Update cahngelog

Fixes https://github.com/matrix-org/synapse/issues/9572

When a SSO user logs in for the first time, we create a local Matrix user for them. This goes through the register_user flow, which ends up triggering the spam checker. Spam checker modules don't currently have any way to differentiate between a user trying to sign up initially, versus an SSO user (whom has presumably already been approved elsewhere) trying to log in for the first time.

This PR passes `auth_provider_id` as an argument to the `check_registration_for_spam` function. This argument will contain an ID of an SSO provider (`"saml"`, `"cas"`, etc.) if one was used, else `None`.

.git-blame-ignore-revs

@@ -0,0 +1,8 @@
+# Black reformatting (#5482).
+32e7c9e7f20b57dd081023ac42d6931a8da9b3a3
+
+# Target Python 3.5 with black (#8664).
+aff1eb7c671b0a3813407321d2702ec46c71fa56
+
+# Update black to 20.8b1 (#9381).
+0a00b7ff14890987f09112a2ae696c61001e6cf1

.gitignore

@@ -6,13 +6,14 @@
 *.egg
 *.egg-info
 *.lock
-*.pyc
+*.py[cod]
 *.snap
 *.tac
 _trial_temp/
 _trial_temp*/
 /out
 .DS_Store
+__pycache__/
 
 # stuff that is likely to exist when you run a server locally
 /*.db

CHANGES.md

@@ -1,10 +1,197 @@
-Synapse 1.28.0rc1 (2021-02-19)
+Synapse 1.30.1 (2021-03-26)
+===========================
+
+This release is identical to Synapse 1.30.0, with the exception of explicitly
+setting a minimum version of Python's Cryptography library to ensure that users
+of Synapse are protected from the recent [OpenSSL security advisories](https://mta.openssl.org/pipermail/openssl-announce/2021-March/000198.html),
+especially CVE-2021-3449.
+
+Note that Cryptography defaults to bundling its own statically linked copy of
+OpenSSL, which means that you may not be protected by your operating system's
+security updates.
+
+It's also worth noting that Cryptography no longer supports Python 3.5, so
+admins deploying to older environments may not be protected against this or
+future vulnerabilities. Synapse will be dropping support for Python 3.5 at the
+end of March.
+
+
+Updates to the Docker image
+---------------------------
+
+- Ensure that the docker container has up to date versions of openssl. ([\#9697](https://github.com/matrix-org/synapse/issues/9697))
+
+
+Internal Changes
+----------------
+
+- Enforce that `cryptography` dependency is up to date to ensure it has the most recent openssl patches. ([\#9697](https://github.com/matrix-org/synapse/issues/9697))
+
+
+Synapse 1.30.0 (2021-03-22)
+===========================
+
+Note that this release deprecates the ability for appservices to
+call `POST /_matrix/client/r0/register`  without the body parameter `type`. Appservice
+developers should use a `type` value of `m.login.application_service` as
+per [the spec](https://matrix.org/docs/spec/application_service/r0.1.2#server-admin-style-permissions).
+In future releases, calling this endpoint with an access token - but without a `m.login.application_service`
+type - will fail.
+
+
+No significant changes.
+
+
+Synapse 1.30.0rc1 (2021-03-16)
 ==============================
 
+Features
+--------
+
+- Add prometheus metrics for number of users successfully registering and logging in. ([\#9510](https://github.com/matrix-org/synapse/issues/9510), [\#9511](https://github.com/matrix-org/synapse/issues/9511), [\#9573](https://github.com/matrix-org/synapse/issues/9573))
+- Add `synapse_federation_last_sent_pdu_time` and `synapse_federation_last_received_pdu_time` prometheus metrics, which monitor federation delays by reporting the timestamps of messages sent and received to a set of remote servers. ([\#9540](https://github.com/matrix-org/synapse/issues/9540))
+- Add support for generating JSON Web Tokens dynamically for use as OIDC client secrets. ([\#9549](https://github.com/matrix-org/synapse/issues/9549))
+- Optimise handling of incomplete room history for incoming federation. ([\#9601](https://github.com/matrix-org/synapse/issues/9601))
+- Finalise support for allowing clients to pick an SSO Identity Provider ([MSC2858](https://github.com/matrix-org/matrix-doc/pull/2858)). ([\#9617](https://github.com/matrix-org/synapse/issues/9617))
+- Tell spam checker modules about the SSO IdP a user registered through if one was used. ([\#9626](https://github.com/matrix-org/synapse/issues/9626))
+
+
+Bugfixes
+--------
+
+- Fix long-standing bug when generating thumbnails for some images with transparency: `TypeError: cannot unpack non-iterable int object`. ([\#9473](https://github.com/matrix-org/synapse/issues/9473))
+- Purge chain cover indexes for events that were purged prior to Synapse v1.29.0. ([\#9542](https://github.com/matrix-org/synapse/issues/9542), [\#9583](https://github.com/matrix-org/synapse/issues/9583))
+- Fix bug where federation requests were not correctly retried on 5xx responses. ([\#9567](https://github.com/matrix-org/synapse/issues/9567))
+- Fix re-activating an account via the admin API when local passwords are disabled. ([\#9587](https://github.com/matrix-org/synapse/issues/9587))
+- Fix a bug introduced in Synapse 1.20 which caused incoming federation transactions to stack up, causing slow recovery from outages. ([\#9597](https://github.com/matrix-org/synapse/issues/9597))
+- Fix a bug introduced in v1.28.0 where the OpenID Connect callback endpoint could error with a `MacaroonInitException`. ([\#9620](https://github.com/matrix-org/synapse/issues/9620))
+- Fix Internal Server Error on `GET /_synapse/client/saml2/authn_response` request. ([\#9623](https://github.com/matrix-org/synapse/issues/9623))
+
+
+Updates to the Docker image
+---------------------------
+
+- Make use of an improved malloc implementation (`jemalloc`) in the docker image. ([\#8553](https://github.com/matrix-org/synapse/issues/8553))
+
+
+Improved Documentation
+----------------------
+
+- Add relayd entry to reverse proxy example configurations. ([\#9508](https://github.com/matrix-org/synapse/issues/9508))
+- Improve the SAML2 upgrade notes for 1.27.0. ([\#9550](https://github.com/matrix-org/synapse/issues/9550))
+- Link to the "List user's media" admin API from the media admin API docs. ([\#9571](https://github.com/matrix-org/synapse/issues/9571))
+- Clarify the spam checker modules documentation example to mention that `parse_config` is a required method. ([\#9580](https://github.com/matrix-org/synapse/issues/9580))
+- Clarify the sample configuration for `stats` settings. ([\#9604](https://github.com/matrix-org/synapse/issues/9604))
+
+
+Deprecations and Removals
+-------------------------
+
+- The `synapse_federation_last_sent_pdu_age` and `synapse_federation_last_received_pdu_age` prometheus metrics have been removed. They are replaced by `synapse_federation_last_sent_pdu_time` and `synapse_federation_last_received_pdu_time`. ([\#9540](https://github.com/matrix-org/synapse/issues/9540))
+- Registering an Application Service user without using the `m.login.application_service` login type will be unsupported in an upcoming Synapse release. ([\#9559](https://github.com/matrix-org/synapse/issues/9559))
+
+
+Internal Changes
+----------------
+
+- Add tests to ResponseCache. ([\#9458](https://github.com/matrix-org/synapse/issues/9458))
+- Add type hints to purge room and server notice admin API. ([\#9520](https://github.com/matrix-org/synapse/issues/9520))
+- Add extra logging to ObservableDeferred when callbacks throw exceptions. ([\#9523](https://github.com/matrix-org/synapse/issues/9523))
+- Fix incorrect type hints. ([\#9528](https://github.com/matrix-org/synapse/issues/9528), [\#9543](https://github.com/matrix-org/synapse/issues/9543), [\#9591](https://github.com/matrix-org/synapse/issues/9591), [\#9608](https://github.com/matrix-org/synapse/issues/9608), [\#9618](https://github.com/matrix-org/synapse/issues/9618))
+- Add an additional test for purging a room. ([\#9541](https://github.com/matrix-org/synapse/issues/9541))
+- Add a `.git-blame-ignore-revs` file with the hashes of auto-formatting. ([\#9560](https://github.com/matrix-org/synapse/issues/9560))
+- Increase the threshold before which outbound federation to a server goes into "catch up" mode, which is expensive for the remote server to handle. ([\#9561](https://github.com/matrix-org/synapse/issues/9561))
+- Fix spurious errors reported by the `config-lint.sh` script. ([\#9562](https://github.com/matrix-org/synapse/issues/9562))
+- Fix type hints and tests for BlacklistingAgentWrapper and BlacklistingReactorWrapper. ([\#9563](https://github.com/matrix-org/synapse/issues/9563))
+- Do not have mypy ignore type hints from unpaddedbase64. ([\#9568](https://github.com/matrix-org/synapse/issues/9568))
+- Improve efficiency of calculating the auth chain in large rooms. ([\#9576](https://github.com/matrix-org/synapse/issues/9576))
+- Convert `synapse.types.Requester` to an `attrs` class. ([\#9586](https://github.com/matrix-org/synapse/issues/9586))
+- Add logging for redis connection setup. ([\#9590](https://github.com/matrix-org/synapse/issues/9590))
+- Improve logging when processing incoming transactions. ([\#9596](https://github.com/matrix-org/synapse/issues/9596))
+- Remove unused `stats.retention` setting, and emit a warning if stats are disabled. ([\#9604](https://github.com/matrix-org/synapse/issues/9604))
+- Prevent attempting to bundle aggregations for state events in /context APIs. ([\#9619](https://github.com/matrix-org/synapse/issues/9619))
+
+
+Synapse 1.29.0 (2021-03-08)
+===========================
+
+Note that synapse now expects an `X-Forwarded-Proto` header when used with a reverse proxy. Please see [UPGRADE.rst](UPGRADE.rst#upgrading-to-v1290) for more details on this change.
+
+
+No significant changes.
+
+
+Synapse 1.29.0rc1 (2021-03-04)
+==============================
+
+Features
+--------
+
+- Add rate limiters to cross-user key sharing requests. ([\#8957](https://github.com/matrix-org/synapse/issues/8957))
+- Add `order_by` to the admin API `GET /_synapse/admin/v1/users/<user_id>/media`. Contributed by @dklimpel. ([\#8978](https://github.com/matrix-org/synapse/issues/8978))
+- Add some configuration settings to make users' profile data more private. ([\#9203](https://github.com/matrix-org/synapse/issues/9203))
+- The `no_proxy` and `NO_PROXY` environment variables are now respected in proxied HTTP clients with the lowercase form taking precedence if both are present. Additionally, the lowercase `https_proxy` environment variable is now respected in proxied HTTP clients on top of existing support for the uppercase `HTTPS_PROXY` form and takes precedence if both are present. Contributed by Timothy Leung. ([\#9372](https://github.com/matrix-org/synapse/issues/9372))
+- Add a configuration option, `user_directory.prefer_local_users`, which when enabled will make it more likely for users on the same server as you to appear above other users. ([\#9383](https://github.com/matrix-org/synapse/issues/9383), [\#9385](https://github.com/matrix-org/synapse/issues/9385))
+- Add support for regenerating thumbnails if they have been deleted but the original image is still stored. ([\#9438](https://github.com/matrix-org/synapse/issues/9438))
+- Add support for `X-Forwarded-Proto` header when using a reverse proxy. ([\#9472](https://github.com/matrix-org/synapse/issues/9472), [\#9501](https://github.com/matrix-org/synapse/issues/9501), [\#9512](https://github.com/matrix-org/synapse/issues/9512), [\#9539](https://github.com/matrix-org/synapse/issues/9539))
+
+
+Bugfixes
+--------
+
+- Fix a bug where users' pushers were not all deleted when they deactivated their account. ([\#9285](https://github.com/matrix-org/synapse/issues/9285), [\#9516](https://github.com/matrix-org/synapse/issues/9516))
+- Fix a bug where a lot of unnecessary presence updates were sent when joining a room. ([\#9402](https://github.com/matrix-org/synapse/issues/9402))
+- Fix a bug that caused multiple calls to the experimental `shared_rooms` endpoint to return stale results. ([\#9416](https://github.com/matrix-org/synapse/issues/9416))
+- Fix a bug in single sign-on which could cause a "No session cookie found" error. ([\#9436](https://github.com/matrix-org/synapse/issues/9436))
+- Fix bug introduced in v1.27.0 where allowing a user to choose their own username when logging in via single sign-on did not work unless an `idp_icon` was defined. ([\#9440](https://github.com/matrix-org/synapse/issues/9440))
+- Fix a bug introduced in v1.26.0 where some sequences were not properly configured when running `synapse_port_db`. ([\#9449](https://github.com/matrix-org/synapse/issues/9449))
+- Fix deleting pushers when using sharded pushers. ([\#9465](https://github.com/matrix-org/synapse/issues/9465), [\#9466](https://github.com/matrix-org/synapse/issues/9466), [\#9479](https://github.com/matrix-org/synapse/issues/9479), [\#9536](https://github.com/matrix-org/synapse/issues/9536))
+- Fix missing startup checks for the consistency of certain PostgreSQL sequences. ([\#9470](https://github.com/matrix-org/synapse/issues/9470))
+- Fix a long-standing bug where the media repository could leak file descriptors while previewing media. ([\#9497](https://github.com/matrix-org/synapse/issues/9497))
+- Properly purge the event chain cover index when purging history. ([\#9498](https://github.com/matrix-org/synapse/issues/9498))
+- Fix missing chain cover index due to a schema delta not being applied correctly. Only affected servers that ran development versions. ([\#9503](https://github.com/matrix-org/synapse/issues/9503))
+- Fix a bug introduced in v1.25.0 where `/_synapse/admin/join/` would fail when given a room alias. ([\#9506](https://github.com/matrix-org/synapse/issues/9506))
+- Prevent presence background jobs from running when presence is disabled. ([\#9530](https://github.com/matrix-org/synapse/issues/9530))
+- Fix rare edge case that caused a background update to fail if the server had rejected an event that had duplicate auth events. ([\#9537](https://github.com/matrix-org/synapse/issues/9537))
+
+
+Improved Documentation
+----------------------
+
+- Update the example systemd config to propagate reloads to individual units. ([\#9463](https://github.com/matrix-org/synapse/issues/9463))
+
+
+Internal Changes
+----------------
+
+- Add documentation and type hints to `parse_duration`. ([\#9432](https://github.com/matrix-org/synapse/issues/9432))
+- Remove vestiges of `uploads_path` configuration setting. ([\#9462](https://github.com/matrix-org/synapse/issues/9462))
+- Add a comment about systemd-python. ([\#9464](https://github.com/matrix-org/synapse/issues/9464))
+- Test that we require validated email for email pushers. ([\#9496](https://github.com/matrix-org/synapse/issues/9496))
+- Allow python to generate bytecode for synapse. ([\#9502](https://github.com/matrix-org/synapse/issues/9502))
+- Fix incorrect type hints. ([\#9515](https://github.com/matrix-org/synapse/issues/9515), [\#9518](https://github.com/matrix-org/synapse/issues/9518))
+- Add type hints to device and event report admin API. ([\#9519](https://github.com/matrix-org/synapse/issues/9519))
+- Add type hints to user admin API. ([\#9521](https://github.com/matrix-org/synapse/issues/9521))
+- Bump the versions of mypy and mypy-zope used for static type checking. ([\#9529](https://github.com/matrix-org/synapse/issues/9529))
+
+
+Synapse 1.28.0 (2021-02-25)
+===========================
+
 Note that this release drops support for ARMv7 in the official Docker images, due to repeated problems building for ARMv7 (and the associated maintenance burden this entails).
 
 This release also fixes the documentation included in v1.27.0 around the callback URI for SAML2 identity providers. If your server is configured to use single sign-on via a SAML2 IdP, you may need to make configuration changes. Please review [UPGRADE.rst](UPGRADE.rst) for more details on these changes.
 
+
+Internal Changes
+----------------
+
+- Revert change in v1.28.0rc1 to remove the deprecated SAML endpoint. ([\#9474](https://github.com/matrix-org/synapse/issues/9474))
+
+
+Synapse 1.28.0rc1 (2021-02-19)
+==============================
+
 Removal warning
 ---------------
 
@@ -31,7 +218,7 @@ Bugfixes
 --------
 
 - Fix long-standing bug where sending email notifications would fail for rooms that the server had since left. ([\#9257](https://github.com/matrix-org/synapse/issues/9257))
-- Fix bug in Synapse 1.27.0rc1 which meant the "session expired" error page during SSO registration was badly formatted. ([\#9296](https://github.com/matrix-org/synapse/issues/9296))
+- Fix bug introduced in Synapse 1.27.0rc1 which meant the "session expired" error page during SSO registration was badly formatted. ([\#9296](https://github.com/matrix-org/synapse/issues/9296))
 - Assert a maximum length for some parameters for spec compliance. ([\#9321](https://github.com/matrix-org/synapse/issues/9321), [\#9393](https://github.com/matrix-org/synapse/issues/9393))
 - Fix additional errors when previewing URLs: "AttributeError 'NoneType' object has no attribute 'xpath'" and "ValueError: Unicode strings with encoding declaration are not supported. Please use bytes input or XML fragments without declaration.". ([\#9333](https://github.com/matrix-org/synapse/issues/9333))
 - Fix a bug causing Synapse to impose the wrong type constraints on fields when processing responses from appservices to `/_matrix/app/v1/thirdparty/user/{protocol}`. ([\#9361](https://github.com/matrix-org/synapse/issues/9361))

MANIFEST.in

@@ -20,9 +20,10 @@ recursive-include scripts *
 recursive-include scripts-dev *
 recursive-include synapse *.pyi
 recursive-include tests *.py
-include tests/http/ca.crt
-include tests/http/ca.key
-include tests/http/server.key
+recursive-include tests *.pem
+recursive-include tests *.p8
+recursive-include tests *.crt
+recursive-include tests *.key
 
 recursive-include synapse/res *
 recursive-include synapse/static *.css

README.rst

@@ -183,8 +183,9 @@ Using a reverse proxy with Synapse
 It is recommended to put a reverse proxy such as
 `nginx <https://nginx.org/en/docs/http/ngx_http_proxy_module.html>`_,
 `Apache <https://httpd.apache.org/docs/current/mod/mod_proxy_http.html>`_,
-`Caddy <https://caddyserver.com/docs/quick-starts/reverse-proxy>`_ or
-`HAProxy <https://www.haproxy.org/>`_ in front of Synapse. One advantage of
+`Caddy <https://caddyserver.com/docs/quick-starts/reverse-proxy>`_,
+`HAProxy <https://www.haproxy.org/>`_ or
+`relayd <https://man.openbsd.org/relayd.8>`_ in front of Synapse. One advantage of
 doing so is that it means that you can expose the default https port (443) to
 Matrix clients without needing to run Synapse with root privileges.
 

UPGRADE.rst

@@ -85,6 +85,26 @@ for example:
      wget https://packages.matrix.org/debian/pool/main/m/matrix-synapse-py3/matrix-synapse-py3_1.3.0+stretch1_amd64.deb
      dpkg -i matrix-synapse-py3_1.3.0+stretch1_amd64.deb
 
+Upgrading to v1.29.0
+====================
+
+Requirement for X-Forwarded-Proto header
+----------------------------------------
+
+When using Synapse with a reverse proxy (in particular, when using the
+`x_forwarded` option on an HTTP listener), Synapse now expects to receive an
+`X-Forwarded-Proto` header on incoming HTTP requests. If it is not set, Synapse
+will log a warning on each received request.
+
+To avoid the warning, administrators using a reverse proxy should ensure that
+the reverse proxy sets `X-Forwarded-Proto` header to `https` or `http` to
+indicate the protocol used by the client. See the `reverse proxy documentation
+<docs/reverse_proxy.md>`_, where the example configurations have been updated to
+show how to set this header.
+
+(Users of `Caddy <https://caddyserver.com/>`_ are unaffected, since we believe it
+sets `X-Forwarded-Proto` by default.)
+
 Upgrading to v1.27.0
 ====================
 
@@ -104,6 +124,13 @@ This version changes the URI used for callbacks from OAuth2 and SAML2 identity p
   need to add ``[synapse public baseurl]/_synapse/client/saml2/authn_response`` as a permitted
   "ACS location" (also known as "allowed callback URLs") at the identity provider.
 
+  The "Issuer" in the "AuthnRequest" to the SAML2 identity provider is also updated to
+  ``[synapse public baseurl]/_synapse/client/saml2/metadata.xml``. If your SAML2 identity
+  provider uses this property to validate or otherwise identify Synapse, its configuration
+  will need to be updated to use the new URL. Alternatively you could create a new, separate
+  "EntityDescriptor" in your SAML2 identity provider with the new URLs and leave the URLs in
+  the existing "EntityDescriptor" as they were.
+
 Changes to HTML templates
 -------------------------
 

debian/build_virtualenv

@@ -58,10 +58,10 @@ trap "rm -r $tmpdir" EXIT
 cp -r tests "$tmpdir"
 
 PYTHONPATH="$tmpdir" \
-    "${TARGET_PYTHON}" -B -m twisted.trial --reporter=text -j2 tests
+    "${TARGET_PYTHON}" -m twisted.trial --reporter=text -j2 tests
 
 # build the config file
-"${TARGET_PYTHON}" -B "${VIRTUALENV_DIR}/bin/generate_config" \
+"${TARGET_PYTHON}" "${VIRTUALENV_DIR}/bin/generate_config" \
         --config-dir="/etc/matrix-synapse" \
         --data-dir="/var/lib/matrix-synapse" |
     perl -pe '
@@ -87,7 +87,7 @@ PYTHONPATH="$tmpdir" \
 ' > "${PACKAGE_BUILD_DIR}/etc/matrix-synapse/homeserver.yaml"
 
 # build the log config file
-"${TARGET_PYTHON}" -B "${VIRTUALENV_DIR}/bin/generate_log_config" \
+"${TARGET_PYTHON}" "${VIRTUALENV_DIR}/bin/generate_log_config" \
         --output-file="${PACKAGE_BUILD_DIR}/etc/matrix-synapse/log.yaml"
 
 # add a dependency on the right version of python to substvars.

debian/changelog

@@ -1,3 +1,31 @@
+matrix-synapse-py3 (1.30.1) stable; urgency=medium
+
+  * New synapse release 1.30.1.
+
+ -- Synapse Packaging team <packages@matrix.org>  Fri, 26 Mar 2021 12:01:28 +0000
+
+matrix-synapse-py3 (1.30.0) stable; urgency=medium
+
+  * New synapse release 1.30.0.
+
+ -- Synapse Packaging team <packages@matrix.org>  Mon, 22 Mar 2021 13:15:34 +0000
+
+matrix-synapse-py3 (1.29.0) stable; urgency=medium
+
+  [ Jonathan de Jong ]
+  * Remove the python -B flag (don't generate bytecode) in scripts and documentation.
+
+  [ Synapse Packaging team ]
+  * New synapse release 1.29.0.
+
+ -- Synapse Packaging team <packages@matrix.org>  Mon, 08 Mar 2021 13:51:50 +0000
+
+matrix-synapse-py3 (1.28.0) stable; urgency=medium
+
+  * New synapse release 1.28.0.
+
+ -- Synapse Packaging team <packages@matrix.org>  Thu, 25 Feb 2021 10:21:57 +0000
+
 matrix-synapse-py3 (1.27.0) stable; urgency=medium
 
   [ Dan Callahan ]

debian/synctl.1

@@ -44,7 +44,7 @@ Configuration file may be generated as follows:
 .
 .nf
 
-$ python \-B \-m synapse\.app\.homeserver \-c config\.yaml \-\-generate\-config \-\-server\-name=<server name>
+$ python \-m synapse\.app\.homeserver \-c config\.yaml \-\-generate\-config \-\-server\-name=<server name>
 .
 .fi
 .

debian/synctl.ronn

@@ -41,7 +41,7 @@ process.
 
 Configuration file may be generated as follows:
 
-    $ python -B -m synapse.app.homeserver -c config.yaml --generate-config --server-name=<server name>
+    $ python -m synapse.app.homeserver -c config.yaml --generate-config --server-name=<server name>
 
 ## ENVIRONMENT
 

docker/Dockerfile

@@ -20,17 +20,18 @@ FROM docker.io/python:${PYTHON_VERSION}-slim as builder
 
 # install the OS build deps
 RUN apt-get update && apt-get install -y \
-    build-essential \
-    libffi-dev \
-    libjpeg-dev \
-    libpq-dev \
-    libssl-dev \
-    libwebp-dev \
-    libxml++2.6-dev \
-    libxslt1-dev \
-    rustc \
-    zlib1g-dev \
- && rm -rf /var/lib/apt/lists/*
+        build-essential \
+        libffi-dev \
+        libjpeg-dev \
+        libpq-dev \
+        libssl-dev \
+        libwebp-dev \
+        libxml++2.6-dev \
+        libxslt1-dev \
+        openssl \
+        rustc \
+        zlib1g-dev \
+        && rm -rf /var/lib/apt/lists/*
 
 # Build dependencies that are not available as wheels, to speed up rebuilds
 RUN pip install --prefix="/install" --no-warn-script-location \
@@ -63,13 +64,16 @@ RUN pip install --prefix="/install" --no-warn-script-location \
 FROM docker.io/python:${PYTHON_VERSION}-slim
 
 RUN apt-get update && apt-get install -y \
-    curl \
-    gosu \
-    libjpeg62-turbo \
-    libpq5 \
-    libwebp6 \
-    xmlsec1 \
- && rm -rf /var/lib/apt/lists/*
+        curl \
+        gosu \
+        libjpeg62-turbo \
+        libpq5 \
+        libwebp6 \
+        xmlsec1 \
+        libjemalloc2 \
+        libssl-dev \
+        openssl \
+        && rm -rf /var/lib/apt/lists/*
 
 COPY --from=builder /install /usr/local
 COPY ./docker/start.py /start.py
@@ -82,4 +86,4 @@ EXPOSE 8008/tcp 8009/tcp 8448/tcp
 ENTRYPOINT ["/start.py"]
 
 HEALTHCHECK --interval=1m --timeout=5s \
-  CMD curl -fSs http://localhost:8008/health || exit 1
+        CMD curl -fSs http://localhost:8008/health || exit 1

docker/README.md

@@ -11,7 +11,6 @@ The image also does *not* provide a TURN server.
 By default, the image expects a single volume, located at ``/data``, that will hold:
 
 * configuration files;
-* temporary files during uploads;
 * uploaded media and thumbnails;
 * the SQLite database if you do not configure postgres;
 * the appservices configuration.
@@ -205,3 +204,8 @@ healthcheck:
   timeout: 10s
   retries: 3
 ```
+
+## Using jemalloc
+
+Jemalloc is embedded in the image and will be used instead of the default allocator.
+You can read about jemalloc by reading the Synapse [README](../README.md)

docker/conf/homeserver.yaml

@@ -89,7 +89,6 @@ federation_rc_concurrent: 3
 ## Files ##
 
 media_store_path: "/data/media"
-uploads_path: "/data/uploads"
 max_upload_size: "{{ SYNAPSE_MAX_UPLOAD_SIZE or "50M" }}"
 max_image_pixels: "32M"
 dynamic_thumbnails: false

docker/start.py

@@ -3,6 +3,7 @@
 import codecs
 import glob
 import os
+import platform
 import subprocess
 import sys
 
@@ -213,6 +214,13 @@ def main(args, environ):
     if "-m" not in args:
         args = ["-m", synapse_worker] + args
 
+    jemallocpath = "/usr/lib/%s-linux-gnu/libjemalloc.so.2" % (platform.machine(),)
+
+    if os.path.isfile(jemallocpath):
+        environ["LD_PRELOAD"] = jemallocpath
+    else:
+        log("Could not find %s, will not use" % (jemallocpath,))
+
     # if there are no config files passed to synapse, try adding the default file
     if not any(p.startswith("--config-path") or p.startswith("-c") for p in args):
         config_dir = environ.get("SYNAPSE_CONFIG_DIR", "/data")
@@ -248,9 +256,9 @@ running with 'migrate_config'. See the README for more details.
     args = ["python"] + args
     if ownership is not None:
         args = ["gosu", ownership] + args
-        os.execv("/usr/sbin/gosu", args)
+        os.execve("/usr/sbin/gosu", args, environ)
     else:
-        os.execv("/usr/local/bin/python", args)
+        os.execve("/usr/local/bin/python", args, environ)
 
 
 if __name__ == "__main__":

docs/admin_api/media_admin_api.md

@@ -1,5 +1,7 @@
 # Contents
-- [List all media in a room](#list-all-media-in-a-room)
+- [Querying media](#querying-media)
+  * [List all media in a room](#list-all-media-in-a-room)
+  * [List all media uploaded by a user](#list-all-media-uploaded-by-a-user)
 - [Quarantine media](#quarantine-media)
   * [Quarantining media by ID](#quarantining-media-by-id)
   * [Quarantining media in a room](#quarantining-media-in-a-room)
@@ -10,7 +12,11 @@
   * [Delete local media by date or size](#delete-local-media-by-date-or-size)
 - [Purge Remote Media API](#purge-remote-media-api)
 
-# List all media in a room
+# Querying media
+
+These APIs allow extracting media information from the homeserver.
+
+## List all media in a room
 
 This API gets a list of known media in a room.
 However, it only shows media from unencrypted events or rooms.
@@ -36,6 +42,12 @@ The API returns a JSON body like the following:
 }
 ```
 
+## List all media uploaded by a user
+
+Listing all media that has been uploaded by a local user can be achieved through
+the use of the [List media of a user](user_admin_api.rst#list-media-of-a-user)
+Admin API.
+
 # Quarantine media
 
 Quarantining media means that it is marked as inaccessible by users. It applies

docs/admin_api/user_admin_api.rst

@@ -379,11 +379,12 @@ The following fields are returned in the JSON response body:
 - ``total`` - Number of rooms.
 
 
-List media of an user
-================================
+List media of a user
+====================
 Gets a list of all local media that a specific ``user_id`` has created.
-The response is ordered by creation date descending and media ID descending.
-The newest media is on top.
+By default, the response is ordered by descending creation date and ascending media ID.
+The newest media is on top. You can change the order with parameters
+``order_by`` and ``dir``.
 
