils import random_string
 
@@ -151,7 +151,7 @@ class DeviceMessageHandler:
         # Notify listeners that there are new to-device messages to process,
         # handing them the latest stream id.
         self.notifier.on_new_event(
-            "to_device_key", last_stream_id, users=local_messages.keys()
+            StreamKeyType.TO_DEVICE, last_stream_id, users=local_messages.keys()
         )
 
     async def _check_for_unknown_devices(
@@ -285,7 +285,7 @@ class DeviceMessageHandler:
         # Notify listeners that there are new to-device messages to process,
         # handing them the latest stream id.
         self.notifier.on_new_event(
-            "to_device_key", last_stream_id, users=local_messages.keys()
+            StreamKeyType.TO_DEVICE, last_stream_id, users=local_messages.keys()
         )
 
         if self.federation_sender:

synapse/handlers/directory.py

@@ -73,6 +73,9 @@ class DirectoryHandler:
                 if wchar in room_alias.localpart:
                     raise SynapseError(400, "Invalid characters in room alias")
 
+        if ":" in room_alias.localpart:
+            raise SynapseError(400, "Invalid character in room alias localpart: ':'.")
+
         if not self.hs.is_mine(room_alias):
             raise SynapseError(400, "Room alias must be local")
             # TODO(erikj): Change this.

synapse/handlers/e2e_keys.py

@@ -15,7 +15,7 @@
 # limitations under the License.
 
 import logging
-from typing import TYPE_CHECKING, Dict, Iterable, List, Optional, Tuple
+from typing import TYPE_CHECKING, Any, Dict, Iterable, List, Optional, Tuple
 
 import attr
 from canonicaljson import encode_canonical_json
@@ -1105,22 +1105,19 @@ class E2eKeysHandler:
             # can request over federation
             raise NotFoundError("No %s key found for %s" % (key_type, user_id))
 
-        (
+        cross_signing_keys = await self._retrieve_cross_signing_keys_for_remote_user(
-            key,
+            user, key_type
-            key_id,
+        )
-            verify_key,
+        if cross_signing_keys is None:
-        ) = await self._retrieve_cross_signing_keys_for_remote_user(user, key_type)
-
-        if key is None:
             raise NotFoundError("No %s key found for %s" % (key_type, user_id))
 
-        return key, key_id, verify_key
+        return cross_signing_keys
 
     async def _retrieve_cross_signing_keys_for_remote_user(
         self,
         user: UserID,
         desired_key_type: str,
-    ) -> Tuple[Optional[dict], Optional[str], Optional[VerifyKey]]:
+    ) -> Optional[Tuple[Dict[str, Any], str, VerifyKey]]:
         """Queries cross-signing keys for a remote user and saves them to the database
 
         Only the key specified by `key_type` will be returned, while all retrieved keys
@@ -1146,12 +1143,10 @@ class E2eKeysHandler:
                 type(e),
                 e,
             )
-            return None, None, None
+            return None
 
         # Process each of the retrieved cross-signing keys
-        desired_key = None
+        desired_key_data = None
-        desired_key_id = None
-        desired_verify_key = None
         retrieved_device_ids = []
         for key_type in ["master", "self_signing"]:
             key_content = remote_result.get(key_type + "_key")
@@ -1196,9 +1191,7 @@ class E2eKeysHandler:
 
             # If this is the desired key type, save it and its ID/VerifyKey
             if key_type == desired_key_type:
-                desired_key = key_content
+                desired_key_data = key_content, key_id, verify_key
-                desired_verify_key = verify_key
-                desired_key_id = key_id
 
             # At the same time, store this key in the db for subsequent queries
             await self.store.set_e2e_cross_signing_key(
@@ -1212,7 +1205,7 @@ class E2eKeysHandler:
                 user.to_string(), retrieved_device_ids
             )
 
-        return desired_key, desired_key_id, desired_verify_key
+        return desired_key_data
 
 
 def _check_cross_signing_key(

synapse/handlers/event_auth.py

@@ -241,7 +241,15 @@ class EventAuthHandler:
 
         # If the join rule is not restricted, this doesn't apply.
         join_rules_event = await self._store.get_event(join_rules_event_id)
-        return join_rules_event.content.get("join_rule") == JoinRules.RESTRICTED
+        content_join_rule = join_rules_event.content.get("join_rule")
+        if content_join_rule == JoinRules.RESTRICTED:
+            return True
+
+        # also check for MSC3787 behaviour
+        if room_version.msc3787_knock_restricted_join_rule:
+            return content_join_rule == JoinRules.KNOCK_RESTRICTED
+
+        return False
 
     async def get_rooms_that_allow_join(
         self, state_ids: StateMap[str]

synapse/handlers/federation.py

@@ -54,6 +54,7 @@ from synapse.replication.http.federation import (
     ReplicationStoreRoomOnOutlierMembershipRestServlet,
 )
 from synapse.storage.databases.main.events_worker import EventRedactBehaviour
+from synapse.storage.state import StateFilter
 from synapse.types import JsonDict, StateMap, get_domain_from_id
 from synapse.util.async_helpers import Linearizer
 from synapse.util.retryutils import NotRetryingDestination
@@ -659,7 +660,7 @@ class FederationHandler:
         # in the invitee's sync stream. It is stripped out for all other local users.
         event.unsigned["knock_room_state"] = stripped_room_state["knock_state_events"]
 
-        context = EventContext.for_outlier()
+        context = EventContext.for_outlier(self.storage)
         stream_id = await self._federation_event_handler.persist_events_and_notify(
             event.room_id, [(event, context)]
         )
@@ -848,7 +849,7 @@ class FederationHandler:
             )
         )
 
-        context = EventContext.for_outlier()
+        context = EventContext.for_outlier(self.storage)
         await self._federation_event_handler.persist_events_and_notify(
             event.room_id, [(event, context)]
         )
@@ -877,7 +878,7 @@ class FederationHandler:
 
         await self.federation_client.send_leave(host_list, event)
 
-        context = EventContext.for_outlier()
+        context = EventContext.for_outlier(self.storage)
         stream_id = await self._federation_event_handler.persist_events_and_notify(
             event.room_id, [(event, context)]
         )
@@ -1259,7 +1260,9 @@ class FederationHandler:
             event.content["third_party_invite"]["signed"]["token"],
         )
         original_invite = None
-        prev_state_ids = await context.get_prev_state_ids()
+        prev_state_ids = await context.get_prev_state_ids(
+            StateFilter.from_types([(EventTypes.ThirdPartyInvite, None)])
+        )
         original_invite_id = prev_state_ids.get(key)
         if original_invite_id:
             original_invite = await self.store.get_event(
@@ -1308,7 +1311,9 @@ class FederationHandler:
         signed = event.content["third_party_invite"]["signed"]
         token = signed["token"]
 
-        prev_state_ids = await context.get_prev_state_ids()
+        prev_state_ids = await context.get_prev_state_ids(
+            StateFilter.from_types([(EventTypes.ThirdPartyInvite, None)])
+        )
         invite_event_id = prev_state_ids.get((EventTypes.ThirdPartyInvite, token))
 
         invite_event = None

synapse/handlers/federation_event.py

@@ -30,6 +30,7 @@ from typing import (
 
 from prometheus_client import Counter
 
+from synapse import event_auth
 from synapse.api.constants import (
     EventContentFields,
     EventTypes,
@@ -63,6 +64,7 @@ from synapse.replication.http.federation import (
 )
 from synapse.state import StateResolutionStore
 from synapse.storage.databases.main.events_worker import EventRedactBehaviour
+from synapse.storage.state import StateFilter
 from synapse.types import (
     PersistedEventPosition,
     RoomStreamToken,
@@ -103,7 +105,7 @@ class FederationEventHandler:
         self._event_creation_handler = hs.get_event_creation_handler()
         self._event_auth_handler = hs.get_event_auth_handler()
         self._message_handler = hs.get_message_handler()
-        self._action_generator = hs.get_action_generator()
+        self._bulk_push_rule_evaluator = hs.get_bulk_push_rule_evaluator()
         self._state_resolution_handler = hs.get_state_resolution_handler()
         # avoid a circular dependency by deferring execution here
         self._get_room_member_handler = hs.get_room_member_handler
@@ -475,7 +477,23 @@ class FederationEventHandler:
             # and discover that we do not have it.
             event.internal_metadata.proactively_send = False
 
-            return await self.persist_events_and_notify(room_id, [(event, context)])
+            stream_id_after_persist = await self.persist_events_and_notify(
+                room_id, [(event, context)]
+            )
+
+            # If we're joining the room again, check if there is new marker
+            # state indicating that there is new history imported somewhere in
+            # the DAG. Multiple markers can exist in the current state with
+            # unique state_keys.
+            #
+            # Do this after the state from the remote join was persisted (via
+            # `persist_events_and_notify`). Otherwise we can run into a
+            # situation where the create event doesn't exist yet in the
+            # `current_state_events`
+            for e in state:
+                await self._handle_marker_event(origin, e)
+
+            return stream_id_after_persist
 
     async def update_state_for_partial_state_event(
         self, destination: str, event: EventBase
@@ -1228,6 +1246,14 @@ class FederationEventHandler:
             # Nothing to retrieve then (invalid marker)
             return
 
+        already_seen_insertion_event = await self._store.have_seen_event(
+            marker_event.room_id, insertion_event_id
+        )
+        if already_seen_insertion_event:
+            # No need to process a marker again if we have already seen the
+            # insertion event that it was pointing to
+            return
+
         logger.debug(
             "_handle_marker_event: backfilling insertion event %s", insertion_event_id
         )
@@ -1423,7 +1449,7 @@ class FederationEventHandler:
                 # we're not bothering about room state, so flag the event as an outlier.
                 event.internal_metadata.outlier = True
 
-                context = EventContext.for_outlier()
+                context = EventContext.for_outlier(self._storage)
                 try:
                     validate_event_for_room_version(room_version_obj, event)
                     check_auth_rules_for_event(room_version_obj, event, auth)
@@ -1500,7 +1526,11 @@ class FederationEventHandler:
             return context
 
         # now check auth against what we think the auth events *should* be.
-        prev_state_ids = await context.get_prev_state_ids()
+        event_types = event_auth.auth_types_for_event(event.room_version, event)
+        prev_state_ids = await context.get_prev_state_ids(
+            StateFilter.from_types(event_types)
+        )
+
         auth_events_ids = self._event_auth_handler.compute_auth_events(
             event, prev_state_ids, for_verification=True
         )
@@ -1874,10 +1904,10 @@ class FederationEventHandler:
         )
 
         return EventContext.with_state(
+            storage=self._storage,
             state_group=state_group,
             state_group_before_event=context.state_group_before_event,
-            current_state_ids=current_state_ids,
+            state_delta_due_to_event=state_updates,
-            prev_state_ids=prev_state_ids,
             prev_group=prev_group,
             delta_ids=state_updates,
             partial_state=context.partial_state,
@@ -1913,7 +1943,7 @@ class FederationEventHandler:
                     min_depth,
                 )
             else:
-                await self._action_generator.handle_push_actions_for_event(
+                await self._bulk_push_rule_evaluator.action_for_event_by_user(
                     event, context
                 )
 

synapse/handlers/initial_sync.py

@@ -30,6 +30,7 @@ from synapse.types import (
     Requester,
     RoomStreamToken,
     StateMap,
+    StreamKeyType,
     StreamToken,
     UserID,
 )
@@ -143,7 +144,7 @@ class InitialSyncHandler:
             to_key=int(now_token.receipt_key),
         )
         if self.hs.config.experimental.msc2285_enabled:
-            receipt = ReceiptEventSource.filter_out_private(receipt, user_id)
+            receipt = ReceiptEventSource.filter_out_private_receipts(receipt, user_id)
 
         tags_by_room = await self.store.get_tags_for_user(user_id)
 
@@ -220,8 +221,10 @@ class InitialSyncHandler:
                     self.storage, user_id, messages
                 )
 
-                start_token = now_token.copy_and_replace("room_key", token)
+                start_token = now_token.copy_and_replace(StreamKeyType.ROOM, token)
-                end_token = now_token.copy_and_replace("room_key", room_end_token)
+                end_token = now_token.copy_and_replace(
+                    StreamKeyType.ROOM, room_end_token
+                )
                 time_now = self.clock.time_msec()
 
                 d["messages"] = {
@@ -369,8 +372,8 @@ class InitialSyncHandler:
             self.storage, user_id, messages, is_peeking=is_peeking
         )
 
-        start_token = StreamToken.START.copy_and_replace("room_key", token)
+        start_token = StreamToken.START.copy_and_replace(StreamKeyType.ROOM, token)
-        end_token = StreamToken.START.copy_and_replace("room_key", stream_token)
+        end_token = StreamToken.START.copy_and_replace(StreamKeyType.ROOM, stream_token)
 
         time_now = self.clock.time_msec()
 
@@ -449,7 +452,9 @@ class InitialSyncHandler:
             if not receipts:
                 return []
             if self.hs.config.experimental.msc2285_enabled:
-                receipts = ReceiptEventSource.filter_out_private(receipts, user_id)
+                receipts = ReceiptEventSource.filter_out_private_receipts(
+                    receipts, user_id
+                )
             return receipts
 
         presence, receipts, (messages, token) = await make_deferred_yieldable(
@@ -472,7 +477,7 @@ class InitialSyncHandler:
             self.storage, user_id, messages, is_peeking=is_peeking
         )
 
