fix regexp for log redaction to actually redact AS access_tokens etc

Improve logging of failures in matrixfederationclient

.github/ISSUE_TEMPLATE.md

@@ -0,0 +1,47 @@
+<!-- 
+
+**IF YOU HAVE SUPPORT QUESTIONS ABOUT RUNNING OR CONFIGURING YOUR OWN HOME SERVER**: 
+You will likely get better support more quickly if you ask in ** #matrix:matrix.org ** ;)
+
+
+This is a bug report template. By following the instructions below and
+filling out the sections with your information, you will help the us to get all
+the necessary data to fix your issue.
+
+You can also preview your report before submitting it. You may remove sections
+that aren't relevant to your particular case.
+
+Text between <!-- and --​> marks will be invisible in the report.
+
+-->
+
+### Description
+
+Describe here the problem that you are experiencing, or the feature you are requesting.
+
+### Steps to reproduce
+
+- For bugs, list the steps
+- that reproduce the bug
+- using hyphens as bullet points
+
+Describe how what happens differs from what you expected.
+
+If you can identify any relevant log snippets from _homeserver.log_, please include
+those here (please be careful to remove any personal or private data):
+
+### Version information
+
+<!-- IMPORTANT: please answer the following questions, to help us narrow down the problem -->
+
+- **Homeserver**: Was this issue identified on matrix.org or another homeserver?
+
+If not matrix.org:
+- **Version**:        What version of Synapse is running? <!-- 
+You can find the Synapse version by inspecting the server headers (replace matrix.org with
+your own homeserver domain):
+$ curl -v https://matrix.org/_matrix/client/versions 2>&1 | grep "Server:"
+-->
+- **Install method**: package manager/git clone/pip      
+- **Platform**:       Tell us about the environment in which your homeserver is operating
+                      - distro, hardware, if it's running in a vm/container, etc.

CHANGES.rst

@@ -1,3 +1,122 @@
+Changes in synapse v0.23.0 (2017-10-02)
+=======================================
+
+No changes since v0.23.0-rc2
+
+
+Changes in synapse v0.23.0-rc2 (2017-09-26)
+===========================================
+
+Bug fixes:
+
+* Fix regression in performance of syncs (PR #2470)
+
+
+Changes in synapse v0.23.0-rc1 (2017-09-25)
+===========================================
+
+Features:
+
+* Add a frontend proxy worker (PR #2344)
+* Add support for event_id_only push format (PR #2450)
+* Add a PoC for filtering spammy events (PR #2456)
+* Add a config option to block all room invites (PR #2457)
+
+
+Changes:
+
+* Use bcrypt module instead of py-bcrypt (PR #2288) Thanks to @kyrias!
+* Improve performance of generating push notifications (PR #2343, #2357, #2365,
+  #2366, #2371)
+* Improve DB performance for device list handling in sync (PR #2362)
+* Include a sample prometheus config (PR #2416)
+* Document known to work postgres version (PR #2433) Thanks to @ptman!
+
+
+Bug fixes:
+
+* Fix caching error in the push evaluator (PR #2332)
+* Fix bug where pusherpool didn't start and broke some rooms (PR #2342)
+* Fix port script for user directory tables (PR #2375)
+* Fix device lists notifications when user rejoins a room (PR #2443, #2449)
+* Fix sync to always send down current state events in timeline (PR #2451)
+* Fix bug where guest users were incorrectly kicked (PR #2453)
+* Fix bug talking to IPv6 only servers using SRV records (PR #2462)
+
+
+Changes in synapse v0.22.1 (2017-07-06)
+=======================================
+
+Bug fixes:
+
+* Fix bug where pusher pool didn't start and caused issues when
+  interacting with some rooms (PR #2342)
+
+
+Changes in synapse v0.22.0 (2017-07-06)
+=======================================
+
+No changes since v0.22.0-rc2
+
+
+Changes in synapse v0.22.0-rc2 (2017-07-04)
+===========================================
+
+Changes:
+
+* Improve performance of storing user IPs (PR #2307, #2308)
+* Slightly improve performance of verifying access tokens (PR #2320)
+* Slightly improve performance of event persistence (PR #2321)
+* Increase default cache factor size from 0.1 to 0.5 (PR #2330)
+
+Bug fixes:
+
+* Fix bug with storing registration sessions that caused frequent CPU churn
+  (PR #2319)
+
+
+Changes in synapse v0.22.0-rc1 (2017-06-26)
+===========================================
+
+Features:
+
+* Add a user directory API (PR #2252, and many more)
+* Add shutdown room API to remove room from local server (PR #2291)
+* Add API to quarantine media (PR #2292)
+* Add new config option to not send event contents to push servers (PR #2301)
+  Thanks to @cjdelisle!
+
+Changes:
+
+* Various performance fixes (PR #2177, #2233, #2230, #2238, #2248, #2256,
+  #2274)
+* Deduplicate sync filters (PR #2219) Thanks to @krombel!
+* Correct a typo in UPGRADE.rst (PR #2231) Thanks to @aaronraimist!
+* Add count of one time keys to sync stream (PR #2237)
+* Only store event_auth for state events (PR #2247)
+* Store URL cache preview downloads separately (PR #2299)
+
+Bug fixes:
+
+* Fix users not getting notifications when AS listened to that user_id (PR
+  #2216) Thanks to @slipeer!
+* Fix users without push set up not getting notifications after joining rooms
+  (PR #2236)
+* Fix preview url API to trim long descriptions (PR #2243)
+* Fix bug where we used cached but unpersisted state group as prev group,
+  resulting in broken state of restart (PR #2263)
+* Fix removing of pushers when using workers (PR #2267)
+* Fix CORS headers to allow Authorization header (PR #2285) Thanks to @krombel!
+
+
+Changes in synapse v0.21.1 (2017-06-15)
+=======================================
+
+Bug fixes:
+
+* Fix bug in anonymous usage statistic reporting (PR #2281)
+
+
 Changes in synapse v0.21.0 (2017-05-18)
 =======================================
 

MANIFEST.in

@@ -27,4 +27,5 @@ exclude jenkins*.sh
 exclude jenkins*
 recursive-exclude jenkins *.sh
 
+prune .github
 prune demo/etc

README.rst

@@ -200,11 +200,11 @@ different. See `the spec`__ for more information on key management.)
 .. __: `key_management`_
 
 The default configuration exposes two HTTP ports: 8008 and 8448. Port 8008 is
-configured without TLS; it is not recommended this be exposed outside your
-local network. Port 8448 is configured to use TLS with a self-signed
-certificate. This is fine for testing with but, to avoid your clients
-complaining about the certificate, you will almost certainly want to use
-another certificate for production purposes. (Note that a self-signed
+configured without TLS; it should be behind a reverse proxy for TLS/SSL
+termination on port 443 which in turn should be used for clients. Port 8448
+is configured to use TLS with a self-signed certificate. If you would like
+to do initial test with a client without having to setup a reverse proxy,
+you can temporarly use another certificate. (Note that a self-signed
 certificate is fine for `Federation`_). You can do so by changing
 ``tls_certificate_path``, ``tls_private_key_path`` and ``tls_dh_params_path``
 in ``homeserver.yaml``; alternatively, you can use a reverse-proxy, but be sure
@@ -283,10 +283,16 @@ Connecting to Synapse from a client
 The easiest way to try out your new Synapse installation is by connecting to it
 from a web client. The easiest option is probably the one at
 http://riot.im/app. You will need to specify a "Custom server" when you log on
-or register: set this to ``https://localhost:8448`` - remember to specify the
-port (``:8448``) unless you changed the configuration. (Leave the identity
+or register: set this to ``https://domain.tld`` if you setup a reverse proxy
+following the recommended setup, or ``https://localhost:8448`` - remember to specify the
+port (``:8448``) if not ``:443`` unless you changed the configuration. (Leave the identity
 server as the default - see `Identity servers`_.)
 
+If using port 8448 you will run into errors until you accept the self-signed
+certificate. You can easily do this by going to ``https://localhost:8448``
+directly with your browser and accept the presented certificate. You can then
+go back in your web client and proceed further.
+
 If all goes well you should at least be able to log in, create a room, and
 start sending messages.
 
@@ -359,7 +365,7 @@ https://www.archlinux.org/packages/community/any/matrix-synapse/, which should p
 the necessary dependencies. If the default web client is to be served (enabled by default in
 the generated config),
 https://www.archlinux.org/packages/community/any/python2-matrix-angular-sdk/ will also need to
-be installed. 
+be installed.
 
 Alternatively, to install using pip a few changes may be needed as ArchLinux
 defaults to python 3, but synapse currently assumes python 2.7 by default:
@@ -528,6 +534,30 @@ fix try re-installing from PyPI or directly from
     # Install from github
     pip install --user https://github.com/pyca/pynacl/tarball/master
 