 The API is::
 
@@ -440,6 +441,35 @@ The following parameters should be set in the URL:
   denoting the offset in the returned results. This should be treated as an opaque value and
   not explicitly set to anything other than the return value of ``next_token`` from a previous call.
   Defaults to ``0``.
+- ``order_by`` - The method by which to sort the returned list of media.
+  If the ordered field has duplicates, the second order is always by ascending ``media_id``,
+  which guarantees a stable ordering. Valid values are:
+
+  - ``media_id`` - Media are ordered alphabetically by ``media_id``.
+  - ``upload_name`` - Media are ordered alphabetically by name the media was uploaded with.
+  - ``created_ts`` - Media are ordered by when the content was uploaded in ms.
+    Smallest to largest. This is the default.
+  - ``last_access_ts`` - Media are ordered by when the content was last accessed in ms.
+    Smallest to largest.
+  - ``media_length`` - Media are ordered by length of the media in bytes.
+    Smallest to largest.
+  - ``media_type`` - Media are ordered alphabetically by MIME-type.
+  - ``quarantined_by`` - Media are ordered alphabetically by the user ID that
+    initiated the quarantine request for this media.
+  - ``safe_from_quarantine`` - Media are ordered by the status if this media is safe
+    from quarantining.
+
+- ``dir`` - Direction of media order. Either ``f`` for forwards or ``b`` for backwards.
+  Setting this value to ``b`` will reverse the above sort order. Defaults to ``f``.
+
+If neither ``order_by`` nor ``dir`` is set, the default order is newest media on top
+(corresponds to ``order_by`` = ``created_ts`` and ``dir`` = ``b``).
+
+Caution. The database only has indexes on the columns ``media_id``,
+``user_id`` and ``created_ts``. This means that if a different sort order is used
+(``upload_name``, ``last_access_ts``, ``media_length``, ``media_type``,
+``quarantined_by`` or ``safe_from_quarantine``), this can cause a large load on the
+database, especially for large environments.
 
 **Response**
 

docs/openid.md

@@ -226,7 +226,7 @@ Synapse config:
 oidc_providers:
   - idp_id: github
     idp_name: Github
-    idp_brand: "org.matrix.github"  # optional: styling hint for clients
+    idp_brand: "github"  # optional: styling hint for clients
     discover: false
     issuer: "https://github.com/"
     client_id: "your-client-id" # TO BE FILLED
@@ -252,7 +252,7 @@ oidc_providers:
    oidc_providers:
      - idp_id: google
        idp_name: Google
-       idp_brand: "org.matrix.google"  # optional: styling hint for clients
+       idp_brand: "google"  # optional: styling hint for clients
        issuer: "https://accounts.google.com/"
        client_id: "your-client-id" # TO BE FILLED
        client_secret: "your-client-secret" # TO BE FILLED
@@ -299,7 +299,7 @@ Synapse config:
 oidc_providers:
   - idp_id: gitlab
     idp_name: Gitlab
-    idp_brand: "org.matrix.gitlab"  # optional: styling hint for clients
+    idp_brand: "gitlab"  # optional: styling hint for clients
     issuer: "https://gitlab.com/"
     client_id: "your-client-id" # TO BE FILLED
     client_secret: "your-client-secret" # TO BE FILLED
@@ -334,7 +334,7 @@ Synapse config:
 ```yaml
   - idp_id: facebook
     idp_name: Facebook
-    idp_brand: "org.matrix.facebook"  # optional: styling hint for clients
+    idp_brand: "facebook"  # optional: styling hint for clients
     discover: false
     issuer: "https://facebook.com"
     client_id: "your-client-id" # TO BE FILLED
@@ -386,7 +386,7 @@ oidc_providers:
       config:
         subject_claim: "id"
         localpart_template: "{{ user.login }}"
-        display_name_template: "{{ user.full_name }}" 
+        display_name_template: "{{ user.full_name }}"
 ```
 
 ### XWiki
@@ -401,8 +401,7 @@ oidc_providers:
     idp_name: "XWiki"
     issuer: "https://myxwikihost/xwiki/oidc/"
     client_id: "your-client-id" # TO BE FILLED
-    # Needed until https://github.com/matrix-org/synapse/issues/9212 is fixed
-    client_secret: "dontcare"
+    client_auth_method: none
     scopes: ["openid", "profile"]
     user_profile_method: "userinfo_endpoint"
     user_mapping_provider:
@@ -410,3 +409,40 @@ oidc_providers:
         localpart_template: "{{ user.preferred_username }}"
         display_name_template: "{{ user.name }}"
 ```
+
+## Apple
+
+Configuring "Sign in with Apple" (SiWA) requires an Apple Developer account.
+
+You will need to create a new "Services ID" for SiWA, and create and download a
+private key with "SiWA" enabled.
+
+As well as the private key file, you will need:
+ * Client ID: the "identifier" you gave the "Services ID"
+ * 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
+documentation on setting up SiWA.
+
+The synapse config will look like this:
+
+```yaml
+  - idp_id: apple
+    idp_name: Apple
+    issuer: "https://appleid.apple.com"
+    client_id: "your-client-id" # Set to the "identifier" for your "ServicesID"
+    client_auth_method: "client_secret_post"
+    client_secret_jwt_key:
+      key_file: "/path/to/AuthKey_KEYIDCODE.p8"  # point to your key file
+      jwt_header:
+        alg: ES256
+        kid: "KEYIDCODE"   # Set to the 10-char Key ID
+      jwt_payload:
+        iss: TEAMIDCODE    # Set to the 10-char Team ID
+    scopes: ["name", "email", "openid"]
+    authorization_endpoint: https://appleid.apple.com/auth/authorize?response_mode=form_post
+    user_mapping_provider:
+      config:
+        email_template: "{{ user.email }}"
+```

docs/reverse_proxy.md

@@ -3,30 +3,31 @@
 It is recommended to put a reverse proxy such as
 [nginx](https://nginx.org/en/docs/http/ngx_http_proxy_module.html),
 [Apache](https://httpd.apache.org/docs/current/mod/mod_proxy_http.html),
-[Caddy](https://caddyserver.com/docs/quick-starts/reverse-proxy) or
-[HAProxy](https://www.haproxy.org/) in front of Synapse. One advantage
+[Caddy](https://caddyserver.com/docs/quick-starts/reverse-proxy),
+[HAProxy](https://www.haproxy.org/) or
+[relayd](https://man.openbsd.org/relayd.8) in front of Synapse. One advantage
 of doing so is that it means that you can expose the default https port
 (443) to Matrix clients without needing to run Synapse with root
 privileges.
 
+You should configure your reverse proxy to forward requests to `/_matrix` or
+`/_synapse/client` to Synapse, and have it set the `X-Forwarded-For` and
+`X-Forwarded-Proto` request headers.
+
+You should remember that Matrix clients and other Matrix servers do not
+necessarily need to connect to your server via the same server name or
+port. Indeed, clients will use port 443 by default, whereas servers default to
+port 8448. Where these are different, we refer to the 'client port' and the
+'federation port'. See [the Matrix
+specification](https://matrix.org/docs/spec/server_server/latest#resolving-server-names)
+for more details of the algorithm used for federation connections, and
+[delegate.md](<delegate.md>) for instructions on setting up delegation.
+
 **NOTE**: Your reverse proxy must not `canonicalise` or `normalise`
 the requested URI in any way (for example, by decoding `%xx` escapes).
 Beware that Apache *will* canonicalise URIs unless you specify
 `nocanon`.
 
-When setting up a reverse proxy, remember that Matrix clients and other
-Matrix servers do not necessarily need to connect to your server via the
-same server name or port. Indeed, clients will use port 443 by default,
-whereas servers default to port 8448. Where these are different, we
-refer to the 'client port' and the 'federation port'. See [the Matrix
-specification](https://matrix.org/docs/spec/server_server/latest#resolving-server-names)
-for more details of the algorithm used for federation connections, and
-[delegate.md](<delegate.md>) for instructions on setting up delegation.
-
-Endpoints that are part of the standardised Matrix specification are
-located under `/_matrix`, whereas endpoints specific to Synapse are
-located under `/_synapse/client`.
-
 Let's assume that we expect clients to connect to our server at
 `https://matrix.example.com`, and other servers to connect at
 `https://example.com:8448`.  The following sections detail the configuration of
@@ -52,6 +53,9 @@ server {
     location ~* ^(\/_matrix|\/_synapse\/client) {
         proxy_pass http://localhost:8008;
         proxy_set_header X-Forwarded-For $remote_addr;
+        proxy_set_header X-Forwarded-Proto $scheme;
+        proxy_set_header Host $host;
+
         # Nginx by default only allows file uploads up to 1M in size
         # Increase client_max_body_size to match max_upload_size defined in homeserver.yaml
         client_max_body_size 50M;
@@ -102,6 +106,7 @@ example.com:8448 {
     SSLEngine on
     ServerName matrix.example.com;
 
+    RequestHeader set "X-Forwarded-Proto" expr=%{REQUEST_SCHEME}
     AllowEncodedSlashes NoDecode
     ProxyPass /_matrix http://127.0.0.1:8008/_matrix nocanon
     ProxyPassReverse /_matrix http://127.0.0.1:8008/_matrix
@@ -113,6 +118,7 @@ example.com:8448 {
     SSLEngine on
     ServerName example.com;
 
+    RequestHeader set "X-Forwarded-Proto" expr=%{REQUEST_SCHEME}
     AllowEncodedSlashes NoDecode
     ProxyPass /_matrix http://127.0.0.1:8008/_matrix nocanon
     ProxyPassReverse /_matrix http://127.0.0.1:8008/_matrix
@@ -134,6 +140,9 @@ example.com:8448 {
 ```
 frontend https
   bind :::443 v4v6 ssl crt /etc/ssl/haproxy/ strict-sni alpn h2,http/1.1
+  http-request set-header X-Forwarded-Proto https if { ssl_fc }
+  http-request set-header X-Forwarded-Proto http if !{ ssl_fc }
+  http-request set-header X-Forwarded-For %[src]
 
   # Matrix client traffic
   acl matrix-host hdr(host) -i matrix.example.com
@@ -144,12 +153,62 @@ frontend https
 
 frontend matrix-federation
   bind :::8448 v4v6 ssl crt /etc/ssl/haproxy/synapse.pem alpn h2,http/1.1
+  http-request set-header X-Forwarded-Proto https if { ssl_fc }
+  http-request set-header X-Forwarded-Proto http if !{ ssl_fc }
+  http-request set-header X-Forwarded-For %[src]
+
   default_backend matrix
 
 backend matrix
   server matrix 127.0.0.1:8008
 ```
 
+### Relayd
+
+```
+table <webserver>    { 127.0.0.1 }
+table <matrixserver> { 127.0.0.1 }
+
+http protocol "https" {
+    tls { no tlsv1.0, ciphers "HIGH" }
+    tls keypair "example.com"
+    match header set "X-Forwarded-For"   value "$REMOTE_ADDR"
+    match header set "X-Forwarded-Proto" value "https"
+
+    # set CORS header for .well-known/matrix/server, .well-known/matrix/client
+    # httpd does not support setting headers, so do it here
+    match request path "/.well-known/matrix/*" tag "matrix-cors"
+    match response tagged "matrix-cors" header set "Access-Control-Allow-Origin" value "*"
+
+    pass quick path "/_matrix/*"         forward to <matrixserver>
+    pass quick path "/_synapse/client/*" forward to <matrixserver>
+
+    # pass on non-matrix traffic to webserver
+    pass                                 forward to <webserver>
+}
+
+relay "https_traffic" {
+    listen on egress port 443 tls
+    protocol "https"
+    forward to <matrixserver> port 8008 check tcp
+    forward to <webserver>    port 8080 check tcp
+}
+
+http protocol "matrix" {
+    tls { no tlsv1.0, ciphers "HIGH" }
+    tls keypair "example.com"
+    block
+    pass quick path "/_matrix/*"         forward to <matrixserver>
+    pass quick path "/_synapse/client/*" forward to <matrixserver>
+}
+
+relay "matrix_federation" {
+    listen on egress port 8448 tls
+    protocol "matrix"
+    forward to <matrixserver> port 8008 check tcp
+}
+```
+
 ## Homeserver Configuration
 
 You will also want to set `bind_addresses: ['127.0.0.1']` and

docs/sample_config.yaml

@@ -89,8 +89,7 @@ pid_file: DATADIR/homeserver.pid
 # Whether to require authentication to retrieve profile data (avatars,
 # display names) of other users through the client API. Defaults to
 # 'false'. Note that profile data is also available via the federation
-# API, so this setting is of limited value if federation is enabled on
-# the server.
+# API, unless allow_profile_lookup_over_federation is set to false.
 #
 #require_auth_for_profile_requests: true
 
@@ -101,6 +100,14 @@ pid_file: DATADIR/homeserver.pid
 #
 #limit_profile_requests_to_users_who_share_rooms: true
 
+# Uncomment to prevent a user's profile data from being retrieved and
+# displayed in a room until they have joined it. By default, a user's
+# profile data is included in an invite event, regardless of the values
+# of the above two settings, and whether or not the users share a server.
+# Defaults to 'true'.
+#
+#include_profile_data_on_invite: false
+
 # If set to 'true', removes the need for authentication to access the server's
 # public rooms directory through the client API, meaning that anyone can
 # query the room directory. Defaults to 'false'.
@@ -699,6 +706,12 @@ acme:
 #  - matrix.org
 #  - example.com
 
+# Uncomment to disable profile lookup over federation. By default, the
+# Federation API allows other homeservers to obtain profile data of any user
+# on this homeserver. Defaults to 'true'.
+#
+#allow_profile_lookup_over_federation: false
+
 
 ## Caching ##
 
@@ -1766,7 +1779,26 @@ saml2_config:
 #
 #   client_id: Required. oauth2 client id to use.
 #
-#   client_secret: Required. oauth2 client secret to use.
+#   client_secret: oauth2 client secret to use. May be omitted if
+#        client_secret_jwt_key is given, or if client_auth_method is 'none'.
+#
+#   client_secret_jwt_key: Alternative to client_secret: details of a key used
+#      to create a JSON Web Token to be used as an OAuth2 client secret. If
+#      given, must be a dictionary with the following properties:
+#
+#          key: a pem-encoded signing key. Must be a suitable key for the
+#              algorithm specified. Required unless 'key_file' is given.
+#
+#          key_file: the path to file containing a pem-encoded signing key file.
+#              Required unless 'key' is given.
+#
+#          jwt_header: a dictionary giving properties to include in the JWT
+#              header. Must include the key 'alg', giving the algorithm used to
+#              sign the JWT, such as "ES256", using the JWA identifiers in
+#              RFC7518.
+#
+#          jwt_payload: an optional dictionary giving properties to include in
+#              the JWT payload. Normally this should include an 'iss' key.
 #
 #   client_auth_method: auth method to use when exchanging the token. Valid
 #       values are 'client_secret_basic' (default), 'client_secret_post' and
@@ -1887,7 +1919,7 @@ oidc_providers:
   #
   #- idp_id: github
   #  idp_name: Github
-  #  idp_brand: org.matrix.github
+  #  idp_brand: github
   #  discover: false
   #  issuer: "https://github.com/"
   #  client_id: "your-client-id" # TO BE FILLED
@@ -2530,19 +2562,35 @@ spam_checker:
 
 # User Directory configuration
 #
-# 'enabled' defines whether users can search the user directory. If
-# false then empty responses are returned to all queries. Defaults to
-# true.
-#
-# 'search_all_users' defines whether to search all users visible to your HS
-# when searching the user directory, rather than limiting to users visible
-# in public rooms.  Defaults to false.  If you set it True, you'll have to
-# rebuild the user_directory search indexes, see
-# https://github.com/matrix-org/synapse/blob/master/docs/user_directory.md
-#
-#user_directory:
-#  enabled: true
-#  search_all_users: false
+user_directory:
+    # Defines whether users can search the user directory. If false then
+    # empty responses are returned to all queries. Defaults to true.
+    #
+    # Uncomment to disable the user directory.
+    #
+    #enabled: false
+
+    # Defines whether to search all users visible to your HS when searching
+    # the user directory, rather than limiting to users visible in public
+    # rooms. Defaults to false.
+    #
+    # If you set it true, you'll have to rebuild the user_directory search
+    # indexes, see:
+    # https://github.com/matrix-org/synapse/blob/master/docs/user_directory.md
+    #
+    # Uncomment to return search results containing all known users, even if that
+    # user does not share a room with the requester.
+    #
+    #search_all_users: true
+
+    # Defines whether to prefer local users in search query results.
+    # If True, local users are more likely to appear above remote users
+    # when searching the user directory. Defaults to false.
+    #
+    # Uncomment to prefer local over remote users in user directory search
+    # results.
+    #
+    #prefer_local_users: true
 
 
 # User Consent configuration
@@ -2597,19 +2645,20 @@ spam_checker:
 
 
 
-# Local statistics collection. Used in populating the room directory.
+# Settings for local room and user statistics collection. See
+# docs/room_and_user_statistics.md.
 #
-# 'bucket_size' controls how large each statistics timeslice is. It can
-# be defined in a human readable short form -- e.g. "1d", "1y".
-#
-# 'retention' controls how long historical statistics will be kept for.
-# It can be defined in a human readable short form -- e.g. "1d", "1y".
-#
-#
-#stats:
-#   enabled: true
-#   bucket_size: 1d
-#   retention: 1y
+stats:
+  # Uncomment the following to disable room and user statistics. Note that doing
+  # so may cause certain features (such as the room directory) not to work
+  # correctly.
+  #
+  #enabled: false
+
+  # The size of each timeslice in the room_stats_historical and
+  # user_stats_historical tables, as a time period. Defaults to "1d".
+  #
+  #bucket_size: 1h
 
 
 # Server Notices room configuration

docs/spam_checker.md

@@ -14,6 +14,7 @@ The Python class is instantiated with two objects:
 * An instance of `synapse.module_api.ModuleApi`.
 
 It then implements methods which return a boolean to alter behavior in Synapse.
+All the methods must be defined.
 
 There's a generic method for checking every event (`check_event_for_spam`), as
 well as some specific methods:
@@ -24,13 +25,18 @@ well as some specific methods:
 * `user_may_publish_room`
 * `check_username_for_spam`
 * `check_registration_for_spam`
+* `check_media_file_for_spam`
 
-The details of the each of these methods (as well as their inputs and outputs)
+The details of each of these methods (as well as their inputs and outputs)
 are documented in the `synapse.events.spamcheck.SpamChecker` class.
 
 The `ModuleApi` class provides a way for the custom spam checker class to
 call back into the homeserver internals.
 
+Additionally, a `parse_config` method is mandatory and receives the plugin config
+dictionary. After parsing, It must return an object which will be
+passed to `__init__` later.
+
 ### Example
 
 ```python
@@ -41,6 +47,10 @@ class ExampleSpamChecker:
         self.config = config
         self.api = api
 
+    @staticmethod
+    def parse_config(config):
+        return config
+        
     async def check_event_for_spam(self, foo):
         return False  # allow all events
 
@@ -59,7 +69,13 @@ class ExampleSpamChecker:
     async def check_username_for_spam(self, user_profile):
         return False  # allow all usernames
 
-    async def check_registration_for_spam(self, email_threepid, username, request_info):
+    async def check_registration_for_spam(
+        self,
+        email_threepid,
+        username,
+        request_info,
+        auth_provider_id,
+    ):
         return RegistrationBehaviour.ALLOW  # allow all registrations
 
     async def check_media_file_for_spam(self, file_wrapper, file_info):

docs/systemd-with-workers/system/matrix-synapse-worker@.service

@@ -4,6 +4,7 @@ AssertPathExists=/etc/matrix-synapse/workers/%i.yaml
 
 # This service should be restarted when the synapse target is restarted.
 PartOf=matrix-synapse.target
+ReloadPropagatedFrom=matrix-synapse.target
 
 # if this is started at the same time as the main, let the main process start
 # first, to initialise the database schema.

docs/systemd-with-workers/system/matrix-synapse.service

@@ -3,6 +3,7 @@ Description=Synapse master
 
 # This service should be restarted when the synapse target is restarted.
 PartOf=matrix-synapse.target
+ReloadPropagatedFrom=matrix-synapse.target
 
 [Service]
 Type=notify

docs/tcp_replication.md

@@ -220,10 +220,6 @@ Asks the server for the current position of all streams.
 
    Acknowledge receipt of some federation data
 
-#### REMOVE_PUSHER (C)
-
-   Inform the server a pusher should be removed
-
 ### REMOTE_SERVER_UP (S, C)
 
    Inform other processes that a remote server may have come back online.

mypy.ini

@@ -69,6 +69,7 @@ files =
   synapse/util/async_helpers.py,
   synapse/util/caches,
   synapse/util/metrics.py,
+  synapse/util/macaroons.py,
   synapse/util/stringutils.py,
   tests/replication,
   tests/test_utils,
@@ -116,9 +117,6 @@ ignore_missing_imports = True
 [mypy-saml2.*]
 ignore_missing_imports = True
 
-[mypy-unpaddedbase64]
-ignore_missing_imports = True
-
 [mypy-canonicaljson]
 ignore_missing_imports = True
 

scripts-dev/config-lint.sh

@@ -2,9 +2,14 @@
 # Find linting errors in Synapse's default config file.
 # Exits with 0 if there are no problems, or another code otherwise.
 
+# cd to the root of the repository
+cd `dirname $0`/..
+
+# Restore backup of sample config upon script exit
+trap "mv docs/sample_config.yaml.bak docs/sample_config.yaml" EXIT
+
 # Fix non-lowercase true/false values
 sed -i.bak -E "s/: +True/: true/g; s/: +False/: false/g;" docs/sample_config.yaml
-rm docs/sample_config.yaml.bak
 
 # Check if anything changed
-git diff --exit-code docs/sample_config.yaml
+diff docs/sample_config.yaml docs/sample_config.yaml.bak

scripts/synapse_port_db

@@ -22,7 +22,7 @@ import logging
 import sys
 import time
 import traceback
-from typing import Dict, Optional, Set
+from typing import Dict, Iterable, Optional, Set
 
 import yaml
 
@@ -47,6 +47,7 @@ from synapse.storage.databases.main.events_bg_updates import (
 from synapse.storage.databases.main.media_repository import (
     MediaRepositoryBackgroundUpdateStore,
 )
+from synapse.storage.databases.main.pusher import PusherWorkerStore
 from synapse.storage.databases.main.registration import (
     RegistrationBackgroundUpdateStore,
     find_max_generated_user_id_localpart,
@@ -177,6 +178,7 @@ class Store(
     UserDirectoryBackgroundUpdateStore,
     EndToEndKeyBackgroundStore,
     StatsStore,
+    PusherWorkerStore,
 ):
     def execute(self, f, *args, **kwargs):
         return self.db_pool.runInteraction(f.__name__, f, *args, **kwargs)
@@ -629,7 +631,13 @@ class Porter(object):
             await self._setup_state_group_id_seq()
             await self._setup_user_id_seq()
             await self._setup_events_stream_seqs()
-            await self._setup_device_inbox_seq()
+            await self._setup_sequence(
+                "device_inbox_sequence", ("device_inbox", "device_federation_outbox")
+            )
+            await self._setup_sequence(
+                "account_data_sequence", ("room_account_data", "room_tags_revisions", "account_data"))
+            await self._setup_sequence("receipts_sequence", ("receipts_linearized", ))
+            await self._setup_auth_chain_sequence()
 
             # Step 3. Get tables.
             self.progress.set_state("Fetching tables")
@@ -854,7 +862,7 @@ class Porter(object):
 
         return done, remaining + done
 
-    async def _setup_state_group_id_seq(self):
+    async def _setup_state_group_id_seq(self) -> None:
         curr_id = await self.sqlite_store.db_pool.simple_select_one_onecol(
             table="state_groups", keyvalues={}, retcol="MAX(id)", allow_none=True
         )
@@ -868,7 +876,7 @@ class Porter(object):
 
         await self.postgres_store.db_pool.runInteraction("setup_state_group_id_seq", r)
 
-    async def _setup_user_id_seq(self):
+    async def _setup_user_id_seq(self) -> None:
         curr_id = await self.sqlite_store.db_pool.runInteraction(
             "setup_user_id_seq", find_max_generated_user_id_localpart
         )
@@ -877,9 +885,9 @@ class Porter(object):
             next_id = curr_id + 1
             txn.execute("ALTER SEQUENCE user_id_seq RESTART WITH %s", (next_id,))
 
-        return self.postgres_store.db_pool.runInteraction("setup_user_id_seq", r)
+        await self.postgres_store.db_pool.runInteraction("setup_user_id_seq", r)
 
-    async def _setup_events_stream_seqs(self):
+    async def _setup_events_stream_seqs(self) -> None:
         """Set the event stream sequences to the correct values.
         """
 
@@ -908,35 +916,46 @@ class Porter(object):
                 (curr_backward_id + 1,),
             )
 
-        return await self.postgres_store.db_pool.runInteraction(
+        await self.postgres_store.db_pool.runInteraction(
             "_setup_events_stream_seqs", _setup_events_stream_seqs_set_pos,
         )
 
-    async def _setup_device_inbox_seq(self):
-        """Set the device inbox sequence to the correct value.
+    async def _setup_sequence(self, sequence_name: str, stream_id_tables: Iterable[str]) -> None:
+        """Set a sequence to the correct value.
         """
-        curr_local_id = await self.sqlite_store.db_pool.simple_select_one_onecol(
-            table="device_inbox",
-            keyvalues={},
-            retcol="COALESCE(MAX(stream_id), 1)",
-            allow_none=True,
-        )
+        current_stream_ids = []
+        for stream_id_table in stream_id_tables:
+            max_stream_id = await self.sqlite_store.db_pool.simple_select_one_onecol(
+                table=stream_id_table,
+                keyvalues={},
+                retcol="COALESCE(MAX(stream_id), 1)",
+                allow_none=True,
+            )
+            current_stream_ids.append(max_stream_id)
 
-        curr_federation_id = await self.sqlite_store.db_pool.simple_select_one_onecol(
-            table="device_federation_outbox",
-            keyvalues={},
-            retcol="COALESCE(MAX(stream_id), 1)",
-            allow_none=True,
-        )
+        next_id = max(current_stream_ids) + 1
 
-        next_id = max(curr_local_id, curr_federation_id) + 1
+        def r(txn):
+            sql = "ALTER SEQUENCE %s RESTART WITH" % (sequence_name, )
+            txn.execute(sql + " %s", (next_id, ))
+
+        await self.postgres_store.db_pool.runInteraction("_setup_%s" % (sequence_name,), r)
+
+    async def _setup_auth_chain_sequence(self) -> None:
+        curr_chain_id = await self.sqlite_store.db_pool.simple_select_one_onecol(
+            table="event_auth_chains", keyvalues={}, retcol="MAX(chain_id)", allow_none=True
+        )
 
         def r(txn):
             txn.execute(
-                "ALTER SEQUENCE device_inbox_sequence RESTART WITH %s", (next_id,)
+                "ALTER SEQUENCE event_auth_chain_id RESTART WITH %s",
+                (curr_chain_id,),
             )
 
-        return self.postgres_store.db_pool.runInteraction("_setup_device_inbox_seq", r)
+        await self.postgres_store.db_pool.runInteraction(
+            "_setup_event_auth_chain_id", r,
+        )
+
 
 
 ##############################################

setup.cfg

@@ -3,6 +3,7 @@ test_suite = tests
 
 [check-manifest]
 ignore =
+    .git-blame-ignore-revs
     contrib
     contrib/*
     docs/*

setup.py

@@ -102,7 +102,7 @@ CONDITIONAL_REQUIREMENTS["lint"] = [
     "flake8",
 ]
 
-CONDITIONAL_REQUIREMENTS["mypy"] = ["mypy==0.790", "mypy-zope==0.2.8"]
+CONDITIONAL_REQUIREMENTS["mypy"] = ["mypy==0.812", "mypy-zope==0.2.11"]
 
 # Dependencies which are exclusively required by unit test code. This is
 # NOT a list of all modules that are necessary to run the unit tests.

stubs/txredisapi.pyi

@@ -17,7 +17,9 @@
 """
 from typing import Any, List, Optional, Type, Union
 
-class RedisProtocol:
+from twisted.internet import protocol
+
+class RedisProtocol(protocol.Protocol):
     def publish(self, channel: str, message: bytes): ...
     async def ping(self) -> None: ...
     async def set(
@@ -52,7 +54,7 @@ def lazyConnection(
 
 class ConnectionHandler: ...
 