-        start_token = now_token.copy_and_replace("room_key", token)
+        start_token = now_token.copy_and_replace(StreamKeyType.ROOM, token)
         end_token = now_token
 
         time_now = self.clock.time_msec()

synapse/handlers/message.py

@@ -23,6 +23,7 @@ from canonicaljson import encode_canonical_json
 
 from twisted.internet.interfaces import IDelayedCall
 
+import synapse
 from synapse import event_auth
 from synapse.api.constants import (
     EventContentFields,
@@ -44,7 +45,7 @@ from synapse.api.errors import (
 from synapse.api.room_versions import KNOWN_ROOM_VERSIONS, RoomVersions
 from synapse.api.urls import ConsentURIBuilder
 from synapse.event_auth import validate_event_for_room_version
-from synapse.events import EventBase
+from synapse.events import EventBase, relation_from_event
 from synapse.events.builder import EventBuilder
 from synapse.events.snapshot import EventContext
 from synapse.events.validator import EventValidator
@@ -428,7 +429,7 @@ class EventCreationHandler:
         #       it just creates unnecessary state resolution without any performance improvements.
         self.limiter = Linearizer(max_count=1, name="room_event_creation_limit")
 
-        self.action_generator = hs.get_action_generator()
+        self._bulk_push_rule_evaluator = hs.get_bulk_push_rule_evaluator()
 
         self.spam_checker = hs.get_spam_checker()
         self.third_party_event_rules: "ThirdPartyEventRules" = (
@@ -636,7 +637,9 @@ class EventCreationHandler:
             # federation as well as those created locally. As of room v3, aliases events
             # can be created by users that are not in the room, therefore we have to
             # tolerate them in event_auth.check().
-            prev_state_ids = await context.get_prev_state_ids()
+            prev_state_ids = await context.get_prev_state_ids(
+                StateFilter.from_types([(EventTypes.Member, None)])
+            )
             prev_event_id = prev_state_ids.get((EventTypes.Member, event.sender))
             prev_event = (
                 await self.store.get_event(prev_event_id, allow_none=True)
@@ -759,7 +762,13 @@ class EventCreationHandler:
             The previous version of the event is returned, if it is found in the
             event context. Otherwise, None is returned.
         """
-        prev_state_ids = await context.get_prev_state_ids()
+        if event.internal_metadata.is_outlier():
+            # This can happen due to out of band memberships
+            return None
+
+        prev_state_ids = await context.get_prev_state_ids(
+            StateFilter.from_types([(event.type, None)])
+        )
         prev_event_id = prev_state_ids.get((event.type, event.state_key))
         if not prev_event_id:
             return None
@@ -879,11 +888,11 @@ class EventCreationHandler:
                 event.sender,
             )
 
-            spam_error = await self.spam_checker.check_event_for_spam(event)
+            spam_check = await self.spam_checker.check_event_for_spam(event)
-            if spam_error:
+            if spam_check is not synapse.spam_checker_api.Allow.ALLOW:
-                if not isinstance(spam_error, str):
+                raise SynapseError(
-                    spam_error = "Spam is not permitted here"
+                    403, "This message had been rejected as probable spam", spam_check
-                raise SynapseError(403, spam_error, Codes.FORBIDDEN)
+                )
 
             ev = await self.handle_new_client_event(
                 requester=requester,
@@ -1003,7 +1012,7 @@ class EventCreationHandler:
         # after it is created
         if builder.internal_metadata.outlier:
             event.internal_metadata.outlier = True
-            context = EventContext.for_outlier()
+            context = EventContext.for_outlier(self.storage)
         elif (
             event.type == EventTypes.MSC2716_INSERTION
             and state_event_ids
@@ -1058,20 +1067,11 @@ class EventCreationHandler:
             SynapseError if the event is invalid.
         """
 
-        relation = event.content.get("m.relates_to")
+        relation = relation_from_event(event)
         if not relation:
             return
 
-        relation_type = relation.get("rel_type")
+        parent_event = await self.store.get_event(relation.parent_id, allow_none=True)
-        if not relation_type:
-            return
-
-        # Ensure the parent is real.
-        relates_to = relation.get("event_id")
-        if not relates_to:
-            return
-
-        parent_event = await self.store.get_event(relates_to, allow_none=True)
         if parent_event:
             # And in the same room.
             if parent_event.room_id != event.room_id:
@@ -1080,28 +1080,31 @@ class EventCreationHandler:
         else:
             # There must be some reason that the client knows the event exists,
             # see if there are existing relations. If so, assume everything is fine.
-            if not await self.store.event_is_target_of_relation(relates_to):
+            if not await self.store.event_is_target_of_relation(relation.parent_id):
                 # Otherwise, the client can't know about the parent event!
                 raise SynapseError(400, "Can't send relation to unknown event")
 
         # If this event is an annotation then we check that that the sender
         # can't annotate the same way twice (e.g. stops users from liking an
         # event multiple times).
-        if relation_type == RelationTypes.ANNOTATION:
+        if relation.rel_type == RelationTypes.ANNOTATION:
-            aggregation_key = relation["key"]
+            aggregation_key = relation.aggregation_key
+
+            if aggregation_key is None:
+                raise SynapseError(400, "Missing aggregation key")
 
             if len(aggregation_key) > 500:
                 raise SynapseError(400, "Aggregation key is too long")
 
             already_exists = await self.store.has_user_annotated_event(
-                relates_to, event.type, aggregation_key, event.sender
+                relation.parent_id, event.type, aggregation_key, event.sender
             )
             if already_exists:
                 raise SynapseError(400, "Can't send same reaction twice")
 
         # Don't attempt to start a thread if the parent event is a relation.
-        elif relation_type == RelationTypes.THREAD:
+        elif relation.rel_type == RelationTypes.THREAD:
-            if await self.store.event_includes_relation(relates_to):
+            if await self.store.event_includes_relation(relation.parent_id):
                 raise SynapseError(
                     400, "Cannot start threads from an event with a relation"
                 )
@@ -1252,7 +1255,9 @@ class EventCreationHandler:
         # and `state_groups` because they have `prev_events` that aren't persisted yet
         # (historical messages persisted in reverse-chronological order).
         if not event.internal_metadata.is_historical() and not event.content.get(EventContentFields.MSC2716_HISTORICAL):
-            await self.action_generator.handle_push_actions_for_event(event, context)
+            await self._bulk_push_rule_evaluator.action_for_event_by_user(
+                event, context
+            )
 
         try:
             # If we're a worker we need to hit out to the master.
@@ -1559,7 +1564,11 @@ class EventCreationHandler:
                         "Redacting MSC2716 events is not supported in this room version",
                     )
 
-            prev_state_ids = await context.get_prev_state_ids()
+            event_types = event_auth.auth_types_for_event(event.room_version, event)
+            prev_state_ids = await context.get_prev_state_ids(
+                StateFilter.from_types(event_types)
+            )
+
             auth_events_ids = self._event_auth_handler.compute_auth_events(
                 event, prev_state_ids, for_verification=True
             )

synapse/handlers/oidc.py

@@ -224,7 +224,7 @@ class OidcHandler:
             self._sso_handler.render_error(request, "invalid_session", str(e))
             return
         except MacaroonInvalidSignatureException as e:
-            logger.exception("Could not verify session for OIDC callback")
+            logger.warning("Could not verify session for OIDC callback: %s", e)
             self._sso_handler.render_error(request, "mismatching_session", str(e))
             return
 
@@ -827,7 +827,7 @@ class OidcProvider:
             logger.debug("Exchanging OAuth2 code for a token")
             token = await self._exchange_code(code)
         except OidcError as e:
-            logger.exception("Could not exchange OAuth2 code")
+            logger.warning("Could not exchange OAuth2 code: %s", e)
             self._sso_handler.render_error(request, e.error, e.error_description)
             return
 

synapse/handlers/pagination.py

@@ -27,7 +27,7 @@ from synapse.handlers.room import ShutdownRoomResponse
 from synapse.metrics.background_process_metrics import run_as_background_process
 from synapse.storage.state import StateFilter
 from synapse.streams.config import PaginationConfig
-from synapse.types import JsonDict, Requester
+from synapse.types import JsonDict, Requester, StreamKeyType
 from synapse.util.async_helpers import ReadWriteLock
 from synapse.util.stringutils import random_string
 from synapse.visibility import filter_events_for_client
@@ -239,7 +239,7 @@ class PaginationHandler:
             # defined in the server's configuration, we can safely assume that's the
             # case and use it for this room.
             max_lifetime = (
-                retention_policy["max_lifetime"] or self._retention_default_max_lifetime
+                retention_policy.max_lifetime or self._retention_default_max_lifetime
             )
 
             # Cap the effective max_lifetime to be within the range allowed in the
@@ -448,7 +448,7 @@ class PaginationHandler:
             )
             # We expect `/messages` to use historic pagination tokens by default but
             # `/messages` should still works with live tokens when manually provided.
-            assert from_token.room_key.topological
+            assert from_token.room_key.topological is not None
 
         if pagin_config.limit is None:
             # This shouldn't happen as we've set a default limit before this
@@ -491,7 +491,7 @@ class PaginationHandler:
 
                     if leave_token.topological < curr_topo:
                         from_token = from_token.copy_and_replace(
-                            "room_key", leave_token
+                            StreamKeyType.ROOM, leave_token
                         )
 
                 await self.hs.get_federation_handler().maybe_backfill(
@@ -513,7 +513,7 @@ class PaginationHandler:
                 event_filter=event_filter,
             )
 
-            next_token = from_token.copy_and_replace("room_key", next_key)
+            next_token = from_token.copy_and_replace(StreamKeyType.ROOM, next_key)
 
         if events:
             if event_filter:

synapse/handlers/presence.py

@@ -66,7 +66,7 @@ from synapse.replication.tcp.commands import ClearUserSyncsCommand
 from synapse.replication.tcp.streams import PresenceFederationStream, PresenceStream
 from synapse.storage.databases.main import DataStore
 from synapse.streams import EventSource
-from synapse.types import JsonDict, UserID, get_domain_from_id
+from synapse.types import JsonDict, StreamKeyType, UserID, get_domain_from_id
 from synapse.util.async_helpers import Linearizer
 from synapse.util.caches.descriptors import _CacheContext, cached
 from synapse.util.metrics import Measure
@@ -522,7 +522,7 @@ class WorkerPresenceHandler(BasePresenceHandler):
         room_ids_to_states, users_to_states = parties
 
         self.notifier.on_new_event(
-            "presence_key",
+            StreamKeyType.PRESENCE,
             stream_id,
             rooms=room_ids_to_states.keys(),
             users=users_to_states.keys(),
@@ -1145,7 +1145,7 @@ class PresenceHandler(BasePresenceHandler):
         room_ids_to_states, users_to_states = parties
 
         self.notifier.on_new_event(
-            "presence_key",
+            StreamKeyType.PRESENCE,
             stream_id,
             rooms=room_ids_to_states.keys(),
             users=[UserID.from_string(u) for u in users_to_states],

synapse/handlers/receipts.py

@@ -17,7 +17,13 @@ from typing import TYPE_CHECKING, Iterable, List, Optional, Tuple
 from synapse.api.constants import ReceiptTypes
 from synapse.appservice import ApplicationService
 from synapse.streams import EventSource
-from synapse.types import JsonDict, ReadReceipt, UserID, get_domain_from_id
+from synapse.types import (
+    JsonDict,
+    ReadReceipt,
+    StreamKeyType,
+    UserID,
+    get_domain_from_id,
+)
 
 if TYPE_CHECKING:
     from synapse.server import HomeServer
@@ -129,7 +135,9 @@ class ReceiptsHandler:
 
         affected_room_ids = list({r.room_id for r in receipts})
 
-        self.notifier.on_new_event("receipt_key", max_batch_id, rooms=affected_room_ids)
+        self.notifier.on_new_event(
+            StreamKeyType.RECEIPT, max_batch_id, rooms=affected_room_ids
+        )
         # Note that the min here shouldn't be relied upon to be accurate.
         await self.hs.get_pusherpool().on_new_receipts(
             min_batch_id, max_batch_id, affected_room_ids
@@ -166,43 +174,69 @@ class ReceiptEventSource(EventSource[int, JsonDict]):
         self.config = hs.config
 
     @staticmethod
-    def filter_out_private(events: List[JsonDict], user_id: str) -> List[JsonDict]:
+    def filter_out_private_receipts(
+        rooms: List[JsonDict], user_id: str
+    ) -> List[JsonDict]:
         """
-        This method takes in what is returned by
+        Filters a list of serialized receipts (as returned by /sync and /initialSync)
-        get_linearized_receipts_for_rooms() and goes through read receipts
+        and removes private read receipts of other users.
-        filtering out m.read.private receipts if they were not sent by the
+
-        current user.
+        This operates on the return value of get_linearized_receipts_for_rooms(),
+        which is wrapped in a cache. Care must be taken to ensure that the input
+        values are not modified.
+
+        Args:
+            rooms: A list of mappings, each mapping has a `content` field, which
+                is a map of event ID -> receipt type -> user ID -> receipt information.
+
+        Returns:
+            The same as rooms, but filtered.
         """
 