+Running out of File Handles
+~~~~~~~~~~~~~~~~~~~~~~~~~~~
+
+If synapse runs out of filehandles, it typically fails badly - live-locking
+at 100% CPU, and/or failing to accept new TCP connections (blocking the
+connecting client).  Matrix currently can legitimately use a lot of file handles,
+thanks to busy rooms like #matrix:matrix.org containing hundreds of participating
+servers.  The first time a server talks in a room it will try to connect
+simultaneously to all participating servers, which could exhaust the available
+file descriptors between DNS queries & HTTPS sockets, especially if DNS is slow
+to respond.  (We need to improve the routing algorithm used to be better than
+full mesh, but as of June 2017 this hasn't happened yet).
+
+If you hit this failure mode, we recommend increasing the maximum number of
+open file handles to be at least 4096 (assuming a default of 1024 or 256).
+This is typically done by editing ``/etc/security/limits.conf``
+
+Separately, Synapse may leak file handles if inbound HTTP requests get stuck
+during processing - e.g. blocked behind a lock or talking to a remote server etc.
+This is best diagnosed by matching up the 'Received request' and 'Processed request'
+log lines and looking for any 'Processed request' lines which take more than
+a few seconds to execute.  Please let us know at #matrix-dev:matrix.org if
+you see this failure mode so we can help debug it, however.
+
 ArchLinux
 ~~~~~~~~~
 
@@ -569,8 +599,9 @@ you to run your server on a machine that might not have the same name as your
 domain name. For example, you might want to run your server at
 ``synapse.example.com``, but have your Matrix user-ids look like
 ``@user:example.com``. (A SRV record also allows you to change the port from
-the default 8448. However, if you are thinking of using a reverse-proxy, be
-sure to read `Reverse-proxying the federation port`_ first.)
+the default 8448. However, if you are thinking of using a reverse-proxy on the
+federation port, which is not recommended, be sure to read
+`Reverse-proxying the federation port`_ first.)
 
 To use a SRV record, first create your SRV record and publish it in DNS. This
 should have the format ``_matrix._tcp.<yourdomain.com> <ttl> IN SRV 10 0 <port>
@@ -650,7 +681,7 @@ For information on how to install and use PostgreSQL, please see
 Using a reverse proxy with Synapse
 ==================================
 
-It is possible to put a reverse proxy such as
+It is recommended to put a reverse proxy such as
 `nginx <https://nginx.org/en/docs/http/ngx_http_proxy_module.html>`_,
 `Apache <https://httpd.apache.org/docs/current/mod/mod_proxy_http.html>`_ or
 `HAProxy <http://www.haproxy.org/>`_ in front of Synapse. One advantage of
@@ -668,9 +699,9 @@ federation port has a number of pitfalls. It is possible, but be sure to read
 `Reverse-proxying the federation port`_.
 
 The recommended setup is therefore to configure your reverse-proxy on port 443
-for client connections, but to also expose port 8448 for server-server
-connections. All the Matrix endpoints begin ``/_matrix``, so an example nginx
-configuration might look like::
+to port 8008 of synapse for client connections, but to also directly expose port
+8448 for server-server connections. All the Matrix endpoints begin ``/_matrix``,
+so an example nginx configuration might look like::
 
   server {
       listen 443 ssl;
@@ -875,12 +906,9 @@ cache a lot of recent room data and metadata in RAM in order to speed up
 common requests.  We'll improve this in future, but for now the easiest
 way to either reduce the RAM usage (at the risk of slowing things down)
 is to set the almost-undocumented ``SYNAPSE_CACHE_FACTOR`` environment
-variable.  Roughly speaking, a SYNAPSE_CACHE_FACTOR of 1.0 will max out
-at around 3-4GB of resident memory - this is what we currently run the
-matrix.org on.  The default setting is currently 0.1, which is probably
-around a ~700MB footprint.  You can dial it down further to 0.02 if
-desired, which targets roughly ~512MB.  Conversely you can dial it up if
-you need performance for lots of users and have a box with a lot of RAM.
+variable.  The default is 0.5, which can be decreased to reduce RAM usage
+in memory constrained enviroments, or increased if performance starts to
+degrade.
 
 
 .. _`key_management`: https://matrix.org/docs/spec/server_server/unstable.html#retrieving-server-keys

UPGRADE.rst

@@ -5,39 +5,48 @@ Before upgrading check if any special steps are required to upgrade from the
 what you currently have installed to current version of synapse. The extra
 instructions that may be required are listed later in this document.
 
-If synapse was installed in a virtualenv then active that virtualenv before
-upgrading. If synapse is installed in a virtualenv in ``~/.synapse/`` then run:
+1. If synapse was installed in a virtualenv then active that virtualenv before
+   upgrading. If synapse is installed in a virtualenv in ``~/.synapse/`` then
+   run:
+
+   .. code:: bash
+
+       source ~/.synapse/bin/activate
+
+2. If synapse was installed using pip then upgrade to the latest version by
+   running:
+
+   .. code:: bash
+
+       pip install --upgrade --process-dependency-links https://github.com/matrix-org/synapse/tarball/master
+
+       # restart synapse
+       synctl restart
+
+
+   If synapse was installed using git then upgrade to the latest version by
+   running:
+
+   .. code:: bash
+
+       # Pull the latest version of the master branch.
+       git pull
+       # Update the versions of synapse's python dependencies.
+       python synapse/python_dependencies.py | xargs pip install --upgrade
+
+       # restart synapse
+       ./synctl restart
+
+
+To check whether your update was sucessful, you can check the Server header
+returned by the Client-Server API:
 
 .. code:: bash
 
-    source ~/.synapse/bin/activate
-
-If synapse was installed using pip then upgrade to the latest version by
-running:
-
-.. code:: bash
-
-    pip install --upgrade --process-dependency-links https://github.com/matrix-org/synapse/tarball/master
-
-If synapse was installed using git then upgrade to the latest version by
-running:
-
-.. code:: bash
-
-    # Pull the latest version of the master branch.
-    git pull
-    # Update the versions of synapse's python dependencies.
-    python synapse/python_dependencies.py | xargs -n1 pip install --upgrade
-	
-To check whether your update was sucessfull, run:
-
-.. code:: bash
-
-	 # replace your.server.domain with ther domain of your synaspe homeserver
-	 curl https://<your.server.domain>/_matrix/federation/v1/version 
-
-So for the Matrix.org HS server the URL would be: https://matrix.org/_matrix/federation/v1/version.
-
+    # replace <host.name> with the hostname of your synapse homeserver.
+    # You may need to specify a port (eg, :8448) if your server is not
+    # configured on port 443.
+    curl -kv https://<host.name>/_matrix/client/versions 2>&1 | grep "Server:"
 
 Upgrading to v0.15.0
 ====================
@@ -77,7 +86,7 @@ It has been replaced by specifying a list of application service registrations i
 ``homeserver.yaml``::
 
   app_service_config_files: ["registration-01.yaml", "registration-02.yaml"]
-  
+
 Where ``registration-01.yaml`` looks like::
 
   url: <String>  # e.g. "https://my.application.service.com"
@@ -166,7 +175,7 @@ This release completely changes the database schema and so requires upgrading
 it before starting the new version of the homeserver.
 
 The script "database-prepare-for-0.5.0.sh" should be used to upgrade the
-database. This will save all user information, such as logins and profiles, 
+database. This will save all user information, such as logins and profiles,
 but will otherwise purge the database. This includes messages, which
 rooms the home server was a member of and room alias mappings.
 
@@ -175,18 +184,18 @@ file and ask for help in #matrix:matrix.org. The upgrade process is,
 unfortunately, non trivial and requires human intervention to resolve any
 resulting conflicts during the upgrade process.
 
-Before running the command the homeserver should be first completely 
+Before running the command the homeserver should be first completely
 shutdown. To run it, simply specify the location of the database, e.g.:
 
   ./scripts/database-prepare-for-0.5.0.sh "homeserver.db"
 
-Once this has successfully completed it will be safe to restart the 
-homeserver. You may notice that the homeserver takes a few seconds longer to 
+Once this has successfully completed it will be safe to restart the
+homeserver. You may notice that the homeserver takes a few seconds longer to
 restart than usual as it reinitializes the database.
 
 On startup of the new version, users can either rejoin remote rooms using room
 aliases or by being reinvited. Alternatively, if any other homeserver sends a
-message to a room that the homeserver was previously in the local HS will 
+message to a room that the homeserver was previously in the local HS will
 automatically rejoin the room.
 
 Upgrading to v0.4.0
@@ -245,7 +254,7 @@ automatically generate default config use::
         --config-path homeserver.config \
         --generate-config
 
-This config can be edited if desired, for example to specify a different SSL 
+This config can be edited if desired, for example to specify a different SSL
 certificate to use. Once done you can run the home server using::
 
     $ python synapse/app/homeserver.py --config-path homeserver.config
@@ -266,20 +275,20 @@ This release completely changes the database schema and so requires upgrading
 it before starting the new version of the homeserver.
 
 The script "database-prepare-for-0.0.1.sh" should be used to upgrade the
-database. This will save all user information, such as logins and profiles, 
+database. This will save all user information, such as logins and profiles,
 but will otherwise purge the database. This includes messages, which
 rooms the home server was a member of and room alias mappings.
 
-Before running the command the homeserver should be first completely 
+Before running the command the homeserver should be first completely
 shutdown. To run it, simply specify the location of the database, e.g.:
 
   ./scripts/database-prepare-for-0.0.1.sh "homeserver.db"
 
-Once this has successfully completed it will be safe to restart the 
-homeserver. You may notice that the homeserver takes a few seconds longer to 
+Once this has successfully completed it will be safe to restart the
+homeserver. You may notice that the homeserver takes a few seconds longer to
 restart than usual as it reinitializes the database.
 
 On startup of the new version, users can either rejoin remote rooms using room
 aliases or by being reinvited. Alternatively, if any other homeserver sends a
-message to a room that the homeserver was previously in the local HS will 
+message to a room that the homeserver was previously in the local HS will
 automatically rejoin the room.

contrib/prometheus/README

@@ -0,0 +1,20 @@
+This directory contains some sample monitoring config for using the
+'Prometheus' monitoring server against synapse.
+
+To use it, first install prometheus by following the instructions at
+
+  http://prometheus.io/
+
+Then add a new job to the main prometheus.conf file:
+
+  job: {
+    name: "synapse"
+
+    target_group: {
+      target: "http://SERVER.LOCATION.HERE:PORT/_synapse/metrics"
+    }
+  }
+
+Metrics are disabled by default when running synapse; they must be enabled
+with the 'enable-metrics' option, either in the synapse config file or as a
+command-line option.

contrib/prometheus/consoles/synapse.html

@@ -0,0 +1,395 @@
+{{ template "head" . }}
+
+{{ template "prom_content_head" . }}
+<h1>System Resources</h1>
+
+<h3>CPU</h3>
+<div id="process_resource_utime"></div>
+<script>
+new PromConsole.Graph({
+  node: document.querySelector("#process_resource_utime"),
+  expr: "rate(process_cpu_seconds_total[2m]) * 100",
+  name: "[[job]]",  
+  min: 0,
+  max: 100,
+  renderer: "line",
+  height: 150,
+  yAxisFormatter: PromConsole.NumberFormatter.humanize,
+  yHoverFormatter: PromConsole.NumberFormatter.humanize,
+  yUnits: "%",
+  yTitle: "CPU Usage"
+})
+</script>
+
+<h3>Memory</h3>
+<div id="process_resource_maxrss"></div>
+<script>
+new PromConsole.Graph({
+  node: document.querySelector("#process_resource_maxrss"),
+  expr: "process_psutil_rss:max",
+  name: "Maxrss",
+  min: 0,
+  renderer: "line",