-class RedisFactory:
+class RedisFactory(protocol.ReconnectingClientFactory):
     continueTrying: bool
     handler: RedisProtocol
     pool: List[RedisProtocol]

synapse/__init__.py

@@ -48,7 +48,7 @@ try:
 except ImportError:
     pass
 
-__version__ = "1.28.0rc1"
+__version__ = "1.30.1"
 
 if bool(os.environ.get("SYNAPSE_TEST_PATCH_LOG_CONTEXTS", False)):
     # We import here so that we don't have to install a bunch of deps when

synapse/api/auth.py

@@ -39,6 +39,7 @@ from synapse.logging import opentracing as opentracing
 from synapse.storage.databases.main.registration import TokenLookupResult
 from synapse.types import StateMap, UserID
 from synapse.util.caches.lrucache import LruCache
+from synapse.util.macaroons import get_value_from_macaroon, satisfy_expiry
 from synapse.util.metrics import Measure
 
 logger = logging.getLogger(__name__)
@@ -163,7 +164,7 @@ class Auth:
 
     async def get_user_by_req(
         self,
-        request: Request,
+        request: SynapseRequest,
         allow_guest: bool = False,
         rights: str = "access",
         allow_expired: bool = False,
@@ -408,7 +409,7 @@ class Auth:
             raise _InvalidMacaroonException()
 
         try:
-            user_id = self.get_user_id_from_macaroon(macaroon)
+            user_id = get_value_from_macaroon(macaroon, "user_id")
 
             guest = False
             for caveat in macaroon.caveats:
@@ -416,7 +417,12 @@ class Auth:
                     guest = True
 
             self.validate_macaroon(macaroon, rights, user_id=user_id)
-        except (pymacaroons.exceptions.MacaroonException, TypeError, ValueError):
+        except (
+            pymacaroons.exceptions.MacaroonException,
+            KeyError,
+            TypeError,
+            ValueError,
+        ):
             raise InvalidClientTokenError("Invalid macaroon passed.")
 
         if rights == "access":
@@ -424,27 +430,6 @@ class Auth:
 
         return user_id, guest
 
-    def get_user_id_from_macaroon(self, macaroon):
-        """Retrieve the user_id given by the caveats on the macaroon.
-
-        Does *not* validate the macaroon.
-
-        Args:
-            macaroon (pymacaroons.Macaroon): The macaroon to validate
-
-        Returns:
-            (str) user id
-
-        Raises:
-            InvalidClientCredentialsError if there is no user_id caveat in the
-                macaroon
-        """
-        user_prefix = "user_id = "
-        for caveat in macaroon.caveats:
-            if caveat.caveat_id.startswith(user_prefix):
-                return caveat.caveat_id[len(user_prefix) :]
-        raise InvalidClientTokenError("No user caveat in macaroon")
-
     def validate_macaroon(self, macaroon, type_string, user_id):
         """
         validate that a Macaroon is understood by and was signed by this server.
@@ -465,21 +450,13 @@ class Auth:
         v.satisfy_exact("type = " + type_string)
         v.satisfy_exact("user_id = %s" % user_id)
         v.satisfy_exact("guest = true")
-        v.satisfy_general(self._verify_expiry)
+        satisfy_expiry(v, self.clock.time_msec)
 
         # access_tokens include a nonce for uniqueness: any value is acceptable
         v.satisfy_general(lambda c: c.startswith("nonce = "))
 
         v.verify(macaroon, self._macaroon_secret_key)
 
-    def _verify_expiry(self, caveat):
-        prefix = "time < "
-        if not caveat.startswith(prefix):
-            return False
-        expiry = int(caveat[len(prefix) :])
-        now = self.hs.get_clock().time_msec()
-        return now < expiry
-
     def get_appservice_by_req(self, request: SynapseRequest) -> ApplicationService:
         token = self.get_access_token_from_request(request)
         service = self.store.get_app_service_by_token(token)

synapse/api/constants.py

@@ -98,11 +98,14 @@ class EventTypes:
 
     Retention = "m.room.retention"
 
-    Presence = "m.presence"
-
     Dummy = "org.matrix.dummy_event"
 
 
+class EduTypes:
+    Presence = "m.presence"
+    RoomKeyRequest = "m.room_key_request"
+
+
 class RejectedReason:
     AUTH_ERROR = "auth_error"
 

synapse/api/ratelimiting.py

@@ -14,7 +14,7 @@
 # limitations under the License.
 
 from collections import OrderedDict
-from typing import Any, Optional, Tuple
+from typing import Hashable, Optional, Tuple
 
 from synapse.api.errors import LimitExceededError
 from synapse.types import Requester
@@ -42,7 +42,9 @@ class Ratelimiter:
         #   * How many times an action has occurred since a point in time
         #   * The point in time
         #   * The rate_hz of this particular entry. This can vary per request
-        self.actions = OrderedDict()  # type: OrderedDict[Any, Tuple[float, int, float]]
+        self.actions = (
+            OrderedDict()
+        )  # type: OrderedDict[Hashable, Tuple[float, int, float]]
 
     def can_requester_do_action(
         self,
@@ -82,7 +84,7 @@ class Ratelimiter:
 
     def can_do_action(
         self,
-        key: Any,
+        key: Hashable,
         rate_hz: Optional[float] = None,
         burst_count: Optional[int] = None,
         update: bool = True,
@@ -175,7 +177,7 @@ class Ratelimiter:
 
     def ratelimit(
         self,
-        key: Any,
+        key: Hashable,
         rate_hz: Optional[float] = None,
         burst_count: Optional[int] = None,
         update: bool = True,

synapse/app/__init__.py

@@ -17,8 +17,6 @@ import sys
 
 from synapse import python_dependencies  # noqa: E402
 
-sys.dont_write_bytecode = True
-
 logger = logging.getLogger(__name__)
 
 try:

synapse/app/admin_cmd.py

@@ -210,7 +210,9 @@ def start(config_options):
     config.update_user_directory = False
     config.run_background_tasks = False
     config.start_pushers = False
+    config.pusher_shard_config.instances = []
     config.send_federation = False
+    config.federation_shard_config.instances = []
 
     synapse.events.USE_FROZEN_DICTS = config.use_frozen_dicts
 

synapse/app/generic_worker.py

@@ -23,6 +23,7 @@ from typing_extensions import ContextManager
 
 from twisted.internet import address
 from twisted.web.resource import IResource
+from twisted.web.server import Request
 
 import synapse
 import synapse.events
@@ -190,7 +191,7 @@ class KeyUploadServlet(RestServlet):
         self.http_client = hs.get_simple_http_client()
         self.main_uri = hs.config.worker_main_http_uri
 
-    async def on_POST(self, request, device_id):
+    async def on_POST(self, request: Request, device_id: Optional[str]):
         requester = await self.auth.get_user_by_req(request, allow_guest=True)
         user_id = requester.user.to_string()
         body = parse_json_object_from_request(request)
@@ -223,10 +224,12 @@ class KeyUploadServlet(RestServlet):
                 header: request.requestHeaders.getRawHeaders(header, [])
                 for header in (b"Authorization", b"User-Agent")
             }
-            # Add the previous hop the the X-Forwarded-For header.
+            # Add the previous hop to the X-Forwarded-For header.
             x_forwarded_for = request.requestHeaders.getRawHeaders(
                 b"X-Forwarded-For", []
             )
+            # we use request.client here, since we want the previous hop, not the
+            # original client (as returned by request.getClientAddress()).
             if isinstance(request.client, (address.IPv4Address, address.IPv6Address)):
                 previous_host = request.client.host.encode("ascii")
                 # If the header exists, add to the comma-separated list of the first
@@ -239,6 +242,14 @@ class KeyUploadServlet(RestServlet):
                     x_forwarded_for = [previous_host]
             headers[b"X-Forwarded-For"] = x_forwarded_for
 
+            # Replicate the original X-Forwarded-Proto header. Note that
+            # XForwardedForRequest overrides isSecure() to give us the original protocol
+            # used by the client, as opposed to the protocol used by our upstream proxy
+            # - which is what we want here.
+            headers[b"X-Forwarded-Proto"] = [
+                b"https" if request.isSecure() else b"http"
+            ]
+
             try:
                 result = await self.http_client.post_json_get_json(
                     self.main_uri + request.uri.decode("ascii"), body, headers=headers
@@ -645,9 +656,6 @@ class GenericWorkerServer(HomeServer):
 
         self.get_tcp_replication().start_replication(self)
 
-    async def remove_pusher(self, app_id, push_key, user_id):
-        self.get_tcp_replication().send_remove_pusher(app_id, push_key, user_id)
-
     @cache_in_self
     def get_replication_data_handler(self):
         return GenericWorkerReplicationHandler(self)
@@ -922,22 +930,6 @@ def start(config_options):
         # For other worker types we force this to off.
         config.appservice.notify_appservices = False
 
-    if config.worker_app == "synapse.app.pusher":
-        if config.server.start_pushers:
-            sys.stderr.write(
-                "\nThe pushers must be disabled in the main synapse process"
-                "\nbefore they can be run in a separate worker."
-                "\nPlease add ``start_pushers: false`` to the main config"
-                "\n"
-            )
-            sys.exit(1)
-
-        # Force the pushers to start since they will be disabled in the main config
-        config.server.start_pushers = True
-    else:
-        # For other worker types we force this to off.
-        config.server.start_pushers = False
-
     if config.worker_app == "synapse.app.user_dir":
         if config.server.update_user_directory:
             sys.stderr.write(
@@ -954,22 +946,6 @@ def start(config_options):
         # For other worker types we force this to off.
         config.server.update_user_directory = False
 
-    if config.worker_app == "synapse.app.federation_sender":
-        if config.worker.send_federation:
-            sys.stderr.write(
-                "\nThe send_federation must be disabled in the main synapse process"
-                "\nbefore they can be run in a separate worker."
-                "\nPlease add ``send_federation: false`` to the main config"
-                "\n"
-            )
-            sys.exit(1)
-
-        # Force the pushers to start since they will be disabled in the main config
-        config.worker.send_federation = True
-    else:
-        # For other worker types we force this to off.
-        config.worker.send_federation = False
-
     synapse.events.USE_FROZEN_DICTS = config.use_frozen_dicts
 
     hs = GenericWorkerServer(

synapse/appservice/api.py

@@ -90,7 +90,7 @@ class ApplicationServiceApi(SimpleHttpClient):
         self.clock = hs.get_clock()
 
         self.protocol_meta_cache = ResponseCache(
-            hs, "as_protocol_meta", timeout_ms=HOUR_IN_MS
+            hs.get_clock(), "as_protocol_meta", timeout_ms=HOUR_IN_MS
         )  # type: ResponseCache[Tuple[str, str]]
 
     async def query_user(self, service, user_id):

synapse/config/_base.py

@@ -21,7 +21,7 @@ import os
 from collections import OrderedDict
 from hashlib import sha256
 from textwrap import dedent
-from typing import Any, Iterable, List, MutableMapping, Optional
+from typing import Any, Iterable, List, MutableMapping, Optional, Union
 
 import attr
 import jinja2
@@ -147,7 +147,20 @@ class Config:
         return int(value) * size
 
     @staticmethod
-    def parse_duration(value):
+    def parse_duration(value: Union[str, int]) -> int:
+        """Convert a duration as a string or integer to a number of milliseconds.
+
+        If an integer is provided it is treated as milliseconds and is unchanged.
+
+        String durations can have a suffix of 's', 'm', 'h', 'd', 'w', or 'y'.
+        No suffix is treated as milliseconds.
+
+        Args:
+            value: The duration to parse.
+
+        Returns:
+            The number of milliseconds in the duration.
+        """
         if isinstance(value, int):
             return value
         second = 1000
@@ -199,9 +212,8 @@ class Config:
 
     @classmethod
     def read_file(cls, file_path, config_name):
-        cls.check_file(file_path, config_name)
-        with open(file_path) as file_stream:
-            return file_stream.read()
+        """Deprecated: call read_file directly"""
+        return read_file(file_path, (config_name,))
 
     def read_template(self, filename: str) -> jinja2.Template:
         """Load a template file from disk.
@@ -831,22 +843,23 @@ class ShardedWorkerHandlingConfig:
 
     def should_handle(self, instance_name: str, key: str) -> bool:
         """Whether this instance is responsible for handling the given key."""
-        # If multiple instances are not defined we always return true
-        if not self.instances or len(self.instances) == 1:
-            return True
+        # If no instances are defined we assume some other worker is handling
+        # this.
+        if not self.instances:
+            return False
 
-        return self.get_instance(key) == instance_name
+        return self._get_instance(key) == instance_name
 
-    def get_instance(self, key: str) -> str:
+    def _get_instance(self, key: str) -> str:
         """Get the instance responsible for handling the given key.
 
-        Note: For things like federation sending the config for which instance
-        is sending is known only to the sender instance if there is only one.
-        Therefore `should_handle` should be used where possible.
+        Note: For federation sending and pushers the config for which instance
+        is sending is known only to the sender instance, so we don't expose this
+        method by default.
         """
 
         if not self.instances:
-            return "master"
+            raise Exception("Unknown worker")
 
         if len(self.instances) == 1:
             return self.instances[0]
@@ -863,4 +876,52 @@ class ShardedWorkerHandlingConfig:
         return self.instances[remainder]
 
 
-__all__ = ["Config", "RootConfig", "ShardedWorkerHandlingConfig"]
+@attr.s
+class RoutableShardedWorkerHandlingConfig(ShardedWorkerHandlingConfig):
+    """A version of `ShardedWorkerHandlingConfig` that is used for config
+    options where all instances know which instances are responsible for the
+    sharded work.
+    """
+
+    def __attrs_post_init__(self):
+        # We require that `self.instances` is non-empty.
+        if not self.instances:
+            raise Exception("Got empty list of instances for shard config")
+
+    def get_instance(self, key: str) -> str:
+        """Get the instance responsible for handling the given key."""
+        return self._get_instance(key)
+
+
+def read_file(file_path: Any, config_path: Iterable[str]) -> str:
+    """Check the given file exists, and read it into a string
+
+    If it does not, emit an error indicating the problem
+
+    Args:
+        file_path: the file to be read
+        config_path: where in the configuration file_path came from, so that a useful
+           error can be emitted if it does not exist.
+    Returns:
+        content of the file.
+    Raises:
+        ConfigError if there is a problem reading the file.
+    """
+    if not isinstance(file_path, str):
+        raise ConfigError("%r is not a string", config_path)
+
+    try:
+        os.stat(file_path)
+        with open(file_path) as file_stream:
+            return file_stream.read()
+    except OSError as e:
+        raise ConfigError("Error accessing file %r" % (file_path,), config_path) from e
+
+
+__all__ = [
+    "Config",
+    "RootConfig",
+    "ShardedWorkerHandlingConfig",
+    "RoutableShardedWorkerHandlingConfig",
+    "read_file",
+]

synapse/config/_base.pyi

@@ -149,4 +149,8 @@ class ShardedWorkerHandlingConfig:
     instances: List[str]
     def __init__(self, instances: List[str]) -> None: ...
     def should_handle(self, instance_name: str, key: str) -> bool: ...
+
+class RoutableShardedWorkerHandlingConfig(ShardedWorkerHandlingConfig):
     def get_instance(self, key: str) -> str: ...
+
+def read_file(file_path: Any, config_path: Iterable[str]) -> str: ...

synapse/config/federation.py

@@ -41,6 +41,10 @@ class FederationConfig(Config):
         )
         self.federation_metrics_domains = set(federation_metrics_domains)
 
+        self.allow_profile_lookup_over_federation = config.get(
+            "allow_profile_lookup_over_federation", True
+        )
+
     def generate_config_section(self, config_dir_path, server_name, **kwargs):
         return """\
         ## Federation ##
@@ -66,6 +70,12 @@ class FederationConfig(Config):
         #federation_metrics_domains:
         #  - matrix.org
         #  - example.com
+
+        # Uncomment to disable profile lookup over federation. By default, the
+        # Federation API allows other homeservers to obtain profile data of any user
+        # on this homeserver. Defaults to 'true'.
+        #
+        #allow_profile_lookup_over_federation: false
         """
 
 

synapse/config/logger.py

@@ -21,8 +21,10 @@ import threading
 from string import Template
 
 import yaml
+from zope.interface import implementer
 
 from twisted.logger import (
+    ILogObserver,
     LogBeginner,
     STDLibLogObserver,
     eventAsText,
@@ -227,7 +229,8 @@ def _setup_stdlib_logging(config, log_config_path, logBeginner: LogBeginner) ->
 
     threadlocal = threading.local()
 
-    def _log(event):
+    @implementer(ILogObserver)
+    def _log(event: dict) -> None:
         if "log_text" in event:
             if event["log_text"].startswith("DNSDatagramProtocol starting on "):
                 return

synapse/config/oidc_config.py

@@ -15,7 +15,7 @@
 # limitations under the License.
 
 from collections import Counter
-from typing import Iterable, Optional, Tuple, Type
+from typing import Iterable, Mapping, Optional, Tuple, Type
 
 import attr
 
@@ -25,7 +25,7 @@ from synapse.types import Collection, JsonDict
 from synapse.util.module_loader import load_module
 from synapse.util.stringutils import parse_and_validate_mxc_uri
 
-from ._base import Config, ConfigError
+from ._base import Config, ConfigError, read_file
 
 DEFAULT_USER_MAPPING_PROVIDER = "synapse.handlers.oidc_handler.JinjaOidcMappingProvider"
 
@@ -97,7 +97,26 @@ class OIDCConfig(Config):
         #
         #   client_id: Required. oauth2 client id to use.
         #
-        #   client_secret: Required. oauth2 client secret to use.
+        #   client_secret: oauth2 client secret to use. May be omitted if
+        #        client_secret_jwt_key is given, or if client_auth_method is 'none'.
+        #
+        #   client_secret_jwt_key: Alternative to client_secret: details of a key used
+        #      to create a JSON Web Token to be used as an OAuth2 client secret. If
+        #      given, must be a dictionary with the following properties:
+        #
+        #          key: a pem-encoded signing key. Must be a suitable key for the
+        #              algorithm specified. Required unless 'key_file' is given.
+        #
+        #          key_file: the path to file containing a pem-encoded signing key file.
+        #              Required unless 'key' is given.
+        #
+        #          jwt_header: a dictionary giving properties to include in the JWT
+        #              header. Must include the key 'alg', giving the algorithm used to
+        #              sign the JWT, such as "ES256", using the JWA identifiers in
+        #              RFC7518.
+        #
+        #          jwt_payload: an optional dictionary giving properties to include in
+        #              the JWT payload. Normally this should include an 'iss' key.
         #
         #   client_auth_method: auth method to use when exchanging the token. Valid
         #       values are 'client_secret_basic' (default), 'client_secret_post' and
@@ -218,7 +237,7 @@ class OIDCConfig(Config):
           #
           #- idp_id: github
           #  idp_name: Github
-          #  idp_brand: org.matrix.github
+          #  idp_brand: github
           #  discover: false
           #  issuer: "https://github.com/"
           #  client_id: "your-client-id" # TO BE FILLED
@@ -240,7 +259,7 @@ class OIDCConfig(Config):
 # jsonschema definition of the configuration settings for an oidc identity provider
 OIDC_PROVIDER_CONFIG_SCHEMA = {
     "type": "object",
-    "required": ["issuer", "client_id", "client_secret"],
+    "required": ["issuer", "client_id"],
     "properties": {
         "idp_id": {
             "type": "string",
@@ -253,7 +272,12 @@ OIDC_PROVIDER_CONFIG_SCHEMA = {
         "idp_icon": {"type": "string"},
         "idp_brand": {
             "type": "string",
-            # MSC2758-style namespaced identifier
+            "minLength": 1,
+            "maxLength": 255,
+            "pattern": "^[a-z][a-z0-9_.-]*$",
+        },
+        "idp_unstable_brand": {
+            "type": "string",
             "minLength": 1,
             "maxLength": 255,
             "pattern": "^[a-z][a-z0-9_.-]*$",
@@ -262,6 +286,30 @@ OIDC_PROVIDER_CONFIG_SCHEMA = {
         "issuer": {"type": "string"},
         "client_id": {"type": "string"},
         "client_secret": {"type": "string"},
+        "client_secret_jwt_key": {
+            "type": "object",
+            "required": ["jwt_header"],
+            "oneOf": [
+                {"required": ["key"]},
+                {"required": ["key_file"]},
+            ],
+            "properties": {
+                "key": {"type": "string"},
+                "key_file": {"type": "string"},
+                "jwt_header": {
+                    "type": "object",
+                    "required": ["alg"],
+                    "properties": {
+                        "alg": {"type": "string"},
+                    },
+                    "additionalProperties": {"type": "string"},
+                },
+                "jwt_payload": {
+                    "type": "object",
+                    "additionalProperties": {"type": "string"},
+                },
+            },
+        },
         "client_auth_method": {
             "type": "string",
             # the following list is the same as the keys of
@@ -404,15 +452,31 @@ def _parse_oidc_config_dict(
                 "idp_icon must be a valid MXC URI", config_path + ("idp_icon",)
             ) from e
 
+    client_secret_jwt_key_config = oidc_config.get("client_secret_jwt_key")
+    client_secret_jwt_key = None  # type: Optional[OidcProviderClientSecretJwtKey]
+    if client_secret_jwt_key_config is not None:
+        keyfile = client_secret_jwt_key_config.get("key_file")
+        if keyfile:
+            key = read_file(keyfile, config_path + ("client_secret_jwt_key",))
+        else:
+            key = client_secret_jwt_key_config["key"]
+        client_secret_jwt_key = OidcProviderClientSecretJwtKey(
+            key=key,
+            jwt_header=client_secret_jwt_key_config["jwt_header"],
+            jwt_payload=client_secret_jwt_key_config.get("jwt_payload", {}),
+        )
+
     return OidcProviderConfig(
         idp_id=idp_id,
         idp_name=oidc_config.get("idp_name", "OIDC"),
         idp_icon=idp_icon,
         idp_brand=oidc_config.get("idp_brand"),
+        unstable_idp_brand=oidc_config.get("unstable_idp_brand"),
         discover=oidc_config.get("discover", True),
         issuer=oidc_config["issuer"],
         client_id=oidc_config["client_id"],
-        client_secret=oidc_config["client_secret"],
+        client_secret=oidc_config.get("client_secret"),
+        client_secret_jwt_key=client_secret_jwt_key,
         client_auth_method=oidc_config.get("client_auth_method", "client_secret_basic"),
         scopes=oidc_config.get("scopes", ["openid"]),
         authorization_endpoint=oidc_config.get("authorization_endpoint"),
@@ -427,6 +491,18 @@ def _parse_oidc_config_dict(
     )
 
 
+@attr.s(slots=True, frozen=True)
+class OidcProviderClientSecretJwtKey:
+    # a pem-encoded signing key
+    key = attr.ib(type=str)
+
+    # properties to include in the JWT header
+    jwt_header = attr.ib(type=Mapping[str, str])
+
+    # properties to include in the JWT payload.
+    jwt_payload = attr.ib(type=Mapping[str, str])
+
+
 @attr.s(slots=True, frozen=True)
 class OidcProviderConfig:
     # a unique identifier for this identity provider. Used in the 'user_external_ids'
@@ -442,6 +518,9 @@ class OidcProviderConfig:
     # Optional brand identifier for this IdP.
     idp_brand = attr.ib(type=Optional[str])
 
+    # Optional brand identifier for the unstable API (see MSC2858).
+    unstable_idp_brand = attr.ib(type=Optional[str])
+
     # whether the OIDC discovery mechanism is used to discover endpoints
     discover = attr.ib(type=bool)
 
@@ -452,8 +531,13 @@ class OidcProviderConfig:
     # oauth2 client id to use
     client_id = attr.ib(type=str)
 
-    # oauth2 client secret to use
-    client_secret = attr.ib(type=str)
+    # oauth2 client secret to use. if `None`, use client_secret_jwt_key to generate
+    # a secret.
+    client_secret = attr.ib(type=Optional[str])
+
+    # key to use to construct a JWT to use as a client secret. May be `None` if
+    # `client_secret` is set.
+    client_secret_jwt_key = attr.ib(type=Optional[OidcProviderClientSecretJwtKey])
 
     # auth method to use when exchanging the token.
     # Valid values are 'client_secret_basic', 'client_secret_post' and

synapse/config/push.py

@@ -14,7 +14,7 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 
-from ._base import Config, ShardedWorkerHandlingConfig
+from ._base import Config
 
 
 class PushConfig(Config):
@@ -27,9 +27,6 @@ class PushConfig(Config):
             "group_unread_count_by_room", True
         )
 
-        pusher_instances = config.get("pusher_instances") or []
-        self.pusher_shard_config = ShardedWorkerHandlingConfig(pusher_instances)
-
         # There was a a 'redact_content' setting but mistakenly read from the
         # 'email'section'. Check for the flag in the 'push' section, and log,
         # but do not honour it to avoid nasty surprises when people upgrade.

synapse/config/ratelimiting.py

@@ -102,6 +102,16 @@ class RatelimitConfig(Config):
             defaults={"per_second": 0.01, "burst_count": 3},
         )
 
+        # Ratelimit cross-user key requests:
+        # * For local requests this is keyed by the sending device.
+        # * For requests received over federation this is keyed by the origin.
+        #
+        # Note that this isn't exposed in the configuration as it is obscure.
+        self.rc_key_requests = RateLimitConfig(
+            config.get("rc_key_requests", {}),
+            defaults={"per_second": 20, "burst_count": 100},
+        )
+
         self.rc_3pid_validation = RateLimitConfig(
             config.get("rc_3pid_validation") or {},
             defaults={"per_second": 0.003, "burst_count": 5},

synapse/config/repository.py

@@ -206,7 +206,6 @@ class ContentRepositoryConfig(Config):
 
     def generate_config_section(self, data_dir_path, **kwargs):
         media_store = os.path.join(data_dir_path, "media_store")
-        uploads_path = os.path.join(data_dir_path, "uploads")
 
         formatted_thumbnail_sizes = "".join(
             THUMBNAIL_SIZE_YAML % s for s in DEFAULT_THUMBNAIL_SIZES

synapse/config/server.py

@@ -263,6 +263,12 @@ class ServerConfig(Config):
             False,
         )
 
+        # Whether to retrieve and display profile data for a user when they
+        # are invited to a room
+        self.include_profile_data_on_invite = config.get(
+            "include_profile_data_on_invite", True
+        )
+
         if "restrict_public_rooms_to_local_users" in config and (
             "allow_public_rooms_without_auth" in config
             or "allow_public_rooms_over_federation" in config
@@ -391,7 +397,6 @@ class ServerConfig(Config):
         if self.public_baseurl is not None:
             if self.public_baseurl[-1] != "/":
                 self.public_baseurl += "/"
-        self.start_pushers = config.get("start_pushers", True)
 
         # (undocumented) option for torturing the worker-mode replication a bit,
         # for testing. The value defines the number of milliseconds to pause before
@@ -836,8 +841,7 @@ class ServerConfig(Config):
         # Whether to require authentication to retrieve profile data (avatars,
         # display names) of other users through the client API. Defaults to
         # 'false'. Note that profile data is also available via the federation
-        # API, so this setting is of limited value if federation is enabled on
-        # the server.
+        # API, unless allow_profile_lookup_over_federation is set to false.
         #
         #require_auth_for_profile_requests: true
 
@@ -848,6 +852,14 @@ class ServerConfig(Config):
         #
         #limit_profile_requests_to_users_who_share_rooms: true
 
+        # Uncomment to prevent a user's profile data from being retrieved and
+        # displayed in a room until they have joined it. By default, a user's
+        # profile data is included in an invite event, regardless of the values
+        # of the above two settings, and whether or not the users share a server.
+        # Defaults to 'true'.
+        #
+        #include_profile_data_on_invite: false
+
         # If set to 'true', removes the need for authentication to access the server's
         # public rooms directory through the client API, meaning that anyone can
         # query the room directory. Defaults to 'false'.

synapse/config/stats.py

@@ -13,10 +13,22 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 
-import sys
+import logging
 
 from ._base import Config
 
+ROOM_STATS_DISABLED_WARN = """\
+WARNING: room/user statistics have been disabled via the stats.enabled
+configuration setting. This means that certain features (such as the room
+directory) will not operate correctly. Future versions of Synapse may ignore
+this setting.
+
+To fix this warning, remove the stats.enabled setting from your configuration
+file.
+--------------------------------------------------------------------------------"""
+
+logger = logging.getLogger(__name__)
+
 
 class StatsConfig(Config):
     """Stats Configuration
@@ -28,30 +40,29 @@ class StatsConfig(Config):
     def read_config(self, config, **kwargs):
         self.stats_enabled = True
         self.stats_bucket_size = 86400 * 1000
-        self.stats_retention = sys.maxsize
         stats_config = config.get("stats", None)
         if stats_config:
             self.stats_enabled = stats_config.get("enabled", self.stats_enabled)
             self.stats_bucket_size = self.parse_duration(
                 stats_config.get("bucket_size", "1d")
             )
-            self.stats_retention = self.parse_duration(
-                stats_config.get("retention", "%ds" % (sys.maxsize,))
-            )
+        if not self.stats_enabled:
+            logger.warning(ROOM_STATS_DISABLED_WARN)
 
     def generate_config_section(self, config_dir_path, server_name, **kwargs):
         return """
-        # Local statistics collection. Used in populating the room directory.
+        # Settings for local room and user statistics collection. See
+        # docs/room_and_user_statistics.md.
         #
-        # 'bucket_size' controls how large each statistics timeslice is. It can
-        # be defined in a human readable short form -- e.g. "1d", "1y".
-        #
-        # 'retention' controls how long historical statistics will be kept for.
-        # It can be defined in a human readable short form -- e.g. "1d", "1y".
-        #
-        #
-        #stats:
-        #   enabled: true
-        #   bucket_size: 1d
-        #   retention: 1y
+        stats:
+          # Uncomment the following to disable room and user statistics. Note that doing
+          # so may cause certain features (such as the room directory) not to work
+          # correctly.
+          #
+          #enabled: false
+
+          # The size of each timeslice in the room_stats_historical and
+          # user_stats_historical tables, as a time period. Defaults to "1d".
+          #
+          #bucket_size: 1h
         """

synapse/config/user_directory.py

@@ -24,32 +24,46 @@ class UserDirectoryConfig(Config):
     section = "userdirectory"
 
     def read_config(self, config, **kwargs):
-        self.user_directory_search_enabled = True
-        self.user_directory_search_all_users = False
-        user_directory_config = config.get("user_directory", None)
-        if user_directory_config:
-            self.user_directory_search_enabled = user_directory_config.get(
-                "enabled", True
-            )
-            self.user_directory_search_all_users = user_directory_config.get(
-                "search_all_users", False
-            )
+        user_directory_config = config.get("user_directory") or {}
+        self.user_directory_search_enabled = user_directory_config.get("enabled", True)
+        self.user_directory_search_all_users = user_directory_config.get(
+            "search_all_users", False
+        )
+        self.user_directory_search_prefer_local_users = user_directory_config.get(
+            "prefer_local_users", False
+        )
 
     def generate_config_section(self, config_dir_path, server_name, **kwargs):
         return """
         # User Directory configuration
         #
-        # 'enabled' defines whether users can search the user directory. If
-        # false then empty responses are returned to all queries. Defaults to
-        # true.
-        #
-        # 'search_all_users' defines whether to search all users visible to your HS
-        # when searching the user directory, rather than limiting to users visible
-        # in public rooms.  Defaults to false.  If you set it True, you'll have to
-        # rebuild the user_directory search indexes, see
-        # https://github.com/matrix-org/synapse/blob/master/docs/user_directory.md
-        #
-        #user_directory:
-        #  enabled: true
-        #  search_all_users: false
+        user_directory:
+            # Defines whether users can search the user directory. If false then
+            # empty responses are returned to all queries. Defaults to true.
+            #
+            # Uncomment to disable the user directory.
+            #
+            #enabled: false
+
+            # Defines whether to search all users visible to your HS when searching
+            # the user directory, rather than limiting to users visible in public
+            # rooms. Defaults to false.
+            #
+            # If you set it true, you'll have to rebuild the user_directory search
+            # indexes, see:
+            # https://github.com/matrix-org/synapse/blob/master/docs/user_directory.md
+            #
+            # Uncomment to return search results containing all known users, even if that
+            # user does not share a room with the requester.
+            #
+            #search_all_users: true
+
+            # Defines whether to prefer local users in search query results.
+            # If True, local users are more likely to appear above remote users
+            # when searching the user directory. Defaults to false.
+            #
+            # Uncomment to prefer local over remote users in user directory search
+            # results.
+            #
+            #prefer_local_users: true
         """

synapse/config/workers.py

@@ -17,9 +17,28 @@ from typing import List, Union
 
 import attr
 
-from ._base import Config, ConfigError, ShardedWorkerHandlingConfig
+from ._base import (
+    Config,
+    ConfigError,
+    RoutableShardedWorkerHandlingConfig,
+    ShardedWorkerHandlingConfig,
+)
 from .server import ListenerConfig, parse_listener_def
 
+_FEDERATION_SENDER_WITH_SEND_FEDERATION_ENABLED_ERROR = """
+The send_federation config option must be disabled in the main
+synapse process before they can be run in a separate worker.
+
+Please add ``send_federation: false`` to the main config
+"""
+
+_PUSHER_WITH_START_PUSHERS_ENABLED_ERROR = """
+The start_pushers config option must be disabled in the main
+synapse process before they can be run in a separate worker.
+
+Please add ``start_pushers: false`` to the main config
+"""
+
 
 def _instance_to_list_converter(obj: Union[str, List[str]]) -> List[str]:
     """Helper for allowing parsing a string or list of strings to a config
@@ -103,6 +122,7 @@ class WorkerConfig(Config):
         self.worker_replication_secret = config.get("worker_replication_secret", None)
 
         self.worker_name = config.get("worker_name", self.worker_app)
+        self.instance_name = self.worker_name or "master"
 
         self.worker_main_http_uri = config.get("worker_main_http_uri", None)
 
@@ -118,12 +138,41 @@ class WorkerConfig(Config):
                 )
             )
 
-        # Whether to send federation traffic out in this process. This only
-        # applies to some federation traffic, and so shouldn't be used to
-        # "disable" federation
-        self.send_federation = config.get("send_federation", True)
+        # Handle federation sender configuration.
+        #
+        # There are two ways of configuring which instances handle federation
+        # sending:
+        #   1. The old way where "send_federation" is set to false and running a
+        #      `synapse.app.federation_sender` worker app.
+        #   2. Specifying the workers sending federation in
+        #      `federation_sender_instances`.
+        #
 
-        federation_sender_instances = config.get("federation_sender_instances") or []
+        send_federation = config.get("send_federation", True)
+
+        federation_sender_instances = config.get("federation_sender_instances")
+        if federation_sender_instances is None:
+            # Default to an empty list, which means "another, unknown, worker is
+            # responsible for it".
+            federation_sender_instances = []
+
+            # If no federation sender instances are set we check if
+            # `send_federation` is set, which means use master
+            if send_federation:
+                federation_sender_instances = ["master"]
+
+            if self.worker_app == "synapse.app.federation_sender":
+                if send_federation:
+                    # If we're running federation senders, and not using
+                    # `federation_sender_instances`, then we should have
+                    # explicitly set `send_federation` to false.
+                    raise ConfigError(
+                        _FEDERATION_SENDER_WITH_SEND_FEDERATION_ENABLED_ERROR
+                    )
+
+                federation_sender_instances = [self.worker_name]
+
+        self.send_federation = self.instance_name in federation_sender_instances
         self.federation_shard_config = ShardedWorkerHandlingConfig(
             federation_sender_instances
         )
@@ -164,7 +213,37 @@ class WorkerConfig(Config):
                 "Must only specify one instance to handle `receipts` messages."
             )
 
-        self.events_shard_config = ShardedWorkerHandlingConfig(self.writers.events)
+        if len(self.writers.events) == 0:
+            raise ConfigError("Must specify at least one instance to handle `events`.")
+
+        self.events_shard_config = RoutableShardedWorkerHandlingConfig(
+            self.writers.events
+        )
+
+        # Handle sharded push
+        start_pushers = config.get("start_pushers", True)
+        pusher_instances = config.get("pusher_instances")
+        if pusher_instances is None:
+            # Default to an empty list, which means "another, unknown, worker is
+            # responsible for it".
+            pusher_instances = []
+
+            # If no pushers instances are set we check if `start_pushers` is
+            # set, which means use master
+            if start_pushers:
+                pusher_instances = ["master"]
+
+            if self.worker_app == "synapse.app.pusher":
+                if start_pushers:
+                    # If we're running pushers, and not using
+                    # `pusher_instances`, then we should have explicitly set
+                    # `start_pushers` to false.
+                    raise ConfigError(_PUSHER_WITH_START_PUSHERS_ENABLED_ERROR)
+
+                pusher_instances = [self.instance_name]
+
+        self.start_pushers = self.instance_name in pusher_instances
+        self.pusher_shard_config = ShardedWorkerHandlingConfig(pusher_instances)
 
         # Whether this worker should run background tasks or not.
         #

synapse/events/spamcheck.py

@@ -15,6 +15,7 @@
 # limitations under the License.
 
 import inspect
+import logging
 from typing import TYPE_CHECKING, Any, Dict, List, Optional, Tuple, Union
 
 from synapse.rest.media.v1._base import FileInfo
@@ -27,6 +28,8 @@ if TYPE_CHECKING:
     import synapse.events
     import synapse.server
 
+logger = logging.getLogger(__name__)
+
 
 class SpamChecker:
     def __init__(self, hs: "synapse.server.HomeServer"):
@@ -190,6 +193,7 @@ class SpamChecker:
         email_threepid: Optional[dict],
         username: Optional[str],
         request_info: Collection[Tuple[str, str]],
+        auth_provider_id: Optional[str] = None,
     ) -> RegistrationBehaviour:
         """Checks if we should allow the given registration request.
 