-        visible_events = []
+        result = []
 
-        # filter out private receipts the user shouldn't see
+        # Iterate through each room's receipt content.
-        for event in events:
+        for room in rooms:
-            content = event.get("content", {})
+            # The receipt content with other user's private read receipts removed.
-            new_event = event.copy()
+            content = {}
-            new_event["content"] = {}
 
-            for event_id, event_content in content.items():
+            # Iterate over each event ID / receipts for that event.
-                receipt_event = {}
+            for event_id, orig_event_content in room.get("content", {}).items():
-                for receipt_type, receipt_content in event_content.items():
+                event_content = orig_event_content
-                    if receipt_type == ReceiptTypes.READ_PRIVATE:
+                # If there are private read receipts, additional logic is necessary.
-                        user_rr = receipt_content.get(user_id, None)
+                if ReceiptTypes.READ_PRIVATE in event_content:
-                        if user_rr:
+                    # Make a copy without private read receipts to avoid leaking
-                            receipt_event[ReceiptTypes.READ_PRIVATE] = {
+                    # other user's private read receipts..
-                                user_id: user_rr.copy()
+                    event_content = {
-                            }
+                        receipt_type: receipt_value
-                    else:
+                        for receipt_type, receipt_value in event_content.items()
-                        receipt_event[receipt_type] = receipt_content.copy()
+                        if receipt_type != ReceiptTypes.READ_PRIVATE
+                    }
 
-                # Only include the receipt event if it is non-empty.
+                    # Copy the current user's private read receipt from the
-                if receipt_event:
+                    # original content, if it exists.
-                    new_event["content"][event_id] = receipt_event
+                    user_private_read_receipt = orig_event_content[
+                        ReceiptTypes.READ_PRIVATE
+                    ].get(user_id, None)
+                    if user_private_read_receipt:
+                        event_content[ReceiptTypes.READ_PRIVATE] = {
+                            user_id: user_private_read_receipt
+                        }
 
-            # Append new_event to visible_events unless empty
+                # Include the event if there is at least one non-private read
-            if len(new_event["content"].keys()) > 0:
+                # receipt or the current user has a private read receipt.
-                visible_events.append(new_event)
+                if event_content:
+                    content[event_id] = event_content
 
-        return visible_events
+            # Include the event if there is at least one non-private read receipt
+            # or the current user has a private read receipt.
+            if content:
+                # Build a new event to avoid mutating the cache.
+                new_room = {k: v for k, v in room.items() if k != "content"}
+                new_room["content"] = content
+                result.append(new_room)
+
+        return result
 
     async def get_new_events(
         self,
@@ -224,7 +258,9 @@ class ReceiptEventSource(EventSource[int, JsonDict]):
         )
 
         if self.config.experimental.msc2285_enabled:
-            events = ReceiptEventSource.filter_out_private(events, user.to_string())
+            events = ReceiptEventSource.filter_out_private_receipts(
+                events, user.to_string()
+            )
 
         return events, to_key
 

synapse/handlers/relations.py

@@ -11,7 +11,6 @@
 # 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 collections.abc
 import logging
 from typing import (
     TYPE_CHECKING,
@@ -28,7 +27,7 @@ import attr
 
 from synapse.api.constants import RelationTypes
 from synapse.api.errors import SynapseError
-from synapse.events import EventBase
+from synapse.events import EventBase, relation_from_event
 from synapse.storage.databases.main.relations import _RelatedEvent
 from synapse.types import JsonDict, Requester, StreamToken, UserID
 from synapse.visibility import filter_events_for_client
@@ -373,20 +372,21 @@ class RelationsHandler:
             if event.is_state():
                 continue
 
-            relates_to = event.content.get("m.relates_to")
+            relates_to = relation_from_event(event)
-            relation_type = None
+            if relates_to:
-            if isinstance(relates_to, collections.abc.Mapping):
-                relation_type = relates_to.get("rel_type")
                 # An event which is a replacement (ie edit) or annotation (ie,
                 # reaction) may not have any other event related to it.
-                if relation_type in (RelationTypes.ANNOTATION, RelationTypes.REPLACE):
+                if relates_to.rel_type in (
+                    RelationTypes.ANNOTATION,
+                    RelationTypes.REPLACE,
+                ):
                     continue
 
+                # Track the event's relation information for later.
+                relations_by_id[event.event_id] = relates_to.rel_type
+
             # The event should get bundled aggregations.
             events_by_id[event.event_id] = event
-            # Track the event's relation information for later.
-            if isinstance(relation_type, str):
-                relations_by_id[event.event_id] = relation_type
 
         # event ID -> bundled aggregation in non-serialized form.
         results: Dict[str, BundledAggregations] = {}

synapse/handlers/room.py

@@ -33,6 +33,7 @@ from typing import (
 import attr
 from typing_extensions import TypedDict
 
+import synapse.events.snapshot
 from synapse.api.constants import (
     EventContentFields,
     EventTypes,
@@ -72,12 +73,12 @@ from synapse.types import (
     RoomID,
     RoomStreamToken,
     StateMap,
+    StreamKeyType,
     StreamToken,
     UserID,
     create_requester,
 )
 from synapse.util import stringutils
-from synapse.util.async_helpers import Linearizer
 from synapse.util.caches.response_cache import ResponseCache
 from synapse.util.stringutils import parse_and_validate_server_name
 from synapse.visibility import filter_events_for_client
@@ -149,10 +150,11 @@ class RoomCreationHandler:
             )
             preset_config["encrypted"] = encrypted
 
-        self._replication = hs.get_replication_data_handler()
+        self._default_power_level_content_override = (
+            self.config.room.default_power_level_content_override
+        )
 
-        # linearizer to stop two upgrades happening at once
+        self._replication = hs.get_replication_data_handler()
-        self._upgrade_linearizer = Linearizer("room_upgrade_linearizer")
 
         # If a user tries to update the same room multiple times in quick
         # succession, only process the first attempt and return its result to
@@ -196,50 +198,17 @@ class RoomCreationHandler:
                     400, "An upgrade for this room is currently in progress"
                 )
 
-        # Upgrade the room
+        # Check whether the room exists and 404 if it doesn't.
-        #
+        # We could go straight for the auth check, but that will raise a 403 instead.
-        # If this user has sent multiple upgrade requests for the same room
+        old_room = await self.store.get_room(old_room_id)
-        # and one of them is not complete yet, cache the response and
+        if old_room is None:
-        # return it to all subsequent requests
-        ret = await self._upgrade_response_cache.wrap(
-            (old_room_id, user_id),
-            self._upgrade_room,
-            requester,
-            old_room_id,
-            new_version,  # args for _upgrade_room
-        )
-
-        return ret
-
-    async def _upgrade_room(
-        self, requester: Requester, old_room_id: str, new_version: RoomVersion
-    ) -> str:
-        """
-        Args:
-            requester: the user requesting the upgrade
-            old_room_id: the id of the room to be replaced
-            new_versions: the version to upgrade the room to
-
-        Raises:
-            ShadowBanError if the requester is shadow-banned.
-        """
-        user_id = requester.user.to_string()
-        assert self.hs.is_mine_id(user_id), "User must be our own: %s" % (user_id,)
-
-        # start by allocating a new room id
-        r = await self.store.get_room(old_room_id)
-        if r is None:
             raise NotFoundError("Unknown room id %s" % (old_room_id,))
-        new_room_id = await self._generate_room_id(
-            creator_id=user_id,
-            is_public=r["is_public"],
-            room_version=new_version,
-        )
 
-        logger.info("Creating new room %s to replace %s", new_room_id, old_room_id)
+        new_room_id = self._generate_room_id()
 
-        # we create and auth the tombstone event before properly creating the new
+        # Check whether the user has the power level to carry out the upgrade.
-        # room, to check our user has perms in the old room.
+        # `check_auth_rules_from_context` will check that they are in the room and have
+        # the required power level to send the tombstone event.
         (
             tombstone_event,
             tombstone_context,
@@ -262,6 +231,63 @@ class RoomCreationHandler:
             old_room_version, tombstone_event, tombstone_context
         )
 
+        # Upgrade the room
+        #
+        # If this user has sent multiple upgrade requests for the same room
+        # and one of them is not complete yet, cache the response and
+        # return it to all subsequent requests
+        ret = await self._upgrade_response_cache.wrap(
+            (old_room_id, user_id),
+            self._upgrade_room,
+            requester,
+            old_room_id,
+            old_room,  # args for _upgrade_room
+            new_room_id,
+            new_version,
+            tombstone_event,
+            tombstone_context,
+        )
+
+        return ret
+
+    async def _upgrade_room(
+        self,
+        requester: Requester,
+        old_room_id: str,
+        old_room: Dict[str, Any],
+        new_room_id: str,
+        new_version: RoomVersion,
+        tombstone_event: EventBase,
+        tombstone_context: synapse.events.snapshot.EventContext,
+    ) -> str:
+        """
+        Args:
+            requester: the user requesting the upgrade
+            old_room_id: the id of the room to be replaced
+            old_room: a dict containing room information for the room to be replaced,
+                as returned by `RoomWorkerStore.get_room`.
+            new_room_id: the id of the replacement room
+            new_version: the version to upgrade the room to
+            tombstone_event: the tombstone event to send to the old room
+            tombstone_context: the context for the tombstone event
+
+        Raises:
+            ShadowBanError if the requester is shadow-banned.
+        """
+        user_id = requester.user.to_string()
+        assert self.hs.is_mine_id(user_id), "User must be our own: %s" % (user_id,)
+
+        logger.info("Creating new room %s to replace %s", new_room_id, old_room_id)
+
+        # create the new room. may raise a `StoreError` in the exceedingly unlikely
+        # event of a room ID collision.
+        await self.store.store_room(
+            room_id=new_room_id,
+            room_creator_user_id=user_id,
+            is_public=old_room["is_public"],
+            room_version=new_version,
+        )
+
         await self.clone_existing_room(
             requester,
             old_room_id=old_room_id,
@@ -277,7 +303,10 @@ class RoomCreationHandler:
             context=tombstone_context,
         )
 
-        old_room_state = await tombstone_context.get_current_state_ids()
+        state_filter = StateFilter.from_types(
+            [(EventTypes.CanonicalAlias, ""), (EventTypes.PowerLevels, "")]
+        )
+        old_room_state = await tombstone_context.get_current_state_ids(state_filter)
 
         # We know the tombstone event isn't an outlier so it has current state.
         assert old_room_state is not None
@@ -401,7 +430,7 @@ class RoomCreationHandler:
             requester: the user requesting the upgrade
             old_room_id : the id of the room to be replaced
             new_room_id: the id to give the new room (should already have been
-                created with _gemerate_room_id())
+                created with _generate_room_id())
             new_room_version: the new room version to use
             tombstone_event_id: the ID of the tombstone event in the old room.
         """
@@ -443,14 +472,14 @@ class RoomCreationHandler:
             (EventTypes.PowerLevels, ""),
         ]
 
-        # If the old room was a space, copy over the room type and the rooms in
+        # Copy the room type as per MSC3818.
-        # the space.
+        room_type = old_room_create_event.content.get(EventContentFields.ROOM_TYPE)
-        if (
+        if room_type is not None:
-            old_room_create_event.content.get(EventContentFields.ROOM_TYPE)
+            creation_content[EventContentFields.ROOM_TYPE] = room_type
-            == RoomTypes.SPACE
+
-        ):
+            # If the old room was a space, copy over the rooms in the space.
-            creation_content[EventContentFields.ROOM_TYPE] = RoomTypes.SPACE
+            if room_type == RoomTypes.SPACE:
-            types_to_copy.append((EventTypes.SpaceChild, None))
+                types_to_copy.append((EventTypes.SpaceChild, None))
 
         old_room_state_ids = await self.store.get_filtered_current_state_ids(
             old_room_id, StateFilter.from_types(types_to_copy)
@@ -725,6 +754,21 @@ class RoomCreationHandler:
                 if wchar in config["room_alias_name"]:
                     raise SynapseError(400, "Invalid characters in room alias")
 
+            if ":" in config["room_alias_name"]:
+                # Prevent someone from trying to pass in a full alias here.
+                # Note that it's permissible for a room alias to have multiple
+                # hash symbols at the start (notably bridged over from IRC, too),
+                # but the first colon in the alias is defined to separate the local
+                # part from the server name.
+                # (remember server names can contain port numbers, also separated
+                # by a colon. But under no circumstances should the local part be
+                # allowed to contain a colon!)
+                raise SynapseError(
+                    400,
+                    "':' is not permitted in the room alias name. "
+                    "Please note this expects a local part — 'wombat', not '#wombat:example.com'.",
+                )
+
             room_alias = RoomAlias(config["room_alias_name"], self.hs.hostname)
             mapping = await self.store.get_association_from_room_alias(room_alias)
 
@@ -790,8 +834,10 @@ class RoomCreationHandler:
             except StoreError:
                 raise SynapseError(409, "Room ID already in use", errcode="M_CONFLICT")
         else:
-            room_id = await self._generate_room_id(
+            room_id = await self._generate_and_create_room_id(
-                creator_id=user_id, is_public=is_public, room_version=room_version,
+                creator_id=user_id,
+                is_public=is_public,
+                room_version=room_version,
             )
 