+  height: 150,
+  yAxisFormatter: PromConsole.NumberFormatter.humanizeNoSmallPrefix,
+  yHoverFormatter: PromConsole.NumberFormatter.humanizeNoSmallPrefix,
+  yUnits: "bytes",
+  yTitle: "Usage"
+})
+</script>
+
+<h3>File descriptors</h3>
+<div id="process_fds"></div>
+<script>
+new PromConsole.Graph({
+  node: document.querySelector("#process_fds"),
+  expr: "process_open_fds{job='synapse'}",
+  name: "FDs",
+  min: 0,
+  renderer: "line",
+  height: 150,
+  yAxisFormatter: PromConsole.NumberFormatter.humanize,
+  yHoverFormatter: PromConsole.NumberFormatter.humanize,
+  yUnits: "",
+  yTitle: "Descriptors"
+})
+</script>
+
+<h1>Reactor</h1>
+
+<h3>Total reactor time</h3>
+<div id="reactor_total_time"></div>
+<script>
+new PromConsole.Graph({
+  node: document.querySelector("#reactor_total_time"),
+  expr: "rate(python_twisted_reactor_tick_time:total[2m]) / 1000",
+  name: "time",
+  max: 1,
+  min: 0,
+  renderer: "area",
+  height: 150,
+  yAxisFormatter: PromConsole.NumberFormatter.humanize,
+  yHoverFormatter: PromConsole.NumberFormatter.humanize,
+  yUnits: "s/s",
+  yTitle: "Usage"
+})
+</script>
+
+<h3>Average reactor tick time</h3>
+<div id="reactor_average_time"></div>
+<script>
+new PromConsole.Graph({
+  node: document.querySelector("#reactor_average_time"),
+  expr: "rate(python_twisted_reactor_tick_time:total[2m]) / rate(python_twisted_reactor_tick_time:count[2m]) / 1000",
+  name: "time",
+  min: 0,
+  renderer: "line",
+  height: 150,
+  yAxisFormatter: PromConsole.NumberFormatter.humanize,
+  yHoverFormatter: PromConsole.NumberFormatter.humanize,
+  yUnits: "s",
+  yTitle: "Time"
+})
+</script>
+
+<h3>Pending calls per tick</h3>
+<div id="reactor_pending_calls"></div>
+<script>
+new PromConsole.Graph({
+  node: document.querySelector("#reactor_pending_calls"),
+  expr: "rate(python_twisted_reactor_pending_calls:total[30s])/rate(python_twisted_reactor_pending_calls:count[30s])",
+  name: "calls",
+  min: 0,
+  renderer: "line",
+  height: 150,
+  yAxisFormatter: PromConsole.NumberFormatter.humanize,
+  yHoverFormatter: PromConsole.NumberFormatter.humanize,
+  yTitle: "Pending Cals"
+})
+</script>
+
+<h1>Storage</h1>
+
+<h3>Queries</h3>
+<div id="synapse_storage_query_time"></div>
+<script>
+new PromConsole.Graph({
+  node: document.querySelector("#synapse_storage_query_time"),
+  expr: "rate(synapse_storage_query_time:count[2m])",
+  name: "[[verb]]",
+  yAxisFormatter: PromConsole.NumberFormatter.humanizeNoSmallPrefix,
+  yHoverFormatter: PromConsole.NumberFormatter.humanizeNoSmallPrefix,
+  yUnits: "queries/s",
+  yTitle: "Queries"
+})
+</script>
+
+<h3>Transactions</h3>
+<div id="synapse_storage_transaction_time"></div>
+<script>
+new PromConsole.Graph({
+  node: document.querySelector("#synapse_storage_transaction_time"),
+  expr: "rate(synapse_storage_transaction_time:count[2m])",
+  name: "[[desc]]",
+  min: 0,
+  yAxisFormatter: PromConsole.NumberFormatter.humanizeNoSmallPrefix,
+  yHoverFormatter: PromConsole.NumberFormatter.humanizeNoSmallPrefix,
+  yUnits: "txn/s",
+  yTitle: "Transactions"
+})
+</script>
+
+<h3>Transaction execution time</h3>
+<div id="synapse_storage_transactions_time_msec"></div>
+<script>
+new PromConsole.Graph({
+  node: document.querySelector("#synapse_storage_transactions_time_msec"),
+  expr: "rate(synapse_storage_transaction_time:total[2m]) / 1000",
+  name: "[[desc]]",
+  min: 0,
+  yAxisFormatter: PromConsole.NumberFormatter.humanize,
+  yHoverFormatter: PromConsole.NumberFormatter.humanize,
+  yUnits: "s/s",
+  yTitle: "Usage"
+})
+</script>
+
+<h3>Database scheduling latency</h3>
+<div id="synapse_storage_schedule_time"></div>
+<script>
+new PromConsole.Graph({
+  node: document.querySelector("#synapse_storage_schedule_time"),
+  expr: "rate(synapse_storage_schedule_time:total[2m]) / 1000",
+  name: "Total latency",
+  min: 0,
+  yAxisFormatter: PromConsole.NumberFormatter.humanize,
+  yHoverFormatter: PromConsole.NumberFormatter.humanize,
+  yUnits: "s/s",
+  yTitle: "Usage"
+})
+</script>
+
+<h3>Cache hit ratio</h3>
+<div id="synapse_cache_ratio"></div>
+<script>
+new PromConsole.Graph({
+  node: document.querySelector("#synapse_cache_ratio"),
+  expr: "rate(synapse_util_caches_cache:total[2m]) * 100",
+  name: "[[name]]",
+  min: 0,
+  max: 100,
+  yAxisFormatter: PromConsole.NumberFormatter.humanizeNoSmallPrefix,
+  yHoverFormatter: PromConsole.NumberFormatter.humanizeNoSmallPrefix,
+  yUnits: "%",
+  yTitle: "Percentage"
+})
+</script>
+
+<h3>Cache size</h3>
+<div id="synapse_cache_size"></div>
+<script>
+new PromConsole.Graph({
+  node: document.querySelector("#synapse_cache_size"),
+  expr: "synapse_util_caches_cache:size",
+  name: "[[name]]",
+  yAxisFormatter: PromConsole.NumberFormatter.humanizeNoSmallPrefix,
+  yHoverFormatter: PromConsole.NumberFormatter.humanizeNoSmallPrefix,
+  yUnits: "",
+  yTitle: "Items"
+})
+</script>
+
+<h1>Requests</h1>
+
+<h3>Requests by Servlet</h3>
+<div id="synapse_http_server_requests_servlet"></div>
+<script>
+new PromConsole.Graph({
+  node: document.querySelector("#synapse_http_server_requests_servlet"),
+  expr: "rate(synapse_http_server_requests:servlet[2m])",
+  name: "[[servlet]]",
+  yAxisFormatter: PromConsole.NumberFormatter.humanize,
+  yHoverFormatter: PromConsole.NumberFormatter.humanize,
+  yUnits: "req/s",
+  yTitle: "Requests"
+})
+</script>
+<h4>&nbsp;(without <tt>EventStreamRestServlet</tt> or <tt>SyncRestServlet</tt>)</h4>
+<div id="synapse_http_server_requests_servlet_minus_events"></div>
+<script>
+new PromConsole.Graph({
+  node: document.querySelector("#synapse_http_server_requests_servlet_minus_events"),
+  expr: "rate(synapse_http_server_requests:servlet{servlet!=\"EventStreamRestServlet\", servlet!=\"SyncRestServlet\"}[2m])",
+  name: "[[servlet]]",
+  yAxisFormatter: PromConsole.NumberFormatter.humanize,
+  yHoverFormatter: PromConsole.NumberFormatter.humanize,
+  yUnits: "req/s",
+  yTitle: "Requests"
+})
+</script>
+
+<h3>Average response times</h3>
+<div id="synapse_http_server_response_time_avg"></div>
+<script>
+new PromConsole.Graph({
+  node: document.querySelector("#synapse_http_server_response_time_avg"),
+  expr: "rate(synapse_http_server_response_time:total[2m]) / rate(synapse_http_server_response_time:count[2m]) / 1000",
+  name: "[[servlet]]",
+  yAxisFormatter: PromConsole.NumberFormatter.humanize,
+  yHoverFormatter: PromConsole.NumberFormatter.humanize,
+  yUnits: "s/req",
+  yTitle: "Response time"
+})
+</script>
+
+<h3>All responses by code</h3>
+<div id="synapse_http_server_responses"></div>
+<script>
+new PromConsole.Graph({
+  node: document.querySelector("#synapse_http_server_responses"),
+  expr: "rate(synapse_http_server_responses[2m])",
+  name: "[[method]] / [[code]]",
+  yAxisFormatter: PromConsole.NumberFormatter.humanize,
+  yHoverFormatter: PromConsole.NumberFormatter.humanize,
+  yUnits: "req/s",
+  yTitle: "Requests"
+})
+</script>
+
+<h3>Error responses by code</h3>
+<div id="synapse_http_server_responses_err"></div>
+<script>
+new PromConsole.Graph({
+  node: document.querySelector("#synapse_http_server_responses_err"),
+  expr: "rate(synapse_http_server_responses{code=~\"[45]..\"}[2m])",
+  name: "[[method]] / [[code]]",
+  yAxisFormatter: PromConsole.NumberFormatter.humanize,
+  yHoverFormatter: PromConsole.NumberFormatter.humanize,
+  yUnits: "req/s",
+  yTitle: "Requests"
+})
+</script>
+
+
+<h3>CPU Usage</h3>
+<div id="synapse_http_server_response_ru_utime"></div>
+<script>
+new PromConsole.Graph({
+  node: document.querySelector("#synapse_http_server_response_ru_utime"),
+  expr: "rate(synapse_http_server_response_ru_utime:total[2m])",
+  name: "[[servlet]]",
+  yAxisFormatter: PromConsole.NumberFormatter.humanize,
+  yHoverFormatter: PromConsole.NumberFormatter.humanize,
+  yUnits: "s/s",
+  yTitle: "CPU Usage"
+})
+</script>
+
+
+<h3>DB Usage</h3>
+<div id="synapse_http_server_response_db_txn_duration"></div>
+<script>
+new PromConsole.Graph({
+  node: document.querySelector("#synapse_http_server_response_db_txn_duration"),
+  expr: "rate(synapse_http_server_response_db_txn_duration:total[2m])",
+  name: "[[servlet]]",
+  yAxisFormatter: PromConsole.NumberFormatter.humanize,
+  yHoverFormatter: PromConsole.NumberFormatter.humanize,
+  yUnits: "s/s",
+  yTitle: "DB Usage"
+})
+</script>
+
+
+<h3>Average event send times</h3>
+<div id="synapse_http_server_send_time_avg"></div>
+<script>
+new PromConsole.Graph({
+  node: document.querySelector("#synapse_http_server_send_time_avg"),
+  expr: "rate(synapse_http_server_response_time:total{servlet='RoomSendEventRestServlet'}[2m]) / rate(synapse_http_server_response_time:count{servlet='RoomSendEventRestServlet'}[2m]) / 1000",
+  name: "[[servlet]]",
+  yAxisFormatter: PromConsole.NumberFormatter.humanize,
+  yHoverFormatter: PromConsole.NumberFormatter.humanize,
+  yUnits: "s/req",
+  yTitle: "Response time"
+})
+</script>
+
+<h1>Federation</h1>
+
+<h3>Sent Messages</h3>
+<div id="synapse_federation_client_sent"></div>
+<script>
+new PromConsole.Graph({
+  node: document.querySelector("#synapse_federation_client_sent"),
+  expr: "rate(synapse_federation_client_sent[2m])",
+  name: "[[type]]",
+  yAxisFormatter: PromConsole.NumberFormatter.humanize,
+  yHoverFormatter: PromConsole.NumberFormatter.humanize,
+  yUnits: "req/s",
+  yTitle: "Requests"
+})
+</script>
+
+<h3>Received Messages</h3>
+<div id="synapse_federation_server_received"></div>
+<script>
+new PromConsole.Graph({
+  node: document.querySelector("#synapse_federation_server_received"),
+  expr: "rate(synapse_federation_server_received[2m])",
+  name: "[[type]]",
+  yAxisFormatter: PromConsole.NumberFormatter.humanize,
+  yHoverFormatter: PromConsole.NumberFormatter.humanize,
+  yUnits: "req/s",
+  yTitle: "Requests"
+})
+</script>
+
+<h3>Pending</h3>
+<div id="synapse_federation_transaction_queue_pending"></div>
+<script>
+new PromConsole.Graph({
+  node: document.querySelector("#synapse_federation_transaction_queue_pending"),
+  expr: "synapse_federation_transaction_queue_pending",
+  name: "[[type]]",
+  yAxisFormatter: PromConsole.NumberFormatter.humanizeNoSmallPrefix,
+  yHoverFormatter: PromConsole.NumberFormatter.humanizeNoSmallPrefix,
+  yUnits: "",
+  yTitle: "Units"
+})
+</script>
+
+<h1>Clients</h1>
+
+<h3>Notifiers</h3>
+<div id="synapse_notifier_listeners"></div>
+<script>
+new PromConsole.Graph({
+  node: document.querySelector("#synapse_notifier_listeners"),
+  expr: "synapse_notifier_listeners",
+  name: "listeners",
+  min: 0,
+  yAxisFormatter: PromConsole.NumberFormatter.humanizeNoSmallPrefix,
+  yHoverFormatter: PromConsole.NumberFormatter.humanizeNoSmallPrefix,
+  yUnits: "",
+  yTitle: "Listeners"
+})
+</script>
+
+<h3>Notified Events</h3>
+<div id="synapse_notifier_notified_events"></div>
+<script>
+new PromConsole.Graph({
+  node: document.querySelector("#synapse_notifier_notified_events"),
+  expr: "rate(synapse_notifier_notified_events[2m])",
+  name: "events",
+  yAxisFormatter: PromConsole.NumberFormatter.humanize,
+  yHoverFormatter: PromConsole.NumberFormatter.humanize,
+  yUnits: "events/s",
+  yTitle: "Event rate"
+})
+</script>
+
+{{ template "prom_content_tail" . }}
+
+{{ template "tail" }}

contrib/prometheus/synapse.rules

@@ -0,0 +1,21 @@
+synapse_federation_transaction_queue_pendingEdus:total = sum(synapse_federation_transaction_queue_pendingEdus or absent(synapse_federation_transaction_queue_pendingEdus)*0)
+synapse_federation_transaction_queue_pendingPdus:total = sum(synapse_federation_transaction_queue_pendingPdus or absent(synapse_federation_transaction_queue_pendingPdus)*0)
+
+synapse_http_server_requests:method{servlet=""} = sum(synapse_http_server_requests) by (method)
+synapse_http_server_requests:servlet{method=""} = sum(synapse_http_server_requests) by (servlet)
+
+synapse_http_server_requests:total{servlet=""} = sum(synapse_http_server_requests:by_method) by (servlet)
+
+synapse_cache:hit_ratio_5m = rate(synapse_util_caches_cache:hits[5m]) / rate(synapse_util_caches_cache:total[5m])
+synapse_cache:hit_ratio_30s = rate(synapse_util_caches_cache:hits[30s]) / rate(synapse_util_caches_cache:total[30s])
+
+synapse_federation_client_sent{type="EDU"} = synapse_federation_client_sent_edus + 0
+synapse_federation_client_sent{type="PDU"} = synapse_federation_client_sent_pdu_destinations:count + 0
+synapse_federation_client_sent{type="Query"} = sum(synapse_federation_client_sent_queries) by (job)
+