@@ -198,6 +202,9 @@ class SpamChecker:
             username: The request user name, if any
             request_info: List of tuples of user agent and IP that
                 were used during the registration process.
+            auth_provider_id: The SSO IdP the user used, e.g "oidc", "saml",
+                "cas". If any. Note this does not include users registered
+                via a password provider.
 
         Returns:
             Enum for how the request should be handled
@@ -208,9 +215,25 @@ class SpamChecker:
             # spam checker
             checker = getattr(spam_checker, "check_registration_for_spam", None)
             if checker:
-                behaviour = await maybe_awaitable(
-                    checker(email_threepid, username, request_info)
-                )
+                # Provide auth_provider_id if the function supports it
+                checker_args = inspect.signature(checker)
+                if len(checker_args.parameters) == 4:
+                    d = checker(
+                        email_threepid,
+                        username,
+                        request_info,
+                        auth_provider_id,
+                    )
+                elif len(checker_args.parameters) == 3:
+                    d = checker(email_threepid, username, request_info)
+                else:
+                    logger.error(
+                        "Invalid signature for %s.check_registration_for_spam. Denying registration",
+                        spam_checker.__module__,
+                    )
+                    return RegistrationBehaviour.DENY
+
+                behaviour = await maybe_awaitable(d)
                 assert isinstance(behaviour, RegistrationBehaviour)
                 if behaviour != RegistrationBehaviour.ALLOW:
                     return behaviour

synapse/federation/federation_server.py

@@ -22,6 +22,7 @@ from typing import (
     Awaitable,
     Callable,
     Dict,
+    Iterable,
     List,
     Optional,
     Tuple,
@@ -34,7 +35,7 @@ from twisted.internet import defer
 from twisted.internet.abstract import isIPAddress
 from twisted.python import failure
 
-from synapse.api.constants import EventTypes, Membership
+from synapse.api.constants import EduTypes, EventTypes, Membership
 from synapse.api.errors import (
     AuthError,
     Codes,
@@ -44,6 +45,7 @@ from synapse.api.errors import (
     SynapseError,
     UnsupportedRoomVersionError,
 )
+from synapse.api.ratelimiting import Ratelimiter
 from synapse.api.room_versions import KNOWN_ROOM_VERSIONS
 from synapse.events import EventBase
 from synapse.federation.federation_base import FederationBase, event_from_pdu_json
@@ -89,16 +91,15 @@ pdu_process_time = Histogram(
     "Time taken to process an event",
 )
 
-
-last_pdu_age_metric = Gauge(
-    "synapse_federation_last_received_pdu_age",
-    "The age (in seconds) of the last PDU successfully received from the given domain",
+last_pdu_ts_metric = Gauge(
+    "synapse_federation_last_received_pdu_time",
+    "The timestamp of the last PDU which was successfully received from the given domain",
     labelnames=("server_name",),
 )
 
 
 class FederationServer(FederationBase):
-    def __init__(self, hs):
+    def __init__(self, hs: "HomeServer"):
         super().__init__(hs)
 
         self.auth = hs.get_auth()
@@ -111,14 +112,15 @@ class FederationServer(FederationBase):
         # with FederationHandlerRegistry.
         hs.get_directory_handler()
 
-        self._federation_ratelimiter = hs.get_federation_ratelimiter()
-
         self._server_linearizer = Linearizer("fed_server")
-        self._transaction_linearizer = Linearizer("fed_txn_handler")
+
+        # origins that we are currently processing a transaction from.
+        # a dict from origin to txn id.
+        self._active_transactions = {}  # type: Dict[str, str]
 
         # We cache results for transaction with the same ID
         self._transaction_resp_cache = ResponseCache(
-            hs, "fed_txn_handler", timeout_ms=30000
+            hs.get_clock(), "fed_txn_handler", timeout_ms=30000
         )  # type: ResponseCache[Tuple[str, str]]
 
         self.transaction_actions = TransactionActions(self.store)
@@ -128,10 +130,10 @@ class FederationServer(FederationBase):
         # We cache responses to state queries, as they take a while and often
         # come in waves.
         self._state_resp_cache = ResponseCache(
-            hs, "state_resp", timeout_ms=30000
+            hs.get_clock(), "state_resp", timeout_ms=30000
         )  # type: ResponseCache[Tuple[str, str]]
         self._state_ids_resp_cache = ResponseCache(
-            hs, "state_ids_resp", timeout_ms=30000
+            hs.get_clock(), "state_ids_resp", timeout_ms=30000
         )  # type: ResponseCache[Tuple[str, str]]
 
         self._federation_metrics_domains = (
@@ -168,6 +170,33 @@ class FederationServer(FederationBase):
 
         logger.debug("[%s] Got transaction", transaction_id)
 
+        # Reject malformed transactions early: reject if too many PDUs/EDUs
+        if len(transaction.pdus) > 50 or (  # type: ignore
+            hasattr(transaction, "edus") and len(transaction.edus) > 100  # type: ignore
+        ):
+            logger.info("Transaction PDU or EDU count too large. Returning 400")
+            return 400, {}
+
+        # we only process one transaction from each origin at a time. We need to do
+        # this check here, rather than in _on_incoming_transaction_inner so that we
+        # don't cache the rejection in _transaction_resp_cache (so that if the txn
+        # arrives again later, we can process it).
+        current_transaction = self._active_transactions.get(origin)
+        if current_transaction and current_transaction != transaction_id:
+            logger.warning(
+                "Received another txn %s from %s while still processing %s",
+                transaction_id,
+                origin,
+                current_transaction,
+            )
+            return 429, {
+                "errcode": Codes.UNKNOWN,
+                "error": "Too many concurrent transactions",
+            }
+
+        # CRITICAL SECTION: we must now not await until we populate _active_transactions
+        # in _on_incoming_transaction_inner.
+
         # We wrap in a ResponseCache so that we de-duplicate retried
         # transactions.
         return await self._transaction_resp_cache.wrap(
@@ -181,26 +210,18 @@ class FederationServer(FederationBase):
     async def _on_incoming_transaction_inner(
         self, origin: str, transaction: Transaction, request_time: int
     ) -> Tuple[int, Dict[str, Any]]:
-        # Use a linearizer to ensure that transactions from a remote are
-        # processed in order.
-        with await self._transaction_linearizer.queue(origin):
-            # We rate limit here *after* we've queued up the incoming requests,
-            # so that we don't fill up the ratelimiter with blocked requests.
-            #
-            # This is important as the ratelimiter allows N concurrent requests
-            # at a time, and only starts ratelimiting if there are more requests
-            # than that being processed at a time. If we queued up requests in
-            # the linearizer/response cache *after* the ratelimiting then those
-            # queued up requests would count as part of the allowed limit of N
-            # concurrent requests.
-            with self._federation_ratelimiter.ratelimit(origin) as d:
-                await d
+        # CRITICAL SECTION: the first thing we must do (before awaiting) is
+        # add an entry to _active_transactions.
+        assert origin not in self._active_transactions
+        self._active_transactions[origin] = transaction.transaction_id  # type: ignore
 
-                result = await self._handle_incoming_transaction(
-                    origin, transaction, request_time
-                )
-
-        return result
+        try:
+            result = await self._handle_incoming_transaction(
+                origin, transaction, request_time
+            )
+            return result
+        finally:
+            del self._active_transactions[origin]
 
     async def _handle_incoming_transaction(
         self, origin: str, transaction: Transaction, request_time: int
@@ -226,19 +247,6 @@ class FederationServer(FederationBase):
 
         logger.debug("[%s] Transaction is new", transaction.transaction_id)  # type: ignore
 
-        # Reject if PDU count > 50 or EDU count > 100
-        if len(transaction.pdus) > 50 or (  # type: ignore
-            hasattr(transaction, "edus") and len(transaction.edus) > 100  # type: ignore
-        ):
-
-            logger.info("Transaction PDU or EDU count too large. Returning 400")
-
-            response = {}
-            await self.transaction_actions.set_response(
-                origin, transaction, 400, response
-            )
-            return 400, response
-
         # We process PDUs and EDUs in parallel. This is important as we don't
         # want to block things like to device messages from reaching clients
         # behind the potentially expensive handling of PDUs.
@@ -334,42 +342,48 @@ class FederationServer(FederationBase):
         # impose a limit to avoid going too crazy with ram/cpu.
 
         async def process_pdus_for_room(room_id: str):
-            logger.debug("Processing PDUs for %s", room_id)
-            try:
-                await self.check_server_matches_acl(origin_host, room_id)
-            except AuthError as e:
-                logger.warning("Ignoring PDUs for room %s from banned server", room_id)
-                for pdu in pdus_by_room[room_id]:
-                    event_id = pdu.event_id
-                    pdu_results[event_id] = e.error_dict()
-                return
+            with nested_logging_context(room_id):
+                logger.debug("Processing PDUs for %s", room_id)
 
-            for pdu in pdus_by_room[room_id]:
-                event_id = pdu.event_id
-                with pdu_process_time.time():
-                    with nested_logging_context(event_id):
-                        try:
-                            await self._handle_received_pdu(origin, pdu)
-                            pdu_results[event_id] = {}
-                        except FederationError as e:
-                            logger.warning("Error handling PDU %s: %s", event_id, e)
-                            pdu_results[event_id] = {"error": str(e)}
-                        except Exception as e:
-                            f = failure.Failure()
-                            pdu_results[event_id] = {"error": str(e)}
-                            logger.error(
-                                "Failed to handle PDU %s",
-                                event_id,
-                                exc_info=(f.type, f.value, f.getTracebackObject()),
-                            )
+                try:
+                    await self.check_server_matches_acl(origin_host, room_id)
+                except AuthError as e:
+                    logger.warning(
+                        "Ignoring PDUs for room %s from banned server", room_id
+                    )
+                    for pdu in pdus_by_room[room_id]:
+                        event_id = pdu.event_id
+                        pdu_results[event_id] = e.error_dict()
+                    return
+
+                for pdu in pdus_by_room[room_id]:
+                    pdu_results[pdu.event_id] = await process_pdu(pdu)
+
+        async def process_pdu(pdu: EventBase) -> JsonDict:
+            event_id = pdu.event_id
+            with pdu_process_time.time():
+                with nested_logging_context(event_id):
+                    try:
+                        await self._handle_received_pdu(origin, pdu)
+                        return {}
+                    except FederationError as e:
+                        logger.warning("Error handling PDU %s: %s", event_id, e)
+                        return {"error": str(e)}
+                    except Exception as e:
+                        f = failure.Failure()
+                        logger.error(
+                            "Failed to handle PDU %s",
+                            event_id,
+                            exc_info=(f.type, f.value, f.getTracebackObject()),  # type: ignore
+                        )
+                        return {"error": str(e)}
 
         await concurrently_execute(
             process_pdus_for_room, pdus_by_room.keys(), TRANSACTION_CONCURRENCY_LIMIT
         )
 
         if newest_pdu_ts and origin in self._federation_metrics_domains:
-            newest_pdu_age = self._clock.time_msec() - newest_pdu_ts
-            last_pdu_age_metric.labels(server_name=origin).set(newest_pdu_age / 1000)
+            last_pdu_ts_metric.labels(server_name=origin).set(newest_pdu_ts / 1000)
 
         return pdu_results
 
@@ -447,18 +461,22 @@ class FederationServer(FederationBase):
 
     async def _on_state_ids_request_compute(self, room_id, event_id):
         state_ids = await self.handler.get_state_ids_for_pdu(room_id, event_id)
-        auth_chain_ids = await self.store.get_auth_chain_ids(state_ids)
+        auth_chain_ids = await self.store.get_auth_chain_ids(room_id, state_ids)
         return {"pdu_ids": state_ids, "auth_chain_ids": auth_chain_ids}
 
     async def _on_context_state_request_compute(
         self, room_id: str, event_id: str
     ) -> Dict[str, list]:
         if event_id:
-            pdus = await self.handler.get_state_for_pdu(room_id, event_id)
+            pdus = await self.handler.get_state_for_pdu(
+                room_id, event_id
+            )  # type: Iterable[EventBase]
         else:
             pdus = (await self.state.get_current_state(room_id)).values()
 
-        auth_chain = await self.store.get_auth_chain([pdu.event_id for pdu in pdus])
+        auth_chain = await self.store.get_auth_chain(
+            room_id, [pdu.event_id for pdu in pdus]
+        )
 
         return {
             "pdus": [pdu.get_pdu_json() for pdu in pdus],
@@ -862,13 +880,22 @@ class FederationHandlerRegistry:
         self.edu_handlers = (
             {}
         )  # type: Dict[str, Callable[[str, dict], Awaitable[None]]]
-        self.query_handlers = {}  # type: Dict[str, Callable[[dict], Awaitable[None]]]
+        self.query_handlers = (
+            {}
+        )  # type: Dict[str, Callable[[dict], Awaitable[JsonDict]]]
 
         # Map from type to instance names that we should route EDU handling to.
         # We randomly choose one instance from the list to route to for each new
         # EDU received.
         self._edu_type_to_instance = {}  # type: Dict[str, List[str]]
 
+        # A rate limiter for incoming room key requests per origin.
+        self._room_key_request_rate_limiter = Ratelimiter(
+            clock=self.clock,
+            rate_hz=self.config.rc_key_requests.per_second,
+            burst_count=self.config.rc_key_requests.burst_count,
+        )
+
     def register_edu_handler(
         self, edu_type: str, handler: Callable[[str, JsonDict], Awaitable[None]]
     ):
@@ -889,7 +916,7 @@ class FederationHandlerRegistry:
         self.edu_handlers[edu_type] = handler
 
     def register_query_handler(
-        self, query_type: str, handler: Callable[[dict], defer.Deferred]
+        self, query_type: str, handler: Callable[[dict], Awaitable[JsonDict]]
     ):
         """Sets the handler callable that will be used to handle an incoming
         federation query of the given type.
@@ -917,7 +944,15 @@ class FederationHandlerRegistry:
         self._edu_type_to_instance[edu_type] = instance_names
 
     async def on_edu(self, edu_type: str, origin: str, content: dict):
-        if not self.config.use_presence and edu_type == "m.presence":
+        if not self.config.use_presence and edu_type == EduTypes.Presence:
+            return
+
+        # If the incoming room key requests from a particular origin are over
+        # the limit, drop them.
+        if (
+            edu_type == EduTypes.RoomKeyRequest
+            and not self._room_key_request_rate_limiter.can_do_action(origin)
+        ):
             return
 
         # Check if we have a handler on this instance
@@ -954,7 +989,7 @@ class FederationHandlerRegistry:
         # Oh well, let's just log and move on.
         logger.warning("No handler registered for EDU type %s", edu_type)
 
-    async def on_query(self, query_type: str, args: dict):
+    async def on_query(self, query_type: str, args: dict) -> JsonDict:
         handler = self.query_handlers.get(query_type)
         if handler:
             return await handler(args)

synapse/federation/sender/__init__.py

@@ -474,7 +474,7 @@ class FederationSender:
             self._processing_pending_presence = False
 
     def send_presence_to_destinations(
-        self, states: List[UserPresenceState], destinations: List[str]
+        self, states: Iterable[UserPresenceState], destinations: Iterable[str]
     ) -> None:
         """Send the given presence states to the given destinations.
         destinations (list[str])

synapse/federation/sender/per_destination_queue.py

@@ -17,6 +17,7 @@ import datetime
 import logging
 from typing import TYPE_CHECKING, Dict, Hashable, Iterable, List, Optional, Tuple, cast
 
+import attr
 from prometheus_client import Counter
 
 from synapse.api.errors import (
@@ -93,6 +94,10 @@ class PerDestinationQueue:
         self._destination = destination
         self.transmission_loop_running = False
 
+        # Flag to signal to any running transmission loop that there is new data
+        # queued up to be sent.
+        self._new_data_to_send = False
+
         # True whilst we are sending events that the remote homeserver missed
         # because it was unreachable. We start in this state so we can perform
         # catch-up at startup.
@@ -108,7 +113,7 @@ class PerDestinationQueue:
         # destination (we are the only updater so this is safe)
         self._last_successful_stream_ordering = None  # type: Optional[int]
 
-        # a list of pending PDUs
+        # a queue of pending PDUs
         self._pending_pdus = []  # type: List[EventBase]
 
         # XXX this is never actually used: see
@@ -208,6 +213,10 @@ class PerDestinationQueue:
         transaction in the background.
         """
 
+        # Mark that we (may) have new things to send, so that any running
+        # transmission loop will recheck whether there is stuff to send.
+        self._new_data_to_send = True
+
         if self.transmission_loop_running:
             # XXX: this can get stuck on by a never-ending
             # request at which point pending_pdus just keeps growing.
@@ -250,125 +259,41 @@ class PerDestinationQueue:
 
             pending_pdus = []
             while True:
-                # We have to keep 2 free slots for presence and rr_edus
-                limit = MAX_EDUS_PER_TRANSACTION - 2
+                self._new_data_to_send = False
 
-                device_update_edus, dev_list_id = await self._get_device_update_edus(
-                    limit
-                )
-
-                limit -= len(device_update_edus)
-
-                (
-                    to_device_edus,
-                    device_stream_id,
-                ) = await self._get_to_device_message_edus(limit)
-
-                pending_edus = device_update_edus + to_device_edus
-
-                # BEGIN CRITICAL SECTION
-                #
-                # In order to avoid a race condition, we need to make sure that
-                # the following code (from popping the queues up to the point
-                # where we decide if we actually have any pending messages) is
-                # atomic - otherwise new PDUs or EDUs might arrive in the
-                # meantime, but not get sent because we hold the
-                # transmission_loop_running flag.
-
-                pending_pdus = self._pending_pdus
-
-                # We can only include at most 50 PDUs per transactions
-                pending_pdus, self._pending_pdus = pending_pdus[:50], pending_pdus[50:]
-
-                pending_edus.extend(self._get_rr_edus(force_flush=False))
-                pending_presence = self._pending_presence
-                self._pending_presence = {}
-                if pending_presence:
-                    pending_edus.append(
-                        Edu(
-                            origin=self._server_name,
-                            destination=self._destination,
-                            edu_type="m.presence",
-                            content={
-                                "push": [
-                                    format_user_presence_state(
-                                        presence, self._clock.time_msec()
-                                    )
-                                    for presence in pending_presence.values()
-                                ]
-                            },
-                        )
-                    )
-
-                pending_edus.extend(
-                    self._pop_pending_edus(MAX_EDUS_PER_TRANSACTION - len(pending_edus))
-                )
-                while (
-                    len(pending_edus) < MAX_EDUS_PER_TRANSACTION
-                    and self._pending_edus_keyed
+                async with _TransactionQueueManager(self) as (
+                    pending_pdus,
+                    pending_edus,
                 ):
-                    _, val = self._pending_edus_keyed.popitem()
-                    pending_edus.append(val)
+                    if not pending_pdus and not pending_edus:
+                        logger.debug("TX [%s] Nothing to send", self._destination)
 
-                if pending_pdus:
-                    logger.debug(
-                        "TX [%s] len(pending_pdus_by_dest[dest]) = %d",
-                        self._destination,
-                        len(pending_pdus),
+                        # If we've gotten told about new things to send during
+                        # checking for things to send, we try looking again.
+                        # Otherwise new PDUs or EDUs might arrive in the meantime,
+                        # but not get sent because we hold the
+                        # `transmission_loop_running` flag.
+                        if self._new_data_to_send:
+                            continue
+                        else:
+                            return
+
+                    if pending_pdus:
+                        logger.debug(
+                            "TX [%s] len(pending_pdus_by_dest[dest]) = %d",
+                            self._destination,
+                            len(pending_pdus),
+                        )
+
+                    await self._transaction_manager.send_new_transaction(
+                        self._destination, pending_pdus, pending_edus
                     )
 
-                if not pending_pdus and not pending_edus:
-                    logger.debug("TX [%s] Nothing to send", self._destination)
-                    self._last_device_stream_id = device_stream_id
-                    return
-
-                # if we've decided to send a transaction anyway, and we have room, we
-                # may as well send any pending RRs
-                if len(pending_edus) < MAX_EDUS_PER_TRANSACTION:
-                    pending_edus.extend(self._get_rr_edus(force_flush=True))
-
-                # END CRITICAL SECTION
-
-                success = await self._transaction_manager.send_new_transaction(
-                    self._destination, pending_pdus, pending_edus
-                )
-                if success:
                     sent_transactions_counter.inc()
                     sent_edus_counter.inc(len(pending_edus))
                     for edu in pending_edus:
                         sent_edus_by_type.labels(edu.edu_type).inc()
-                    # Remove the acknowledged device messages from the database
-                    # Only bother if we actually sent some device messages
-                    if to_device_edus:
-                        await self._store.delete_device_msgs_for_remote(
-                            self._destination, device_stream_id
-                        )
 
-                    # also mark the device updates as sent
-                    if device_update_edus:
-                        logger.info(
-                            "Marking as sent %r %r", self._destination, dev_list_id
-                        )
-                        await self._store.mark_as_sent_devices_by_remote(
-                            self._destination, dev_list_id
-                        )
-
-                    self._last_device_stream_id = device_stream_id
-                    self._last_device_list_stream_id = dev_list_id
-
-                    if pending_pdus:
-                        # we sent some PDUs and it was successful, so update our
-                        # last_successful_stream_ordering in the destinations table.
-                        final_pdu = pending_pdus[-1]
-                        last_successful_stream_ordering = (
-                            final_pdu.internal_metadata.stream_ordering
-                        )
-                        assert last_successful_stream_ordering
-                        await self._store.set_destination_last_successful_stream_ordering(
-                            self._destination, last_successful_stream_ordering
-                        )
-                else:
-                    break
         except NotRetryingDestination as e:
             logger.debug(
                 "TX [%s] not ready for retry yet (next retry at %s) - "
@@ -401,7 +326,7 @@ class PerDestinationQueue:
                 self._pending_presence = {}
                 self._pending_rrs = {}
 
-            self._start_catching_up()
+                self._start_catching_up()
         except FederationDeniedError as e:
             logger.info(e)
         except HttpResponseException as e:
@@ -412,7 +337,6 @@ class PerDestinationQueue:
                 e,
             )
 
-            self._start_catching_up()
         except RequestSendFailed as e:
             logger.warning(
                 "TX [%s] Failed to send transaction: %s", self._destination, e
@@ -422,16 +346,12 @@ class PerDestinationQueue:
                 logger.info(
                     "Failed to send event %s to %s", p.event_id, self._destination
                 )
-
-            self._start_catching_up()
         except Exception:
             logger.exception("TX [%s] Failed to send transaction", self._destination)
             for p in pending_pdus:
                 logger.info(
                     "Failed to send event %s to %s", p.event_id, self._destination
                 )
-
-            self._start_catching_up()
         finally:
             # We want to be *very* sure we clear this after we stop processing
             self.transmission_loop_running = False
@@ -499,13 +419,10 @@ class PerDestinationQueue:
                 rooms = [p.room_id for p in catchup_pdus]
                 logger.info("Catching up rooms to %s: %r", self._destination, rooms)
 
-            success = await self._transaction_manager.send_new_transaction(
+            await self._transaction_manager.send_new_transaction(
                 self._destination, catchup_pdus, []
             )
 
-            if not success:
-                return
-
             sent_transactions_counter.inc()
             final_pdu = catchup_pdus[-1]
             self._last_successful_stream_ordering = cast(
@@ -584,3 +501,135 @@ class PerDestinationQueue:
         """
         self._catching_up = True
         self._pending_pdus = []
+
+
+@attr.s(slots=True)
+class _TransactionQueueManager:
+    """A helper async context manager for pulling stuff off the queues and
+    tracking what was last successfully sent, etc.
+    """
+
+    queue = attr.ib(type=PerDestinationQueue)
+
+    _device_stream_id = attr.ib(type=Optional[int], default=None)
+    _device_list_id = attr.ib(type=Optional[int], default=None)
+    _last_stream_ordering = attr.ib(type=Optional[int], default=None)
+    _pdus = attr.ib(type=List[EventBase], factory=list)
+
+    async def __aenter__(self) -> Tuple[List[EventBase], List[Edu]]:
+        # First we calculate the EDUs we want to send, if any.
+
+        # We start by fetching device related EDUs, i.e device updates and to
+        # device messages. We have to keep 2 free slots for presence and rr_edus.
+        limit = MAX_EDUS_PER_TRANSACTION - 2
+
+        device_update_edus, dev_list_id = await self.queue._get_device_update_edus(
+            limit
+        )
+
+        if device_update_edus:
+            self._device_list_id = dev_list_id
+        else:
+            self.queue._last_device_list_stream_id = dev_list_id
+
+        limit -= len(device_update_edus)
+
+        (
+            to_device_edus,
+            device_stream_id,
+        ) = await self.queue._get_to_device_message_edus(limit)
+
+        if to_device_edus:
+            self._device_stream_id = device_stream_id
+        else:
+            self.queue._last_device_stream_id = device_stream_id
+
+        pending_edus = device_update_edus + to_device_edus
+
+        # Now add the read receipt EDU.
+        pending_edus.extend(self.queue._get_rr_edus(force_flush=False))
+
+        # And presence EDU.
+        if self.queue._pending_presence:
+            pending_edus.append(
+                Edu(
+                    origin=self.queue._server_name,
+                    destination=self.queue._destination,
+                    edu_type="m.presence",
+                    content={
+                        "push": [
+                            format_user_presence_state(
+                                presence, self.queue._clock.time_msec()
+                            )
+                            for presence in self.queue._pending_presence.values()
+                        ]
+                    },
+                )
+            )
+            self.queue._pending_presence = {}
+
+        # Finally add any other types of EDUs if there is room.
+        pending_edus.extend(
+            self.queue._pop_pending_edus(MAX_EDUS_PER_TRANSACTION - len(pending_edus))
+        )
+        while (
+            len(pending_edus) < MAX_EDUS_PER_TRANSACTION
+            and self.queue._pending_edus_keyed
+        ):
+            _, val = self.queue._pending_edus_keyed.popitem()
+            pending_edus.append(val)
+
+        # Now we look for any PDUs to send, by getting up to 50 PDUs from the
+        # queue
+        self._pdus = self.queue._pending_pdus[:50]
+
+        if not self._pdus and not pending_edus:
+            return [], []
+
+        # if we've decided to send a transaction anyway, and we have room, we
+        # may as well send any pending RRs
+        if len(pending_edus) < MAX_EDUS_PER_TRANSACTION:
+            pending_edus.extend(self.queue._get_rr_edus(force_flush=True))
+
+        if self._pdus:
+            self._last_stream_ordering = self._pdus[
+                -1
+            ].internal_metadata.stream_ordering
+            assert self._last_stream_ordering
+
+        return self._pdus, pending_edus
+
+    async def __aexit__(self, exc_type, exc, tb):
+        if exc_type is not None:
+            # Failed to send transaction, so we bail out.
+            return
+
+        # Successfully sent transactions, so we remove pending PDUs from the queue
+        if self._pdus:
+            self.queue._pending_pdus = self.queue._pending_pdus[len(self._pdus) :]
+
+        # Succeeded to send the transaction so we record where we have sent up
+        # to in the various streams
+
+        if self._device_stream_id:
+            await self.queue._store.delete_device_msgs_for_remote(
+                self.queue._destination, self._device_stream_id
+            )
+            self.queue._last_device_stream_id = self._device_stream_id
+
+        # also mark the device updates as sent
+        if self._device_list_id:
+            logger.info(
+                "Marking as sent %r %r", self.queue._destination, self._device_list_id
+            )
+            await self.queue._store.mark_as_sent_devices_by_remote(
+                self.queue._destination, self._device_list_id
+            )
+            self.queue._last_device_list_stream_id = self._device_list_id
+
+        if self._last_stream_ordering:
+            # we sent some PDUs and it was successful, so update our
+            # last_successful_stream_ordering in the destinations table.
+            await self.queue._store.set_destination_last_successful_stream_ordering(
+                self.queue._destination, self._last_stream_ordering
+            )

synapse/federation/sender/transaction_manager.py

@@ -36,9 +36,9 @@ if TYPE_CHECKING:
 
 logger = logging.getLogger(__name__)
 