         # Check whether this visibility value is blocked by a third party module
@@ -1052,9 +1098,19 @@ class RoomCreationHandler:
                 for invitee in invite_list:
                     power_level_content["users"][invitee] = 100
 
-            # Power levels overrides are defined per chat preset
+            # If the user supplied a preset name e.g. "private_chat",
+            # we apply that preset
             power_level_content.update(config["power_level_content_override"])
 
+            # If the server config contains default_power_level_content_override,
+            # and that contains information for this room preset, apply it.
+            if self._default_power_level_content_override:
+                override = self._default_power_level_content_override.get(preset_config)
+                if override is not None:
+                    power_level_content.update(override)
+
+            # Finally, if the user supplied specific permissions for this room,
+            # apply those.
             if power_level_content_override:
                 power_level_content.update(power_level_content_override)
 
@@ -1100,7 +1156,26 @@ class RoomCreationHandler:
 
         return last_sent_stream_id
 
-    async def _generate_room_id(
+    def _generate_room_id(self) -> str:
+        """Generates a random room ID.
+
+        Room IDs look like "!opaque_id:domain" and are case-sensitive as per the spec
+        at https://spec.matrix.org/v1.2/appendices/#room-ids-and-event-ids.
+
+        Does not check for collisions with existing rooms or prevent future calls from
+        returning the same room ID. To ensure the uniqueness of a new room ID, use
+        `_generate_and_create_room_id` instead.
+
+        Synapse's room IDs are 18 [a-zA-Z] characters long, which comes out to around
+        102 bits.
+
+        Returns:
+            A random room ID of the form "!opaque_id:domain".
+        """
+        random_string = stringutils.random_string(18)
+        return RoomID(random_string, self.hs.hostname).to_string()
+
+    async def _generate_and_create_room_id(
         self,
         creator_id: str,
         is_public: bool,
@@ -1111,8 +1186,7 @@ class RoomCreationHandler:
         attempts = 0
         while attempts < 5:
             try:
-                random_string = stringutils.random_string(18)
+                gen_room_id = self._generate_room_id()
-                gen_room_id = RoomID(random_string, self.hs.hostname).to_string()
                 await self.store.store_room(
                     room_id=gen_room_id,
                     room_creator_user_id=creator_id,
@@ -1249,10 +1323,10 @@ class RoomContextHandler:
             events_after=events_after,
             state=await filter_evts(state_events),
             aggregations=aggregations,
-            start=await token.copy_and_replace("room_key", results.start).to_string(
+            start=await token.copy_and_replace(
-                self.store
+                StreamKeyType.ROOM, results.start
-            ),
+            ).to_string(self.store),
-            end=await token.copy_and_replace("room_key", results.end).to_string(
+            end=await token.copy_and_replace(StreamKeyType.ROOM, results.end).to_string(
                 self.store
             ),
         )

synapse/handlers/room_batch.py

@@ -54,6 +54,7 @@ class RoomBatchHandler:
         # We want to use the successor event depth so they appear after `prev_event` because
         # it has a larger `depth` but before the successor event because the `stream_ordering`
         # is negative before the successor event.
+        assert most_recent_prev_event_id is not None
         successor_event_ids = await self.store.get_successor_events(
             most_recent_prev_event_id
         )
@@ -141,6 +142,7 @@ class RoomBatchHandler:
             _,
         ) = await self.store.get_max_depth_of(event_ids)
         # mapping from (type, state_key) -> state_event_id
+        assert most_recent_event_id is not None
         prev_state_map = await self.state_store.get_state_ids_for_event(
             most_recent_event_id
         )

synapse/handlers/room_member.py

@@ -38,6 +38,7 @@ from synapse.event_auth import get_named_level, get_power_level_event
 from synapse.events import EventBase
 from synapse.events.snapshot import EventContext
 from synapse.handlers.profile import MAX_AVATAR_URL_LEN, MAX_DISPLAYNAME_LEN
+from synapse.storage.state import StateFilter
 from synapse.types import (
     JsonDict,
     Requester,
@@ -362,7 +363,9 @@ class RoomMemberHandler(metaclass=abc.ABCMeta):
             historical=historical,
         )
 
-        prev_state_ids = await context.get_prev_state_ids()
+        prev_state_ids = await context.get_prev_state_ids(
+            StateFilter.from_types([(EventTypes.Member, None)])
+        )
 
         prev_member_event_id = prev_state_ids.get((EventTypes.Member, user_id), None)
 
@@ -1160,7 +1163,9 @@ class RoomMemberHandler(metaclass=abc.ABCMeta):
         else:
             requester = types.create_requester(target_user)
 
-        prev_state_ids = await context.get_prev_state_ids()
+        prev_state_ids = await context.get_prev_state_ids(
+            StateFilter.from_types([(EventTypes.GuestAccess, None)])
+        )
         if event.membership == Membership.JOIN:
             if requester.is_guest:
                 guest_can_join = await self._can_guest_join(prev_state_ids)

synapse/handlers/room_summary.py

@@ -562,8 +562,13 @@ class RoomSummaryHandler:
         if join_rules_event_id:
             join_rules_event = await self._store.get_event(join_rules_event_id)
             join_rule = join_rules_event.content.get("join_rule")
-            if join_rule == JoinRules.PUBLIC or (
+            if (
-                room_version.msc2403_knocking and join_rule == JoinRules.KNOCK
+                join_rule == JoinRules.PUBLIC
+                or (room_version.msc2403_knocking and join_rule == JoinRules.KNOCK)
+                or (
+                    room_version.msc3787_knock_restricted_join_rule
+                    and join_rule == JoinRules.KNOCK_RESTRICTED
+                )
             ):
                 return True
 

synapse/handlers/search.py

@@ -24,7 +24,7 @@ from synapse.api.errors import NotFoundError, SynapseError
 from synapse.api.filtering import Filter
 from synapse.events import EventBase
 from synapse.storage.state import StateFilter
-from synapse.types import JsonDict, UserID
+from synapse.types import JsonDict, StreamKeyType, UserID
 from synapse.visibility import filter_events_for_client
 
 if TYPE_CHECKING:
@@ -655,11 +655,11 @@ class SearchHandler:
                 "events_before": events_before,
                 "events_after": events_after,
                 "start": await now_token.copy_and_replace(
-                    "room_key", res.start
+                    StreamKeyType.ROOM, res.start
+                ).to_string(self.store),
+                "end": await now_token.copy_and_replace(
+                    StreamKeyType.ROOM, res.end
                 ).to_string(self.store),
-                "end": await now_token.copy_and_replace("room_key", res.end).to_string(
-                    self.store
-                ),
             }
 
             if include_profile:

synapse/handlers/sync.py

@@ -37,6 +37,7 @@ from synapse.types import (
     Requester,
     RoomStreamToken,
     StateMap,
+    StreamKeyType,
     StreamToken,
     UserID,
 )
@@ -410,10 +411,10 @@ class SyncHandler:
             set_tag(SynapseTags.SYNC_RESULT, bool(sync_result))
             return sync_result
 
-    async def push_rules_for_user(self, user: UserID) -> JsonDict:
+    async def push_rules_for_user(self, user: UserID) -> Dict[str, Dict[str, list]]:
         user_id = user.to_string()
-        rules = await self.store.get_push_rules_for_user(user_id)
+        rules_raw = await self.store.get_push_rules_for_user(user_id)
-        rules = format_push_rules_for_user(user, rules)
+        rules = format_push_rules_for_user(user, rules_raw)
         return rules
 
     async def ephemeral_by_room(
@@ -449,7 +450,7 @@ class SyncHandler:
                 room_ids=room_ids,
                 is_guest=sync_config.is_guest,
             )
-            now_token = now_token.copy_and_replace("typing_key", typing_key)
+            now_token = now_token.copy_and_replace(StreamKeyType.TYPING, typing_key)
 
             ephemeral_by_room: JsonDict = {}
 
@@ -471,7 +472,7 @@ class SyncHandler:
                 room_ids=room_ids,
                 is_guest=sync_config.is_guest,
             )
-            now_token = now_token.copy_and_replace("receipt_key", receipt_key)
+            now_token = now_token.copy_and_replace(StreamKeyType.RECEIPT, receipt_key)
 
             for event in receipts:
                 room_id = event["room_id"]
@@ -537,7 +538,9 @@ class SyncHandler:
                 prev_batch_token = now_token
                 if recents:
                     room_key = recents[0].internal_metadata.before
-                    prev_batch_token = now_token.copy_and_replace("room_key", room_key)
+                    prev_batch_token = now_token.copy_and_replace(
+                        StreamKeyType.ROOM, room_key
+                    )
 
                 return TimelineBatch(
                     events=recents, prev_batch=prev_batch_token, limited=False
@@ -611,7 +614,7 @@ class SyncHandler:
                 recents = recents[-timeline_limit:]
                 room_key = recents[0].internal_metadata.before
 
-            prev_batch_token = now_token.copy_and_replace("room_key", room_key)
+            prev_batch_token = now_token.copy_and_replace(StreamKeyType.ROOM, room_key)
 
         # Don't bother to bundle aggregations if the timeline is unlimited,
         # as clients will have all the necessary information.
@@ -1397,7 +1400,7 @@ class SyncHandler:
                 now_token.to_device_key,
             )
             sync_result_builder.now_token = now_token.copy_and_replace(
-                "to_device_key", stream_id
+                StreamKeyType.TO_DEVICE, stream_id
             )
             sync_result_builder.to_device = messages
         else:
@@ -1502,7 +1505,7 @@ class SyncHandler:
         )
         assert presence_key
         sync_result_builder.now_token = now_token.copy_and_replace(
-            "presence_key", presence_key
+            StreamKeyType.PRESENCE, presence_key
         )
 
         extra_users_ids = set(newly_joined_or_invited_users)
@@ -1825,7 +1828,7 @@ class SyncHandler:
                 # stream token as it'll only be used in the context of this
                 # room. (c.f. the docstring of `to_room_stream_token`).
                 leave_token = since_token.copy_and_replace(
-                    "room_key", leave_position.to_room_stream_token()
+                    StreamKeyType.ROOM, leave_position.to_room_stream_token()
                 )
 
                 # If this is an out of band message, like a remote invite
@@ -1874,7 +1877,9 @@ class SyncHandler:
             if room_entry:
                 events, start_key = room_entry
 
-                prev_batch_token = now_token.copy_and_replace("room_key", start_key)
+                prev_batch_token = now_token.copy_and_replace(
+                    StreamKeyType.ROOM, start_key
+                )
 
                 entry = RoomSyncResultBuilder(
                     room_id=room_id,
@@ -1971,7 +1976,7 @@ class SyncHandler:
                             continue
 
                 leave_token = now_token.copy_and_replace(
-                    "room_key", RoomStreamToken(None, event.stream_ordering)
+                    StreamKeyType.ROOM, RoomStreamToken(None, event.stream_ordering)
                 )
                 room_entries.append(
                     RoomSyncResultBuilder(

synapse/handlers/typing.py

@@ -25,7 +25,7 @@ from synapse.metrics.background_process_metrics import (
 )
 from synapse.replication.tcp.streams import TypingStream
 from synapse.streams import EventSource
-from synapse.types import JsonDict, Requester, UserID, get_domain_from_id
+from synapse.types import JsonDict, Requester, StreamKeyType, UserID, get_domain_from_id
 from synapse.util.caches.stream_change_cache import StreamChangeCache
 from synapse.util.metrics import Measure
 from synapse.util.wheel_timer import WheelTimer
@@ -382,7 +382,7 @@ class TypingWriterHandler(FollowerTypingHandler):
         )
 
         self.notifier.on_new_event(
-            "typing_key", self._latest_room_serial, rooms=[member.room_id]
+            StreamKeyType.TYPING, self._latest_room_serial, rooms=[member.room_id]
         )
 
     async def get_all_typing_updates(

synapse/http/client.py

@@ -43,8 +43,10 @@ from twisted.internet import defer, error as twisted_error, protocol, ssl
 from twisted.internet.address import IPv4Address, IPv6Address
 from twisted.internet.interfaces import (
     IAddress,
+    IDelayedCall,
     IHostResolution,
     IReactorPluggableNameResolver,
+    IReactorTime,
     IResolutionReceiver,
     ITCPTransport,
 )
@@ -121,13 +123,15 @@ def check_against_blacklist(
 _EPSILON = 0.00000001
 
 
-def _make_scheduler(reactor):
+def _make_scheduler(
+    reactor: IReactorTime,
+) -> Callable[[Callable[[], object]], IDelayedCall]:
     """Makes a schedular suitable for a Cooperator using the given reactor.
 