+synapse_federation_server_received{type="EDU"} = synapse_federation_server_received_edus + 0
+synapse_federation_server_received{type="PDU"} = synapse_federation_server_received_pdus + 0
+synapse_federation_server_received{type="Query"} = sum(synapse_federation_server_received_queries) by (job)
+
+synapse_federation_transaction_queue_pending{type="EDU"} = synapse_federation_transaction_queue_pending_edus + 0
+synapse_federation_transaction_queue_pending{type="PDU"} = synapse_federation_transaction_queue_pending_pdus + 0

contrib/systemd/synapse.service

@@ -9,9 +9,10 @@ Description=Synapse Matrix homeserver
 Type=simple
 User=synapse
 Group=synapse
-EnvironmentFile=-/etc/sysconfig/synapse
 WorkingDirectory=/var/lib/synapse
-ExecStart=/usr/bin/python2.7 -m synapse.app.homeserver --config-path=/etc/synapse/homeserver.yaml --log-config=/etc/synapse/log_config.yaml
+ExecStart=/usr/bin/python2.7 -m synapse.app.homeserver --config-path=/etc/synapse/homeserver.yaml
+ExecStop=/usr/bin/synctl stop /etc/synapse/homeserver.yaml
 
 [Install]
 WantedBy=multi-user.target
+

docs/postgres.rst

@@ -1,6 +1,8 @@
 Using Postgres
 --------------
 
+Postgres version 9.4 or later is known to work.
+
 Set up database
 ===============
 

jenkins-dendron-haproxy-postgres.sh

@@ -17,6 +17,7 @@ export HAPROXY_BIN=/home/haproxy/haproxy-1.6.11/haproxy
 ./sytest/jenkins/prep_sytest_for_postgres.sh
 
 ./sytest/jenkins/install_and_run.sh \
+    --python $WORKSPACE/.tox/py27/bin/python \
     --synapse-directory $WORKSPACE \
     --dendron $WORKSPACE/dendron/bin/dendron \
     --haproxy \

jenkins-dendron-postgres.sh

@@ -15,5 +15,6 @@ export SYNAPSE_CACHE_FACTOR=1
 ./sytest/jenkins/prep_sytest_for_postgres.sh
 
 ./sytest/jenkins/install_and_run.sh \
+    --python $WORKSPACE/.tox/py27/bin/python \
     --synapse-directory $WORKSPACE \
     --dendron $WORKSPACE/dendron/bin/dendron \

jenkins-postgres.sh

@@ -14,4 +14,5 @@ export SYNAPSE_CACHE_FACTOR=1
 ./sytest/jenkins/prep_sytest_for_postgres.sh
 
 ./sytest/jenkins/install_and_run.sh \
+    --python $WORKSPACE/.tox/py27/bin/python \
     --synapse-directory $WORKSPACE \

jenkins-sqlite.sh

@@ -12,4 +12,5 @@ export SYNAPSE_CACHE_FACTOR=1
 ./jenkins/clone.sh sytest https://github.com/matrix-org/sytest.git
 
 ./sytest/jenkins/install_and_run.sh \
+    --python $WORKSPACE/.tox/py27/bin/python \
     --synapse-directory $WORKSPACE \

scripts-dev/federation_client.py

@@ -1,10 +1,30 @@
+#!/usr/bin/env python
+#
+# Copyright 2015, 2016 OpenMarket Ltd
+# Copyright 2017 New Vector Ltd
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+from __future__ import print_function
+
+import argparse
 import nacl.signing
 import json
 import base64
 import requests
 import sys
 import srvlookup
-
+import yaml
 
 def encode_base64(input_bytes):
     """Encode bytes as a base64 string without any padding."""
@@ -120,11 +140,13 @@ def get_json(origin_name, origin_key, destination, path):
             origin_name, key, sig,
         )
         authorization_headers.append(bytes(header))
-        sys.stderr.write(header)
-        sys.stderr.write("\n")
+        print ("Authorization: %s" % header, file=sys.stderr)
+
+    dest = lookup(destination, path)
+    print ("Requesting %s" % dest, file=sys.stderr)
 
     result = requests.get(
-        lookup(destination, path),
+        dest,
         headers={"Authorization": authorization_headers[0]},
         verify=False,
     )
@@ -133,17 +155,66 @@ def get_json(origin_name, origin_key, destination, path):
 
 
 def main():
-    origin_name, keyfile, destination, path = sys.argv[1:]
+    parser = argparse.ArgumentParser(
+        description=
+            "Signs and sends a federation request to a matrix homeserver",
+    )
 
-    with open(keyfile) as f:
+    parser.add_argument(
+        "-N", "--server-name",
+        help="Name to give as the local homeserver. If unspecified, will be "
+             "read from the config file.",
+    )
+
+    parser.add_argument(
+        "-k", "--signing-key-path",
+        help="Path to the file containing the private ed25519 key to sign the "
+             "request with.",
+    )
+
+    parser.add_argument(
+        "-c", "--config",
+        default="homeserver.yaml",
+        help="Path to server config file. Ignored if --server-name and "
+             "--signing-key-path are both given.",
+    )
+
+    parser.add_argument(
+        "-d", "--destination",
+        default="matrix.org",
+        help="name of the remote homeserver. We will do SRV lookups and "
+             "connect appropriately.",
+    )
+
+    parser.add_argument(
+        "path",
+        help="request path. We will add '/_matrix/federation/v1/' to this."
+    )
+
+    args = parser.parse_args()
+
+    if not args.server_name or not args.signing_key_path:
+        read_args_from_config(args)
+
+    with open(args.signing_key_path) as f:
         key = read_signing_keys(f)[0]
 
     result = get_json(
-        origin_name, key, destination, "/_matrix/federation/v1/" + path
+        args.server_name, key, args.destination, "/_matrix/federation/v1/" + args.path
     )
 
     json.dump(result, sys.stdout)
-    print ""
+    print ("")
+
+
+def read_args_from_config(args):
+    with open(args.config, 'r') as fh:
+        config = yaml.safe_load(fh)
+        if not args.server_name:
+            args.server_name = config['server_name']
+        if not args.signing_key_path:
+            args.signing_key_path = config['signing_key_path']
+
 
 if __name__ == "__main__":
     main()

scripts/synapse_port_db

@@ -41,6 +41,7 @@ BOOLEAN_COLUMNS = {
     "presence_stream": ["currently_active"],
     "public_room_list_stream": ["visibility"],
     "device_lists_outbound_pokes": ["sent"],
+    "users_who_share_rooms": ["share_private"],
 }
 
 
@@ -121,7 +122,7 @@ class Store(object):
                     try:
                         txn = conn.cursor()
                         return func(
-                            LoggingTransaction(txn, desc, self.database_engine, []),
+                            LoggingTransaction(txn, desc, self.database_engine, [], []),
                             *args, **kwargs
                         )
                     except self.database_engine.module.DatabaseError as e:
@@ -251,6 +252,25 @@ class Porter(object):
             )
             return
 
+        if table in (
+            "user_directory", "user_directory_search", "users_who_share_rooms",
+            "users_in_pubic_room",
+        ):
+            # We don't port these tables, as they're a faff and we can regenreate
+            # them anyway.
+            self.progress.update(table, table_size)  # Mark table as done
+            return
+
+        if table == "user_directory_stream_pos":
+            # We need to make sure there is a single row, `(X, null), as that is
+            # what synapse expects to be there.
+            yield self.postgres_store._simple_insert(
+                table=table,
+                values={"stream_id": None},
+            )
+            self.progress.update(table, table_size)  # Mark table as done
+            return
+
         forward_select = (
             "SELECT rowid, * FROM %s WHERE rowid >= ? ORDER BY rowid LIMIT ?"
             % (table,)

synapse/__init__.py

@@ -16,4 +16,4 @@
 """ This is a reference implementation of a Matrix home server.
 """
 
-__version__ = "0.21.0"
+__version__ = "0.23.0"

synapse/api/auth.py

@@ -23,7 +23,8 @@ from synapse import event_auth
 from synapse.api.constants import EventTypes, Membership, JoinRules
 from synapse.api.errors import AuthError, Codes
 from synapse.types import UserID
-from synapse.util import logcontext
+from synapse.util.caches import register_cache, CACHE_SIZE_FACTOR
+from synapse.util.caches.lrucache import LruCache
 from synapse.util.metrics import Measure
 
 logger = logging.getLogger(__name__)
@@ -39,6 +40,10 @@ AuthEventTypes = (
 GUEST_DEVICE_ID = "guest_device"
 
 
+class _InvalidMacaroonException(Exception):
+    pass
+
+
 class Auth(object):
     """
     FIXME: This class contains a mix of functions for authenticating users
@@ -51,6 +56,9 @@ class Auth(object):
         self.state = hs.get_state_handler()
         self.TOKEN_NOT_FOUND_HTTP_STATUS = 401
 
+        self.token_cache = LruCache(CACHE_SIZE_FACTOR * 10000)
+        register_cache("token_cache", self.token_cache)
+
     @defer.inlineCallbacks
     def check_from_context(self, event, context, do_sig_check=True):
         auth_events_ids = yield self.compute_auth_events(
@@ -144,17 +152,8 @@ class Auth(object):
     @defer.inlineCallbacks
     def check_host_in_room(self, room_id, host):
         with Measure(self.clock, "check_host_in_room"):
-            latest_event_ids = yield self.store.get_latest_event_ids_in_room(room_id)
-
-            logger.debug("calling resolve_state_groups from check_host_in_room")
-            entry = yield self.state.resolve_state_groups(
-                room_id, latest_event_ids
-            )
-
-            ret = yield self.store.is_host_joined(
-                room_id, host, entry.state_group, entry.state
-            )
-            defer.returnValue(ret)
+            latest_event_ids = yield self.store.is_host_joined(room_id, host)
+            defer.returnValue(latest_event_ids)
 
     def _check_joined_room(self, member, user_id, room_id):
         if not member or member.membership != Membership.JOIN:
@@ -209,8 +208,8 @@ class Auth(object):
                 default=[""]
             )[0]
             if user and access_token and ip_addr:
-                logcontext.preserve_fn(self.store.insert_client_ip)(
-                    user=user,
+                self.store.insert_client_ip(
+                    user_id=user.to_string(),
                     access_token=access_token,
                     ip=ip_addr,
                     user_agent=user_agent,
@@ -276,8 +275,8 @@ class Auth(object):
             AuthError if no user by that token exists or the token is invalid.
         """
         try:
-            macaroon = pymacaroons.Macaroon.deserialize(token)
-        except Exception:  # deserialize can throw more-or-less anything
+            user_id, guest = self._parse_and_validate_macaroon(token, rights)
+        except _InvalidMacaroonException:
             # doesn't look like a macaroon: treat it as an opaque token which
             # must be in the database.
             # TODO: it would be nice to get rid of this, but apparently some
@@ -286,19 +285,8 @@ class Auth(object):
             defer.returnValue(r)
 
         try:
-            user_id = self.get_user_id_from_macaroon(macaroon)
             user = UserID.from_string(user_id)
 
-            self.validate_macaroon(
-                macaroon, rights, self.hs.config.expire_access_token,
-                user_id=user_id,
-            )
-
-            guest = False
-            for caveat in macaroon.caveats:
-                if caveat.caveat_id == "guest = true":
-                    guest = True
-
             if guest:
                 # Guest access tokens are not stored in the database (there can
                 # only be one access token per guest, anyway).
@@ -370,6 +358,55 @@ class Auth(object):
                 errcode=Codes.UNKNOWN_TOKEN
             )
 
+    def _parse_and_validate_macaroon(self, token, rights="access"):
+        """Takes a macaroon and tries to parse and validate it. This is cached
+        if and only if rights == access and there isn't an expiry.
+
+        On invalid macaroon raises _InvalidMacaroonException
+
+        Returns:
+            (user_id, is_guest)
+        """
+        if rights == "access":
+            cached = self.token_cache.get(token, None)
+            if cached:
+                return cached
+
+        try:
+            macaroon = pymacaroons.Macaroon.deserialize(token)
+        except Exception:  # deserialize can throw more-or-less anything
+            # doesn't look like a macaroon: treat it as an opaque token which
+            # must be in the database.
+            # TODO: it would be nice to get rid of this, but apparently some
+            # people use access tokens which aren't macaroons
+            raise _InvalidMacaroonException()
+
+        try:
+            user_id = self.get_user_id_from_macaroon(macaroon)
+
+            has_expiry = False
+            guest = False
+            for caveat in macaroon.caveats:
+                if caveat.caveat_id.startswith("time "):
+                    has_expiry = True