-last_pdu_age_metric = Gauge(
-    "synapse_federation_last_sent_pdu_age",
-    "The age (in seconds) of the last PDU successfully sent to the given domain",
+last_pdu_ts_metric = Gauge(
+    "synapse_federation_last_sent_pdu_time",
+    "The timestamp of the last PDU which was successfully sent to the given domain",
     labelnames=("server_name",),
 )
 
@@ -69,15 +69,12 @@ class TransactionManager:
         destination: str,
         pdus: List[EventBase],
         edus: List[Edu],
-    ) -> bool:
+    ) -> None:
         """
         Args:
             destination: The destination to send to (e.g. 'example.org')
             pdus: In-order list of PDUs to send
             edus: List of EDUs to send
-
-        Returns:
-            True iff the transaction was successful
         """
 
         # Make a transaction-sending opentracing span. This span follows on from
@@ -96,8 +93,6 @@ class TransactionManager:
                 edu.strip_context()
 
         with start_active_span_follows_from("send_transaction", span_contexts):
-            success = True
-
             logger.debug("TX [%s] _attempt_new_transaction", destination)
 
             txn_id = str(self._next_txn_id)
@@ -152,45 +147,29 @@ class TransactionManager:
                 response = await self._transport_layer.send_transaction(
                     transaction, json_data_cb
                 )
-                code = 200
             except HttpResponseException as e:
                 code = e.code
                 response = e.response
 
-                if e.code in (401, 404, 429) or 500 <= e.code:
-                    logger.info(
-                        "TX [%s] {%s} got %d response", destination, txn_id, code
-                    )
-                    raise e
+                set_tag(tags.ERROR, True)
 
-            logger.info("TX [%s] {%s} got %d response", destination, txn_id, code)
+                logger.info("TX [%s] {%s} got %d response", destination, txn_id, code)
+                raise
 
-            if code == 200:
-                for e_id, r in response.get("pdus", {}).items():
-                    if "error" in r:
-                        logger.warning(
-                            "TX [%s] {%s} Remote returned error for %s: %s",
-                            destination,
-                            txn_id,
-                            e_id,
-                            r,
-                        )
-            else:
-                for p in pdus:
+            logger.info("TX [%s] {%s} got 200 response", destination, txn_id)
+
+            for e_id, r in response.get("pdus", {}).items():
+                if "error" in r:
                     logger.warning(
-                        "TX [%s] {%s} Failed to send event %s",
+                        "TX [%s] {%s} Remote returned error for %s: %s",
                         destination,
                         txn_id,
-                        p.event_id,
+                        e_id,
+                        r,
                     )
-                success = False
 
-            if success and pdus and destination in self._federation_metrics_domains:
+            if pdus and destination in self._federation_metrics_domains:
                 last_pdu = pdus[-1]
-                last_pdu_age = self.clock.time_msec() - last_pdu.origin_server_ts
-                last_pdu_age_metric.labels(server_name=destination).set(
-                    last_pdu_age / 1000
+                last_pdu_ts_metric.labels(server_name=destination).set(
+                    last_pdu.origin_server_ts / 1000
                 )
-
-            set_tag(tags.ERROR, not success)
-            return success

synapse/federation/transport/server.py

@@ -484,10 +484,9 @@ class FederationQueryServlet(BaseFederationServlet):
 
     # This is when we receive a server-server Query
     async def on_GET(self, origin, content, query, query_type):
-        return await self.handler.on_query_request(
-            query_type,
-            {k.decode("utf8"): v[0].decode("utf-8") for k, v in query.items()},
-        )
+        args = {k.decode("utf8"): v[0].decode("utf-8") for k, v in query.items()}
+        args["origin"] = origin
+        return await self.handler.on_query_request(query_type, args)
 
 
 class FederationMakeJoinServlet(BaseFederationServlet):

synapse/handlers/acme.py

@@ -73,7 +73,9 @@ class AcmeHandler:
                 "Listening for ACME requests on %s:%i", host, self.hs.config.acme_port
             )
             try:
-                self.reactor.listenTCP(self.hs.config.acme_port, srv, interface=host)
+                self.reactor.listenTCP(
+                    self.hs.config.acme_port, srv, backlog=50, interface=host
+                )
             except twisted.internet.error.CannotListenError as e:
                 check_bind_error(e, host, bind_addresses)
 

synapse/handlers/auth.py

@@ -36,7 +36,7 @@ import attr
 import bcrypt
 import pymacaroons
 
-from twisted.web.http import Request
+from twisted.web.server import Request
 
 from synapse.api.constants import LoginType
 from synapse.api.errors import (
@@ -65,6 +65,7 @@ from synapse.storage.roommember import ProfileInfo
 from synapse.types import JsonDict, Requester, UserID
 from synapse.util import stringutils as stringutils
 from synapse.util.async_helpers import maybe_awaitable
+from synapse.util.macaroons import get_value_from_macaroon, satisfy_expiry
 from synapse.util.msisdn import phone_number_to_msisdn
 from synapse.util.threepids import canonicalise_email
 
@@ -170,6 +171,16 @@ class SsoLoginExtraAttributes:
     extra_attributes = attr.ib(type=JsonDict)
 
 
+@attr.s(slots=True, frozen=True)
+class LoginTokenAttributes:
+    """Data we store in a short-term login token"""
+
+    user_id = attr.ib(type=str)
+
+    # the SSO Identity Provider that the user authenticated with, to get this token
+    auth_provider_id = attr.ib(type=str)
+
+
 class AuthHandler(BaseHandler):
     SESSION_EXPIRE_MS = 48 * 60 * 60 * 1000
 
@@ -326,7 +337,8 @@ class AuthHandler(BaseHandler):
                 user is too high to proceed
 
         """
-
+        if not requester.access_token_id:
+            raise ValueError("Cannot validate a user without an access token")
         if self._ui_auth_session_timeout:
             last_validated = await self.store.get_access_token_last_validated(
                 requester.access_token_id
@@ -481,7 +493,7 @@ class AuthHandler(BaseHandler):
             sid = authdict["session"]
 
         # Convert the URI and method to strings.
-        uri = request.uri.decode("utf-8")
+        uri = request.uri.decode("utf-8")  # type: ignore
         method = request.method.decode("utf-8")
 
         # If there's no session ID, create a new session.
@@ -1164,18 +1176,16 @@ class AuthHandler(BaseHandler):
             return None
         return user_id
 
-    async def validate_short_term_login_token_and_get_user_id(self, login_token: str):
-        auth_api = self.hs.get_auth()
-        user_id = None
+    async def validate_short_term_login_token(
+        self, login_token: str
+    ) -> LoginTokenAttributes:
         try:
-            macaroon = pymacaroons.Macaroon.deserialize(login_token)
-            user_id = auth_api.get_user_id_from_macaroon(macaroon)
-            auth_api.validate_macaroon(macaroon, "login", user_id)
+            res = self.macaroon_gen.verify_short_term_login_token(login_token)
         except Exception:
             raise AuthError(403, "Invalid token", errcode=Codes.FORBIDDEN)
 
-        await self.auth.check_auth_blocking(user_id)
-        return user_id
+        await self.auth.check_auth_blocking(res.user_id)
+        return res
 
     async def delete_access_token(self, access_token: str):
         """Invalidate a single access token
@@ -1204,7 +1214,7 @@ class AuthHandler(BaseHandler):
     async def delete_access_tokens_for_user(
         self,
         user_id: str,
-        except_token_id: Optional[str] = None,
+        except_token_id: Optional[int] = None,
         device_id: Optional[str] = None,
     ):
         """Invalidate access tokens belonging to a user
@@ -1397,6 +1407,7 @@ class AuthHandler(BaseHandler):
     async def complete_sso_login(
         self,
         registered_user_id: str,
+        auth_provider_id: str,
         request: Request,
         client_redirect_url: str,
         extra_attributes: Optional[JsonDict] = None,
@@ -1406,6 +1417,9 @@ class AuthHandler(BaseHandler):
 
         Args:
             registered_user_id: The registered user ID to complete SSO login for.
+            auth_provider_id: The id of the SSO Identity provider that was used for
+                login. This will be stored in the login token for future tracking in
+                prometheus metrics.
             request: The request to complete.
             client_redirect_url: The URL to which to redirect the user at the end of the
                 process.
@@ -1427,6 +1441,7 @@ class AuthHandler(BaseHandler):
 
         self._complete_sso_login(
             registered_user_id,
+            auth_provider_id,
             request,
             client_redirect_url,
             extra_attributes,
@@ -1437,6 +1452,7 @@ class AuthHandler(BaseHandler):
     def _complete_sso_login(
         self,
         registered_user_id: str,
+        auth_provider_id: str,
         request: Request,
         client_redirect_url: str,
         extra_attributes: Optional[JsonDict] = None,
@@ -1463,7 +1479,7 @@ class AuthHandler(BaseHandler):
 
         # Create a login token
         login_token = self.macaroon_gen.generate_short_term_login_token(
-            registered_user_id
+            registered_user_id, auth_provider_id=auth_provider_id
         )
 
         # Append the login token to the original redirect URL (i.e. with its query
@@ -1569,15 +1585,48 @@ class MacaroonGenerator:
         return macaroon.serialize()
 
     def generate_short_term_login_token(
-        self, user_id: str, duration_in_ms: int = (2 * 60 * 1000)
+        self,
+        user_id: str,
+        auth_provider_id: str,
+        duration_in_ms: int = (2 * 60 * 1000),
     ) -> str:
         macaroon = self._generate_base_macaroon(user_id)
         macaroon.add_first_party_caveat("type = login")
         now = self.hs.get_clock().time_msec()
         expiry = now + duration_in_ms
         macaroon.add_first_party_caveat("time < %d" % (expiry,))
+        macaroon.add_first_party_caveat("auth_provider_id = %s" % (auth_provider_id,))
         return macaroon.serialize()
 
+    def verify_short_term_login_token(self, token: str) -> LoginTokenAttributes:
+        """Verify a short-term-login macaroon
+
+        Checks that the given token is a valid, unexpired short-term-login token
+        minted by this server.
+
+        Args:
+            token: the login token to verify
+
+        Returns:
+            the user_id that this token is valid for
+
+        Raises:
+            MacaroonVerificationFailedException if the verification failed
+        """
+        macaroon = pymacaroons.Macaroon.deserialize(token)
+        user_id = get_value_from_macaroon(macaroon, "user_id")
+        auth_provider_id = get_value_from_macaroon(macaroon, "auth_provider_id")
+
+        v = pymacaroons.Verifier()
+        v.satisfy_exact("gen = 1")
+        v.satisfy_exact("type = login")
+        v.satisfy_general(lambda c: c.startswith("user_id = "))
+        v.satisfy_general(lambda c: c.startswith("auth_provider_id = "))
+        satisfy_expiry(v, self.hs.get_clock().time_msec)
+        v.verify(macaroon, self.hs.config.key.macaroon_secret_key)
+
+        return LoginTokenAttributes(user_id=user_id, auth_provider_id=auth_provider_id)
+
     def generate_delete_pusher_token(self, user_id: str) -> str:
         macaroon = self._generate_base_macaroon(user_id)
         macaroon.add_first_party_caveat("type = delete_pusher")

synapse/handlers/cas_handler.py

@@ -83,6 +83,7 @@ class CasHandler:
         # the SsoIdentityProvider protocol type.
         self.idp_icon = None
         self.idp_brand = None
+        self.unstable_idp_brand = None
 
         self._sso_handler = hs.get_sso_handler()
 

synapse/handlers/deactivate_account.py

@@ -120,6 +120,11 @@ class DeactivateAccountHandler(BaseHandler):
 
         await self.store.user_set_password_hash(user_id, None)
 
+        # Most of the pushers will have been deleted when we logged out the
+        # associated devices above, but we still need to delete pushers not
+        # associated with devices, e.g. email pushers.
+        await self.store.delete_all_pushers_for_user(user_id)
+
         # Add the user to a table of users pending deactivation (ie.
         # removal from all the rooms they're a member of)
         await self.store.add_user_pending_deactivation(user_id)

synapse/handlers/devicemessage.py

@@ -16,7 +16,9 @@
 import logging
 from typing import TYPE_CHECKING, Any, Dict
 
+from synapse.api.constants import EduTypes
 from synapse.api.errors import SynapseError
+from synapse.api.ratelimiting import Ratelimiter
 from synapse.logging.context import run_in_background
 from synapse.logging.opentracing import (
     get_active_span_text_map,
@@ -25,7 +27,7 @@ from synapse.logging.opentracing import (
     start_active_span,
 )
 from synapse.replication.http.devices import ReplicationUserDevicesResyncRestServlet
-from synapse.types import JsonDict, UserID, get_domain_from_id
+from synapse.types import JsonDict, Requester, UserID, get_domain_from_id
 from synapse.util import json_encoder
 from synapse.util.stringutils import random_string
 
@@ -78,6 +80,12 @@ class DeviceMessageHandler:
                 ReplicationUserDevicesResyncRestServlet.make_client(hs)
             )
 
+        self._ratelimiter = Ratelimiter(
+            clock=hs.get_clock(),
+            rate_hz=hs.config.rc_key_requests.per_second,
+            burst_count=hs.config.rc_key_requests.burst_count,
+        )
+
     async def on_direct_to_device_edu(self, origin: str, content: JsonDict) -> None:
         local_messages = {}
         sender_user_id = content["sender"]
@@ -168,15 +176,27 @@ class DeviceMessageHandler:
 
     async def send_device_message(
         self,
-        sender_user_id: str,
+        requester: Requester,
         message_type: str,
         messages: Dict[str, Dict[str, JsonDict]],
     ) -> None:
+        sender_user_id = requester.user.to_string()
+
         set_tag("number_of_messages", len(messages))
         set_tag("sender", sender_user_id)
         local_messages = {}
         remote_messages = {}  # type: Dict[str, Dict[str, Dict[str, JsonDict]]]
         for user_id, by_device in messages.items():
+            # Ratelimit local cross-user key requests by the sending device.
+            if (
+                message_type == EduTypes.RoomKeyRequest
+                and user_id != sender_user_id
+                and self._ratelimiter.can_do_action(
+                    (sender_user_id, requester.device_id)
+                )
+            ):
+                continue
+
             # we use UserID.from_string to catch invalid user ids
             if self.is_mine(UserID.from_string(user_id)):
                 messages_by_device = {

synapse/handlers/events.py

@@ -17,7 +17,7 @@ import logging
 import random
 from typing import TYPE_CHECKING, Iterable, List, Optional
 
-from synapse.api.constants import EventTypes, Membership
+from synapse.api.constants import EduTypes, EventTypes, Membership
 from synapse.api.errors import AuthError, SynapseError
 from synapse.events import EventBase
 from synapse.handlers.presence import format_user_presence_state
@@ -113,7 +113,7 @@ class EventStreamHandler(BaseHandler):
                     states = await presence_handler.get_states(users)
                     to_add.extend(
                         {
-                            "type": EventTypes.Presence,
+                            "type": EduTypes.Presence,
                             "content": format_user_presence_state(state, time_now),
                         }
                         for state in states

synapse/handlers/federation.py

@@ -201,7 +201,7 @@ class FederationHandler(BaseHandler):
             or pdu.internal_metadata.is_outlier()
         )
         if already_seen:
-            logger.debug("[%s %s]: Already seen pdu", room_id, event_id)
+            logger.debug("Already seen pdu")
             return
 
         # do some initial sanity-checking of the event. In particular, make
@@ -210,18 +210,14 @@ class FederationHandler(BaseHandler):
         try:
             self._sanity_check_event(pdu)
         except SynapseError as err:
-            logger.warning(
-                "[%s %s] Received event failed sanity checks", room_id, event_id
-            )
+            logger.warning("Received event failed sanity checks")
             raise FederationError("ERROR", err.code, err.msg, affected=pdu.event_id)
 
         # If we are currently in the process of joining this room, then we
         # queue up events for later processing.
         if room_id in self.room_queues:
             logger.info(
-                "[%s %s] Queuing PDU from %s for now: join in progress",
-                room_id,
-                event_id,
+                "Queuing PDU from %s for now: join in progress",
                 origin,
             )
             self.room_queues[room_id].append((pdu, origin))
@@ -236,9 +232,7 @@ class FederationHandler(BaseHandler):
         is_in_room = await self.auth.check_host_in_room(room_id, self.server_name)
         if not is_in_room:
             logger.info(
-                "[%s %s] Ignoring PDU from %s as we're not in the room",
-                room_id,
-                event_id,
+                "Ignoring PDU from %s as we're not in the room",
                 origin,
             )
             return None
@@ -250,7 +244,7 @@ class FederationHandler(BaseHandler):
             # We only backfill backwards to the min depth.
             min_depth = await self.get_min_depth_for_context(pdu.room_id)
 
-            logger.debug("[%s %s] min_depth: %d", room_id, event_id, min_depth)
+            logger.debug("min_depth: %d", min_depth)
 
             prevs = set(pdu.prev_event_ids())
             seen = await self.store.have_events_in_timeline(prevs)
@@ -267,17 +261,13 @@ class FederationHandler(BaseHandler):
                     # If we're missing stuff, ensure we only fetch stuff one
                     # at a time.
                     logger.info(
-                        "[%s %s] Acquiring room lock to fetch %d missing prev_events: %s",
-                        room_id,
-                        event_id,
+                        "Acquiring room lock to fetch %d missing prev_events: %s",
                         len(missing_prevs),
                         shortstr(missing_prevs),
                     )
                     with (await self._room_pdu_linearizer.queue(pdu.room_id)):
                         logger.info(
-                            "[%s %s] Acquired room lock to fetch %d missing prev_events",
-                            room_id,
-                            event_id,
+                            "Acquired room lock to fetch %d missing prev_events",
                             len(missing_prevs),
                         )
 
@@ -297,9 +287,7 @@ class FederationHandler(BaseHandler):
 
                         if not prevs - seen:
                             logger.info(
-                                "[%s %s] Found all missing prev_events",
-                                room_id,
-                                event_id,
+                                "Found all missing prev_events",
                             )
 
             if prevs - seen:
@@ -329,9 +317,7 @@ class FederationHandler(BaseHandler):
 
                 if sent_to_us_directly:
                     logger.warning(
-                        "[%s %s] Rejecting: failed to fetch %d prev events: %s",
-                        room_id,
-                        event_id,
+                        "Rejecting: failed to fetch %d prev events: %s",
                         len(prevs - seen),
                         shortstr(prevs - seen),
                     )
@@ -367,17 +353,16 @@ class FederationHandler(BaseHandler):
                     # Ask the remote server for the states we don't
                     # know about
                     for p in prevs - seen:
-                        logger.info(
-                            "Requesting state at missing prev_event %s",
-                            event_id,
-                        )
+                        logger.info("Requesting state after missing prev_event %s", p)
 
                         with nested_logging_context(p):
                             # note that if any of the missing prevs share missing state or
                             # auth events, the requests to fetch those events are deduped
                             # by the get_pdu_cache in federation_client.
-                            (remote_state, _,) = await self._get_state_for_room(
-                                origin, room_id, p, include_event_in_state=True
+                            remote_state = (
+                                await self._get_state_after_missing_prev_event(
+                                    origin, room_id, p
+                                )
                             )
 
                             remote_state_map = {
@@ -414,10 +399,7 @@ class FederationHandler(BaseHandler):
                     state = [event_map[e] for e in state_map.values()]
                 except Exception:
                     logger.warning(
-                        "[%s %s] Error attempting to resolve state at missing "
-                        "prev_events",
-                        room_id,
-                        event_id,
+                        "Error attempting to resolve state at missing " "prev_events",
                         exc_info=True,
                     )
                     raise FederationError(
@@ -454,9 +436,7 @@ class FederationHandler(BaseHandler):
         latest |= seen
 
         logger.info(
-            "[%s %s]: Requesting missing events between %s and %s",
-            room_id,
-            event_id,
+            "Requesting missing events between %s and %s",
             shortstr(latest),
             event_id,
         )
@@ -523,15 +503,11 @@ class FederationHandler(BaseHandler):
             # We failed to get the missing events, but since we need to handle
             # the case of `get_missing_events` not returning the necessary
             # events anyway, it is safe to simply log the error and continue.
-            logger.warning(
-                "[%s %s]: Failed to get prev_events: %s", room_id, event_id, e
-            )
+            logger.warning("Failed to get prev_events: %s", e)
             return
 
         logger.info(
-            "[%s %s]: Got %d prev_events: %s",
-            room_id,
-            event_id,
+            "Got %d prev_events: %s",
             len(missing_events),
             shortstr(missing_events),
         )
@@ -542,9 +518,7 @@ class FederationHandler(BaseHandler):
 
         for ev in missing_events:
             logger.info(
-                "[%s %s] Handling received prev_event %s",
-                room_id,
-                event_id,
+                "Handling received prev_event %s",
                 ev.event_id,
             )
             with nested_logging_context(ev.event_id):
@@ -553,9 +527,7 @@ class FederationHandler(BaseHandler):
                 except FederationError as e:
                     if e.code == 403:
                         logger.warning(
-                            "[%s %s] Received prev_event %s failed history check.",
-                            room_id,
-                            event_id,
+                            "Received prev_event %s failed history check.",
                             ev.event_id,
                         )
                     else:
@@ -566,7 +538,6 @@ class FederationHandler(BaseHandler):
         destination: str,
         room_id: str,
         event_id: str,
-        include_event_in_state: bool = False,
     ) -> Tuple[List[EventBase], List[EventBase]]:
         """Requests all of the room state at a given event from a remote homeserver.
 
@@ -574,11 +545,9 @@ class FederationHandler(BaseHandler):
             destination: The remote homeserver to query for the state.
             room_id: The id of the room we're interested in.
             event_id: The id of the event we want the state at.
-            include_event_in_state: if true, the event itself will be included in the
-                returned state event list.
 
         Returns:
-            A list of events in the state, possibly including the event itself, and
+            A list of events in the state, not including the event itself, and
             a list of events in the auth chain for the given event.
         """
         (
@@ -590,9 +559,6 @@ class FederationHandler(BaseHandler):
 
         desired_events = set(state_event_ids + auth_event_ids)
 
-        if include_event_in_state:
-            desired_events.add(event_id)
-
         event_map = await self._get_events_from_store_or_dest(
             destination, room_id, desired_events
         )
@@ -609,13 +575,6 @@ class FederationHandler(BaseHandler):
             event_map[e_id] for e_id in state_event_ids if e_id in event_map
         ]
 
-        if include_event_in_state:
-            remote_event = event_map.get(event_id)
-            if not remote_event:
-                raise Exception("Unable to get missing prev_event %s" % (event_id,))
-            if remote_event.is_state() and remote_event.rejected_reason is None:
-                remote_state.append(remote_event)
-
         auth_chain = [event_map[e_id] for e_id in auth_event_ids if e_id in event_map]
         auth_chain.sort(key=lambda e: e.depth)
 
@@ -689,6 +648,131 @@ class FederationHandler(BaseHandler):
 
         return fetched_events
 
+    async def _get_state_after_missing_prev_event(
+        self,
+        destination: str,
+        room_id: str,
+        event_id: str,
+    ) -> List[EventBase]:
+        """Requests all of the room state at a given event from a remote homeserver.
+
+        Args:
+            destination: The remote homeserver to query for the state.
+            room_id: The id of the room we're interested in.
+            event_id: The id of the event we want the state at.
+
+        Returns:
+            A list of events in the state, including the event itself
+        """
+        # TODO: This function is basically the same as _get_state_for_room. Can
+        #   we make backfill() use it, rather than having two code paths? I think the
+        #   only difference is that backfill() persists the prev events separately.
+
+        (
+            state_event_ids,
+            auth_event_ids,
+        ) = await self.federation_client.get_room_state_ids(
+            destination, room_id, event_id=event_id
+        )
+
+        logger.debug(
+            "state_ids returned %i state events, %i auth events",
+            len(state_event_ids),
+            len(auth_event_ids),
+        )
+
+        # start by just trying to fetch the events from the store
+        desired_events = set(state_event_ids)
+        desired_events.add(event_id)
+        logger.debug("Fetching %i events from cache/store", len(desired_events))
+        fetched_events = await self.store.get_events(
+            desired_events, allow_rejected=True
+        )
+
+        missing_desired_events = desired_events - fetched_events.keys()
+        logger.debug(
+            "We are missing %i events (got %i)",
+            len(missing_desired_events),
+            len(fetched_events),
+        )
+
+        # We probably won't need most of the auth events, so let's just check which
+        # we have for now, rather than thrashing the event cache with them all
+        # unnecessarily.
+
+        # TODO: we probably won't actually need all of the auth events, since we
+        #   already have a bunch of the state events. It would be nice if the
+        #   federation api gave us a way of finding out which we actually need.
+
+        missing_auth_events = set(auth_event_ids) - fetched_events.keys()
+        missing_auth_events.difference_update(
+            await self.store.have_seen_events(missing_auth_events)
+        )
+        logger.debug("We are also missing %i auth events", len(missing_auth_events))
+
+        missing_events = missing_desired_events | missing_auth_events
+        logger.debug("Fetching %i events from remote", len(missing_events))
+        await self._get_events_and_persist(
+            destination=destination, room_id=room_id, events=missing_events
+        )
+
+        # we need to make sure we re-load from the database to get the rejected
+        # state correct.
+        fetched_events.update(
+            (await self.store.get_events(missing_desired_events, allow_rejected=True))
+        )
+
+        # check for events which were in the wrong room.
+        #
+        # this can happen if a remote server claims that the state or
+        # auth_events at an event in room A are actually events in room B
+
+        bad_events = [
+            (event_id, event.room_id)
+            for event_id, event in fetched_events.items()
+            if event.room_id != room_id
+        ]
+
+        for bad_event_id, bad_room_id in bad_events:
+            # This is a bogus situation, but since we may only discover it a long time
+            # after it happened, we try our best to carry on, by just omitting the
+            # bad events from the returned state set.
+            logger.warning(
+                "Remote server %s claims event %s in room %s is an auth/state "
+                "event in room %s",
+                destination,
+                bad_event_id,
+                bad_room_id,
+                room_id,
+            )
+
+            del fetched_events[bad_event_id]
+
+        # if we couldn't get the prev event in question, that's a problem.
+        remote_event = fetched_events.get(event_id)
+        if not remote_event:
+            raise Exception("Unable to get missing prev_event %s" % (event_id,))
+
+        # missing state at that event is a warning, not a blocker
+        # XXX: this doesn't sound right? it means that we'll end up with incomplete
+        #   state.
+        failed_to_fetch = desired_events - fetched_events.keys()
+        if failed_to_fetch:
+            logger.warning(
+                "Failed to fetch missing state events for %s %s",
+                event_id,
+                failed_to_fetch,
+            )
+
+        remote_state = [
+            fetched_events[e_id] for e_id in state_event_ids if e_id in fetched_events
+        ]
+
+        if remote_event.is_state() and remote_event.rejected_reason is None:
+            remote_state.append(remote_event)
+
+        return remote_state
+
     async def _process_received_pdu(
         self,
         origin: str,
@@ -707,10 +791,7 @@ class FederationHandler(BaseHandler):
                 (ie, we are missing one or more prev_events), the resolved state at the
                 event
         """
-        room_id = event.room_id
-        event_id = event.event_id
-
-        logger.debug("[%s %s] Processing event: %s", room_id, event_id, event)
+        logger.debug("Processing event: %s", event)
 
         try:
             await self._handle_new_event(origin, event, state=state)
@@ -871,7 +952,6 @@ class FederationHandler(BaseHandler):
                 destination=dest,
                 room_id=room_id,
                 event_id=e_id,
-                include_event_in_state=False,
             )
             auth_events.update({a.event_id: a for a in auth})
             auth_events.update({s.event_id: s for s in state})
@@ -1317,7 +1397,7 @@ class FederationHandler(BaseHandler):
     async def on_event_auth(self, event_id: str) -> List[EventBase]:
         event = await self.store.get_event(event_id)
         auth = await self.store.get_auth_chain(
-            list(event.auth_event_ids()), include_given=True
+            event.room_id, list(event.auth_event_ids()), include_given=True
         )
         return list(auth)
 
@@ -1580,7 +1660,7 @@ class FederationHandler(BaseHandler):
         prev_state_ids = await context.get_prev_state_ids()
 
         state_ids = list(prev_state_ids.values())
-        auth_chain = await self.store.get_auth_chain(state_ids)
+        auth_chain = await self.store.get_auth_chain(event.room_id, state_ids)
 
         state = await self.store.get_events(list(prev_state_ids.values()))
 
@@ -2219,7 +2299,7 @@ class FederationHandler(BaseHandler):
 
         # Now get the current auth_chain for the event.
         local_auth_chain = await self.store.get_auth_chain(
-            list(event.auth_event_ids()), include_given=True
+            room_id, list(event.auth_event_ids()), include_given=True
         )
 