     (This is effectively just a copy from `twisted.internet.task`)
     """
 
-    def _scheduler(x):
+    def _scheduler(x: Callable[[], object]) -> IDelayedCall:
         return reactor.callLater(_EPSILON, x)
 
     return _scheduler
@@ -349,7 +353,7 @@ class SimpleHttpClient:
         # XXX: The justification for using the cache factor here is that larger instances
         # will need both more cache and more connections.
         # Still, this should probably be a separate dial
-        pool.maxPersistentPerHost = max((100 * hs.config.caches.global_factor, 5))
+        pool.maxPersistentPerHost = max(int(100 * hs.config.caches.global_factor), 5)
         pool.cachedConnectionTimeout = 2 * 60
 
         self.agent: IAgent = ProxyAgent(
@@ -776,7 +780,7 @@ class SimpleHttpClient:
         )
 
 
-def _timeout_to_request_timed_out_error(f: Failure):
+def _timeout_to_request_timed_out_error(f: Failure) -> Failure:
     if f.check(twisted_error.TimeoutError, twisted_error.ConnectingCancelledError):
         # The TCP connection has its own timeout (set by the 'connectTimeout' param
         # on the Agent), which raises twisted_error.TimeoutError exception.
@@ -810,7 +814,7 @@ class _DiscardBodyWithMaxSizeProtocol(protocol.Protocol):
     def __init__(self, deferred: defer.Deferred):
         self.deferred = deferred
 
-    def _maybe_fail(self):
+    def _maybe_fail(self) -> None:
         """
         Report a max size exceed error and disconnect the first time this is called.
         """
@@ -934,12 +938,12 @@ class InsecureInterceptableContextFactory(ssl.ContextFactory):
     Do not use this since it allows an attacker to intercept your communications.
     """
 
-    def __init__(self):
+    def __init__(self) -> None:
         self._context = SSL.Context(SSL.SSLv23_METHOD)
         self._context.set_verify(VERIFY_NONE, lambda *_: False)
 
     def getContext(self, hostname=None, port=None):
         return self._context
 
-    def creatorForNetloc(self, hostname, port):
+    def creatorForNetloc(self, hostname: bytes, port: int):
         return self

synapse/http/connectproxyclient.py

@@ -14,15 +14,22 @@
 
 import base64
 import logging
-from typing import Optional
+from typing import Optional, Union
 
 import attr
 from zope.interface import implementer
 
 from twisted.internet import defer, protocol
 from twisted.internet.error import ConnectError
-from twisted.internet.interfaces import IReactorCore, IStreamClientEndpoint
+from twisted.internet.interfaces import (
+    IAddress,
+    IConnector,
+    IProtocol,
+    IReactorCore,
+    IStreamClientEndpoint,
+)
 from twisted.internet.protocol import ClientFactory, Protocol, connectionDone
+from twisted.python.failure import Failure
 from twisted.web import http
 
 logger = logging.getLogger(__name__)
@@ -81,14 +88,14 @@ class HTTPConnectProxyEndpoint:
         self._port = port
         self._proxy_creds = proxy_creds
 
-    def __repr__(self):
+    def __repr__(self) -> str:
         return "<HTTPConnectProxyEndpoint %s>" % (self._proxy_endpoint,)
 
     # Mypy encounters a false positive here: it complains that ClientFactory
     # is incompatible with IProtocolFactory. But ClientFactory inherits from
     # Factory, which implements IProtocolFactory. So I think this is a bug
     # in mypy-zope.
-    def connect(self, protocolFactory: ClientFactory):  # type: ignore[override]
+    def connect(self, protocolFactory: ClientFactory) -> "defer.Deferred[IProtocol]":  # type: ignore[override]
         f = HTTPProxiedClientFactory(
             self._host, self._port, protocolFactory, self._proxy_creds
         )
@@ -125,10 +132,10 @@ class HTTPProxiedClientFactory(protocol.ClientFactory):
         self.proxy_creds = proxy_creds
         self.on_connection: "defer.Deferred[None]" = defer.Deferred()
 
-    def startedConnecting(self, connector):
+    def startedConnecting(self, connector: IConnector) -> None:
         return self.wrapped_factory.startedConnecting(connector)
 
-    def buildProtocol(self, addr):
+    def buildProtocol(self, addr: IAddress) -> "HTTPConnectProtocol":
         wrapped_protocol = self.wrapped_factory.buildProtocol(addr)
         if wrapped_protocol is None:
             raise TypeError("buildProtocol produced None instead of a Protocol")
@@ -141,13 +148,13 @@ class HTTPProxiedClientFactory(protocol.ClientFactory):
             self.proxy_creds,
         )
 
-    def clientConnectionFailed(self, connector, reason):
+    def clientConnectionFailed(self, connector: IConnector, reason: Failure) -> None:
         logger.debug("Connection to proxy failed: %s", reason)
         if not self.on_connection.called:
             self.on_connection.errback(reason)
         return self.wrapped_factory.clientConnectionFailed(connector, reason)
 
-    def clientConnectionLost(self, connector, reason):
+    def clientConnectionLost(self, connector: IConnector, reason: Failure) -> None:
         logger.debug("Connection to proxy lost: %s", reason)
         if not self.on_connection.called:
             self.on_connection.errback(reason)
@@ -191,10 +198,10 @@ class HTTPConnectProtocol(protocol.Protocol):
         )
         self.http_setup_client.on_connected.addCallback(self.proxyConnected)
 
-    def connectionMade(self):
+    def connectionMade(self) -> None:
         self.http_setup_client.makeConnection(self.transport)
 
-    def connectionLost(self, reason=connectionDone):
+    def connectionLost(self, reason: Failure = connectionDone) -> None:
         if self.wrapped_protocol.connected:
             self.wrapped_protocol.connectionLost(reason)
 
@@ -203,7 +210,7 @@ class HTTPConnectProtocol(protocol.Protocol):
         if not self.connected_deferred.called:
             self.connected_deferred.errback(reason)
 
-    def proxyConnected(self, _):
+    def proxyConnected(self, _: Union[None, "defer.Deferred[None]"]) -> None:
         self.wrapped_protocol.makeConnection(self.transport)
 
         self.connected_deferred.callback(self.wrapped_protocol)
@@ -213,7 +220,7 @@ class HTTPConnectProtocol(protocol.Protocol):
         if buf:
             self.wrapped_protocol.dataReceived(buf)
 
-    def dataReceived(self, data: bytes):
+    def dataReceived(self, data: bytes) -> None:
         # if we've set up the HTTP protocol, we can send the data there
         if self.wrapped_protocol.connected:
             return self.wrapped_protocol.dataReceived(data)
@@ -243,7 +250,7 @@ class HTTPConnectSetupClient(http.HTTPClient):
         self.proxy_creds = proxy_creds
         self.on_connected: "defer.Deferred[None]" = defer.Deferred()
 
-    def connectionMade(self):
+    def connectionMade(self) -> None:
         logger.debug("Connected to proxy, sending CONNECT")
         self.sendCommand(b"CONNECT", b"%s:%d" % (self.host, self.port))
 
@@ -257,14 +264,14 @@ class HTTPConnectSetupClient(http.HTTPClient):
 
         self.endHeaders()
 
-    def handleStatus(self, version: bytes, status: bytes, message: bytes):
+    def handleStatus(self, version: bytes, status: bytes, message: bytes) -> None:
         logger.debug("Got Status: %s %s %s", status, message, version)
         if status != b"200":
             raise ProxyConnectError(f"Unexpected status on CONNECT: {status!s}")
 
-    def handleEndHeaders(self):
+    def handleEndHeaders(self) -> None:
         logger.debug("End Headers")
         self.on_connected.callback(None)
 
-    def handleResponse(self, body):
+    def handleResponse(self, body: bytes) -> None:
         pass

synapse/http/federation/matrix_federation_agent.py

@@ -239,7 +239,7 @@ class MatrixHostnameEndpointFactory:
 
         self._srv_resolver = srv_resolver
 
-    def endpointForURI(self, parsed_uri: URI):
+    def endpointForURI(self, parsed_uri: URI) -> "MatrixHostnameEndpoint":
         return MatrixHostnameEndpoint(
             self._reactor,
             self._proxy_reactor,

synapse/http/federation/srv_resolver.py

@@ -16,7 +16,7 @@
 import logging
 import random
 import time
-from typing import Callable, Dict, List
+from typing import Any, Callable, Dict, List
 
 import attr
 
@@ -109,7 +109,7 @@ class SrvResolver:
 
     def __init__(
         self,
-        dns_client=client,
+        dns_client: Any = client,
         cache: Dict[bytes, List[Server]] = SERVER_CACHE,
         get_time: Callable[[], float] = time.time,
     ):

synapse/http/federation/well_known_resolver.py

@@ -74,9 +74,9 @@ _well_known_cache: TTLCache[bytes, Optional[bytes]] = TTLCache("well-known")
 _had_valid_well_known_cache: TTLCache[bytes, bool] = TTLCache("had-valid-well-known")
 
 
-@attr.s(slots=True, frozen=True)
+@attr.s(slots=True, frozen=True, auto_attribs=True)
 class WellKnownLookupResult:
-    delegated_server = attr.ib()
+    delegated_server: Optional[bytes]
 
 
 class WellKnownResolver:
@@ -336,4 +336,4 @@ def _parse_cache_control(headers: Headers) -> Dict[bytes, Optional[bytes]]:
 class _FetchWellKnownFailure(Exception):
     # True if we didn't get a non-5xx HTTP response, i.e. this may or may not be
     # a temporary failure.
-    temporary = attr.ib()
+    temporary: bool = attr.ib()

synapse/http/matrixfederationclient.py

@@ -23,6 +23,8 @@ from http import HTTPStatus
 from io import BytesIO, StringIO
 from typing import (
     TYPE_CHECKING,
+    Any,
+    BinaryIO,
     Callable,
     Dict,
     Generic,
@@ -44,7 +46,7 @@ from typing_extensions import Literal
 from twisted.internet import defer
 from twisted.internet.error import DNSLookupError
 from twisted.internet.interfaces import IReactorTime
-from twisted.internet.task import _EPSILON, Cooperator
+from twisted.internet.task import Cooperator
 from twisted.web.client import ResponseFailed
 from twisted.web.http_headers import Headers
 from twisted.web.iweb import IBodyProducer, IResponse
@@ -58,11 +60,13 @@ from synapse.api.errors import (
     RequestSendFailed,
     SynapseError,
 )
+from synapse.crypto.context_factory import FederationPolicyForHTTPS
 from synapse.http import QuieterFileBodyProducer
 from synapse.http.client import (
     BlacklistingAgentWrapper,
     BodyExceededMaxSize,
     ByteWriteable,
+    _make_scheduler,
     encode_query_args,
     read_body_with_max_size,
 )
@@ -181,7 +185,7 @@ class JsonParser(ByteParser[Union[JsonDict, list]]):
 
     CONTENT_TYPE = "application/json"
 
-    def __init__(self):
+    def __init__(self) -> None:
         self._buffer = StringIO()
         self._binary_wrapper = BinaryIOWrapper(self._buffer)
 
@@ -299,7 +303,9 @@ async def _handle_response(
 class BinaryIOWrapper:
     """A wrapper for a TextIO which converts from bytes on the fly."""
 
-    def __init__(self, file: typing.TextIO, encoding="utf-8", errors="strict"):
+    def __init__(
+        self, file: typing.TextIO, encoding: str = "utf-8", errors: str = "strict"
+    ):
         self.decoder = codecs.getincrementaldecoder(encoding)(errors)
         self.file = file
 
@@ -317,7 +323,11 @@ class MatrixFederationHttpClient:
             requests.
     """
 
-    def __init__(self, hs: "HomeServer", tls_client_options_factory):
+    def __init__(
+        self,
+        hs: "HomeServer",
+        tls_client_options_factory: Optional[FederationPolicyForHTTPS],
+    ):
         self.hs = hs
         self.signing_key = hs.signing_key
         self.server_name = hs.hostname
@@ -348,10 +358,7 @@ class MatrixFederationHttpClient:
         self.version_string_bytes = hs.version_string.encode("ascii")
         self.default_timeout = 60
 
-        def schedule(x):
+        self._cooperator = Cooperator(scheduler=_make_scheduler(self.reactor))
-            self.reactor.callLater(_EPSILON, x)
-
-        self._cooperator = Cooperator(scheduler=schedule)
 
         self._sleeper = AwakenableSleeper(self.reactor)
 
@@ -364,7 +371,7 @@ class MatrixFederationHttpClient:
         self,
         request: MatrixFederationRequest,
         try_trailing_slash_on_400: bool = False,
-        **send_request_args,
+        **send_request_args: Any,
     ) -> IResponse:
         """Wrapper for _send_request which can optionally retry the request
         upon receiving a combination of a 400 HTTP response code and a
@@ -740,7 +747,7 @@ class MatrixFederationHttpClient:
         for key, sig in request["signatures"][self.server_name].items():
             auth_headers.append(
                 (
-                    'X-Matrix origin=%s,key="%s",sig="%s",destination="%s"'
+                    'X-Matrix origin="%s",key="%s",sig="%s",destination="%s"'
                     % (
                         self.server_name,
                         key,
@@ -1159,7 +1166,7 @@ class MatrixFederationHttpClient:
         self,
         destination: str,
         path: str,
-        output_stream,
+        output_stream: BinaryIO,
         args: Optional[QueryParams] = None,
         retry_on_dns_fail: bool = True,
         max_size: Optional[int] = None,
@@ -1250,10 +1257,10 @@ class MatrixFederationHttpClient:
         return length, headers
 