+                elif caveat.caveat_id == "guest = true":
+                    guest = True
+
+            self.validate_macaroon(
+                macaroon, rights, self.hs.config.expire_access_token,
+                user_id=user_id,
+            )
+        except (pymacaroons.exceptions.MacaroonException, TypeError, ValueError):
+            raise AuthError(
+                self.TOKEN_NOT_FOUND_HTTP_STATUS, "Invalid macaroon passed.",
+                errcode=Codes.UNKNOWN_TOKEN
+            )
+
+        if not has_expiry and rights == "access":
+            self.token_cache[token] = (user_id, guest)
+
+        return user_id, guest
+
     def get_user_id_from_macaroon(self, macaroon):
         """Retrieve the user_id given by the caveats on the macaroon.
 
@@ -482,6 +519,14 @@ class Auth(object):
             )
 
     def is_server_admin(self, user):
+        """ Check if the given user is a local server admin.
+
+        Args:
+            user (str): mxid of user to check
+
+        Returns:
+            bool: True if the user is an admin
+        """
         return self.store.is_server_admin(user)
 
     @defer.inlineCallbacks

synapse/app/_base.py

@@ -0,0 +1,99 @@
+# -*- coding: utf-8 -*-
+# Copyright 2017 New Vector Ltd
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+import gc
+import logging
+
+import affinity
+from daemonize import Daemonize
+from synapse.util import PreserveLoggingContext
+from synapse.util.rlimit import change_resource_limit
+from twisted.internet import reactor
+
+
+def start_worker_reactor(appname, config):
+    """ Run the reactor in the main process
+
+    Daemonizes if necessary, and then configures some resources, before starting
+    the reactor. Pulls configuration from the 'worker' settings in 'config'.
+
+    Args:
+        appname (str): application name which will be sent to syslog
+        config (synapse.config.Config): config object
+    """
+
+    logger = logging.getLogger(config.worker_app)
+
+    start_reactor(
+        appname,
+        config.soft_file_limit,
+        config.gc_thresholds,
+        config.worker_pid_file,
+        config.worker_daemonize,
+        config.worker_cpu_affinity,
+        logger,
+    )
+
+
+def start_reactor(
+        appname,
+        soft_file_limit,
+        gc_thresholds,
+        pid_file,
+        daemonize,
+        cpu_affinity,
+        logger,
+):
+    """ Run the reactor in the main process
+
+    Daemonizes if necessary, and then configures some resources, before starting
+    the reactor
+
+    Args:
+        appname (str): application name which will be sent to syslog
+        soft_file_limit (int):
+        gc_thresholds:
+        pid_file (str): name of pid file to write to if daemonize is True
+        daemonize (bool): true to run the reactor in a background process
+        cpu_affinity (int|None): cpu affinity mask
+        logger (logging.Logger): logger instance to pass to Daemonize
+    """
+
+    def run():
+        # make sure that we run the reactor with the sentinel log context,
+        # otherwise other PreserveLoggingContext instances will get confused
+        # and complain when they see the logcontext arbitrarily swapping
+        # between the sentinel and `run` logcontexts.
+        with PreserveLoggingContext():
+            logger.info("Running")
+            if cpu_affinity is not None:
+                logger.info("Setting CPU affinity to %s" % cpu_affinity)
+                affinity.set_process_affinity_mask(0, cpu_affinity)
+            change_resource_limit(soft_file_limit)
+            if gc_thresholds:
+                gc.set_threshold(*gc_thresholds)
+            reactor.run()
+
+    if daemonize:
+        daemon = Daemonize(
+            app=appname,
+            pid=pid_file,
+            action=run,
+            auto_close_fds=False,
+            verbose=True,
+            logger=logger,
+        )
+        daemon.start()
+    else:
+        run()

synapse/app/appservice.py

@@ -13,38 +13,31 @@
 # 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
+import sys
 
 import synapse
-
-from synapse.server import HomeServer
+from synapse import events
+from synapse.app import _base
 from synapse.config._base import ConfigError
-from synapse.config.logger import setup_logging
 from synapse.config.homeserver import HomeServerConfig
+from synapse.config.logger import setup_logging
 from synapse.http.site import SynapseSite
-from synapse.metrics.resource import MetricsResource, METRICS_PREFIX
+from synapse.metrics.resource import METRICS_PREFIX, MetricsResource
+from synapse.replication.slave.storage.appservice import SlavedApplicationServiceStore
 from synapse.replication.slave.storage.directory import DirectoryStore
 from synapse.replication.slave.storage.events import SlavedEventStore
-from synapse.replication.slave.storage.appservice import SlavedApplicationServiceStore
 from synapse.replication.slave.storage.registration import SlavedRegistrationStore
 from synapse.replication.tcp.client import ReplicationClientHandler
+from synapse.server import HomeServer
 from synapse.storage.engines import create_engine
 from synapse.util.httpresourcetree import create_resource_tree
-from synapse.util.logcontext import LoggingContext, PreserveLoggingContext, preserve_fn
+from synapse.util.logcontext import LoggingContext, preserve_fn
 from synapse.util.manhole import manhole
-from synapse.util.rlimit import change_resource_limit
 from synapse.util.versionstring import get_version_string
-
-from synapse import events
-
 from twisted.internet import reactor
 from twisted.web.resource import Resource
 
-from daemonize import Daemonize
-
-import sys
-import logging
-import gc
-
 logger = logging.getLogger("synapse.app.appservice")
 
 
@@ -181,36 +174,13 @@ def start(config_options):
     ps.setup()
     ps.start_listening(config.worker_listeners)
 
-    def run():
-        # make sure that we run the reactor with the sentinel log context,
-        # otherwise other PreserveLoggingContext instances will get confused
-        # and complain when they see the logcontext arbitrarily swapping
-        # between the sentinel and `run` logcontexts.
-        with PreserveLoggingContext():
-            logger.info("Running")
-            change_resource_limit(config.soft_file_limit)
-            if config.gc_thresholds:
-                gc.set_threshold(*config.gc_thresholds)
-            reactor.run()
-
     def start():
         ps.get_datastore().start_profiling()
         ps.get_state_handler().start_caching()
 
     reactor.callWhenRunning(start)
 
-    if config.worker_daemonize:
-        daemon = Daemonize(
-            app="synapse-appservice",
-            pid=config.worker_pid_file,
-            action=run,
-            auto_close_fds=False,
-            verbose=True,
-            logger=logger,
-        )
-        daemon.start()
-    else:
-        run()
+    _base.start_worker_reactor("synapse-appservice", config)
 
 
 if __name__ == '__main__':

synapse/app/client_reader.py

@@ -13,47 +13,39 @@
 # 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
+import sys
 
 import synapse
-
+from synapse import events
+from synapse.app import _base
 from synapse.config._base import ConfigError
 from synapse.config.homeserver import HomeServerConfig
 from synapse.config.logger import setup_logging
-from synapse.http.site import SynapseSite
+from synapse.crypto import context_factory
 from synapse.http.server import JsonResource
-from synapse.metrics.resource import MetricsResource, METRICS_PREFIX
+from synapse.http.site import SynapseSite
+from synapse.metrics.resource import METRICS_PREFIX, MetricsResource
 from synapse.replication.slave.storage._base import BaseSlavedStore
 from synapse.replication.slave.storage.appservice import SlavedApplicationServiceStore
+from synapse.replication.slave.storage.client_ips import SlavedClientIpStore
+from synapse.replication.slave.storage.directory import DirectoryStore
 from synapse.replication.slave.storage.events import SlavedEventStore
 from synapse.replication.slave.storage.keys import SlavedKeyStore
-from synapse.replication.slave.storage.room import RoomStore
-from synapse.replication.slave.storage.directory import DirectoryStore
 from synapse.replication.slave.storage.registration import SlavedRegistrationStore
+from synapse.replication.slave.storage.room import RoomStore
 from synapse.replication.slave.storage.transactions import TransactionStore
 from synapse.replication.tcp.client import ReplicationClientHandler
 from synapse.rest.client.v1.room import PublicRoomListRestServlet
 from synapse.server import HomeServer
-from synapse.storage.client_ips import ClientIpStore
 from synapse.storage.engines import create_engine
 from synapse.util.httpresourcetree import create_resource_tree
-from synapse.util.logcontext import LoggingContext, PreserveLoggingContext
+from synapse.util.logcontext import LoggingContext
 from synapse.util.manhole import manhole
-from synapse.util.rlimit import change_resource_limit
 from synapse.util.versionstring import get_version_string
-from synapse.crypto import context_factory
-
-from synapse import events
-
-
 from twisted.internet import reactor
 from twisted.web.resource import Resource
 
-from daemonize import Daemonize
-
-import sys
-import logging
-import gc
-
 logger = logging.getLogger("synapse.app.client_reader")
 
 
@@ -65,8 +57,8 @@ class ClientReaderSlavedStore(
     SlavedApplicationServiceStore,
     SlavedRegistrationStore,
     TransactionStore,
+    SlavedClientIpStore,
     BaseSlavedStore,
-    ClientIpStore,  # After BaseSlavedStore because the constructor is different
 ):
     pass
 
@@ -183,36 +175,13 @@ def start(config_options):
     ss.get_handlers()
     ss.start_listening(config.worker_listeners)
 
-    def run():
-        # make sure that we run the reactor with the sentinel log context,
-        # otherwise other PreserveLoggingContext instances will get confused
-        # and complain when they see the logcontext arbitrarily swapping
-        # between the sentinel and `run` logcontexts.
-        with PreserveLoggingContext():
-            logger.info("Running")
-            change_resource_limit(config.soft_file_limit)
-            if config.gc_thresholds:
-                gc.set_threshold(*config.gc_thresholds)
-            reactor.run()
-
     def start():
         ss.get_state_handler().start_caching()
         ss.get_datastore().start_profiling()
 
     reactor.callWhenRunning(start)
 
-    if config.worker_daemonize:
-        daemon = Daemonize(
-            app="synapse-client-reader",
-            pid=config.worker_pid_file,
-            action=run,
-            auto_close_fds=False,
-            verbose=True,
-            logger=logger,
-        )
-        daemon.start()
-    else:
-        run()
+    _base.start_worker_reactor("synapse-client-reader", config)
 
 
 if __name__ == '__main__':

synapse/app/federation_reader.py

@@ -13,44 +13,36 @@
 # 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
+import sys
 
 import synapse
-
+from synapse import events
+from synapse.api.urls import FEDERATION_PREFIX
+from synapse.app import _base
 from synapse.config._base import ConfigError
 from synapse.config.homeserver import HomeServerConfig
 from synapse.config.logger import setup_logging
+from synapse.crypto import context_factory
+from synapse.federation.transport.server import TransportLayerServer
 from synapse.http.site import SynapseSite
-from synapse.metrics.resource import MetricsResource, METRICS_PREFIX
+from synapse.metrics.resource import METRICS_PREFIX, MetricsResource
 from synapse.replication.slave.storage._base import BaseSlavedStore
+from synapse.replication.slave.storage.directory import DirectoryStore
 from synapse.replication.slave.storage.events import SlavedEventStore
 from synapse.replication.slave.storage.keys import SlavedKeyStore
 from synapse.replication.slave.storage.room import RoomStore
 from synapse.replication.slave.storage.transactions import TransactionStore
-from synapse.replication.slave.storage.directory import DirectoryStore
 from synapse.replication.tcp.client import ReplicationClientHandler
 from synapse.server import HomeServer
 from synapse.storage.engines import create_engine
 from synapse.util.httpresourcetree import create_resource_tree
-from synapse.util.logcontext import LoggingContext, PreserveLoggingContext
+from synapse.util.logcontext import LoggingContext
 from synapse.util.manhole import manhole