         # TODO: Check if we would now reject event_id. If so we need to tell

synapse/handlers/initial_sync.py

@@ -18,7 +18,7 @@ from typing import TYPE_CHECKING, Optional, Tuple
 
 from twisted.internet import defer
 
-from synapse.api.constants import EventTypes, Membership
+from synapse.api.constants import EduTypes, EventTypes, Membership
 from synapse.api.errors import SynapseError
 from synapse.events.validator import EventValidator
 from synapse.handlers.presence import format_user_presence_state
@@ -48,7 +48,7 @@ class InitialSyncHandler(BaseHandler):
         self.clock = hs.get_clock()
         self.validator = EventValidator()
         self.snapshot_cache = ResponseCache(
-            hs, "initial_sync_cache"
+            hs.get_clock(), "initial_sync_cache"
         )  # type: ResponseCache[Tuple[str, Optional[StreamToken], Optional[StreamToken], str, Optional[int], bool, bool]]
         self._event_serializer = hs.get_event_client_serializer()
         self.storage = hs.get_storage()
@@ -412,7 +412,7 @@ class InitialSyncHandler(BaseHandler):
 
             return [
                 {
-                    "type": EventTypes.Presence,
+                    "type": EduTypes.Presence,
                     "content": format_user_presence_state(s, time_now),
                 }
                 for s in states

synapse/handlers/message.py

@@ -387,6 +387,12 @@ class EventCreationHandler:
 
         self.room_invite_state_types = self.hs.config.room_invite_state_types
 
+        self.membership_types_to_include_profile_data_in = (
+            {Membership.JOIN, Membership.INVITE}
+            if self.hs.config.include_profile_data_on_invite
+            else {Membership.JOIN}
+        )
+
         self.send_event = ReplicationSendEventRestServlet.make_client(hs)
 
         # This is only used to get at ratelimit function, and maybe_kick_guest_users
@@ -500,7 +506,7 @@ class EventCreationHandler:
             membership = builder.content.get("membership", None)
             target = UserID.from_string(builder.state_key)
 
-            if membership in {Membership.JOIN, Membership.INVITE}:
+            if membership in self.membership_types_to_include_profile_data_in:
                 # If event doesn't include a display name, add one.
                 profile = self.profile_handler
                 content = builder.content

synapse/handlers/oidc_handler.py

@@ -1,5 +1,6 @@
 # -*- coding: utf-8 -*-
 # Copyright 2020 Quentin Gliech
+# Copyright 2021 The Matrix.org Foundation C.I.C.
 #
 # Licensed under the Apache License, Version 2.0 (the "License");
 # you may not use this file except in compliance with the License.
@@ -14,13 +15,13 @@
 # limitations under the License.
 import inspect
 import logging
-from typing import TYPE_CHECKING, Dict, Generic, List, Optional, TypeVar
+from typing import TYPE_CHECKING, Dict, Generic, List, Optional, TypeVar, Union
 from urllib.parse import urlencode
 
 import attr
 import pymacaroons
 from authlib.common.security import generate_token
-from authlib.jose import JsonWebToken
+from authlib.jose import JsonWebToken, jwt
 from authlib.oauth2.auth import ClientAuth
 from authlib.oauth2.rfc6749.parameters import prepare_grant_uri
 from authlib.oidc.core import CodeIDToken, ImplicitIDToken, UserInfo
@@ -28,20 +29,26 @@ from authlib.oidc.discovery import OpenIDProviderMetadata, get_well_known_url
 from jinja2 import Environment, Template
 from pymacaroons.exceptions import (
     MacaroonDeserializationException,
+    MacaroonInitException,
     MacaroonInvalidSignatureException,
 )
 from typing_extensions import TypedDict
 
 from twisted.web.client import readBody
+from twisted.web.http_headers import Headers
 
 from synapse.config import ConfigError
-from synapse.config.oidc_config import OidcProviderConfig
+from synapse.config.oidc_config import (
+    OidcProviderClientSecretJwtKey,
+    OidcProviderConfig,
+)
 from synapse.handlers.sso import MappingException, UserAttributes
 from synapse.http.site import SynapseRequest
 from synapse.logging.context import make_deferred_yieldable
 from synapse.types import JsonDict, UserID, map_username_to_mxid_localpart
-from synapse.util import json_decoder
+from synapse.util import Clock, json_decoder
 from synapse.util.caches.cached_call import RetryOnExceptionCachedCall
+from synapse.util.macaroons import get_value_from_macaroon, satisfy_expiry
 
 if TYPE_CHECKING:
     from synapse.server import HomeServer
@@ -211,7 +218,7 @@ class OidcHandler:
             session_data = self._token_generator.verify_oidc_session_token(
                 session, state
             )
-        except (MacaroonDeserializationException, ValueError) as e:
+        except (MacaroonInitException, MacaroonDeserializationException, KeyError) as e:
             logger.exception("Invalid session for OIDC callback")
             self._sso_handler.render_error(request, "invalid_session", str(e))
             return
@@ -275,9 +282,21 @@ class OidcProvider:
 
         self._scopes = provider.scopes
         self._user_profile_method = provider.user_profile_method
+
+        client_secret = None  # type: Union[None, str, JwtClientSecret]
+        if provider.client_secret:
+            client_secret = provider.client_secret
+        elif provider.client_secret_jwt_key:
+            client_secret = JwtClientSecret(
+                provider.client_secret_jwt_key,
+                provider.client_id,
+                provider.issuer,
+                hs.get_clock(),
+            )
+
         self._client_auth = ClientAuth(
             provider.client_id,
-            provider.client_secret,
+            client_secret,
             provider.client_auth_method,
         )  # type: ClientAuth
         self._client_auth_method = provider.client_auth_method
@@ -312,6 +331,9 @@ class OidcProvider:
         # optional brand identifier for this auth provider
         self.idp_brand = provider.idp_brand
 
+        # Optional brand identifier for the unstable API (see MSC2858).
+        self.unstable_idp_brand = provider.unstable_idp_brand
+
         self._sso_handler = hs.get_sso_handler()
 
         self._sso_handler.register_identity_provider(self)
@@ -521,7 +543,7 @@ class OidcProvider:
         """
         metadata = await self.load_metadata()
         token_endpoint = metadata.get("token_endpoint")
-        headers = {
+        raw_headers = {
             "Content-Type": "application/x-www-form-urlencoded",
             "User-Agent": self._http_client.user_agent,
             "Accept": "application/json",
@@ -535,10 +557,10 @@ class OidcProvider:
         body = urlencode(args, True)
 
         # Fill the body/headers with credentials
-        uri, headers, body = self._client_auth.prepare(
-            method="POST", uri=token_endpoint, headers=headers, body=body
+        uri, raw_headers, body = self._client_auth.prepare(
+            method="POST", uri=token_endpoint, headers=raw_headers, body=body
         )
-        headers = {k: [v] for (k, v) in headers.items()}
+        headers = Headers({k: [v] for (k, v) in raw_headers.items()})
 
         # Do the actual request
         # We're not using the SimpleHttpClient util methods as we don't want to
@@ -745,7 +767,7 @@ class OidcProvider:
                 idp_id=self.idp_id,
                 nonce=nonce,
                 client_redirect_url=client_redirect_url.decode(),
-                ui_auth_session_id=ui_auth_session_id,
+                ui_auth_session_id=ui_auth_session_id or "",
             ),
         )
 
@@ -976,6 +998,81 @@ class OidcProvider:
         return str(remote_user_id)
 
 
+# number of seconds a newly-generated client secret should be valid for
+CLIENT_SECRET_VALIDITY_SECONDS = 3600
+
+# minimum remaining validity on a client secret before we should generate a new one
+CLIENT_SECRET_MIN_VALIDITY_SECONDS = 600
+
+
+class JwtClientSecret:
+    """A class which generates a new client secret on demand, based on a JWK
+
+    This implementation is designed to comply with the requirements for Apple Sign in:
+    https://developer.apple.com/documentation/sign_in_with_apple/generate_and_validate_tokens#3262048
+
+    It looks like those requirements are based on https://tools.ietf.org/html/rfc7523,
+    but it's worth noting that we still put the generated secret in the "client_secret"
+    field (or rather, whereever client_auth_method puts it) rather than in a
+    client_assertion field in the body as that RFC seems to require.
+    """
+
+    def __init__(
+        self,
+        key: OidcProviderClientSecretJwtKey,
+        oauth_client_id: str,
+        oauth_issuer: str,
+        clock: Clock,
+    ):
+        self._key = key
+        self._oauth_client_id = oauth_client_id
+        self._oauth_issuer = oauth_issuer
+        self._clock = clock
+        self._cached_secret = b""
+        self._cached_secret_replacement_time = 0
+
+    def __str__(self):
+        # if client_auth_method is client_secret_basic, then ClientAuth.prepare calls
+        # encode_client_secret_basic, which calls "{}".format(secret), which ends up
+        # here.
+        return self._get_secret().decode("ascii")
+
+    def __bytes__(self):
+        # if client_auth_method is client_secret_post, then ClientAuth.prepare calls
+        # encode_client_secret_post, which ends up here.
+        return self._get_secret()
+
+    def _get_secret(self) -> bytes:
+        now = self._clock.time()
+
+        # if we have enough validity on our existing secret, use it
+        if now < self._cached_secret_replacement_time:
+            return self._cached_secret
+
+        issued_at = int(now)
+        expires_at = issued_at + CLIENT_SECRET_VALIDITY_SECONDS
+
+        # we copy the configured header because jwt.encode modifies it.
+        header = dict(self._key.jwt_header)
+
+        # see https://tools.ietf.org/html/rfc7523#section-3
+        payload = {
+            "sub": self._oauth_client_id,
+            "aud": self._oauth_issuer,
+            "iat": issued_at,
+            "exp": expires_at,
+            **self._key.jwt_payload,
+        }
+        logger.info(
+            "Generating new JWT for %s: %s %s", self._oauth_issuer, header, payload
+        )
+        self._cached_secret = jwt.encode(header, payload, self._key.key)
+        self._cached_secret_replacement_time = (
+            expires_at - CLIENT_SECRET_MIN_VALIDITY_SECONDS
+        )
+        return self._cached_secret
+
+
 class OidcSessionTokenGenerator:
     """Methods for generating and checking OIDC Session cookies."""
 
@@ -1020,10 +1117,9 @@ class OidcSessionTokenGenerator:
         macaroon.add_first_party_caveat(
             "client_redirect_url = %s" % (session_data.client_redirect_url,)
         )
-        if session_data.ui_auth_session_id:
-            macaroon.add_first_party_caveat(
-                "ui_auth_session_id = %s" % (session_data.ui_auth_session_id,)
-            )
+        macaroon.add_first_party_caveat(
+            "ui_auth_session_id = %s" % (session_data.ui_auth_session_id,)
+        )
         now = self._clock.time_msec()
         expiry = now + duration_in_ms
         macaroon.add_first_party_caveat("time < %d" % (expiry,))
@@ -1046,7 +1142,7 @@ class OidcSessionTokenGenerator:
             The data extracted from the session cookie
 
         Raises:
-            ValueError if an expected caveat is missing from the macaroon.
+            KeyError if an expected caveat is missing from the macaroon.
         """
         macaroon = pymacaroons.Macaroon.deserialize(session)
 
@@ -1057,26 +1153,16 @@ class OidcSessionTokenGenerator:
         v.satisfy_general(lambda c: c.startswith("nonce = "))
         v.satisfy_general(lambda c: c.startswith("idp_id = "))
         v.satisfy_general(lambda c: c.startswith("client_redirect_url = "))
-        # Sometimes there's a UI auth session ID, it seems to be OK to attempt
-        # to always satisfy this.
         v.satisfy_general(lambda c: c.startswith("ui_auth_session_id = "))
-        v.satisfy_general(self._verify_expiry)
+        satisfy_expiry(v, self._clock.time_msec)
 
         v.verify(macaroon, self._macaroon_secret_key)
 
         # Extract the session data from the token.
-        nonce = self._get_value_from_macaroon(macaroon, "nonce")
-        idp_id = self._get_value_from_macaroon(macaroon, "idp_id")
-        client_redirect_url = self._get_value_from_macaroon(
-            macaroon, "client_redirect_url"
-        )
-        try:
-            ui_auth_session_id = self._get_value_from_macaroon(
-                macaroon, "ui_auth_session_id"
-            )  # type: Optional[str]
-        except ValueError:
-            ui_auth_session_id = None
-
+        nonce = get_value_from_macaroon(macaroon, "nonce")
+        idp_id = get_value_from_macaroon(macaroon, "idp_id")
+        client_redirect_url = get_value_from_macaroon(macaroon, "client_redirect_url")
+        ui_auth_session_id = get_value_from_macaroon(macaroon, "ui_auth_session_id")
         return OidcSessionData(
             nonce=nonce,
             idp_id=idp_id,
@@ -1084,33 +1170,6 @@ class OidcSessionTokenGenerator:
             ui_auth_session_id=ui_auth_session_id,
         )
 
-    def _get_value_from_macaroon(self, macaroon: pymacaroons.Macaroon, key: str) -> str:
-        """Extracts a caveat value from a macaroon token.
-
-        Args:
-            macaroon: the token
-            key: the key of the caveat to extract
-
-        Returns:
-            The extracted value
-
-        Raises:
-            ValueError: if the caveat was not in the macaroon
-        """
-        prefix = key + " = "
-        for caveat in macaroon.caveats:
-            if caveat.caveat_id.startswith(prefix):
-                return caveat.caveat_id[len(prefix) :]
-        raise ValueError("No %s caveat in macaroon" % (key,))
-
-    def _verify_expiry(self, caveat: str) -> bool:
-        prefix = "time < "
-        if not caveat.startswith(prefix):
-            return False
-        expiry = int(caveat[len(prefix) :])
-        now = self._clock.time_msec()
-        return now < expiry
-
 
 @attr.s(frozen=True, slots=True)
 class OidcSessionData:
@@ -1125,8 +1184,8 @@ class OidcSessionData:
     # The URL the client gave when it initiated the flow. ("" if this is a UI Auth)
     client_redirect_url = attr.ib(type=str)
 
-    # The session ID of the ongoing UI Auth (None if this is a login)
-    ui_auth_session_id = attr.ib(type=Optional[str], default=None)
+    # The session ID of the ongoing UI Auth ("" if this is a login)
+    ui_auth_session_id = attr.ib(type=str)
 
 
 UserAttributeDict = TypedDict(

synapse/handlers/pagination.py

@@ -285,7 +285,7 @@ class PaginationHandler:
         except Exception:
             f = Failure()
             logger.error(
-                "[purge] failed", exc_info=(f.type, f.value, f.getTracebackObject())
+                "[purge] failed", exc_info=(f.type, f.value, f.getTracebackObject())  # type: ignore
             )
             self._purges_by_id[purge_id].status = PurgeStatus.STATUS_FAILED
         finally:

synapse/handlers/presence.py

@@ -274,22 +274,25 @@ class PresenceHandler(BasePresenceHandler):
 
         self.external_sync_linearizer = Linearizer(name="external_sync_linearizer")
 
-        # Start a LoopingCall in 30s that fires every 5s.
-        # The initial delay is to allow disconnected clients a chance to
-        # reconnect before we treat them as offline.
-        def run_timeout_handler():
-            return run_as_background_process(
-                "handle_presence_timeouts", self._handle_timeouts
+        if self._presence_enabled:
+            # Start a LoopingCall in 30s that fires every 5s.
+            # The initial delay is to allow disconnected clients a chance to
+            # reconnect before we treat them as offline.
+            def run_timeout_handler():
+                return run_as_background_process(
+                    "handle_presence_timeouts", self._handle_timeouts
+                )
+
+            self.clock.call_later(
+                30, self.clock.looping_call, run_timeout_handler, 5000
             )
 
-        self.clock.call_later(30, self.clock.looping_call, run_timeout_handler, 5000)
+            def run_persister():
+                return run_as_background_process(
+                    "persist_presence_changes", self._persist_unpersisted_changes
+                )
 
-        def run_persister():
-            return run_as_background_process(
-                "persist_presence_changes", self._persist_unpersisted_changes
-            )
-
-        self.clock.call_later(60, self.clock.looping_call, run_persister, 60 * 1000)
+            self.clock.call_later(60, self.clock.looping_call, run_persister, 60 * 1000)
 
         LaterGauge(
             "synapse_handlers_presence_wheel_timer_size",
@@ -299,7 +302,7 @@ class PresenceHandler(BasePresenceHandler):
         )
 
         # Used to handle sending of presence to newly joined users/servers
-        if hs.config.use_presence:
+        if self._presence_enabled:
             self.notifier.add_replication_callback(self.notify_new_event)
 
         # Presence is best effort and quickly heals itself, so lets just always
@@ -849,6 +852,9 @@ class PresenceHandler(BasePresenceHandler):
         """Process current state deltas to find new joins that need to be
         handled.
         """
+        # A map of destination to a set of user state that they should receive
+        presence_destinations = {}  # type: Dict[str, Set[UserPresenceState]]
+
         for delta in deltas:
             typ = delta["type"]
             state_key = delta["state_key"]
@@ -858,6 +864,7 @@ class PresenceHandler(BasePresenceHandler):
 
             logger.debug("Handling: %r %r, %s", typ, state_key, event_id)
 
+            # Drop any event that isn't a membership join
             if typ != EventTypes.Member:
                 continue
 
@@ -880,13 +887,38 @@ class PresenceHandler(BasePresenceHandler):
                     # Ignore changes to join events.
                     continue
 
-            await self._on_user_joined_room(room_id, state_key)
+            # Retrieve any user presence state updates that need to be sent as a result,
+            # and the destinations that need to receive it
+            destinations, user_presence_states = await self._on_user_joined_room(
+                room_id, state_key
+            )
 
-    async def _on_user_joined_room(self, room_id: str, user_id: str) -> None:
+            # Insert the destinations and respective updates into our destinations dict
+            for destination in destinations:
+                presence_destinations.setdefault(destination, set()).update(
+                    user_presence_states
+                )
+
+        # Send out user presence updates for each destination
+        for destination, user_state_set in presence_destinations.items():
+            self.federation.send_presence_to_destinations(
+                destinations=[destination], states=user_state_set
+            )
+
+    async def _on_user_joined_room(
+        self, room_id: str, user_id: str
+    ) -> Tuple[List[str], List[UserPresenceState]]:
         """Called when we detect a user joining the room via the current state
-        delta stream.
-        """
+        delta stream. Returns the destinations that need to be updated and the
+        presence updates to send to them.
 
+        Args:
+            room_id: The ID of the room that the user has joined.
+            user_id: The ID of the user that has joined the room.
+
+        Returns:
+            A tuple of destinations and presence updates to send to them.
+        """
         if self.is_mine_id(user_id):
             # If this is a local user then we need to send their presence
             # out to hosts in the room (who don't already have it)
@@ -894,15 +926,15 @@ class PresenceHandler(BasePresenceHandler):
             # TODO: We should be able to filter the hosts down to those that
             # haven't previously seen the user
 
-            state = await self.current_state_for_user(user_id)
-            hosts = await self.state.get_current_hosts_in_room(room_id)
+            remote_hosts = await self.state.get_current_hosts_in_room(room_id)
 
             # Filter out ourselves.
-            hosts = {host for host in hosts if host != self.server_name}
+            filtered_remote_hosts = [
+                host for host in remote_hosts if host != self.server_name
+            ]
 
-            self.federation.send_presence_to_destinations(
-                states=[state], destinations=hosts
-            )
+            state = await self.current_state_for_user(user_id)
+            return filtered_remote_hosts, [state]
         else:
             # A remote user has joined the room, so we need to:
             #   1. Check if this is a new server in the room
@@ -915,6 +947,8 @@ class PresenceHandler(BasePresenceHandler):
             # TODO: Check that this is actually a new server joining the
             # room.
 
+            remote_host = get_domain_from_id(user_id)
+
             users = await self.state.get_current_users_in_room(room_id)
             user_ids = list(filter(self.is_mine_id, users))
 
@@ -934,10 +968,7 @@ class PresenceHandler(BasePresenceHandler):
                 or state.status_msg is not None
             ]
 
-            if states:
-                self.federation.send_presence_to_destinations(
-                    states=states, destinations=[get_domain_from_id(user_id)]
-                )
+            return [remote_host], states
 
 
 def should_notify(old_state, new_state):

synapse/handlers/profile.py

@@ -310,6 +310,15 @@ class ProfileHandler(BaseHandler):
         await self._update_join_states(requester, target_user)
 
     async def on_profile_query(self, args: JsonDict) -> JsonDict:
+        """Handles federation profile query requests."""
+
+        if not self.hs.config.allow_profile_lookup_over_federation:
+            raise SynapseError(
+                403,
+                "Profile lookup over federation is disabled on this homeserver",
+                Codes.FORBIDDEN,
+            )
+
         user = UserID.from_string(args["user_id"])
         if not self.hs.is_mine(user):
             raise SynapseError(400, "User is not hosted on this homeserver")

synapse/handlers/register.py

@@ -16,7 +16,9 @@
 """Contains functions for registering clients."""
 
 import logging
-from typing import TYPE_CHECKING, Iterable, List, Optional, Tuple
+from typing import TYPE_CHECKING, Dict, Iterable, List, Optional, Tuple
+
+from prometheus_client import Counter
 
 from synapse import types
 from synapse.api.constants import MAX_USERID_LENGTH, EventTypes, JoinRules, LoginType
@@ -41,6 +43,19 @@ if TYPE_CHECKING:
 logger = logging.getLogger(__name__)
 
 
+registration_counter = Counter(
+    "synapse_user_registrations_total",
+    "Number of new users registered (since restart)",
+    ["guest", "shadow_banned", "auth_provider"],
+)
+
+login_counter = Counter(
+    "synapse_user_logins_total",
+    "Number of user logins (since restart)",
+    ["guest", "auth_provider"],
+)
+
+
 class RegistrationHandler(BaseHandler):
     def __init__(self, hs: "HomeServer"):
         super().__init__(hs)
@@ -67,6 +82,7 @@ class RegistrationHandler(BaseHandler):
             )
         else:
             self.device_handler = hs.get_device_handler()
+            self._register_device_client = self.register_device_inner
             self.pusher_pool = hs.get_pusherpool()
 
         self.session_lifetime = hs.config.session_lifetime
@@ -156,6 +172,7 @@ class RegistrationHandler(BaseHandler):
         bind_emails: Iterable[str] = [],
         by_admin: bool = False,
         user_agent_ips: Optional[List[Tuple[str, str]]] = None,
+        auth_provider_id: Optional[str] = None,
     ) -> str:
         """Registers a new client on the server.
 
@@ -181,8 +198,9 @@ class RegistrationHandler(BaseHandler):
               admin api, otherwise False.
             user_agent_ips: Tuples of IP addresses and user-agents used
                 during the registration process.
+            auth_provider_id: The SSO IdP the user used, if any.
         Returns:
-            The registere user_id.
+            The registered user_id.
         Raises:
             SynapseError if there was a problem registering.
         """
@@ -192,6 +210,7 @@ class RegistrationHandler(BaseHandler):
             threepid,
             localpart,
             user_agent_ips or [],
+            auth_provider_id=auth_provider_id,
         )
 
         if result == RegistrationBehaviour.DENY:
@@ -280,6 +299,12 @@ class RegistrationHandler(BaseHandler):
                     # if user id is taken, just generate another
                     fail_count += 1
 
+        registration_counter.labels(
+            guest=make_guest,
+            shadow_banned=shadow_banned,
+            auth_provider=(auth_provider_id or ""),
+        ).inc()
+
         if not self.hs.config.user_consent_at_registration:
             if not self.hs.config.auto_join_rooms_for_guests and make_guest:
                 logger.info(
@@ -638,6 +663,7 @@ class RegistrationHandler(BaseHandler):
         initial_display_name: Optional[str],
         is_guest: bool = False,
         is_appservice_ghost: bool = False,
+        auth_provider_id: Optional[str] = None,
     ) -> Tuple[str, str]:
         """Register a device for a user and generate an access token.
 
@@ -648,21 +674,40 @@ class RegistrationHandler(BaseHandler):
             device_id: The device ID to check, or None to generate a new one.
             initial_display_name: An optional display name for the device.
             is_guest: Whether this is a guest account
-
+            auth_provider_id: The SSO IdP the user used, if any (just used for the
+                prometheus metrics).
         Returns:
             Tuple of device ID and access token
         """
+        res = await self._register_device_client(
+            user_id=user_id,
+            device_id=device_id,
+            initial_display_name=initial_display_name,
+            is_guest=is_guest,
+            is_appservice_ghost=is_appservice_ghost,
+        )
 
-        if self.hs.config.worker_app:
-            r = await self._register_device_client(
-                user_id=user_id,
-                device_id=device_id,
-                initial_display_name=initial_display_name,
-                is_guest=is_guest,
-                is_appservice_ghost=is_appservice_ghost,
-            )
-            return r["device_id"], r["access_token"]
+        login_counter.labels(
+            guest=is_guest,
+            auth_provider=(auth_provider_id or ""),
+        ).inc()
 
+        return res["device_id"], res["access_token"]
+
+    async def register_device_inner(
+        self,
+        user_id: str,
+        device_id: Optional[str],
+        initial_display_name: Optional[str],
+        is_guest: bool = False,
+        is_appservice_ghost: bool = False,
+    ) -> Dict[str, str]:
+        """Helper for register_device
+
+        Does the bits that need doing on the main process. Not for use outside this
+        class and RegisterDeviceReplicationServlet.
+        """
+        assert not self.hs.config.worker_app
         valid_until_ms = None
         if self.session_lifetime is not None:
             if is_guest:
@@ -687,7 +732,7 @@ class RegistrationHandler(BaseHandler):
                 is_appservice_ghost=is_appservice_ghost,
             )
 
-        return (registered_device_id, access_token)
+        return {"device_id": registered_device_id, "access_token": access_token}
 
     async def post_registration_actions(
         self, user_id: str, auth_result: dict, access_token: Optional[str]

synapse/handlers/room.py

@@ -121,7 +121,7 @@ class RoomCreationHandler(BaseHandler):
         # succession, only process the first attempt and return its result to
         # subsequent requests
         self._upgrade_response_cache = ResponseCache(
-            hs, "room_upgrade", timeout_ms=FIVE_MINUTES_IN_MS
+            hs.get_clock(), "room_upgrade", timeout_ms=FIVE_MINUTES_IN_MS
         )  # type: ResponseCache[Tuple[str, str]]
         self._server_notices_mxid = hs.config.server_notices_mxid
 

synapse/handlers/room_list.py

@@ -44,10 +44,10 @@ class RoomListHandler(BaseHandler):
         super().__init__(hs)
         self.enable_room_list_search = hs.config.enable_room_list_search
         self.response_cache = ResponseCache(
-            hs, "room_list"
+            hs.get_clock(), "room_list"
         )  # type: ResponseCache[Tuple[Optional[int], Optional[str], ThirdPartyInstanceID]]
         self.remote_response_cache = ResponseCache(
-            hs, "remote_room_list", timeout_ms=30 * 1000
+            hs.get_clock(), "remote_room_list", timeout_ms=30 * 1000
         )  # type: ResponseCache[Tuple[str, Optional[int], Optional[str], bool, Optional[str]]]
 
     async def get_local_public_room_list(

synapse/handlers/saml_handler.py

@@ -81,6 +81,7 @@ class SamlHandler(BaseHandler):
         # the SsoIdentityProvider protocol type.
         self.idp_icon = None
         self.idp_brand = None
+        self.unstable_idp_brand = None
 
         # a map from saml session id to Saml2SessionData object
         self._outstanding_requests_dict = {}  # type: Dict[str, Saml2SessionData]

synapse/handlers/sso.py

@@ -31,8 +31,8 @@ from urllib.parse import urlencode
 import attr
 from typing_extensions import NoReturn, Protocol
 
-from twisted.web.http import Request
 from twisted.web.iweb import IRequest
+from twisted.web.server import Request
 
 from synapse.api.constants import LoginType
 from synapse.api.errors import Codes, NotFoundError, RedirectException, SynapseError
@@ -98,6 +98,11 @@ class SsoIdentityProvider(Protocol):
         """Optional branding identifier"""
         return None
 
+    @property
+    def unstable_idp_brand(self) -> Optional[str]:
+        """Optional brand identifier for the unstable API (see MSC2858)."""
+        return None
+
     @abc.abstractmethod
     async def handle_redirect_request(
         self,
@@ -456,6 +461,7 @@ class SsoHandler:
 
         await self._auth_handler.complete_sso_login(
             user_id,
+            auth_provider_id,
             request,
             client_redirect_url,
             extra_login_attributes,
@@ -605,6 +611,7 @@ class SsoHandler:
             default_display_name=attributes.display_name,
             bind_emails=attributes.emails,
             user_agent_ips=[(user_agent, ip_address)],
+            auth_provider_id=auth_provider_id,
         )
 
         await self._store.record_user_external_id(
@@ -886,6 +893,7 @@ class SsoHandler:
 
         await self._auth_handler.complete_sso_login(
             user_id,
+            session.auth_provider_id,
             request,
             session.client_redirect_url,
             session.extra_login_attributes,

synapse/handlers/sync.py

@@ -244,7 +244,7 @@ class SyncHandler:
         self.event_sources = hs.get_event_sources()
         self.clock = hs.get_clock()
         self.response_cache = ResponseCache(
-            hs, "sync"
+            hs.get_clock(), "sync"
         )  # type: ResponseCache[Tuple[Any, ...]]
         self.state = hs.get_state_handler()
         self.auth = hs.get_auth()

synapse/http/__init__.py

@@ -14,8 +14,9 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 import re
+from typing import Union
 
-from twisted.internet import task
+from twisted.internet import address, task
 from twisted.web.client import FileBodyProducer
 from twisted.web.iweb import IRequest
 
@@ -53,6 +54,40 @@ class QuieterFileBodyProducer(FileBodyProducer):
             pass
 
 
+def get_request_uri(request: IRequest) -> bytes:
+    """Return the full URI that was requested by the client"""
+    return b"%s://%s%s" % (
+        b"https" if request.isSecure() else b"http",
+        _get_requested_host(request),
+        # despite its name, "request.uri" is only the path and query-string.
+        request.uri,
+    )
+
+
+def _get_requested_host(request: IRequest) -> bytes:
+    hostname = request.getHeader(b"host")
+    if hostname:
+        return hostname
+
+    # no Host header, use the address/port that the request arrived on
+    host = request.getHost()  # type: Union[address.IPv4Address, address.IPv6Address]
+
+    hostname = host.host.encode("ascii")
+
+    if request.isSecure() and host.port == 443:
+        # default port for https
+        return hostname
+
+    if not request.isSecure() and host.port == 80:
+        # default port for http
+        return hostname
+
+    return b"%s:%i" % (
+        hostname,
+        host.port,
+    )
+
+
 def get_request_user_agent(request: IRequest, default: str = "") -> str:
     """Return the last User-Agent header, or the given default."""
     # There could be raw utf-8 bytes in the User-Agent header.