 
-def _flatten_response_never_received(e):
+def _flatten_response_never_received(e: BaseException) -> str:
     if hasattr(e, "reasons"):
         reasons = ", ".join(
-            _flatten_response_never_received(f.value) for f in e.reasons
+            _flatten_response_never_received(f.value) for f in e.reasons  # type: ignore[attr-defined]
         )
 
         return "%s:[%s]" % (type(e).__name__, reasons)

synapse/http/proxyagent.py

@@ -245,7 +245,7 @@ def http_proxy_endpoint(
     proxy: Optional[bytes],
     reactor: IReactorCore,
     tls_options_factory: Optional[IPolicyForHTTPS],
-    **kwargs,
+    **kwargs: object,
 ) -> Tuple[Optional[IStreamClientEndpoint], Optional[ProxyCredentials]]:
     """Parses an http proxy setting and returns an endpoint for the proxy
 

synapse/http/request_metrics.py

@@ -162,7 +162,7 @@ class RequestMetrics:
         with _in_flight_requests_lock:
             _in_flight_requests.add(self)
 
-    def stop(self, time_sec, response_code, sent_bytes):
+    def stop(self, time_sec: float, response_code: int, sent_bytes: int) -> None:
         with _in_flight_requests_lock:
             _in_flight_requests.discard(self)
 
@@ -186,13 +186,13 @@ class RequestMetrics:
             )
             return
 
-        response_code = str(response_code)
+        response_code_str = str(response_code)
 
-        outgoing_responses_counter.labels(self.method, response_code).inc()
+        outgoing_responses_counter.labels(self.method, response_code_str).inc()
 
         response_count.labels(self.method, self.name, tag).inc()
 
-        response_timer.labels(self.method, self.name, tag, response_code).observe(
+        response_timer.labels(self.method, self.name, tag, response_code_str).observe(
             time_sec - self.start_ts
         )
 
@@ -221,7 +221,7 @@ class RequestMetrics:
         # flight.
         self.update_metrics()
 
-    def update_metrics(self):
+    def update_metrics(self) -> None:
         """Updates the in flight metrics with values from this request."""
         if not self.start_context:
             logger.error(

synapse/http/server.py

@@ -33,6 +33,7 @@ from typing import (
     Optional,
     Pattern,
     Tuple,
+    TypeVar,
     Union,
 )
 
@@ -92,6 +93,68 @@ HTML_ERROR_TEMPLATE = """<!DOCTYPE html>
 HTTP_STATUS_REQUEST_CANCELLED = 499
 
 
+F = TypeVar("F", bound=Callable[..., Any])
+
+
+_cancellable_method_names = frozenset(
+    {
+        # `RestServlet`, `BaseFederationServlet` and `BaseFederationServerServlet`
+        # methods
+        "on_GET",
+        "on_PUT",
+        "on_POST",
+        "on_DELETE",
+        # `_AsyncResource`, `DirectServeHtmlResource` and `DirectServeJsonResource`
+        # methods
+        "_async_render_GET",
+        "_async_render_PUT",
+        "_async_render_POST",
+        "_async_render_DELETE",
+        "_async_render_OPTIONS",
+        # `ReplicationEndpoint` methods
+        "_handle_request",
+    }
+)
+
+
+def cancellable(method: F) -> F:
+    """Marks a servlet method as cancellable.
+
+    Methods with this decorator will be cancelled if the client disconnects before we
+    finish processing the request.
+
+    During cancellation, `Deferred.cancel()` will be invoked on the `Deferred` wrapping
+    the method. The `cancel()` call will propagate down to the `Deferred` that is
+    currently being waited on. That `Deferred` will raise a `CancelledError`, which will
+    propagate up, as per normal exception handling.
+
+    Before applying this decorator to a new endpoint, you MUST recursively check
+    that all `await`s in the function are on `async` functions or `Deferred`s that
+    handle cancellation cleanly, otherwise a variety of bugs may occur, ranging from
+    premature logging context closure, to stuck requests, to database corruption.
+
+    Usage:
+        class SomeServlet(RestServlet):
+            @cancellable
+            async def on_GET(self, request: SynapseRequest) -> ...:
+                ...
+    """
+    if method.__name__ not in _cancellable_method_names and not any(
+        method.__name__.startswith(prefix) for prefix in _cancellable_method_names
+    ):
+        raise ValueError(
+            "@cancellable decorator can only be applied to servlet methods."
+        )
+
+    method.cancellable = True  # type: ignore[attr-defined]
+    return method
+
+
+def is_method_cancellable(method: Callable[..., Any]) -> bool:
+    """Checks whether a servlet method has the `@cancellable` flag."""
+    return getattr(method, "cancellable", False)
+
+
 def return_json_error(f: failure.Failure, request: SynapseRequest) -> None:
     """Sends a JSON error response to clients."""
 
@@ -253,6 +316,9 @@ class HttpServer(Protocol):
         If the regex contains groups these gets passed to the callback via
         an unpacked tuple.
 
+        The callback may be marked with the `@cancellable` decorator, which will
+        cause request processing to be cancelled when clients disconnect early.
+
         Args:
             method: The HTTP method to listen to.
             path_patterns: The regex used to match requests.
@@ -283,7 +349,9 @@ class _AsyncResource(resource.Resource, metaclass=abc.ABCMeta):
 
     def render(self, request: SynapseRequest) -> int:
         """This gets called by twisted every time someone sends us a request."""
-        defer.ensureDeferred(self._async_render_wrapper(request))
+        request.render_deferred = defer.ensureDeferred(
+            self._async_render_wrapper(request)
+        )
         return NOT_DONE_YET
 
     @wrap_async_request_handler
@@ -319,6 +387,8 @@ class _AsyncResource(resource.Resource, metaclass=abc.ABCMeta):
 
         method_handler = getattr(self, "_async_render_%s" % (request_method,), None)
         if method_handler:
+            request.is_render_cancellable = is_method_cancellable(method_handler)
+
             raw_callback_return = method_handler(request)
 
             # Is it synchronous? We'll allow this for now.
@@ -479,6 +549,8 @@ class JsonResource(DirectServeJsonResource):
     async def _async_render(self, request: SynapseRequest) -> Tuple[int, Any]:
         callback, servlet_classname, group_dict = self._get_handler_for_request(request)
 
+        request.is_render_cancellable = is_method_cancellable(callback)
+
         # Make sure we have an appropriate name for this handler in prometheus
         # (rather than the default of JsonResource).
         request.request_metrics.name = servlet_classname

synapse/http/site.py

@@ -19,6 +19,7 @@ from typing import TYPE_CHECKING, Any, Generator, Optional, Tuple, Union
 import attr
 from zope.interface import implementer
 
+from twisted.internet.defer import Deferred
 from twisted.internet.interfaces import IAddress, IReactorTime
 from twisted.python.failure import Failure
 from twisted.web.http import HTTPChannel
@@ -91,6 +92,14 @@ class SynapseRequest(Request):
         # we can't yet create the logcontext, as we don't know the method.
         self.logcontext: Optional[LoggingContext] = None
 
+        # The `Deferred` to cancel if the client disconnects early and
+        # `is_render_cancellable` is set. Expected to be set by `Resource.render`.
+        self.render_deferred: Optional["Deferred[None]"] = None
+        # A boolean indicating whether `render_deferred` should be cancelled if the
+        # client disconnects early. Expected to be set by the coroutine started by
+        # `Resource.render`, if rendering is asynchronous.
+        self.is_render_cancellable = False
+
         global _next_request_seq
         self.request_seq = _next_request_seq
         _next_request_seq += 1
@@ -357,7 +366,21 @@ class SynapseRequest(Request):
                     {"event": "client connection lost", "reason": str(reason.value)}
                 )
 
-            if not self._is_processing:
+            if self._is_processing:
+                if self.is_render_cancellable:
+                    if self.render_deferred is not None:
+                        # Throw a cancellation into the request processing, in the hope
+                        # that it will finish up sooner than it normally would.
+                        # The `self.processing()` context manager will call
+                        # `_finished_processing()` when done.
+                        with PreserveLoggingContext():
+                            self.render_deferred.cancel()
+                    else:
+                        logger.error(
+                            "Connection from client lost, but have no Deferred to "
+                            "cancel even though the request is marked as cancellable."
+                        )
+            else:
                 self._finished_processing()
 
     def _started_processing(self, servlet_name: str) -> None:

synapse/logging/_remote.py

@@ -31,7 +31,11 @@ from twisted.internet.endpoints import (
     TCP4ClientEndpoint,
     TCP6ClientEndpoint,
 )
-from twisted.internet.interfaces import IPushProducer, IStreamClientEndpoint
+from twisted.internet.interfaces import (
+    IPushProducer,
+    IReactorTCP,
+    IStreamClientEndpoint,
+)
 from twisted.internet.protocol import Factory, Protocol
 from twisted.internet.tcp import Connection
 from twisted.python.failure import Failure
@@ -59,14 +63,14 @@ class LogProducer:
     _buffer: Deque[logging.LogRecord]
     _paused: bool = attr.ib(default=False, init=False)
 
-    def pauseProducing(self):
+    def pauseProducing(self) -> None:
         self._paused = True
 
-    def stopProducing(self):
+    def stopProducing(self) -> None:
         self._paused = True
         self._buffer = deque()
 
-    def resumeProducing(self):
+    def resumeProducing(self) -> None:
         # If we're already producing, nothing to do.
         self._paused = False
 
@@ -102,8 +106,8 @@ class RemoteHandler(logging.Handler):
         host: str,
         port: int,
         maximum_buffer: int = 1000,
-        level=logging.NOTSET,
+        level: int = logging.NOTSET,
-        _reactor=None,
+        _reactor: Optional[IReactorTCP] = None,
     ):
         super().__init__(level=level)
         self.host = host
@@ -118,7 +122,7 @@ class RemoteHandler(logging.Handler):
         if _reactor is None:
             from twisted.internet import reactor
 
-            _reactor = reactor
+            _reactor = reactor  # type: ignore[assignment]
 
         try:
             ip = ip_address(self.host)
@@ -139,7 +143,7 @@ class RemoteHandler(logging.Handler):
         self._stopping = False
         self._connect()
 
-    def close(self):
+    def close(self) -> None:
         self._stopping = True
         self._service.stopService()
 

synapse/logging/formatter.py

@@ -16,6 +16,8 @@
 import logging
 import traceback
 from io import StringIO
+from types import TracebackType
+from typing import Optional, Tuple, Type
 
 
 class LogFormatter(logging.Formatter):
@@ -28,10 +30,14 @@ class LogFormatter(logging.Formatter):
     where it was caught are logged).
     """
 
-    def __init__(self, *args, **kwargs):
+    def formatException(
-        super().__init__(*args, **kwargs)
+        self,
-
+        ei: Tuple[
-    def formatException(self, ei):
+            Optional[Type[BaseException]],
+            Optional[BaseException],
+            Optional[TracebackType],
+        ],
+    ) -> str:
         sio = StringIO()
         (typ, val, tb) = ei
 

synapse/logging/handlers.py

@@ -49,7 +49,7 @@ class PeriodicallyFlushingMemoryHandler(MemoryHandler):
         )
         self._flushing_thread.start()
 
-        def on_reactor_running():
+        def on_reactor_running() -> None:
             self._reactor_started = True
 
         reactor_to_use: IReactorCore
@@ -74,7 +74,7 @@ class PeriodicallyFlushingMemoryHandler(MemoryHandler):
         else:
             return True
 
-    def _flush_periodically(self):
+    def _flush_periodically(self) -> None:
         """
         Whilst this handler is active, flush the handler periodically.
         """

synapse/logging/scopecontextmanager.py

@@ -13,6 +13,8 @@
 # limitations under the License.import logging
 
 import logging
+from types import TracebackType
+from typing import Optional, Type
 
 from opentracing import Scope, ScopeManager
 
@@ -107,19 +109,26 @@ class _LogContextScope(Scope):
         and - if enter_logcontext was set - the logcontext is finished too.
     """
 
-    def __init__(self, manager, span, logcontext, enter_logcontext, finish_on_close):
+    def __init__(
+        self,
+        manager: LogContextScopeManager,
+        span,
+        logcontext,
+        enter_logcontext: bool,
+        finish_on_close: bool,
+    ):
         """
         Args:
-            manager (LogContextScopeManager):
+            manager:
                 the manager that is responsible for this scope.
             span (Span):
                 the opentracing span which this scope represents the local
                 lifetime for.
             logcontext (LogContext):
                 the logcontext to which this scope is attached.
-            enter_logcontext (Boolean):
+            enter_logcontext:
                 if True the logcontext will be exited when the scope is finished
-            finish_on_close (Boolean):
+            finish_on_close:
                 if True finish the span when the scope is closed
         """
         super().__init__(manager, span)
@@ -127,16 +136,21 @@ class _LogContextScope(Scope):
         self._finish_on_close = finish_on_close
         self._enter_logcontext = enter_logcontext
 
-    def __exit__(self, exc_type, value, traceback):
+    def __exit__(
+        self,
+        exc_type: Optional[Type[BaseException]],
+        value: Optional[BaseException],
+        traceback: Optional[TracebackType],
+    ) -> None:
         if exc_type == twisted.internet.defer._DefGen_Return:
             # filter out defer.returnValue() calls
             exc_type = value = traceback = None
         super().__exit__(exc_type, value, traceback)
 
-    def __str__(self):
+    def __str__(self) -> str:
         return f"Scope<{self.span}>"
 
-    def close(self):
+    def close(self) -> None:
         active_scope = self.manager.active
         if active_scope is not self:
             logger.error(

synapse/metrics/jemalloc.py

@@ -18,6 +18,7 @@ import os
 import re
 from typing import Iterable, Optional, overload
 
+import attr
 from prometheus_client import REGISTRY, Metric
 from typing_extensions import Literal
 
@@ -27,52 +28,24 @@ from synapse.metrics._types import Collector
 logger = logging.getLogger(__name__)
 
 
-def _setup_jemalloc_stats() -> None:
+@attr.s(slots=True, frozen=True, auto_attribs=True)
-    """Checks to see if jemalloc is loaded, and hooks up a collector to record
+class JemallocStats:
-    statistics exposed by jemalloc.
+    jemalloc: ctypes.CDLL
-    """
-
-    # Try to find the loaded jemalloc shared library, if any. We need to
-    # introspect into what is loaded, rather than loading whatever is on the
-    # path, as if we load a *different* jemalloc version things will seg fault.
-
-    # We look in `/proc/self/maps`, which only exists on linux.
-    if not os.path.exists("/proc/self/maps"):
-        logger.debug("Not looking for jemalloc as no /proc/self/maps exist")
-        return
-
-    # We're looking for a path at the end of the line that includes
-    # "libjemalloc".
-    regex = re.compile(r"/\S+/libjemalloc.*$")
-
-    jemalloc_path = None
-    with open("/proc/self/maps") as f:
-        for line in f:
-            match = regex.search(line.strip())
-            if match:
-                jemalloc_path = match.group()
-
-    if not jemalloc_path:
-        # No loaded jemalloc was found.
-        logger.debug("jemalloc not found")
-        return
-
-    logger.debug("Found jemalloc at %s", jemalloc_path)
-
-    jemalloc = ctypes.CDLL(jemalloc_path)
 
     @overload
     def _mallctl(
-        name: str, read: Literal[True] = True, write: Optional[int] = None
+        self, name: str, read: Literal[True] = True, write: Optional[int] = None
     ) -> int:
         ...
 