-from synapse.util.rlimit import change_resource_limit
 from synapse.util.versionstring import get_version_string
-from synapse.api.urls import FEDERATION_PREFIX
-from synapse.federation.transport.server import TransportLayerServer
-from synapse.crypto import context_factory
-
-from synapse import events
-
-
 from twisted.internet import reactor
 from twisted.web.resource import Resource
 
-from daemonize import Daemonize
-
-import sys
-import logging
-import gc
-
 logger = logging.getLogger("synapse.app.federation_reader")
 
 
@@ -172,36 +164,13 @@ def start(config_options):
     ss.get_handlers()
     ss.start_listening(config.worker_listeners)
 
-    def run():
-        # make sure that we run the reactor with the sentinel log context,
-        # otherwise other PreserveLoggingContext instances will get confused
-        # and complain when they see the logcontext arbitrarily swapping
-        # between the sentinel and `run` logcontexts.
-        with PreserveLoggingContext():
-            logger.info("Running")
-            change_resource_limit(config.soft_file_limit)
-            if config.gc_thresholds:
-                gc.set_threshold(*config.gc_thresholds)
-            reactor.run()
-
     def start():
         ss.get_state_handler().start_caching()
         ss.get_datastore().start_profiling()
 
     reactor.callWhenRunning(start)
 
-    if config.worker_daemonize:
-        daemon = Daemonize(
-            app="synapse-federation-reader",
-            pid=config.worker_pid_file,
-            action=run,
-            auto_close_fds=False,
-            verbose=True,
-            logger=logger,
-        )
-        daemon.start()
-    else:
-        run()
+    _base.start_worker_reactor("synapse-federation-reader", config)
 
 
 if __name__ == '__main__':

synapse/app/federation_sender.py

@@ -13,45 +13,38 @@
 # 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
+import sys
 
 import synapse
-
-from synapse.server import HomeServer
+from synapse import events
+from synapse.app import _base
 from synapse.config._base import ConfigError
-from synapse.config.logger import setup_logging
 from synapse.config.homeserver import HomeServerConfig
+from synapse.config.logger import setup_logging
 from synapse.crypto import context_factory
-from synapse.http.site import SynapseSite
 from synapse.federation import send_queue
-from synapse.metrics.resource import MetricsResource, METRICS_PREFIX
+from synapse.http.site import SynapseSite
+from synapse.metrics.resource import METRICS_PREFIX, MetricsResource
 from synapse.replication.slave.storage.deviceinbox import SlavedDeviceInboxStore
+from synapse.replication.slave.storage.devices import SlavedDeviceStore
 from synapse.replication.slave.storage.events import SlavedEventStore
+from synapse.replication.slave.storage.presence import SlavedPresenceStore
 from synapse.replication.slave.storage.receipts import SlavedReceiptsStore
 from synapse.replication.slave.storage.registration import SlavedRegistrationStore
-from synapse.replication.slave.storage.presence import SlavedPresenceStore
 from synapse.replication.slave.storage.transactions import TransactionStore
-from synapse.replication.slave.storage.devices import SlavedDeviceStore
 from synapse.replication.tcp.client import ReplicationClientHandler
+from synapse.server import HomeServer
 from synapse.storage.engines import create_engine
 from synapse.util.async import Linearizer
 from synapse.util.httpresourcetree import create_resource_tree
-from synapse.util.logcontext import LoggingContext, PreserveLoggingContext, preserve_fn
+from synapse.util.logcontext import LoggingContext, preserve_fn
 from synapse.util.manhole import manhole
-from synapse.util.rlimit import change_resource_limit
 from synapse.util.versionstring import get_version_string
-
-from synapse import events
-
-from twisted.internet import reactor, defer
+from twisted.internet import defer, reactor
 from twisted.web.resource import Resource
 
-from daemonize import Daemonize
-
-import sys
-import logging
-import gc
-
-logger = logging.getLogger("synapse.app.appservice")
+logger = logging.getLogger("synapse.app.federation_sender")
 
 
 class FederationSenderSlaveStore(
@@ -213,36 +206,12 @@ def start(config_options):
     ps.setup()
     ps.start_listening(config.worker_listeners)
 
-    def run():
-        # make sure that we run the reactor with the sentinel log context,
-        # otherwise other PreserveLoggingContext instances will get confused
-        # and complain when they see the logcontext arbitrarily swapping
-        # between the sentinel and `run` logcontexts.
-        with PreserveLoggingContext():
-            logger.info("Running")
-            change_resource_limit(config.soft_file_limit)
-            if config.gc_thresholds:
-                gc.set_threshold(*config.gc_thresholds)
-            reactor.run()
-
     def start():
         ps.get_datastore().start_profiling()
         ps.get_state_handler().start_caching()
 
     reactor.callWhenRunning(start)
-
-    if config.worker_daemonize:
-        daemon = Daemonize(
-            app="synapse-federation-sender",
-            pid=config.worker_pid_file,
-            action=run,
-            auto_close_fds=False,
-            verbose=True,
-            logger=logger,
-        )
-        daemon.start()
-    else:
-        run()
+    _base.start_worker_reactor("synapse-federation-sender", config)
 
 
 class FederationSenderHandler(object):

synapse/app/frontend_proxy.py

@@ -0,0 +1,239 @@
+#!/usr/bin/env python
+# -*- coding: utf-8 -*-
+# Copyright 2016 OpenMarket Ltd
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+import logging
+import sys
+
+import synapse
+from synapse import events
+from synapse.api.errors import SynapseError
+from synapse.app import _base
+from synapse.config._base import ConfigError
+from synapse.config.homeserver import HomeServerConfig
+from synapse.config.logger import setup_logging
+from synapse.crypto import context_factory
+from synapse.http.server import JsonResource
+from synapse.http.servlet import (
+    RestServlet, parse_json_object_from_request,
+)
+from synapse.http.site import SynapseSite
+from synapse.metrics.resource import METRICS_PREFIX, MetricsResource
+from synapse.replication.slave.storage._base import BaseSlavedStore
+from synapse.replication.slave.storage.appservice import SlavedApplicationServiceStore
+from synapse.replication.slave.storage.client_ips import SlavedClientIpStore
+from synapse.replication.slave.storage.devices import SlavedDeviceStore
+from synapse.replication.slave.storage.registration import SlavedRegistrationStore
+from synapse.replication.tcp.client import ReplicationClientHandler
+from synapse.rest.client.v2_alpha._base import client_v2_patterns
+from synapse.server import HomeServer
+from synapse.storage.engines import create_engine
+from synapse.util.httpresourcetree import create_resource_tree
+from synapse.util.logcontext import LoggingContext
+from synapse.util.manhole import manhole
+from synapse.util.versionstring import get_version_string
+from twisted.internet import defer, reactor
+from twisted.web.resource import Resource
+
+logger = logging.getLogger("synapse.app.frontend_proxy")
+
+
+class KeyUploadServlet(RestServlet):
+    PATTERNS = client_v2_patterns("/keys/upload(/(?P<device_id>[^/]+))?$",
+                                  releases=())
+
+    def __init__(self, hs):
+        """
+        Args:
+            hs (synapse.server.HomeServer): server
+        """
+        super(KeyUploadServlet, self).__init__()
+        self.auth = hs.get_auth()
+        self.store = hs.get_datastore()
+        self.http_client = hs.get_simple_http_client()
+        self.main_uri = hs.config.worker_main_http_uri
+
+    @defer.inlineCallbacks
+    def on_POST(self, request, device_id):
+        requester = yield self.auth.get_user_by_req(request, allow_guest=True)
+        user_id = requester.user.to_string()
+        body = parse_json_object_from_request(request)
+
+        if device_id is not None:
+            # passing the device_id here is deprecated; however, we allow it
+            # for now for compatibility with older clients.
+            if (requester.device_id is not None and
+                    device_id != requester.device_id):
+                logger.warning("Client uploading keys for a different device "
+                               "(logged in as %s, uploading for %s)",
+                               requester.device_id, device_id)
+        else:
+            device_id = requester.device_id
+
+        if device_id is None:
+            raise SynapseError(
+                400,
+                "To upload keys, you must pass device_id when authenticating"
+            )
+
+        if body:
+            # They're actually trying to upload something, proxy to main synapse.
+            result = yield self.http_client.post_json_get_json(
+                self.main_uri + request.uri,
+                body,
+            )
+
+            defer.returnValue((200, result))
+        else:
+            # Just interested in counts.
+            result = yield self.store.count_e2e_one_time_keys(user_id, device_id)
+            defer.returnValue((200, {"one_time_key_counts": result}))
+
+
+class FrontendProxySlavedStore(
+    SlavedDeviceStore,
+    SlavedClientIpStore,
+    SlavedApplicationServiceStore,
+    SlavedRegistrationStore,
+    BaseSlavedStore,
+):
+    pass
+
+
+class FrontendProxyServer(HomeServer):
+    def get_db_conn(self, run_new_connection=True):
+        # Any param beginning with cp_ is a parameter for adbapi, and should
+        # not be passed to the database engine.
+        db_params = {
+            k: v for k, v in self.db_config.get("args", {}).items()
+            if not k.startswith("cp_")
+        }
+        db_conn = self.database_engine.module.connect(**db_params)
+
+        if run_new_connection:
+            self.database_engine.on_new_connection(db_conn)
+        return db_conn
+
+    def setup(self):
+        logger.info("Setting up.")
+        self.datastore = FrontendProxySlavedStore(self.get_db_conn(), self)
+        logger.info("Finished setting up.")
+
+    def _listen_http(self, listener_config):
+        port = listener_config["port"]
+        bind_addresses = listener_config["bind_addresses"]
+        site_tag = listener_config.get("tag", port)
+        resources = {}
+        for res in listener_config["resources"]:
+            for name in res["names"]:
+                if name == "metrics":
+                    resources[METRICS_PREFIX] = MetricsResource(self)
+                elif name == "client":
+                    resource = JsonResource(self, canonical_json=False)
+                    KeyUploadServlet(self).register(resource)
+                    resources.update({
+                        "/_matrix/client/r0": resource,
+                        "/_matrix/client/unstable": resource,
+                        "/_matrix/client/v2_alpha": resource,
+                        "/_matrix/client/api/v1": resource,
+                    })
+
+        root_resource = create_resource_tree(resources, Resource())
+
+        for address in bind_addresses:
+            reactor.listenTCP(
+                port,
+                SynapseSite(
+                    "synapse.access.http.%s" % (site_tag,),
+                    site_tag,
+                    listener_config,
+                    root_resource,
+                ),
+                interface=address
+            )
+
+        logger.info("Synapse client reader now listening on port %d", port)
+
+    def start_listening(self, listeners):
+        for listener in listeners:
+            if listener["type"] == "http":
+                self._listen_http(listener)
+            elif listener["type"] == "manhole":
+                bind_addresses = listener["bind_addresses"]
+
+                for address in bind_addresses:
+                    reactor.listenTCP(
+                        listener["port"],
+                        manhole(
+                            username="matrix",
+                            password="rabbithole",
+                            globals={"hs": self},
+                        ),
+                        interface=address
+                    )
+            else:
+                logger.warn("Unrecognized listener type: %s", listener["type"])
+
+        self.get_tcp_replication().start_replication(self)
+
+    def build_tcp_replication(self):