synapse/http/client.py

@@ -39,12 +39,15 @@ from zope.interface import implementer, provider
 from OpenSSL import SSL
 from OpenSSL.SSL import VERIFY_NONE
 from twisted.internet import defer, error as twisted_error, protocol, ssl
+from twisted.internet.address import IPv4Address, IPv6Address
 from twisted.internet.interfaces import (
     IAddress,
     IHostResolution,
     IReactorPluggableNameResolver,
     IResolutionReceiver,
+    ITCPTransport,
 )
+from twisted.internet.protocol import connectionDone
 from twisted.internet.task import Cooperator
 from twisted.python.failure import Failure
 from twisted.web._newclient import ResponseDone
@@ -56,13 +59,20 @@ from twisted.web.client import (
 )
 from twisted.web.http import PotentialDataLoss
 from twisted.web.http_headers import Headers
-from twisted.web.iweb import UNKNOWN_LENGTH, IAgent, IBodyProducer, IResponse
+from twisted.web.iweb import (
+    UNKNOWN_LENGTH,
+    IAgent,
+    IBodyProducer,
+    IPolicyForHTTPS,
+    IResponse,
+)
 
 from synapse.api.errors import Codes, HttpResponseException, SynapseError
 from synapse.http import QuieterFileBodyProducer, RequestTimedOutError, redact_uri
 from synapse.http.proxyagent import ProxyAgent
 from synapse.logging.context import make_deferred_yieldable
 from synapse.logging.opentracing import set_tag, start_active_span, tags
+from synapse.types import ISynapseReactor
 from synapse.util import json_decoder
 from synapse.util.async_helpers import timeout_deferred
 
@@ -150,16 +160,17 @@ class _IPBlacklistingResolver:
     def resolveHostName(
         self, recv: IResolutionReceiver, hostname: str, portNumber: int = 0
     ) -> IResolutionReceiver:
-
-        r = recv()
         addresses = []  # type: List[IAddress]
 
         def _callback() -> None:
-            r.resolutionBegan(None)
-
             has_bad_ip = False
-            for i in addresses:
-                ip_address = IPAddress(i.host)
+            for address in addresses:
+                # We only expect IPv4 and IPv6 addresses since only A/AAAA lookups
+                # should go through this path.
+                if not isinstance(address, (IPv4Address, IPv6Address)):
+                    continue
+
+                ip_address = IPAddress(address.host)
 
                 if check_against_blacklist(
                     ip_address, self._ip_whitelist, self._ip_blacklist
@@ -174,15 +185,15 @@ class _IPBlacklistingResolver:
             # request, but all we can really do from here is claim that there were no
             # valid results.
             if not has_bad_ip:
-                for i in addresses:
-                    r.addressResolved(i)
-            r.resolutionComplete()
+                for address in addresses:
+                    recv.addressResolved(address)
+            recv.resolutionComplete()
 
         @provider(IResolutionReceiver)
         class EndpointReceiver:
             @staticmethod
             def resolutionBegan(resolutionInProgress: IHostResolution) -> None:
-                pass
+                recv.resolutionBegan(resolutionInProgress)
 
             @staticmethod
             def addressResolved(address: IAddress) -> None:
@@ -196,10 +207,10 @@ class _IPBlacklistingResolver:
             EndpointReceiver, hostname, portNumber=portNumber
         )
 
-        return r
+        return recv
 
 
-@implementer(IReactorPluggableNameResolver)
+@implementer(ISynapseReactor)
 class BlacklistingReactorWrapper:
     """
     A Reactor wrapper which will prevent DNS resolution to blacklisted IP
@@ -289,8 +300,7 @@ class SimpleHttpClient:
         treq_args: Dict[str, Any] = {},
         ip_whitelist: Optional[IPSet] = None,
         ip_blacklist: Optional[IPSet] = None,
-        http_proxy: Optional[bytes] = None,
-        https_proxy: Optional[bytes] = None,
+        use_proxy: bool = False,
     ):
         """
         Args:
@@ -300,8 +310,8 @@ class SimpleHttpClient:
                 we may not request.
             ip_whitelist: The whitelisted IP addresses, that we can
                request if it were otherwise caught in a blacklist.
-            http_proxy: proxy server to use for http connections. host[:port]
-            https_proxy: proxy server to use for https connections. host[:port]
+            use_proxy: Whether proxy settings should be discovered and used
+                from conventional environment variables.
         """
         self.hs = hs
 
@@ -325,7 +335,7 @@ class SimpleHttpClient:
             # filters out blacklisted IP addresses, to prevent DNS rebinding.
             self.reactor = BlacklistingReactorWrapper(
                 hs.get_reactor(), self._ip_whitelist, self._ip_blacklist
-            )
+            )  # type: ISynapseReactor
         else:
             self.reactor = hs.get_reactor()
 
@@ -345,9 +355,8 @@ class SimpleHttpClient:
             connectTimeout=15,
             contextFactory=self.hs.get_http_client_context_factory(),
             pool=pool,
-            http_proxy=http_proxy,
-            https_proxy=https_proxy,
-        )
+            use_proxy=use_proxy,
+        )  # type: IAgent
 
         if self._ip_blacklist:
             # If we have an IP blacklist, we then install the blacklisting Agent
@@ -750,7 +759,37 @@ class BodyExceededMaxSize(Exception):
     """The maximum allowed size of the HTTP body was exceeded."""
 
 
+class _DiscardBodyWithMaxSizeProtocol(protocol.Protocol):
+    """A protocol which immediately errors upon receiving data."""
+
+    transport = None  # type: Optional[ITCPTransport]
+
+    def __init__(self, deferred: defer.Deferred):
+        self.deferred = deferred
+
+    def _maybe_fail(self):
+        """
+        Report a max size exceed error and disconnect the first time this is called.
+        """
+        if not self.deferred.called:
+            self.deferred.errback(BodyExceededMaxSize())
+            # Close the connection (forcefully) since all the data will get
+            # discarded anyway.
+            assert self.transport is not None
+            self.transport.abortConnection()
+
+    def dataReceived(self, data: bytes) -> None:
+        self._maybe_fail()
+
+    def connectionLost(self, reason: Failure = connectionDone) -> None:
+        self._maybe_fail()
+
+
 class _ReadBodyWithMaxSizeProtocol(protocol.Protocol):
+    """A protocol which reads body to a stream, erroring if the body exceeds a maximum size."""
+
+    transport = None  # type: Optional[ITCPTransport]
+
     def __init__(
         self, stream: BinaryIO, deferred: defer.Deferred, max_size: Optional[int]
     ):
@@ -773,9 +812,10 @@ class _ReadBodyWithMaxSizeProtocol(protocol.Protocol):
             self.deferred.errback(BodyExceededMaxSize())
             # Close the connection (forcefully) since all the data will get
             # discarded anyway.
+            assert self.transport is not None
             self.transport.abortConnection()
 
-    def connectionLost(self, reason: Failure) -> None:
+    def connectionLost(self, reason: Failure = connectionDone) -> None:
         # If the maximum size was already exceeded, there's nothing to do.
         if self.deferred.called:
             return
@@ -807,13 +847,15 @@ def read_body_with_max_size(
     Returns:
         A Deferred which resolves to the length of the read body.
     """
+    d = defer.Deferred()
+
     # If the Content-Length header gives a size larger than the maximum allowed
     # size, do not bother downloading the body.
     if max_size is not None and response.length != UNKNOWN_LENGTH:
         if response.length > max_size:
-            return defer.fail(BodyExceededMaxSize())
+            response.deliverBody(_DiscardBodyWithMaxSizeProtocol(d))
+            return d
 
-    d = defer.Deferred()
     response.deliverBody(_ReadBodyWithMaxSizeProtocol(stream, d, max_size))
     return d
 
@@ -842,6 +884,7 @@ def encode_query_args(args: Optional[Mapping[str, Union[str, List[str]]]]) -> by
     return query_str.encode("utf8")
 
 
+@implementer(IPolicyForHTTPS)
 class InsecureInterceptableContextFactory(ssl.ContextFactory):
     """
     Factory for PyOpenSSL SSL contexts which accepts any certificate for any domain.

synapse/http/federation/matrix_federation_agent.py

@@ -14,7 +14,7 @@
 # limitations under the License.
 import logging
 import urllib.parse
-from typing import List, Optional
+from typing import Any, Generator, List, Optional
 
 from netaddr import AddrFormatError, IPAddress, IPSet
 from zope.interface import implementer
@@ -35,6 +35,7 @@ from synapse.http.client import BlacklistingAgentWrapper
 from synapse.http.federation.srv_resolver import Server, SrvResolver
 from synapse.http.federation.well_known_resolver import WellKnownResolver
 from synapse.logging.context import make_deferred_yieldable, run_in_background
+from synapse.types import ISynapseReactor
 from synapse.util import Clock
 
 logger = logging.getLogger(__name__)
@@ -68,7 +69,7 @@ class MatrixFederationAgent:
 
     def __init__(
         self,
-        reactor: IReactorCore,
+        reactor: ISynapseReactor,
         tls_client_options_factory: Optional[FederationPolicyForHTTPS],
         user_agent: bytes,
         ip_blacklist: IPSet,
@@ -116,7 +117,7 @@ class MatrixFederationAgent:
         uri: bytes,
         headers: Optional[Headers] = None,
         bodyProducer: Optional[IBodyProducer] = None,
-    ) -> defer.Deferred:
+    ) -> Generator[defer.Deferred, Any, defer.Deferred]:
         """
         Args:
             method: HTTP method: GET/POST/etc
@@ -177,17 +178,17 @@ class MatrixFederationAgent:
         # We need to make sure the host header is set to the netloc of the
         # server and that a user-agent is provided.
         if headers is None:
-            headers = Headers()
+            request_headers = Headers()
         else:
-            headers = headers.copy()
+            request_headers = headers.copy()
 
-        if not headers.hasHeader(b"host"):
-            headers.addRawHeader(b"host", parsed_uri.netloc)
-        if not headers.hasHeader(b"user-agent"):
-            headers.addRawHeader(b"user-agent", self.user_agent)
+        if not request_headers.hasHeader(b"host"):
+            request_headers.addRawHeader(b"host", parsed_uri.netloc)
+        if not request_headers.hasHeader(b"user-agent"):
+            request_headers.addRawHeader(b"user-agent", self.user_agent)
 
         res = yield make_deferred_yieldable(
-            self._agent.request(method, uri, headers, bodyProducer)
+            self._agent.request(method, uri, request_headers, bodyProducer)
         )
 
         return res

synapse/http/federation/well_known_resolver.py

@@ -322,7 +322,8 @@ def _cache_period_from_headers(
 
 def _parse_cache_control(headers: Headers) -> Dict[bytes, Optional[bytes]]:
     cache_controls = {}
-    for hdr in headers.getRawHeaders(b"cache-control", []):
+    cache_control_headers = headers.getRawHeaders(b"cache-control") or []
+    for hdr in cache_control_headers:
         for directive in hdr.split(b","):
             splits = [x.strip() for x in directive.split(b"=", 1)]
             k = splits[0].lower()

synapse/http/matrixfederationclient.py

@@ -59,7 +59,7 @@ from synapse.logging.opentracing import (
     start_active_span,
     tags,
 )
-from synapse.types import JsonDict
+from synapse.types import ISynapseReactor, JsonDict
 from synapse.util import json_decoder
 from synapse.util.async_helpers import timeout_deferred
 from synapse.util.metrics import Measure
@@ -237,14 +237,14 @@ class MatrixFederationHttpClient:
         # addresses, to prevent DNS rebinding.
         self.reactor = BlacklistingReactorWrapper(
             hs.get_reactor(), None, hs.config.federation_ip_range_blacklist
-        )
+        )  # type: ISynapseReactor
 
         user_agent = hs.version_string
         if hs.config.user_agent_suffix:
             user_agent = "%s %s" % (user_agent, hs.config.user_agent_suffix)
         user_agent = user_agent.encode("ascii")
 
-        self.agent = MatrixFederationAgent(
+        federation_agent = MatrixFederationAgent(
             self.reactor,
             tls_client_options_factory,
             user_agent,
@@ -254,7 +254,7 @@ class MatrixFederationHttpClient:
         # Use a BlacklistingAgentWrapper to prevent circumventing the IP
         # blacklist via IP literals in server names
         self.agent = BlacklistingAgentWrapper(
-            self.agent,
+            federation_agent,
             ip_blacklist=hs.config.federation_ip_range_blacklist,
         )
 
@@ -534,9 +534,10 @@ class MatrixFederationHttpClient:
                             response.code, response_phrase, body
                         )
 
-                        # Retry if the error is a 429 (Too Many Requests),
-                        # otherwise just raise a standard HttpResponseException
-                        if response.code == 429:
+                        # Retry if the error is a 5xx or a 429 (Too Many
+                        # Requests), otherwise just raise a standard
+                        # `HttpResponseException`
+                        if 500 <= response.code < 600 or response.code == 429:
                             raise RequestSendFailed(exc, can_retry=True) from exc
                         else:
                             raise exc
@@ -1049,14 +1050,14 @@ def check_content_type_is_json(headers: Headers) -> None:
         RequestSendFailed: if the Content-Type header is missing or isn't JSON
 
     """
-    c_type = headers.getRawHeaders(b"Content-Type")
-    if c_type is None:
+    content_type_headers = headers.getRawHeaders(b"Content-Type")
+    if content_type_headers is None:
         raise RequestSendFailed(
             RuntimeError("No Content-Type header received from remote server"),
             can_retry=False,
         )
 
-    c_type = c_type[0].decode("ascii")  # only the first header
+    c_type = content_type_headers[0].decode("ascii")  # only the first header
     val, options = cgi.parse_header(c_type)
     if val != "application/json":
         raise RequestSendFailed(

synapse/http/proxyagent.py

@@ -14,6 +14,7 @@
 # limitations under the License.
 import logging
 import re
+from urllib.request import getproxies_environment, proxy_bypass_environment
 
 from zope.interface import implementer
 
@@ -58,6 +59,9 @@ class ProxyAgent(_AgentBase):
 
         pool (HTTPConnectionPool|None): connection pool to be used. If None, a
             non-persistent pool instance will be created.
+
+        use_proxy (bool): Whether proxy settings should be discovered and used
+            from conventional environment variables.
     """
 
     def __init__(
@@ -68,8 +72,7 @@ class ProxyAgent(_AgentBase):
         connectTimeout=None,
         bindAddress=None,
         pool=None,
-        http_proxy=None,
-        https_proxy=None,
+        use_proxy=False,
     ):
         _AgentBase.__init__(self, reactor, pool)
 
@@ -84,6 +87,15 @@ class ProxyAgent(_AgentBase):
         if bindAddress is not None:
             self._endpoint_kwargs["bindAddress"] = bindAddress
 
+        http_proxy = None
+        https_proxy = None
+        no_proxy = None
+        if use_proxy:
+            proxies = getproxies_environment()
+            http_proxy = proxies["http"].encode() if "http" in proxies else None
+            https_proxy = proxies["https"].encode() if "https" in proxies else None
+            no_proxy = proxies["no"] if "no" in proxies else None
+
         self.http_proxy_endpoint = _http_proxy_endpoint(
             http_proxy, self.proxy_reactor, **self._endpoint_kwargs
         )
@@ -92,6 +104,8 @@ class ProxyAgent(_AgentBase):
             https_proxy, self.proxy_reactor, **self._endpoint_kwargs
         )
 
+        self.no_proxy = no_proxy
+
         self._policy_for_https = contextFactory
         self._reactor = reactor
 
@@ -139,13 +153,28 @@ class ProxyAgent(_AgentBase):
         pool_key = (parsed_uri.scheme, parsed_uri.host, parsed_uri.port)
         request_path = parsed_uri.originForm
 
-        if parsed_uri.scheme == b"http" and self.http_proxy_endpoint:
+        should_skip_proxy = False
+        if self.no_proxy is not None:
+            should_skip_proxy = proxy_bypass_environment(
+                parsed_uri.host.decode(),
+                proxies={"no": self.no_proxy},
+            )
+
+        if (
+            parsed_uri.scheme == b"http"
+            and self.http_proxy_endpoint
+            and not should_skip_proxy
+        ):
             # Cache *all* connections under the same key, since we are only
             # connecting to a single destination, the proxy:
             pool_key = ("http-proxy", self.http_proxy_endpoint)
             endpoint = self.http_proxy_endpoint
             request_path = uri
-        elif parsed_uri.scheme == b"https" and self.https_proxy_endpoint:
+        elif (
+            parsed_uri.scheme == b"https"
+            and self.https_proxy_endpoint
+            and not should_skip_proxy
+        ):
             endpoint = HTTPConnectProxyEndpoint(
                 self.proxy_reactor,
                 self.https_proxy_endpoint,

synapse/http/server.py

@@ -21,6 +21,7 @@ import logging
 import types
 import urllib
 from http import HTTPStatus
+from inspect import isawaitable
 from io import BytesIO
 from typing import (
     Any,
@@ -30,6 +31,7 @@ from typing import (
     Iterable,
     Iterator,
     List,
+    Optional,
     Pattern,
     Tuple,
     Union,
@@ -79,10 +81,12 @@ def return_json_error(f: failure.Failure, request: SynapseRequest) -> None:
     """Sends a JSON error response to clients."""
 
     if f.check(SynapseError):
-        error_code = f.value.code
-        error_dict = f.value.error_dict()
+        # mypy doesn't understand that f.check asserts the type.
+        exc = f.value  # type: SynapseError  # type: ignore
+        error_code = exc.code
+        error_dict = exc.error_dict()
 
-        logger.info("%s SynapseError: %s - %s", request, error_code, f.value.msg)
+        logger.info("%s SynapseError: %s - %s", request, error_code, exc.msg)
     else:
         error_code = 500
         error_dict = {"error": "Internal server error", "errcode": Codes.UNKNOWN}
@@ -91,7 +95,7 @@ def return_json_error(f: failure.Failure, request: SynapseRequest) -> None:
             "Failed handle request via %r: %r",
             request.request_metrics.name,
             request,
-            exc_info=(f.type, f.value, f.getTracebackObject()),
+            exc_info=(f.type, f.value, f.getTracebackObject()),  # type: ignore
         )
 
     # Only respond with an error response if we haven't already started writing,
@@ -128,7 +132,8 @@ def return_html_error(
             `{msg}` placeholders), or a jinja2 template
     """
     if f.check(CodeMessageException):
-        cme = f.value
+        # mypy doesn't understand that f.check asserts the type.
+        cme = f.value  # type: CodeMessageException  # type: ignore
         code = cme.code
         msg = cme.msg
 
@@ -142,7 +147,7 @@ def return_html_error(
             logger.error(
                 "Failed handle request %r",
                 request,
-                exc_info=(f.type, f.value, f.getTracebackObject()),
+                exc_info=(f.type, f.value, f.getTracebackObject()),  # type: ignore
             )
     else:
         code = HTTPStatus.INTERNAL_SERVER_ERROR
@@ -151,7 +156,7 @@ def return_html_error(
         logger.error(
             "Failed handle request %r",
             request,
-            exc_info=(f.type, f.value, f.getTracebackObject()),
+            exc_info=(f.type, f.value, f.getTracebackObject()),  # type: ignore
         )
 
     if isinstance(error_template, str):
@@ -278,7 +283,7 @@ class _AsyncResource(resource.Resource, metaclass=abc.ABCMeta):
             raw_callback_return = method_handler(request)
 
             # Is it synchronous? We'll allow this for now.
-            if isinstance(raw_callback_return, (defer.Deferred, types.CoroutineType)):
+            if isawaitable(raw_callback_return):
                 callback_return = await raw_callback_return
             else:
                 callback_return = raw_callback_return  # type: ignore
@@ -399,8 +404,10 @@ class JsonResource(DirectServeJsonResource):
             A tuple of the callback to use, the name of the servlet, and the
             key word arguments to pass to the callback
         """
+        # At this point the path must be bytes.
+        request_path_bytes = request.path  # type: bytes  # type: ignore
+        request_path = request_path_bytes.decode("ascii")
         # Treat HEAD requests as GET requests.
-        request_path = request.path.decode("ascii")
         request_method = request.method
         if request_method == b"HEAD":
             request_method = b"GET"
@@ -551,7 +558,7 @@ class _ByteProducer:
         request: Request,
         iterator: Iterator[bytes],
     ):
-        self._request = request
+        self._request = request  # type: Optional[Request]
         self._iterator = iterator
         self._paused = False
 
@@ -563,7 +570,7 @@ class _ByteProducer:
         """
         Send a list of bytes as a chunk of a response.
         """
-        if not data:
+        if not data or not self._request:
             return
         self._request.write(b"".join(data))
 

synapse/http/site.py

@@ -14,8 +14,12 @@
 import contextlib
 import logging
 import time
-from typing import Optional, Union
+from typing import Optional, Type, Union
 
+import attr
+from zope.interface import implementer
+
+from twisted.internet.interfaces import IAddress
 from twisted.python.failure import Failure
 from twisted.web.server import Request, Site
 
@@ -53,7 +57,7 @@ class SynapseRequest(Request):
 
     def __init__(self, channel, *args, **kw):
         Request.__init__(self, channel, *args, **kw)
-        self.site = channel.site
+        self.site = channel.site  # type: SynapseSite
         self._channel = channel  # this is used by the tests
         self.start_time = 0.0
 
@@ -92,25 +96,34 @@ class SynapseRequest(Request):
     def get_request_id(self):
         return "%s-%i" % (self.get_method(), self.request_seq)
 
-    def get_redacted_uri(self):
-        uri = self.uri
+    def get_redacted_uri(self) -> str:
+        """Gets the redacted URI associated with the request (or placeholder if the URI
+        has not yet been received).
+
+        Note: This is necessary as the placeholder value in twisted is str
+        rather than bytes, so we need to sanitise `self.uri`.
+
+        Returns:
+            The redacted URI as a string.
+        """
+        uri = self.uri  # type: Union[bytes, str]
         if isinstance(uri, bytes):
-            uri = self.uri.decode("ascii", errors="replace")
+            uri = uri.decode("ascii", errors="replace")
         return redact_uri(uri)
 
-    def get_method(self):
-        """Gets the method associated with the request (or placeholder if not
-        method has yet been received).
+    def get_method(self) -> str:
+        """Gets the method associated with the request (or placeholder if method
+        has not yet been received).
 
         Note: This is necessary as the placeholder value in twisted is str
         rather than bytes, so we need to sanitise `self.method`.
 
         Returns:
-            str
+            The request method as a string.
         """
-        method = self.method
+        method = self.method  # type: Union[bytes, str]
         if isinstance(method, bytes):
-            method = self.method.decode("ascii")
+            return self.method.decode("ascii")
         return method
 
     def render(self, resrc):
@@ -333,27 +346,78 @@ class SynapseRequest(Request):
 
 
 class XForwardedForRequest(SynapseRequest):
-    def __init__(self, *args, **kw):
-        SynapseRequest.__init__(self, *args, **kw)
+    """Request object which honours proxy headers
 
-    """
-    Add a layer on top of another request that only uses the value of an
-    X-Forwarded-For header as the result of C{getClientIP}.
+    Extends SynapseRequest to replace getClientIP, getClientAddress, and isSecure with
+    information from request headers.
     """
 
-    def getClientIP(self):
-        """
-        @return: The client address (the first address) in the value of the
-            I{X-Forwarded-For header}.  If the header is not present, return
-            C{b"-"}.
-        """
-        return (
-            self.requestHeaders.getRawHeaders(b"x-forwarded-for", [b"-"])[0]
-            .split(b",")[0]
-            .strip()
-            .decode("ascii")
+    # the client IP and ssl flag, as extracted from the headers.
+    _forwarded_for = None  # type: Optional[_XForwardedForAddress]
+    _forwarded_https = False  # type: bool
+
+    def requestReceived(self, command, path, version):
+        # this method is called by the Channel once the full request has been
+        # received, to dispatch the request to a resource.
+        # We can use it to set the IP address and protocol according to the
+        # headers.
+        self._process_forwarded_headers()
+        return super().requestReceived(command, path, version)
+
+    def _process_forwarded_headers(self):
+        headers = self.requestHeaders.getRawHeaders(b"x-forwarded-for")
+        if not headers:
+            return
+
+        # for now, we just use the first x-forwarded-for header. Really, we ought
+        # to start from the client IP address, and check whether it is trusted; if it
+        # is, work backwards through the headers until we find an untrusted address.
+        # see https://github.com/matrix-org/synapse/issues/9471
+        self._forwarded_for = _XForwardedForAddress(
+            headers[0].split(b",")[0].strip().decode("ascii")
         )
 
+        # if we got an x-forwarded-for header, also look for an x-forwarded-proto header
+        header = self.getHeader(b"x-forwarded-proto")
+        if header is not None:
+            self._forwarded_https = header.lower() == b"https"
+        else:
+            # this is done largely for backwards-compatibility so that people that
+            # haven't set an x-forwarded-proto header don't get a redirect loop.
+            logger.warning(
+                "forwarded request lacks an x-forwarded-proto header: assuming https"
+            )
+            self._forwarded_https = True
+
+    def isSecure(self):
+        if self._forwarded_https:
+            return True
+        return super().isSecure()
+
+    def getClientIP(self) -> str:
+        """
+        Return the IP address of the client who submitted this request.
+
+        This method is deprecated.  Use getClientAddress() instead.
+        """
+        if self._forwarded_for is not None:
+            return self._forwarded_for.host
+        return super().getClientIP()
+
+    def getClientAddress(self) -> IAddress:
+        """
+        Return the address of the client who submitted this request.
+        """
+        if self._forwarded_for is not None:
+            return self._forwarded_for
+        return super().getClientAddress()
+
+
+@implementer(IAddress)
+@attr.s(frozen=True, slots=True)
+class _XForwardedForAddress:
+    host = attr.ib(type=str)
+
 
 class SynapseSite(Site):
     """
@@ -377,7 +441,9 @@ class SynapseSite(Site):
 
         assert config.http_options is not None
         proxied = config.http_options.x_forwarded
-        self.requestFactory = XForwardedForRequest if proxied else SynapseRequest
+        self.requestFactory = (
+            XForwardedForRequest if proxied else SynapseRequest
+        )  # type: Type[Request]
         self.access_logger = logging.getLogger(logger_name)
         self.server_version_string = server_version_string.encode("ascii")
 

synapse/logging/_remote.py

@@ -32,8 +32,9 @@ from twisted.internet.endpoints import (
     TCP4ClientEndpoint,
     TCP6ClientEndpoint,
 )
-from twisted.internet.interfaces import IPushProducer, ITransport
+from twisted.internet.interfaces import IPushProducer, IStreamClientEndpoint
 from twisted.internet.protocol import Factory, Protocol
+from twisted.internet.tcp import Connection
 from twisted.python.failure import Failure
 
 logger = logging.getLogger(__name__)
@@ -52,7 +53,9 @@ class LogProducer:
         format: A callable to format the log record to a string.
     """
 
-    transport = attr.ib(type=ITransport)
+    # This is essentially ITCPTransport, but that is missing certain fields
+    # (connected and registerProducer) which are part of the implementation.
+    transport = attr.ib(type=Connection)
     _format = attr.ib(type=Callable[[logging.LogRecord], str])
     _buffer = attr.ib(type=deque)
     _paused = attr.ib(default=False, type=bool, init=False)
@@ -121,7 +124,9 @@ class RemoteHandler(logging.Handler):
         try:
             ip = ip_address(self.host)
             if isinstance(ip, IPv4Address):
-                endpoint = TCP4ClientEndpoint(_reactor, self.host, self.port)
+                endpoint = TCP4ClientEndpoint(
+                    _reactor, self.host, self.port
+                )  # type: IStreamClientEndpoint
             elif isinstance(ip, IPv6Address):
                 endpoint = TCP6ClientEndpoint(_reactor, self.host, self.port)
             else:
@@ -147,8 +152,6 @@ class RemoteHandler(logging.Handler):
         if self._connection_waiter:
             return
 
-        self._connection_waiter = self._service.whenConnected(failAfterFailures=1)
-
         def fail(failure: Failure) -> None:
             # If the Deferred was cancelled (e.g. during shutdown) do not try to
             # reconnect (this will cause an infinite loop of errors).
@@ -161,9 +164,13 @@ class RemoteHandler(logging.Handler):
             self._connect()
 
         def writer(result: Protocol) -> None:
+            # Force recognising transport as a Connection and not the more
+            # generic ITransport.
+            transport = result.transport  # type: Connection  # type: ignore
+
             # We have a connection. If we already have a producer, and its
             # transport is the same, just trigger a resumeProducing.
-            if self._producer and result.transport is self._producer.transport:
+            if self._producer and transport is self._producer.transport:
                 self._producer.resumeProducing()
                 self._connection_waiter = None
                 return
@@ -175,14 +182,16 @@ class RemoteHandler(logging.Handler):
             # Make a new producer and start it.
             self._producer = LogProducer(
                 buffer=self._buffer,
-                transport=result.transport,
+                transport=transport,
                 format=self.format,
             )
-            result.transport.registerProducer(self._producer, True)
+            transport.registerProducer(self._producer, True)
             self._producer.resumeProducing()
             self._connection_waiter = None
 
-        self._connection_waiter.addCallbacks(writer, fail)
+        deferred = self._service.whenConnected(failAfterFailures=1)  # type: Deferred
+        deferred.addCallbacks(writer, fail)
+        self._connection_waiter = deferred
 
     def _handle_pressure(self) -> None:
         """

synapse/logging/context.py

@@ -669,7 +669,7 @@ def preserve_fn(f):
     return g
 
 
-def run_in_background(f, *args, **kwargs):
+def run_in_background(f, *args, **kwargs) -> defer.Deferred:
     """Calls a function, ensuring that the current context is restored after
     return from the function, and that the sentinel context is set once the
     deferred returned by the function completes.
@@ -697,8 +697,10 @@ def run_in_background(f, *args, **kwargs):
     if isinstance(res, types.CoroutineType):
         res = defer.ensureDeferred(res)
 
+    # At this point we should have a Deferred, if not then f was a synchronous
+    # function, wrap it in a Deferred for consistency.
     if not isinstance(res, defer.Deferred):
-        return res
+        return defer.succeed(res)
 
     if res.called and not res.paused:
         # The function should have maintained the logcontext, so we can

synapse/metrics/__init__.py

@@ -527,7 +527,7 @@ class ReactorLastSeenMetric:
 REGISTRY.register(ReactorLastSeenMetric())
 
 
-def runUntilCurrentTimer(func):
+def runUntilCurrentTimer(reactor, func):
     @functools.wraps(func)
     def f(*args, **kwargs):
         now = reactor.seconds()
@@ -590,13 +590,14 @@ def runUntilCurrentTimer(func):
 
 try:
     # Ensure the reactor has all the attributes we expect
-    reactor.runUntilCurrent
-    reactor._newTimedCalls
-    reactor.threadCallQueue
+    reactor.seconds  # type: ignore
+    reactor.runUntilCurrent  # type: ignore
+    reactor._newTimedCalls  # type: ignore
+    reactor.threadCallQueue  # type: ignore
 
     # runUntilCurrent is called when we have pending calls. It is called once
     # per iteratation after fd polling.
-    reactor.runUntilCurrent = runUntilCurrentTimer(reactor.runUntilCurrent)
+    reactor.runUntilCurrent = runUntilCurrentTimer(reactor, reactor.runUntilCurrent)  # type: ignore
 
     # We manually run the GC each reactor tick so that we can get some metrics
     # about time spent doing GC,

synapse/module_api/__init__.py

@@ -14,7 +14,7 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 import logging
-from typing import TYPE_CHECKING, Iterable, Optional, Tuple
+from typing import TYPE_CHECKING, Any, Generator, Iterable, Optional, Tuple
 
 from twisted.internet import defer
 
@@ -203,11 +203,26 @@ class ModuleApi:
         )
 
     def generate_short_term_login_token(
-        self, user_id: str, duration_in_ms: int = (2 * 60 * 1000)
+        self,
+        user_id: str,
+        duration_in_ms: int = (2 * 60 * 1000),
+        auth_provider_id: str = "",
     ) -> str:
-        """Generate a login token suitable for m.login.token authentication"""
+        """Generate a login token suitable for m.login.token authentication
+
+        Args:
+            user_id: gives the ID of the user that the token is for
+
+            duration_in_ms: the time that the token will be valid for
+
+            auth_provider_id: the ID of the SSO IdP that the user used to authenticate
+               to get this token, if any. This is encoded in the token so that
+               /login can report stats on number of successful logins by IdP.
+        """
         return self._hs.get_macaroon_generator().generate_short_term_login_token(
-            user_id, duration_in_ms
+            user_id,
+            auth_provider_id,
+            duration_in_ms,
         )
 
     @defer.inlineCallbacks
@@ -276,6 +291,7 @@ class ModuleApi:
         """
         self._auth_handler._complete_sso_login(
             registered_user_id,
+            "<unknown>",
             request,
             client_redirect_url,
         )
@@ -286,6 +302,7 @@ class ModuleApi:
         request: SynapseRequest,
         client_redirect_url: str,
         new_user: bool = False,
+        auth_provider_id: str = "<unknown>",
     ):
         """Complete a SSO login by redirecting the user to a page to confirm whether they
         want their access token sent to `client_redirect_url`, or redirect them to that
@@ -299,15 +316,21 @@ class ModuleApi:
                 redirect them directly if whitelisted).
             new_user: set to true to use wording for the consent appropriate to a user
                 who has just registered.
+            auth_provider_id: the ID of the SSO IdP which was used to log in. This
+                is used to track counts of sucessful logins by IdP.
         """
         await self._auth_handler.complete_sso_login(
-            registered_user_id, request, client_redirect_url, new_user=new_user
+            registered_user_id,
+            auth_provider_id,
+            request,
+            client_redirect_url,
+            new_user=new_user,
         )
 
     @defer.inlineCallbacks
     def get_state_events_in_room(
         self, room_id: str, types: Iterable[Tuple[str, Optional[str]]]
-    ) -> defer.Deferred:
+    ) -> Generator[defer.Deferred, Any, defer.Deferred]:
         """Gets current state events for the given room.
 