     @overload
-    def _mallctl(name: str, read: Literal[False], write: Optional[int] = None) -> None:
+    def _mallctl(
+        self, name: str, read: Literal[False], write: Optional[int] = None
+    ) -> None:
         ...
 
     def _mallctl(
-        name: str, read: bool = True, write: Optional[int] = None
+        self, name: str, read: bool = True, write: Optional[int] = None
     ) -> Optional[int]:
         """Wrapper around `mallctl` for reading and writing integers to
         jemalloc.
@@ -120,7 +93,7 @@ def _setup_jemalloc_stats() -> None:
         # Where oldp/oldlenp is a buffer where the old value will be written to
         # (if not null), and newp/newlen is the buffer with the new value to set
         # (if not null). Note that they're all references *except* newlen.
-        result = jemalloc.mallctl(
+        result = self.jemalloc.mallctl(
             name.encode("ascii"),
             input_var_ref,
             input_len_ref,
@@ -136,21 +109,80 @@ def _setup_jemalloc_stats() -> None:
 
         return input_var.value
 
-    def _jemalloc_refresh_stats() -> None:
+    def refresh_stats(self) -> None:
         """Request that jemalloc updates its internal statistics. This needs to
         be called before querying for stats, otherwise it will return stale
         values.
         """
         try:
-            _mallctl("epoch", read=False, write=1)
+            self._mallctl("epoch", read=False, write=1)
         except Exception as e:
             logger.warning("Failed to reload jemalloc stats: %s", e)
 
+    def get_stat(self, name: str) -> int:
+        """Request the stat of the given name at the time of the last
+        `refresh_stats` call. This may throw if we fail to read
+        the stat.
+        """
+        return self._mallctl(f"stats.{name}")
+
+
+_JEMALLOC_STATS: Optional[JemallocStats] = None
+
+
+def get_jemalloc_stats() -> Optional[JemallocStats]:
+    """Returns an interface to jemalloc, if it is being used.
+
+    Note that this will always return None until `setup_jemalloc_stats` has been
+    called.
+    """
+    return _JEMALLOC_STATS
+
+
+def _setup_jemalloc_stats() -> None:
+    """Checks to see if jemalloc is loaded, and hooks up a collector to record
+    statistics exposed by jemalloc.
+    """
+
+    global _JEMALLOC_STATS
+
+    # Try to find the loaded jemalloc shared library, if any. We need to
+    # introspect into what is loaded, rather than loading whatever is on the
+    # path, as if we load a *different* jemalloc version things will seg fault.
+
+    # We look in `/proc/self/maps`, which only exists on linux.
+    if not os.path.exists("/proc/self/maps"):
+        logger.debug("Not looking for jemalloc as no /proc/self/maps exist")
+        return
+
+    # We're looking for a path at the end of the line that includes
+    # "libjemalloc".
+    regex = re.compile(r"/\S+/libjemalloc.*$")
+
+    jemalloc_path = None
+    with open("/proc/self/maps") as f:
+        for line in f:
+            match = regex.search(line.strip())
+            if match:
+                jemalloc_path = match.group()
+
+    if not jemalloc_path:
+        # No loaded jemalloc was found.
+        logger.debug("jemalloc not found")
+        return
+
+    logger.debug("Found jemalloc at %s", jemalloc_path)
+
+    jemalloc_dll = ctypes.CDLL(jemalloc_path)
+
+    stats = JemallocStats(jemalloc_dll)
+    _JEMALLOC_STATS = stats
+
     class JemallocCollector(Collector):
         """Metrics for internal jemalloc stats."""
 
         def collect(self) -> Iterable[Metric]:
-            _jemalloc_refresh_stats()
+            stats.refresh_stats()
 
             g = GaugeMetricFamily(
                 "jemalloc_stats_app_memory_bytes",
@@ -184,7 +216,7 @@ def _setup_jemalloc_stats() -> None:
                 "metadata",
             ):
                 try:
-                    value = _mallctl(f"stats.{t}")
+                    value = stats.get_stat(t)
                 except Exception as e:
                     # There was an error fetching the value, skip.
                     logger.warning("Failed to read jemalloc stats.%s: %s", t, e)

synapse/module_api/__init__.py

@@ -35,6 +35,7 @@ from typing_extensions import ParamSpec
 from twisted.internet import defer
 from twisted.web.resource import Resource
 
+from synapse import spam_checker_api
 from synapse.api.errors import SynapseError
 from synapse.events import EventBase
 from synapse.events.presence_router import (
@@ -47,6 +48,7 @@ from synapse.events.spamcheck import (
     CHECK_MEDIA_FILE_FOR_SPAM_CALLBACK,
     CHECK_REGISTRATION_FOR_SPAM_CALLBACK,
     CHECK_USERNAME_FOR_SPAM_CALLBACK,
+    SHOULD_DROP_FEDERATED_EVENT_CALLBACK,
     USER_MAY_CREATE_ROOM_ALIAS_CALLBACK,
     USER_MAY_CREATE_ROOM_CALLBACK,
     USER_MAY_INVITE_CALLBACK,
@@ -139,6 +141,9 @@ are loaded into Synapse.
 
 PRESENCE_ALL_USERS = PresenceRouter.ALL_USERS
 
+ALLOW = spam_checker_api.Allow.ALLOW
+# Singleton value used to mark a message as permitted.
+
 __all__ = [
     "errors",
     "make_deferred_yieldable",
@@ -146,6 +151,7 @@ __all__ = [
     "respond_with_html",
     "run_in_background",
     "cached",
+    "Allow",
     "UserID",
     "DatabasePool",
     "LoggingTransaction",
@@ -234,6 +240,9 @@ class ModuleApi:
         self,
         *,
         check_event_for_spam: Optional[CHECK_EVENT_FOR_SPAM_CALLBACK] = None,
+        should_drop_federated_event: Optional[
+            SHOULD_DROP_FEDERATED_EVENT_CALLBACK
+        ] = None,
         user_may_join_room: Optional[USER_MAY_JOIN_ROOM_CALLBACK] = None,
         user_may_invite: Optional[USER_MAY_INVITE_CALLBACK] = None,
         user_may_send_3pid_invite: Optional[USER_MAY_SEND_3PID_INVITE_CALLBACK] = None,
@@ -254,6 +263,7 @@ class ModuleApi:
         """
         return self._spam_checker.register_callbacks(
             check_event_for_spam=check_event_for_spam,
+            should_drop_federated_event=should_drop_federated_event,
             user_may_join_room=user_may_join_room,
             user_may_invite=user_may_invite,
             user_may_send_3pid_invite=user_may_send_3pid_invite,

synapse/module_api/errors.py

@@ -15,6 +15,7 @@
 """Exception types which are exposed as part of the stable module API"""
 
 from synapse.api.errors import (
+    Codes,
     InvalidClientCredentialsError,
     RedirectException,
     SynapseError,
@@ -24,6 +25,7 @@ from synapse.handlers.push_rules import InvalidRuleException
 from synapse.storage.push_rule import RuleNotFoundException
 
 __all__ = [
+    "Codes",
     "InvalidClientCredentialsError",
     "RedirectException",
     "SynapseError",

synapse/notifier.py

@@ -46,6 +46,7 @@ from synapse.types import (
     JsonDict,
     PersistedEventPosition,
     RoomStreamToken,
+    StreamKeyType,
     StreamToken,
     UserID,
 )
@@ -370,7 +371,7 @@ class Notifier:
 
         if users or rooms:
             self.on_new_event(
-                "room_key",
+                StreamKeyType.ROOM,
                 max_room_stream_token,
                 users=users,
                 rooms=rooms,
@@ -440,7 +441,7 @@ class Notifier:
             for room in rooms:
                 user_streams |= self.room_to_user_streams.get(room, set())
 
-            if stream_key == "to_device_key":
+            if stream_key == StreamKeyType.TO_DEVICE:
                 issue9533_logger.debug(
                     "to-device messages stream id %s, awaking streams for %s",
                     new_token,

synapse/push/__init__.py

@@ -12,6 +12,80 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 
+"""
+This module implements the push rules & notifications portion of the Matrix
+specification.
+
+There's a few related features:
+
+* Push notifications (i.e. email or outgoing requests to a Push Gateway).
+* Calculation of unread notifications (for /sync and /notifications).
+
+When Synapse receives a new event (locally, via the Client-Server API, or via
+federation), the following occurs:
+
+1. The push rules get evaluated to generate a set of per-user actions.
+2. The event is persisted into the database.
+3. (In the background) The notifier is notified about the new event.
+
+The per-user actions are initially stored in the event_push_actions_staging table,
+before getting moved into the event_push_actions table when the event is persisted.
+The event_push_actions table is periodically summarised into the event_push_summary
+and event_push_summary_stream_ordering tables.
+
+Since push actions block an event from being persisted the generation of push
+actions is performance sensitive.
+
+The general interaction of the classes are:
+
+        +---------------------------------------------+
+        | FederationEventHandler/EventCreationHandler |
+        +---------------------------------------------+
+                |
+                v
+        +-----------------------+     +---------------------------+
+        | BulkPushRuleEvaluator |---->| PushRuleEvaluatorForEvent |
+        +-----------------------+     +---------------------------+
+                |
+                v
+        +-----------------------------+
+        | EventPushActionsWorkerStore |
+        +-----------------------------+
+
+The notifier notifies the pusher pool of the new event, which checks for affected
+users. Each user-configured pusher of the affected users then performs the
+previously calculated action.
+
+The general interaction of the classes are:
+
+        +----------+
+        | Notifier |
+        +----------+
+                |
+                v
+        +------------+     +--------------+
+        | PusherPool |---->| PusherConfig |
+        +------------+     +--------------+
+                |
+                |     +---------------+
+                +<--->| PusherFactory |
+                |     +---------------+
+                v
+        +------------------------+     +-----------------------------------------------+
+        | EmailPusher/HttpPusher |---->| EventPushActionsWorkerStore/PusherWorkerStore |
+        +------------------------+     +-----------------------------------------------+
+                |
+                v
+        +-------------------------+
+        | Mailer/SimpleHttpClient |
+        +-------------------------+
+
+The Pusher instance also calls out to various utilities for generating payloads
+(or email templates), but those interactions are not detailed in this diagram
+(and are specific to the type of pusher).
+
+"""
+
 import abc
 from typing import TYPE_CHECKING, Any, Dict, Optional
 

synapse/push/action_generator.py

@@ -1,44 +0,0 @@
-# Copyright 2015 OpenMarket 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 typing import TYPE_CHECKING
-
-from synapse.events import EventBase
-from synapse.events.snapshot import EventContext
-from synapse.push.bulk_push_rule_evaluator import BulkPushRuleEvaluator
-from synapse.util.metrics import Measure
-
-if TYPE_CHECKING:
-    from synapse.server import HomeServer
-
-logger = logging.getLogger(__name__)
-
-
-class ActionGenerator:
-    def __init__(self, hs: "HomeServer"):
-        self.clock = hs.get_clock()
-        self.bulk_evaluator = BulkPushRuleEvaluator(hs)
-        # really we want to get all user ids and all profile tags too,
-        # since we want the actions for each profile tag for every user and
-        # also actions for a client with no profile tag for each user.
-        # Currently the event stream doesn't support profile tags on an
-        # event stream, so we just run the rules for a client with no profile
-        # tag (ie. we just need all the users).
-
-    async def handle_push_actions_for_event(
-        self, event: EventBase, context: EventContext
-    ) -> None:
-        with Measure(self.clock, "action_for_event_by_user"):
-            await self.bulk_evaluator.action_for_event_by_user(event, context)

synapse/push/bulk_push_rule_evaluator.py

@@ -20,8 +20,8 @@ import attr
 from prometheus_client import Counter
 
 from synapse.api.constants import EventTypes, Membership, RelationTypes
-from synapse.event_auth import get_user_power_level
+from synapse.event_auth import auth_types_for_event, get_user_power_level
-from synapse.events import EventBase
+from synapse.events import EventBase, relation_from_event
 from synapse.events.snapshot import EventContext
 from synapse.state import POWER_KEY
 from synapse.storage.databases.main.roommember import EventIdMembership
@@ -29,7 +29,9 @@ from synapse.util.async_helpers import Linearizer
 from synapse.util.caches import CacheMetric, register_cache
 from synapse.util.caches.descriptors import lru_cache
 from synapse.util.caches.lrucache import LruCache
+from synapse.util.metrics import measure_func
 