+        return ReplicationClientHandler(self.get_datastore())
+
+
+def start(config_options):
+    try:
+        config = HomeServerConfig.load_config(
+            "Synapse frontend proxy", config_options
+        )
+    except ConfigError as e:
+        sys.stderr.write("\n" + e.message + "\n")
+        sys.exit(1)
+
+    assert config.worker_app == "synapse.app.frontend_proxy"
+
+    assert config.worker_main_http_uri is not None
+
+    setup_logging(config, use_worker_options=True)
+
+    events.USE_FROZEN_DICTS = config.use_frozen_dicts
+
+    database_engine = create_engine(config.database_config)
+
+    tls_server_context_factory = context_factory.ServerContextFactory(config)
+
+    ss = FrontendProxyServer(
+        config.server_name,
+        db_config=config.database_config,
+        tls_server_context_factory=tls_server_context_factory,
+        config=config,
+        version_string="Synapse/" + get_version_string(synapse),
+        database_engine=database_engine,
+    )
+
+    ss.setup()
+    ss.get_handlers()
+    ss.start_listening(config.worker_listeners)
+
+    def start():
+        ss.get_state_handler().start_caching()
+        ss.get_datastore().start_profiling()
+
+    reactor.callWhenRunning(start)
+
+    _base.start_worker_reactor("synapse-frontend-proxy", config)
+
+
+if __name__ == '__main__':
+    with LoggingContext("main"):
+        start(sys.argv[1:])

synapse/app/homeserver.py

@@ -13,61 +13,48 @@
 # 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 synapse
-
 import gc
 import logging
 import os
 import sys
 
+import synapse
 import synapse.config.logger
+from synapse import events
+from synapse.api.urls import CONTENT_REPO_PREFIX, FEDERATION_PREFIX, \
+    LEGACY_MEDIA_PREFIX, MEDIA_PREFIX, SERVER_KEY_PREFIX, SERVER_KEY_V2_PREFIX, \
+    STATIC_PREFIX, WEB_CLIENT_PREFIX
+from synapse.app import _base
 from synapse.config._base import ConfigError
-
-from synapse.python_dependencies import (
-    check_requirements, CONDITIONAL_REQUIREMENTS
-)
-
-from synapse.rest import ClientRestResource
-from synapse.storage.engines import create_engine, IncorrectDatabaseSetup
-from synapse.storage import are_all_users_on_domain
-from synapse.storage.prepare_database import UpgradeDatabaseException, prepare_database
-
-from synapse.server import HomeServer
-
-from twisted.internet import reactor, task, defer
-from twisted.application import service
-from twisted.web.resource import Resource, EncodingResourceWrapper
-from twisted.web.static import File
-from twisted.web.server import GzipEncoderFactory
-from synapse.http.server import RootRedirect
-from synapse.rest.media.v0.content_repository import ContentRepoResource
-from synapse.rest.media.v1.media_repository import MediaRepositoryResource
-from synapse.rest.key.v1.server_key_resource import LocalKey
-from synapse.rest.key.v2 import KeyApiV2Resource
-from synapse.api.urls import (
-    FEDERATION_PREFIX, WEB_CLIENT_PREFIX, CONTENT_REPO_PREFIX,
-    SERVER_KEY_PREFIX, LEGACY_MEDIA_PREFIX, MEDIA_PREFIX, STATIC_PREFIX,
-    SERVER_KEY_V2_PREFIX,
-)
 from synapse.config.homeserver import HomeServerConfig
 from synapse.crypto import context_factory
-from synapse.util.logcontext import LoggingContext, PreserveLoggingContext
-from synapse.metrics import register_memory_metrics, get_metrics_for
-from synapse.metrics.resource import MetricsResource, METRICS_PREFIX
-from synapse.replication.tcp.resource import ReplicationStreamProtocolFactory
 from synapse.federation.transport.server import TransportLayerServer
-
+from synapse.http.server import RootRedirect
+from synapse.http.site import SynapseSite
+from synapse.metrics import register_memory_metrics
+from synapse.metrics.resource import METRICS_PREFIX, MetricsResource
+from synapse.python_dependencies import CONDITIONAL_REQUIREMENTS, \
+    check_requirements
+from synapse.replication.tcp.resource import ReplicationStreamProtocolFactory
+from synapse.rest import ClientRestResource
+from synapse.rest.key.v1.server_key_resource import LocalKey
+from synapse.rest.key.v2 import KeyApiV2Resource
+from synapse.rest.media.v0.content_repository import ContentRepoResource
+from synapse.rest.media.v1.media_repository import MediaRepositoryResource
+from synapse.server import HomeServer
+from synapse.storage import are_all_users_on_domain
+from synapse.storage.engines import IncorrectDatabaseSetup, create_engine
+from synapse.storage.prepare_database import UpgradeDatabaseException, prepare_database
+from synapse.util.httpresourcetree import create_resource_tree
+from synapse.util.logcontext import LoggingContext
+from synapse.util.manhole import manhole
 from synapse.util.rlimit import change_resource_limit
 from synapse.util.versionstring import get_version_string
-from synapse.util.httpresourcetree import create_resource_tree
-from synapse.util.manhole import manhole
-
-from synapse.http.site import SynapseSite
-
-from synapse import events
-
-from daemonize import Daemonize
+from twisted.application import service
+from twisted.internet import defer, reactor
+from twisted.web.resource import EncodingResourceWrapper, Resource
+from twisted.web.server import GzipEncoderFactory
+from twisted.web.static import File
 
 logger = logging.getLogger("synapse.app.homeserver")
 
@@ -398,7 +385,8 @@ def run(hs):
         ThreadPool._worker = profile(ThreadPool._worker)
         reactor.run = profile(reactor.run)
 
-    start_time = hs.get_clock().time()
+    clock = hs.get_clock()
+    start_time = clock.time()
 
     stats = {}
 
@@ -410,41 +398,23 @@ def run(hs):
         if uptime < 0:
             uptime = 0
 
-        # If the stats directory is empty then this is the first time we've
-        # reported stats.
-        first_time = not stats
-
         stats["homeserver"] = hs.config.server_name
         stats["timestamp"] = now
         stats["uptime_seconds"] = uptime
         stats["total_users"] = yield hs.get_datastore().count_all_users()
 
+        total_nonbridged_users = yield hs.get_datastore().count_nonbridged_users()
+        stats["total_nonbridged_users"] = total_nonbridged_users
+
         room_count = yield hs.get_datastore().get_room_count()
         stats["total_room_count"] = room_count
 
         stats["daily_active_users"] = yield hs.get_datastore().count_daily_users()
-        daily_messages = yield hs.get_datastore().count_daily_messages()
-        if daily_messages is not None:
-            stats["daily_messages"] = daily_messages
-        else:
-            stats.pop("daily_messages", None)
+        stats["daily_active_rooms"] = yield hs.get_datastore().count_daily_active_rooms()
+        stats["daily_messages"] = yield hs.get_datastore().count_daily_messages()
 
-        if first_time:
-            # Add callbacks to report the synapse stats as metrics whenever
-            # prometheus requests them, typically every 30s.
-            # As some of the stats are expensive to calculate we only update
-            # them when synapse phones home to matrix.org every 24 hours.
-            metrics = get_metrics_for("synapse.usage")
-            metrics.add_callback("timestamp", lambda: stats["timestamp"])
-            metrics.add_callback("uptime_seconds", lambda: stats["uptime_seconds"])
-            metrics.add_callback("total_users", lambda: stats["total_users"])
-            metrics.add_callback("total_room_count", lambda: stats["total_room_count"])
-            metrics.add_callback(
-                "daily_active_users", lambda: stats["daily_active_users"]
-            )
-            metrics.add_callback(
-                "daily_messages", lambda: stats.get("daily_messages", 0)
-            )
+        daily_sent_messages = yield hs.get_datastore().count_daily_sent_messages()
+        stats["daily_sent_messages"] = daily_sent_messages
 
         logger.info("Reporting stats to matrix.org: %s" % (stats,))
         try:
@@ -456,41 +426,25 @@ def run(hs):
             logger.warn("Error reporting stats: %s", e)
 
     if hs.config.report_stats:
-        phone_home_task = task.LoopingCall(phone_stats_home)
-        logger.info("Scheduling stats reporting for 24 hour intervals")
-        phone_home_task.start(60 * 60 * 24, now=False)
+        logger.info("Scheduling stats reporting for 3 hour intervals")
+        clock.looping_call(phone_stats_home, 3 * 60 * 60 * 1000)
 
-    def in_thread():
-        # Uncomment to enable tracing of log context changes.
-        # sys.settrace(logcontext_tracer)
+        # We wait 5 minutes to send the first set of stats as the server can
+        # be quite busy the first few minutes
+        clock.call_later(5 * 60, phone_stats_home)
 
-        # make sure that we run the reactor with the sentinel log context,
-        # otherwise other PreserveLoggingContext instances will get confused
-        # and complain when they see the logcontext arbitrarily swapping
-        # between the sentinel and `run` logcontexts.
-        with PreserveLoggingContext():
-            change_resource_limit(hs.config.soft_file_limit)
-            if hs.config.gc_thresholds:
-                gc.set_threshold(*hs.config.gc_thresholds)
-            reactor.run()
+    if hs.config.daemonize and hs.config.print_pidfile:
+        print (hs.config.pid_file)
 
-    if hs.config.daemonize:
-
-        if hs.config.print_pidfile:
-            print (hs.config.pid_file)
-
-        daemon = Daemonize(
-            app="synapse-homeserver",
-            pid=hs.config.pid_file,
-            action=lambda: in_thread(),
-            auto_close_fds=False,
-            verbose=True,
-            logger=logger,
-        )
-
-        daemon.start()
-    else:
-        in_thread()
+    _base.start_reactor(
+        "synapse-homeserver",
+        hs.config.soft_file_limit,
+        hs.config.gc_thresholds,
+        hs.config.pid_file,
+        hs.config.daemonize,
+        hs.config.cpu_affinity,
+        logger,
+    )
 
 
 def main():

synapse/app/media_repository.py

@@ -13,57 +13,49 @@
 # 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
+import sys
 
 import synapse
-
+from synapse import events
+from synapse.api.urls import (
+    CONTENT_REPO_PREFIX, LEGACY_MEDIA_PREFIX, MEDIA_PREFIX
+)
+from synapse.app import _base
 from synapse.config._base import ConfigError
 from synapse.config.homeserver import HomeServerConfig
 from synapse.config.logger import setup_logging
+from synapse.crypto import context_factory
 from synapse.http.site import SynapseSite
-from synapse.metrics.resource import MetricsResource, METRICS_PREFIX
+from synapse.metrics.resource import METRICS_PREFIX, MetricsResource
 from synapse.replication.slave.storage._base import BaseSlavedStore
 from synapse.replication.slave.storage.appservice import SlavedApplicationServiceStore
+from synapse.replication.slave.storage.client_ips import SlavedClientIpStore
 from synapse.replication.slave.storage.registration import SlavedRegistrationStore
 from synapse.replication.slave.storage.transactions import TransactionStore
 from synapse.replication.tcp.client import ReplicationClientHandler
 from synapse.rest.media.v0.content_repository import ContentRepoResource
 from synapse.rest.media.v1.media_repository import MediaRepositoryResource
 from synapse.server import HomeServer
-from synapse.storage.client_ips import ClientIpStore
 from synapse.storage.engines import create_engine
 from synapse.storage.media_repository import MediaRepositoryStore
 from synapse.util.httpresourcetree import create_resource_tree
-from synapse.util.logcontext import LoggingContext, PreserveLoggingContext
+from synapse.util.logcontext import LoggingContext
 from synapse.util.manhole import manhole
-from synapse.util.rlimit import change_resource_limit
 from synapse.util.versionstring import get_version_string
-from synapse.api.urls import (
-    CONTENT_REPO_PREFIX, LEGACY_MEDIA_PREFIX, MEDIA_PREFIX
-)
-from synapse.crypto import context_factory
-
-from synapse import events
-
-
 from twisted.internet import reactor
 from twisted.web.resource import Resource
 
-from daemonize import Daemonize
-
-import sys
-import logging
-import gc
-
 logger = logging.getLogger("synapse.app.media_repository")
 
 
 class MediaRepositorySlavedStore(
     SlavedApplicationServiceStore,
     SlavedRegistrationStore,
+    SlavedClientIpStore,
     TransactionStore,
     BaseSlavedStore,
     MediaRepositoryStore,
-    ClientIpStore,
 ):
     pass
 
@@ -180,36 +172,13 @@ def start(config_options):