         (This is exposed for compatibility with the old SpamCheckerApi. We should

synapse/push/emailpusher.py

@@ -16,8 +16,8 @@
 import logging
 from typing import TYPE_CHECKING, Dict, List, Optional
 
-from twisted.internet.base import DelayedCall
 from twisted.internet.error import AlreadyCalled, AlreadyCancelled
+from twisted.internet.interfaces import IDelayedCall
 
 from synapse.metrics.background_process_metrics import run_as_background_process
 from synapse.push import Pusher, PusherConfig, ThrottleParams
@@ -66,7 +66,7 @@ class EmailPusher(Pusher):
 
         self.store = self.hs.get_datastore()
         self.email = pusher_config.pushkey
-        self.timed_call = None  # type: Optional[DelayedCall]
+        self.timed_call = None  # type: Optional[IDelayedCall]
         self.throttle_params = {}  # type: Dict[str, ThrottleParams]
         self._inited = False
 

synapse/push/httppusher.py

@@ -15,11 +15,12 @@
 # limitations under the License.
 import logging
 import urllib.parse
-from typing import TYPE_CHECKING, Any, Dict, Iterable, Union
+from typing import TYPE_CHECKING, Any, Dict, Iterable, Optional, Union
 
 from prometheus_client import Counter
 
 from twisted.internet.error import AlreadyCalled, AlreadyCancelled
+from twisted.internet.interfaces import IDelayedCall
 
 from synapse.api.constants import EventTypes
 from synapse.events import EventBase
@@ -71,9 +72,10 @@ class HttpPusher(Pusher):
         self.data = pusher_config.data
         self.backoff_delay = HttpPusher.INITIAL_BACKOFF_SEC
         self.failing_since = pusher_config.failing_since
-        self.timed_call = None
+        self.timed_call = None  # type: Optional[IDelayedCall]
         self._is_processing = False
         self._group_unread_count_by_room = hs.config.push_group_unread_count_by_room
+        self._pusherpool = hs.get_pusherpool()
 
         self.data = pusher_config.data
         if self.data is None:
@@ -299,7 +301,7 @@ class HttpPusher(Pusher):
                     )
                 else:
                     logger.info("Pushkey %s was rejected: removing", pk)
-                    await self.hs.remove_pusher(self.app_id, pk, self.user_id)
+                    await self._pusherpool.remove_pusher(self.app_id, pk, self.user_id)
         return True
 
     async def _build_notification_dict(

synapse/push/pusherpool.py

@@ -19,12 +19,14 @@ from typing import TYPE_CHECKING, Dict, Iterable, Optional
 
 from prometheus_client import Gauge
 
+from synapse.api.errors import Codes, SynapseError
 from synapse.metrics.background_process_metrics import (
     run_as_background_process,
     wrap_as_background_process,
 )
 from synapse.push import Pusher, PusherConfig, PusherConfigException
 from synapse.push.pusher import PusherFactory
+from synapse.replication.http.push import ReplicationRemovePusherRestServlet
 from synapse.types import JsonDict, RoomStreamToken
 from synapse.util.async_helpers import concurrently_execute
 
@@ -58,7 +60,6 @@ class PusherPool:
     def __init__(self, hs: "HomeServer"):
         self.hs = hs
         self.pusher_factory = PusherFactory(hs)
-        self._should_start_pushers = hs.config.start_pushers
         self.store = self.hs.get_datastore()
         self.clock = self.hs.get_clock()
 
@@ -67,6 +68,16 @@ class PusherPool:
         # We shard the handling of push notifications by user ID.
         self._pusher_shard_config = hs.config.push.pusher_shard_config
         self._instance_name = hs.get_instance_name()
+        self._should_start_pushers = (
+            self._instance_name in self._pusher_shard_config.instances
+        )
+
+        # We can only delete pushers on master.
+        self._remove_pusher_client = None
+        if hs.config.worker.worker_app:
+            self._remove_pusher_client = ReplicationRemovePusherRestServlet.make_client(
+                hs
+            )
 
         # Record the last stream ID that we were poked about so we can get
         # changes since then. We set this to the current max stream ID on
@@ -103,6 +114,11 @@ class PusherPool:
             The newly created pusher.
         """
 
+        if kind == "email":
+            email_owner = await self.store.get_user_id_by_threepid("email", pushkey)
+            if email_owner != user_id:
+                raise SynapseError(400, "Email not found", Codes.THREEPID_NOT_FOUND)
+
         time_now_msec = self.clock.time_msec()
 
         # create the pusher setting last_stream_ordering to the current maximum
@@ -175,9 +191,6 @@ class PusherPool:
             user_id: user to remove pushers for
             access_tokens: access token *ids* to remove pushers for
         """
-        if not self._pusher_shard_config.should_handle(self._instance_name, user_id):
-            return
-
         tokens = set(access_tokens)
         for p in await self.store.get_pushers_by_user_id(user_id):
             if p.access_token in tokens:
@@ -380,6 +393,12 @@ class PusherPool:
 
             synapse_pushers.labels(type(pusher).__name__, pusher.app_id).dec()
 
-        await self.store.delete_pusher_by_app_id_pushkey_user_id(
-            app_id, pushkey, user_id
-        )
+        # We can only delete pushers on master.
+        if self._remove_pusher_client:
+            await self._remove_pusher_client(
+                app_id=app_id, pushkey=pushkey, user_id=user_id
+            )
+        else:
+            await self.store.delete_pusher_by_app_id_pushkey_user_id(
+                app_id, pushkey, user_id
+            )

synapse/python_dependencies.py

@@ -82,6 +82,9 @@ REQUIREMENTS = [
     "Jinja2>=2.9",
     "bleach>=1.4.3",
     "typing-extensions>=3.7.4",
+    # We enforce that we have a `cryptography` version that bundles an `openssl`
+    # with the latest security patches.
+    "cryptography>=3.4.7;python_version>='3.6'",
 ]
 
 CONDITIONAL_REQUIREMENTS = {
@@ -106,6 +109,9 @@ CONDITIONAL_REQUIREMENTS = {
         "pysaml2>=4.5.0;python_version>='3.6'",
     ],
     "oidc": ["authlib>=0.14.0"],
+    # systemd-python is necessary for logging to the systemd journal via
+    # `systemd.journal.JournalHandler`, as is documented in
+    # `contrib/systemd/log_config.yaml`.
     "systemd": ["systemd-python>=231"],
     "url_preview": ["lxml>=3.5.0"],
     "sentry": ["sentry-sdk>=0.7.2"],

synapse/replication/http/__init__.py

@@ -21,6 +21,7 @@ from synapse.replication.http import (
     login,
     membership,
     presence,
+    push,
     register,
     send_event,
     streams,
@@ -42,6 +43,7 @@ class ReplicationRestResource(JsonResource):
         membership.register_servlets(hs, self)
         streams.register_servlets(hs, self)
         account_data.register_servlets(hs, self)
+        push.register_servlets(hs, self)
 
         # The following can't currently be instantiated on workers.
         if hs.config.worker.worker_app is None:

synapse/replication/http/_base.py

@@ -18,7 +18,7 @@ import logging
 import re
 import urllib
 from inspect import signature
-from typing import Dict, List, Tuple
+from typing import TYPE_CHECKING, Dict, List, Tuple
 
 from prometheus_client import Counter, Gauge
 
@@ -28,6 +28,9 @@ from synapse.logging.opentracing import inject_active_span_byte_dict, trace
 from synapse.util.caches.response_cache import ResponseCache
 from synapse.util.stringutils import random_string
 
+if TYPE_CHECKING:
+    from synapse.server import HomeServer
+
 logger = logging.getLogger(__name__)
 
 _pending_outgoing_requests = Gauge(
@@ -88,10 +91,10 @@ class ReplicationEndpoint(metaclass=abc.ABCMeta):
     CACHE = True
     RETRY_ON_TIMEOUT = True
 
-    def __init__(self, hs):
+    def __init__(self, hs: "HomeServer"):
         if self.CACHE:
             self.response_cache = ResponseCache(
-                hs, "repl." + self.NAME, timeout_ms=30 * 60 * 1000
+                hs.get_clock(), "repl." + self.NAME, timeout_ms=30 * 60 * 1000
             )  # type: ResponseCache[str]
 
         # We reserve `instance_name` as a parameter to sending requests, so we

synapse/replication/http/federation.py

@@ -213,8 +213,9 @@ class ReplicationGetQueryRestServlet(ReplicationEndpoint):
             content = parse_json_object_from_request(request)
 
             args = content["args"]
+            args["origin"] = content["origin"]
 
-        logger.info("Got %r query", query_type)
+        logger.info("Got %r query from %s", query_type, args["origin"])
 
         result = await self.registry.on_query(query_type, args)
 

synapse/replication/http/login.py

@@ -61,7 +61,7 @@ class RegisterDeviceReplicationServlet(ReplicationEndpoint):
         is_guest = content["is_guest"]
         is_appservice_ghost = content["is_appservice_ghost"]
 
-        device_id, access_token = await self.registration_handler.register_device(
+        res = await self.registration_handler.register_device_inner(
             user_id,
             device_id,
             initial_display_name,
@@ -69,7 +69,7 @@ class RegisterDeviceReplicationServlet(ReplicationEndpoint):
             is_appservice_ghost=is_appservice_ghost,
         )
 
-        return 200, {"device_id": device_id, "access_token": access_token}
+        return 200, res
 
 
 def register_servlets(hs, http_server):

synapse/replication/http/membership.py

@@ -15,9 +15,10 @@
 import logging
 from typing import TYPE_CHECKING, List, Optional, Tuple
 
-from twisted.web.http import Request
+from twisted.web.server import Request
 
 from synapse.http.servlet import parse_json_object_from_request
+from synapse.http.site import SynapseRequest
 from synapse.replication.http._base import ReplicationEndpoint
 from synapse.types import JsonDict, Requester, UserID
 from synapse.util.distributor import user_left_room
@@ -78,7 +79,7 @@ class ReplicationRemoteJoinRestServlet(ReplicationEndpoint):
         }
 
     async def _handle_request(  # type: ignore
-        self, request: Request, room_id: str, user_id: str
+        self, request: SynapseRequest, room_id: str, user_id: str
     ) -> Tuple[int, JsonDict]:
         content = parse_json_object_from_request(request)
 
@@ -86,7 +87,6 @@ class ReplicationRemoteJoinRestServlet(ReplicationEndpoint):
         event_content = content["content"]
 
         requester = Requester.deserialize(self.store, content["requester"])
-
         request.requester = requester
 
         logger.info("remote_join: %s into room: %s", user_id, room_id)
@@ -147,7 +147,7 @@ class ReplicationRemoteRejectInviteRestServlet(ReplicationEndpoint):
         }
 
     async def _handle_request(  # type: ignore
-        self, request: Request, invite_event_id: str
+        self, request: SynapseRequest, invite_event_id: str
     ) -> Tuple[int, JsonDict]:
         content = parse_json_object_from_request(request)
 
@@ -155,7 +155,6 @@ class ReplicationRemoteRejectInviteRestServlet(ReplicationEndpoint):
         event_content = content["content"]
 
         requester = Requester.deserialize(self.store, content["requester"])
-
         request.requester = requester
 
         # hopefully we're now on the master, so this won't recurse!

synapse/replication/http/push.py

@@ -0,0 +1,72 @@
+# -*- coding: utf-8 -*-
+# Copyright 2021 The Matrix.org Foundation C.I.C.
+#
+# 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.http.servlet import parse_json_object_from_request
+from synapse.replication.http._base import ReplicationEndpoint
+
+if TYPE_CHECKING:
+    from synapse.server import HomeServer
+
+logger = logging.getLogger(__name__)
+
+
+class ReplicationRemovePusherRestServlet(ReplicationEndpoint):
+    """Deletes the given pusher.
+
+    Request format:
+
+        POST /_synapse/replication/remove_pusher/:user_id
+
+        {
+            "app_id": "<some_id>",
+            "pushkey": "<some_key>"
+        }
+
+    """
+
+    NAME = "add_user_account_data"
+    PATH_ARGS = ("user_id",)
+    CACHE = False
+
+    def __init__(self, hs: "HomeServer"):
+        super().__init__(hs)
+
+        self.pusher_pool = hs.get_pusherpool()
+
+    @staticmethod
+    async def _serialize_payload(app_id, pushkey, user_id):
+        payload = {
+            "app_id": app_id,
+            "pushkey": pushkey,
+        }
+
+        return payload
+
+    async def _handle_request(self, request, user_id):
+        content = parse_json_object_from_request(request)
+
+        app_id = content["app_id"]
+        pushkey = content["pushkey"]
+
+        await self.pusher_pool.remove_pusher(app_id, pushkey, user_id)
+
+        return 200, {}
+
+
+def register_servlets(hs, http_server):
+    ReplicationRemovePusherRestServlet(hs).register(http_server)

synapse/replication/tcp/client.py

@@ -108,9 +108,7 @@ class ReplicationDataHandler:
 
         # Map from stream to list of deferreds waiting for the stream to
         # arrive at a particular position. The lists are sorted by stream position.
-        self._streams_to_waiters = (
-            {}
-        )  # type: Dict[str, List[Tuple[int, Deferred[None]]]]
+        self._streams_to_waiters = {}  # type: Dict[str, List[Tuple[int, Deferred]]]
 
     async def on_rdata(
         self, stream_name: str, instance_name: str, token: int, rows: list

synapse/replication/tcp/commands.py

@@ -325,31 +325,6 @@ class FederationAckCommand(Command):
         return "%s %s" % (self.instance_name, self.token)
 
 
-class RemovePusherCommand(Command):
-    """Sent by the client to request the master remove the given pusher.
-
-    Format::
-
-        REMOVE_PUSHER <app_id> <push_key> <user_id>
-    """
-
-    NAME = "REMOVE_PUSHER"
-
-    def __init__(self, app_id, push_key, user_id):
-        self.user_id = user_id
-        self.app_id = app_id
-        self.push_key = push_key
-
-    @classmethod
-    def from_line(cls, line):
-        app_id, push_key, user_id = line.split(" ", 2)
-
-        return cls(app_id, push_key, user_id)
-
-    def to_line(self):
-        return " ".join((self.app_id, self.push_key, self.user_id))
-
-
 class UserIpCommand(Command):
     """Sent periodically when a worker sees activity from a client.
 
@@ -416,7 +391,6 @@ _COMMANDS = (
     ReplicateCommand,
     UserSyncCommand,
     FederationAckCommand,
-    RemovePusherCommand,
     UserIpCommand,
     RemoteServerUpCommand,
     ClearUserSyncsCommand,
@@ -443,7 +417,6 @@ VALID_CLIENT_COMMANDS = (
     UserSyncCommand.NAME,
     ClearUserSyncsCommand.NAME,
     FederationAckCommand.NAME,
-    RemovePusherCommand.NAME,
     UserIpCommand.NAME,
     ErrorCommand.NAME,
     RemoteServerUpCommand.NAME,

synapse/replication/tcp/handler.py

@@ -44,12 +44,11 @@ from synapse.replication.tcp.commands import (
     PositionCommand,
     RdataCommand,
     RemoteServerUpCommand,
-    RemovePusherCommand,
     ReplicateCommand,
     UserIpCommand,
     UserSyncCommand,
 )
-from synapse.replication.tcp.protocol import AbstractConnection
+from synapse.replication.tcp.protocol import IReplicationConnection
 from synapse.replication.tcp.streams import (
     STREAMS_MAP,
     AccountDataStream,
@@ -83,7 +82,7 @@ user_ip_cache_counter = Counter("synapse_replication_tcp_resource_user_ip_cache"
 
 # the type of the entries in _command_queues_by_stream
 _StreamCommandQueue = Deque[
-    Tuple[Union[RdataCommand, PositionCommand], AbstractConnection]
+    Tuple[Union[RdataCommand, PositionCommand], IReplicationConnection]
 ]
 
 
@@ -175,7 +174,7 @@ class ReplicationCommandHandler:
 
         # The currently connected connections. (The list of places we need to send
         # outgoing replication commands to.)
-        self._connections = []  # type: List[AbstractConnection]
+        self._connections = []  # type: List[IReplicationConnection]
 
         LaterGauge(
             "synapse_replication_tcp_resource_total_connections",
@@ -198,7 +197,7 @@ class ReplicationCommandHandler:
 
         # For each connection, the incoming stream names that have received a POSITION
         # from that connection.
-        self._streams_by_connection = {}  # type: Dict[AbstractConnection, Set[str]]
+        self._streams_by_connection = {}  # type: Dict[IReplicationConnection, Set[str]]
 
         LaterGauge(
             "synapse_replication_tcp_command_queue",
@@ -221,7 +220,7 @@ class ReplicationCommandHandler:
             self._server_notices_sender = hs.get_server_notices_sender()
 
     def _add_command_to_stream_queue(
-        self, conn: AbstractConnection, cmd: Union[RdataCommand, PositionCommand]
+        self, conn: IReplicationConnection, cmd: Union[RdataCommand, PositionCommand]
     ) -> None:
         """Queue the given received command for processing
 
@@ -268,7 +267,7 @@ class ReplicationCommandHandler:
     async def _process_command(
         self,
         cmd: Union[PositionCommand, RdataCommand],
-        conn: AbstractConnection,
+        conn: IReplicationConnection,
         stream_name: str,
     ) -> None:
         if isinstance(cmd, PositionCommand):
@@ -303,7 +302,7 @@ class ReplicationCommandHandler:
                 hs, outbound_redis_connection
             )
             hs.get_reactor().connectTCP(
-                hs.config.redis.redis_host,
+                hs.config.redis.redis_host.encode(),
                 hs.config.redis.redis_port,
                 self._factory,
             )
@@ -312,7 +311,7 @@ class ReplicationCommandHandler:
             self._factory = DirectTcpReplicationClientFactory(hs, client_name, self)
             host = hs.config.worker_replication_host
             port = hs.config.worker_replication_port
-            hs.get_reactor().connectTCP(host, port, self._factory)
+            hs.get_reactor().connectTCP(host.encode(), port, self._factory)
 
     def get_streams(self) -> Dict[str, Stream]:
         """Get a map from stream name to all streams."""
@@ -322,10 +321,10 @@ class ReplicationCommandHandler:
         """Get a list of streams that this instances replicates."""
         return self._streams_to_replicate
 
-    def on_REPLICATE(self, conn: AbstractConnection, cmd: ReplicateCommand):
+    def on_REPLICATE(self, conn: IReplicationConnection, cmd: ReplicateCommand):
         self.send_positions_to_connection(conn)
 
-    def send_positions_to_connection(self, conn: AbstractConnection):
+    def send_positions_to_connection(self, conn: IReplicationConnection):
         """Send current position of all streams this process is source of to
         the connection.
         """
@@ -348,7 +347,7 @@ class ReplicationCommandHandler:
             )
 
     def on_USER_SYNC(
-        self, conn: AbstractConnection, cmd: UserSyncCommand
+        self, conn: IReplicationConnection, cmd: UserSyncCommand
     ) -> Optional[Awaitable[None]]:
         user_sync_counter.inc()
 
@@ -360,38 +359,23 @@ class ReplicationCommandHandler:
             return None
 
     def on_CLEAR_USER_SYNC(
-        self, conn: AbstractConnection, cmd: ClearUserSyncsCommand
+        self, conn: IReplicationConnection, cmd: ClearUserSyncsCommand
     ) -> Optional[Awaitable[None]]:
         if self._is_master:
             return self._presence_handler.update_external_syncs_clear(cmd.instance_id)
         else:
             return None
 
-    def on_FEDERATION_ACK(self, conn: AbstractConnection, cmd: FederationAckCommand):
+    def on_FEDERATION_ACK(
+        self, conn: IReplicationConnection, cmd: FederationAckCommand
+    ):
         federation_ack_counter.inc()
 
         if self._federation_sender:
             self._federation_sender.federation_ack(cmd.instance_name, cmd.token)
 
-    def on_REMOVE_PUSHER(
-        self, conn: AbstractConnection, cmd: RemovePusherCommand
-    ) -> Optional[Awaitable[None]]:
-        remove_pusher_counter.inc()
-
-        if self._is_master:
-            return self._handle_remove_pusher(cmd)
-        else:
-            return None
-
-    async def _handle_remove_pusher(self, cmd: RemovePusherCommand):
-        await self._store.delete_pusher_by_app_id_pushkey_user_id(
-            app_id=cmd.app_id, pushkey=cmd.push_key, user_id=cmd.user_id
-        )
-
-        self._notifier.on_new_replication_data()
-
     def on_USER_IP(
-        self, conn: AbstractConnection, cmd: UserIpCommand
+        self, conn: IReplicationConnection, cmd: UserIpCommand
     ) -> Optional[Awaitable[None]]:
         user_ip_cache_counter.inc()
 
@@ -413,7 +397,7 @@ class ReplicationCommandHandler:
         assert self._server_notices_sender is not None
         await self._server_notices_sender.on_user_ip(cmd.user_id)
 
-    def on_RDATA(self, conn: AbstractConnection, cmd: RdataCommand):
+    def on_RDATA(self, conn: IReplicationConnection, cmd: RdataCommand):
         if cmd.instance_name == self._instance_name:
             # Ignore RDATA that are just our own echoes
             return
@@ -430,7 +414,7 @@ class ReplicationCommandHandler:
         self._add_command_to_stream_queue(conn, cmd)
 
     async def _process_rdata(
-        self, stream_name: str, conn: AbstractConnection, cmd: RdataCommand
+        self, stream_name: str, conn: IReplicationConnection, cmd: RdataCommand
     ) -> None:
         """Process an RDATA command
 
@@ -504,7 +488,7 @@ class ReplicationCommandHandler:
             stream_name, instance_name, token, rows
         )
 
-    def on_POSITION(self, conn: AbstractConnection, cmd: PositionCommand):
+    def on_POSITION(self, conn: IReplicationConnection, cmd: PositionCommand):
         if cmd.instance_name == self._instance_name:
             # Ignore POSITION that are just our own echoes
             return
@@ -514,7 +498,7 @@ class ReplicationCommandHandler:
         self._add_command_to_stream_queue(conn, cmd)
 
     async def _process_position(
-        self, stream_name: str, conn: AbstractConnection, cmd: PositionCommand
+        self, stream_name: str, conn: IReplicationConnection, cmd: PositionCommand
     ) -> None:
         """Process a POSITION command
 
@@ -571,7 +555,9 @@ class ReplicationCommandHandler:
 
         self._streams_by_connection.setdefault(conn, set()).add(stream_name)
 
-    def on_REMOTE_SERVER_UP(self, conn: AbstractConnection, cmd: RemoteServerUpCommand):
+    def on_REMOTE_SERVER_UP(
+        self, conn: IReplicationConnection, cmd: RemoteServerUpCommand
+    ):
         """"Called when get a new REMOTE_SERVER_UP command."""
         self._replication_data_handler.on_remote_server_up(cmd.data)
 
@@ -594,7 +580,7 @@ class ReplicationCommandHandler:
         # between two instances, but that is not currently supported).
         self.send_command(cmd, ignore_conn=conn)
 
-    def new_connection(self, connection: AbstractConnection):
+    def new_connection(self, connection: IReplicationConnection):
         """Called when we have a new connection."""
         self._connections.append(connection)
 
@@ -621,7 +607,7 @@ class ReplicationCommandHandler:
                 UserSyncCommand(self._instance_id, user_id, True, now)
             )
 
-    def lost_connection(self, connection: AbstractConnection):
+    def lost_connection(self, connection: IReplicationConnection):
         """Called when a connection is closed/lost."""
         # we no longer need _streams_by_connection for this connection.
         streams = self._streams_by_connection.pop(connection, None)
@@ -642,7 +628,7 @@ class ReplicationCommandHandler:
         return bool(self._connections)
 
     def send_command(
-        self, cmd: Command, ignore_conn: Optional[AbstractConnection] = None
+        self, cmd: Command, ignore_conn: Optional[IReplicationConnection] = None
     ):
         """Send a command to all connected connections.
 
@@ -684,11 +670,6 @@ class ReplicationCommandHandler:
             UserSyncCommand(instance_id, user_id, is_syncing, last_sync_ms)
         )
 
-    def send_remove_pusher(self, app_id: str, push_key: str, user_id: str):
-        """Poke the master to remove a pusher for a user"""
-        cmd = RemovePusherCommand(app_id, push_key, user_id)
-        self.send_command(cmd)
-
     def send_user_ip(
         self,
         user_id: str,

synapse/replication/tcp/protocol.py

@@ -46,7 +46,6 @@ indicate which side is sending, these are *not* included on the wire::
     > ERROR server stopping
     * connection closed by server *
 """
-import abc
 import fcntl
 import logging
 import struct
@@ -54,8 +53,10 @@ from inspect import isawaitable
 from typing import TYPE_CHECKING, List, Optional
 
 from prometheus_client import Counter
+from zope.interface import Interface, implementer
 
 from twisted.internet import task
+from twisted.internet.tcp import Connection
 from twisted.protocols.basic import LineOnlyReceiver
 from twisted.python.failure import Failure
 
@@ -121,6 +122,14 @@ class ConnectionStates:
     CLOSED = "closed"
 
 
+class IReplicationConnection(Interface):
+    """An interface for replication connections."""
+
+    def send_command(cmd: Command):
+        """Send the command down the connection"""
+
+
+@implementer(IReplicationConnection)
 class BaseReplicationStreamProtocol(LineOnlyReceiver):
     """Base replication protocol shared between client and server.
 
@@ -137,6 +146,10 @@ class BaseReplicationStreamProtocol(LineOnlyReceiver):
     (if they send a `PING` command)
     """
 
+    # The transport is going to be an ITCPTransport, but that doesn't have the
+    # (un)registerProducer methods, those are only on the implementation.
+    transport = None  # type: Connection
+
     delimiter = b"\n"
 
     # Valid commands we expect to receive
@@ -181,6 +194,7 @@ class BaseReplicationStreamProtocol(LineOnlyReceiver):
 
         connected_connections.append(self)  # Register connection for metrics
 
+        assert self.transport is not None
         self.transport.registerProducer(self, True)  # For the *Producing callbacks
 
         self._send_pending_commands()
@@ -205,6 +219,7 @@ class BaseReplicationStreamProtocol(LineOnlyReceiver):
                 logger.info(
                     "[%s] Failed to close connection gracefully, aborting", self.id()
                 )
+                assert self.transport is not None
                 self.transport.abortConnection()
         else:
             if now - self.last_sent_command >= PING_TIME:
@@ -294,6 +309,7 @@ class BaseReplicationStreamProtocol(LineOnlyReceiver):
     def close(self):
         logger.warning("[%s] Closing connection", self.id())
         self.time_we_closed = self.clock.time_msec()
+        assert self.transport is not None
         self.transport.loseConnection()
         self.on_connection_closed()
 
@@ -391,6 +407,7 @@ class BaseReplicationStreamProtocol(LineOnlyReceiver):
     def connectionLost(self, reason):
         logger.info("[%s] Replication connection closed: %r", self.id(), reason)
         if isinstance(reason, Failure):
+            assert reason.type is not None
             connection_close_counter.labels(reason.type.__name__).inc()
         else:
             connection_close_counter.labels(reason.__class__.__name__).inc()
@@ -495,20 +512,6 @@ class ClientReplicationStreamProtocol(BaseReplicationStreamProtocol):
         self.send_command(ReplicateCommand())
 
 
-class AbstractConnection(abc.ABC):
-    """An interface for replication connections."""
-
-    @abc.abstractmethod
-    def send_command(self, cmd: Command):
-        """Send the command down the connection"""
-        pass
-
-
-# This tells python that `BaseReplicationStreamProtocol` implements the
-# interface.
-AbstractConnection.register(BaseReplicationStreamProtocol)
-
-
 # The following simply registers metrics for the replication connections
 
 pending_commands = LaterGauge(