+from ..storage.state import StateFilter
 from .push_rule_evaluator import PushRuleEvaluatorForEvent
 
 if TYPE_CHECKING:
@@ -77,8 +79,8 @@ def _should_count_as_unread(event: EventBase, context: EventContext) -> bool:
         return False
 
     # Exclude edits.
-    relates_to = event.content.get("m.relates_to", {})
+    relates_to = relation_from_event(event)
-    if relates_to.get("rel_type") == RelationTypes.REPLACE:
+    if relates_to and relates_to.rel_type == RelationTypes.REPLACE:
         return False
 
     # Mark events that have a non-empty string body as unread.
@@ -105,6 +107,7 @@ class BulkPushRuleEvaluator:
     def __init__(self, hs: "HomeServer"):
         self.hs = hs
         self.store = hs.get_datastores().main
+        self.clock = hs.get_clock()
         self._event_auth_handler = hs.get_event_auth_handler()
 
         # Used by `RulesForRoom` to ensure only one thing mutates the cache at a
@@ -166,8 +169,12 @@ class BulkPushRuleEvaluator:
     async def _get_power_levels_and_sender_level(
         self, event: EventBase, context: EventContext
     ) -> Tuple[dict, int]:
-        prev_state_ids = await context.get_prev_state_ids()
+        event_types = auth_types_for_event(event.room_version, event)
+        prev_state_ids = await context.get_prev_state_ids(
+            StateFilter.from_types(event_types)
+        )
         pl_event_id = prev_state_ids.get(POWER_KEY)
+
         if pl_event_id:
             # fastpath: if there's a power level event, that's all we need, and
             # not having a power level event is an extreme edge case
@@ -185,6 +192,7 @@ class BulkPushRuleEvaluator:
 
         return pl_event.content if pl_event else {}, sender_level
 
+    @measure_func("action_for_event_by_user")
     async def action_for_event_by_user(
         self, event: EventBase, context: EventContext
     ) -> None:
@@ -192,6 +200,10 @@ class BulkPushRuleEvaluator:
         should increment the unread count, and insert the results into the
         event_push_actions_staging table.
         """
+        if event.internal_metadata.is_outlier():
+            # This can happen due to out of band memberships
+            return
+
         count_as_unread = _should_count_as_unread(event, context)
 
         rules_by_user = await self._get_rules_for_event(event, context)
@@ -208,8 +220,6 @@ class BulkPushRuleEvaluator:
             event, len(room_members), sender_power_level, power_levels
         )
 
-        condition_cache: Dict[str, bool] = {}
-
         # If the event is not a state event check if any users ignore the sender.
         if not event.is_state():
             ignorers = await self.store.ignored_by(event.sender)
@@ -247,8 +257,8 @@ class BulkPushRuleEvaluator:
                 if "enabled" in rule and not rule["enabled"]:
                     continue
 
-                matches = _condition_checker(
+                matches = evaluator.check_conditions(
-                    evaluator, rule["conditions"], uid, display_name, condition_cache
+                    rule["conditions"], uid, display_name
                 )
                 if matches:
                     actions = [x for x in rule["actions"] if x != "dont_notify"]
@@ -267,32 +277,6 @@ class BulkPushRuleEvaluator:
         )
 
 
-def _condition_checker(
-    evaluator: PushRuleEvaluatorForEvent,
-    conditions: List[dict],
-    uid: str,
-    display_name: Optional[str],
-    cache: Dict[str, bool],
-) -> bool:
-    for cond in conditions:
-        _cache_key = cond.get("_cache_key", None)
-        if _cache_key:
-            res = cache.get(_cache_key, None)
-            if res is False:
-                return False
-            elif res is True:
-                continue
-
-        res = evaluator.matches(cond, uid, display_name)
-        if _cache_key:
-            cache[_cache_key] = bool(res)
-
-        if not res:
-            return False
-
-    return True
-
-
 MemberMap = Dict[str, Optional[EventIdMembership]]
 Rule = Dict[str, dict]
 RulesByUser = Dict[str, List[Rule]]

synapse/push/httppusher.py

@@ -398,7 +398,7 @@ class HttpPusher(Pusher):
         rejected = []
         if "rejected" in resp:
             rejected = resp["rejected"]
-        else:
+        if not rejected:
             self.badge_count_last_call = badge
         return rejected
 

synapse/push/push_rule_evaluator.py

@@ -129,9 +129,55 @@ class PushRuleEvaluatorForEvent:
         # Maps strings of e.g. 'content.body' -> event["content"]["body"]
         self._value_cache = _flatten_dict(event)
 
+        # Maps cache keys to final values.
+        self._condition_cache: Dict[str, bool] = {}
+
+    def check_conditions(
+        self, conditions: List[dict], uid: str, display_name: Optional[str]
+    ) -> bool:
+        """
+        Returns true if a user's conditions/user ID/display name match the event.
+
+        Args:
+            conditions: The user's conditions to match.
+            uid: The user's MXID.
+            display_name: The display name.
+
+        Returns:
+             True if all conditions match the event, False otherwise.
+        """
+        for cond in conditions:
+            _cache_key = cond.get("_cache_key", None)
+            if _cache_key:
+                res = self._condition_cache.get(_cache_key, None)
+                if res is False:
+                    return False
+                elif res is True:
+                    continue
+
+            res = self.matches(cond, uid, display_name)
+            if _cache_key:
+                self._condition_cache[_cache_key] = bool(res)
+
+            if not res:
+                return False
+
+        return True
+
     def matches(
         self, condition: Dict[str, Any], user_id: str, display_name: Optional[str]
     ) -> bool:
+        """
+        Returns true if a user's condition/user ID/display name match the event.
+
+        Args:
+            condition: The user's condition to match.
+            uid: The user's MXID.
+            display_name: The display name, or None if there is not one.
+
+        Returns:
+             True if the condition matches the event, False otherwise.
+        """
         if condition["kind"] == "event_match":
             return self._event_match(condition, user_id)
         elif condition["kind"] == "contains_display_name":
@@ -146,6 +192,16 @@ class PushRuleEvaluatorForEvent:
             return True
 
     def _event_match(self, condition: dict, user_id: str) -> bool:
+        """
+        Check an "event_match" push rule condition.
+
+        Args:
+            condition: The "event_match" push rule condition to match.
+            user_id: The user's MXID.
+
+        Returns:
+             True if the condition matches the event, False otherwise.
+        """
         pattern = condition.get("pattern", None)
 
         if not pattern:
@@ -167,13 +223,22 @@ class PushRuleEvaluatorForEvent:
 
             return _glob_matches(pattern, body, word_boundary=True)
         else:
-            haystack = self._get_value(condition["key"])
+            haystack = self._value_cache.get(condition["key"], None)
             if haystack is None:
                 return False
 
             return _glob_matches(pattern, haystack)
 
     def _contains_display_name(self, display_name: Optional[str]) -> bool:
+        """
+        Check an "event_match" push rule condition.
+
+        Args:
+            display_name: The display name, or None if there is not one.
+
+        Returns:
+             True if the display name is found in the event body, False otherwise.
+        """
         if not display_name:
             return False
 
@@ -191,9 +256,6 @@ class PushRuleEvaluatorForEvent:
 
         return bool(r.search(body))
 
-    def _get_value(self, dotted_key: str) -> Optional[str]:
-        return self._value_cache.get(dotted_key, None)
-
 
 # Caches (string, is_glob, word_boundary) -> regex for push. See _glob_matches
 regex_cache: LruCache[Tuple[str, bool, bool], Pattern] = LruCache(

synapse/replication/http/_base.py

@@ -26,7 +26,8 @@ from twisted.web.server import Request
 
 from synapse.api.errors import HttpResponseException, SynapseError
 from synapse.http import RequestTimedOutError
-from synapse.http.server import HttpServer
+from synapse.http.server import HttpServer, is_method_cancellable
+from synapse.http.site import SynapseRequest
 from synapse.logging import opentracing
 from synapse.logging.opentracing import trace
 from synapse.types import JsonDict
@@ -310,6 +311,12 @@ class ReplicationEndpoint(metaclass=abc.ABCMeta):
         url_args = list(self.PATH_ARGS)
         method = self.METHOD
 
+        if self.CACHE and is_method_cancellable(self._handle_request):
+            raise Exception(
+                f"{self.__class__.__name__} has been marked as cancellable, but CACHE "
+                "is set. The cancellable flag would have no effect."
+            )
+
         if self.CACHE:
             url_args.append("txn_id")
 
@@ -324,7 +331,7 @@ class ReplicationEndpoint(metaclass=abc.ABCMeta):
         )
 
     async def _check_auth_and_handle(
-        self, request: Request, **kwargs: Any
+        self, request: SynapseRequest, **kwargs: Any
     ) -> Tuple[int, JsonDict]:
         """Called on new incoming requests when caching is enabled. Checks
         if there is a cached response for the request and returns that,
@@ -340,8 +347,18 @@ class ReplicationEndpoint(metaclass=abc.ABCMeta):
         if self.CACHE:
             txn_id = kwargs.pop("txn_id")
 
+            # We ignore the `@cancellable` flag, since cancellation wouldn't interupt
+            # `_handle_request` and `ResponseCache` does not handle cancellation
+            # correctly yet. In particular, there may be issues to do with logging
+            # context lifetimes.
+
             return await self.response_cache.wrap(
                 txn_id, self._handle_request, request, **kwargs
             )
 
+        # The `@cancellable` decorator may be applied to `_handle_request`. But we
+        # told `HttpServer.register_paths` that our handler is `_check_auth_and_handle`,
+        # so we have to set up the cancellable flag ourselves.
+        request.is_render_cancellable = is_method_cancellable(self._handle_request)
+
         return await self._handle_request(request, **kwargs)

synapse/replication/tcp/client.py

@@ -43,7 +43,7 @@ from synapse.replication.tcp.streams.events import (
     EventsStreamEventRow,
     EventsStreamRow,
 )
-from synapse.types import PersistedEventPosition, ReadReceipt, UserID
+from synapse.types import PersistedEventPosition, ReadReceipt, StreamKeyType, UserID
 from synapse.util.async_helpers import Linearizer, timeout_deferred
 from synapse.util.metrics import Measure
 
@@ -153,19 +153,19 @@ class ReplicationDataHandler:
         if stream_name == TypingStream.NAME:
             self._typing_handler.process_replication_rows(token, rows)
             self.notifier.on_new_event(
-                "typing_key", token, rooms=[row.room_id for row in rows]
+                StreamKeyType.TYPING, token, rooms=[row.room_id for row in rows]
             )
         elif stream_name == PushRulesStream.NAME:
             self.notifier.on_new_event(
-                "push_rules_key", token, users=[row.user_id for row in rows]
+                StreamKeyType.PUSH_RULES, token, users=[row.user_id for row in rows]
             )
         elif stream_name in (AccountDataStream.NAME, TagAccountDataStream.NAME):
             self.notifier.on_new_event(
-                "account_data_key", token, users=[row.user_id for row in rows]
+                StreamKeyType.ACCOUNT_DATA, token, users=[row.user_id for row in rows]
             )
         elif stream_name == ReceiptsStream.NAME:
             self.notifier.on_new_event(
-                "receipt_key", token, rooms=[row.room_id for row in rows]
+                StreamKeyType.RECEIPT, token, rooms=[row.room_id for row in rows]
             )
             await self._pusher_pool.on_new_receipts(
                 token, token, {row.room_id for row in rows}
@@ -173,14 +173,18 @@ class ReplicationDataHandler:
         elif stream_name == ToDeviceStream.NAME:
             entities = [row.entity for row in rows if row.entity.startswith("@")]
             if entities:
-                self.notifier.on_new_event("to_device_key", token, users=entities)
+                self.notifier.on_new_event(
+                    StreamKeyType.TO_DEVICE, token, users=entities
+                )
         elif stream_name == DeviceListsStream.NAME:
             all_room_ids: Set[str] = set()
             for row in rows:
                 if row.entity.startswith("@"):
                     room_ids = await self.store.get_rooms_for_user(row.entity)
                     all_room_ids.update(room_ids)
-            self.notifier.on_new_event("device_list_key", token, rooms=all_room_ids)
+            self.notifier.on_new_event(
+                StreamKeyType.DEVICE_LIST, token, rooms=all_room_ids
+            )
         elif stream_name == GroupServerStream.NAME:
             self.notifier.on_new_event(
                 "groups_key", token, users=[row.user_id for row in rows]