     ss.get_handlers()
     ss.start_listening(config.worker_listeners)
 
-    def run():
-        # make sure that we run the reactor with the sentinel log context,
-        # otherwise other PreserveLoggingContext instances will get confused
-        # and complain when they see the logcontext arbitrarily swapping
-        # between the sentinel and `run` logcontexts.
-        with PreserveLoggingContext():
-            logger.info("Running")
-            change_resource_limit(config.soft_file_limit)
-            if config.gc_thresholds:
-                gc.set_threshold(*config.gc_thresholds)
-            reactor.run()
-
     def start():
         ss.get_state_handler().start_caching()
         ss.get_datastore().start_profiling()
 
     reactor.callWhenRunning(start)
 
-    if config.worker_daemonize:
-        daemon = Daemonize(
-            app="synapse-media-repository",
-            pid=config.worker_pid_file,
-            action=run,
-            auto_close_fds=False,
-            verbose=True,
-            logger=logger,
-        )
-        daemon.start()
-    else:
-        run()
+    _base.start_worker_reactor("synapse-media-repository", config)
 
 
 if __name__ == '__main__':

synapse/app/pusher.py

@@ -13,41 +13,33 @@
 # 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
+import sys
 
 import synapse
-
-from synapse.server import HomeServer
+from synapse import events
+from synapse.app import _base
 from synapse.config._base import ConfigError
-from synapse.config.logger import setup_logging
 from synapse.config.homeserver import HomeServerConfig
+from synapse.config.logger import setup_logging
 from synapse.http.site import SynapseSite
-from synapse.metrics.resource import MetricsResource, METRICS_PREFIX
-from synapse.storage.roommember import RoomMemberStore
+from synapse.metrics.resource import METRICS_PREFIX, MetricsResource
+from synapse.replication.slave.storage.account_data import SlavedAccountDataStore
 from synapse.replication.slave.storage.events import SlavedEventStore
 from synapse.replication.slave.storage.pushers import SlavedPusherStore
 from synapse.replication.slave.storage.receipts import SlavedReceiptsStore
-from synapse.replication.slave.storage.account_data import SlavedAccountDataStore
 from synapse.replication.tcp.client import ReplicationClientHandler
-from synapse.storage.engines import create_engine
+from synapse.server import HomeServer
 from synapse.storage import DataStore
+from synapse.storage.engines import create_engine
+from synapse.storage.roommember import RoomMemberStore
 from synapse.util.httpresourcetree import create_resource_tree
-from synapse.util.logcontext import LoggingContext, preserve_fn, \
-    PreserveLoggingContext
+from synapse.util.logcontext import LoggingContext, preserve_fn
 from synapse.util.manhole import manhole
-from synapse.util.rlimit import change_resource_limit
 from synapse.util.versionstring import get_version_string
-
-from synapse import events
-
-from twisted.internet import reactor, defer
+from twisted.internet import defer, reactor
 from twisted.web.resource import Resource
 
-from daemonize import Daemonize
-
-import sys
-import logging
-import gc
-
 logger = logging.getLogger("synapse.app.pusher")
 
 
@@ -244,18 +236,6 @@ def start(config_options):
     ps.setup()
     ps.start_listening(config.worker_listeners)
 
-    def run():
-        # make sure that we run the reactor with the sentinel log context,
-        # otherwise other PreserveLoggingContext instances will get confused
-        # and complain when they see the logcontext arbitrarily swapping
-        # between the sentinel and `run` logcontexts.
-        with PreserveLoggingContext():
-            logger.info("Running")
-            change_resource_limit(config.soft_file_limit)
-            if config.gc_thresholds:
-                gc.set_threshold(*config.gc_thresholds)
-            reactor.run()
-
     def start():
         ps.get_pusherpool().start()
         ps.get_datastore().start_profiling()
@@ -263,18 +243,7 @@ def start(config_options):
 
     reactor.callWhenRunning(start)
 
-    if config.worker_daemonize:
-        daemon = Daemonize(
-            app="synapse-pusher",
-            pid=config.worker_pid_file,
-            action=run,
-            auto_close_fds=False,
-            verbose=True,
-            logger=logger,
-        )
-        daemon.start()
-    else:
-        run()
+    _base.start_worker_reactor("synapse-pusher", config)
 
 
 if __name__ == '__main__':

synapse/app/synchrotron.py

@@ -13,56 +13,50 @@
 # 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 contextlib
+import logging
+import sys
 
 import synapse
-
 from synapse.api.constants import EventTypes
+from synapse.app import _base
 from synapse.config._base import ConfigError
 from synapse.config.homeserver import HomeServerConfig
 from synapse.config.logger import setup_logging
 from synapse.handlers.presence import PresenceHandler, get_interested_parties
-from synapse.http.site import SynapseSite
 from synapse.http.server import JsonResource
-from synapse.metrics.resource import MetricsResource, METRICS_PREFIX
-from synapse.rest.client.v2_alpha import sync
-from synapse.rest.client.v1 import events
-from synapse.rest.client.v1.room import RoomInitialSyncRestServlet
-from synapse.rest.client.v1.initial_sync import InitialSyncRestServlet
+from synapse.http.site import SynapseSite
+from synapse.metrics.resource import METRICS_PREFIX, MetricsResource
 from synapse.replication.slave.storage._base import BaseSlavedStore
-from synapse.replication.slave.storage.events import SlavedEventStore
-from synapse.replication.slave.storage.receipts import SlavedReceiptsStore
 from synapse.replication.slave.storage.account_data import SlavedAccountDataStore
 from synapse.replication.slave.storage.appservice import SlavedApplicationServiceStore
-from synapse.replication.slave.storage.registration import SlavedRegistrationStore
-from synapse.replication.slave.storage.filtering import SlavedFilteringStore
-from synapse.replication.slave.storage.push_rule import SlavedPushRuleStore
-from synapse.replication.slave.storage.presence import SlavedPresenceStore
+from synapse.replication.slave.storage.client_ips import SlavedClientIpStore
 from synapse.replication.slave.storage.deviceinbox import SlavedDeviceInboxStore
 from synapse.replication.slave.storage.devices import SlavedDeviceStore
+from synapse.replication.slave.storage.events import SlavedEventStore
+from synapse.replication.slave.storage.filtering import SlavedFilteringStore
+from synapse.replication.slave.storage.presence import SlavedPresenceStore
+from synapse.replication.slave.storage.push_rule import SlavedPushRuleStore
+from synapse.replication.slave.storage.receipts import SlavedReceiptsStore
+from synapse.replication.slave.storage.registration import SlavedRegistrationStore
 from synapse.replication.slave.storage.room import RoomStore
 from synapse.replication.tcp.client import ReplicationClientHandler
+from synapse.rest.client.v1 import events
+from synapse.rest.client.v1.initial_sync import InitialSyncRestServlet
+from synapse.rest.client.v1.room import RoomInitialSyncRestServlet
+from synapse.rest.client.v2_alpha import sync
 from synapse.server import HomeServer
-from synapse.storage.client_ips import ClientIpStore
 from synapse.storage.engines import create_engine
 from synapse.storage.presence import UserPresenceState
 from synapse.storage.roommember import RoomMemberStore
 from synapse.util.httpresourcetree import create_resource_tree
-from synapse.util.logcontext import LoggingContext, PreserveLoggingContext, preserve_fn
+from synapse.util.logcontext import LoggingContext, preserve_fn
 from synapse.util.manhole import manhole
-from synapse.util.rlimit import change_resource_limit
 from synapse.util.stringutils import random_string
 from synapse.util.versionstring import get_version_string
-
-from twisted.internet import reactor, defer
+from twisted.internet import defer, reactor
 from twisted.web.resource import Resource
 
-from daemonize import Daemonize
-
-import sys
-import logging
-import contextlib
-import gc
-
 logger = logging.getLogger("synapse.app.synchrotron")
 
 
@@ -77,9 +71,9 @@ class SynchrotronSlavedStore(
     SlavedPresenceStore,
     SlavedDeviceInboxStore,
     SlavedDeviceStore,
+    SlavedClientIpStore,
     RoomStore,
     BaseSlavedStore,
-    ClientIpStore,  # After BaseSlavedStore because the constructor is different
 ):
     who_forgot_in_room = (
         RoomMemberStore.__dict__["who_forgot_in_room"]
@@ -440,36 +434,13 @@ def start(config_options):
     ss.setup()
     ss.start_listening(config.worker_listeners)
 
-    def run():
-        # make sure that we run the reactor with the sentinel log context,
-        # otherwise other PreserveLoggingContext instances will get confused
-        # and complain when they see the logcontext arbitrarily swapping
-        # between the sentinel and `run` logcontexts.
-        with PreserveLoggingContext():
-            logger.info("Running")
-            change_resource_limit(config.soft_file_limit)
-            if config.gc_thresholds:
-                gc.set_threshold(*config.gc_thresholds)
-            reactor.run()
-
     def start():
         ss.get_datastore().start_profiling()
         ss.get_state_handler().start_caching()
 
     reactor.callWhenRunning(start)
 
-    if config.worker_daemonize:
-        daemon = Daemonize(
-            app="synapse-synchrotron",
-            pid=config.worker_pid_file,
-            action=run,
-            auto_close_fds=False,
-            verbose=True,
-            logger=logger,
-        )
-        daemon.start()
-    else:
-        run()
+    _base.start_worker_reactor("synapse-synchrotron", config)
 
 
 if __name__ == '__main__':

synapse/app/user_dir.py

@@ -0,0 +1,241 @@
+#!/usr/bin/env python
+# -*- coding: utf-8 -*-
+# Copyright 2017 Vector Creations Ltd
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+import logging
+import sys
+
+import synapse
+from synapse import events
+from synapse.app import _base
+from synapse.config._base import ConfigError
+from synapse.config.homeserver import HomeServerConfig
+from synapse.config.logger import setup_logging
+from synapse.crypto import context_factory
+from synapse.http.server import JsonResource
+from synapse.http.site import SynapseSite
+from synapse.metrics.resource import METRICS_PREFIX, MetricsResource
+from synapse.replication.slave.storage._base import BaseSlavedStore
+from synapse.replication.slave.storage.appservice import SlavedApplicationServiceStore
+from synapse.replication.slave.storage.client_ips import SlavedClientIpStore
+from synapse.replication.slave.storage.events import SlavedEventStore
+from synapse.replication.slave.storage.registration import SlavedRegistrationStore
+from synapse.replication.tcp.client import ReplicationClientHandler
+from synapse.rest.client.v2_alpha import user_directory
+from synapse.server import HomeServer
+from synapse.storage.engines import create_engine
+from synapse.storage.user_directory import UserDirectoryStore
+from synapse.util.caches.stream_change_cache import StreamChangeCache
+from synapse.util.httpresourcetree import create_resource_tree
+from synapse.util.logcontext import LoggingContext, preserve_fn
+from synapse.util.manhole import manhole
+from synapse.util.versionstring import get_version_string
+from twisted.internet import reactor
+from twisted.web.resource import Resource
+
+logger = logging.getLogger("synapse.app.user_dir")
+
+
+class UserDirectorySlaveStore(
+    SlavedEventStore,
+    SlavedApplicationServiceStore,
+    SlavedRegistrationStore,
+    SlavedClientIpStore,
+    UserDirectoryStore,
+    BaseSlavedStore,
+):
+    def __init__(self, db_conn, hs):
+        super(UserDirectorySlaveStore, self).__init__(db_conn, hs)
+
+        events_max = self._stream_id_gen.get_current_token()
+        curr_state_delta_prefill, min_curr_state_delta_id = self._get_cache_dict(
+            db_conn, "current_state_delta_stream",
+            entity_column="room_id",
+            stream_column="stream_id",
+            max_value=events_max,  # As we share the stream id with events token
+            limit=1000,
+        )
+        self._curr_state_delta_stream_cache = StreamChangeCache(
+            "_curr_state_delta_stream_cache", min_curr_state_delta_id,
+            prefilled_cache=curr_state_delta_prefill,
+        )
+
+        self._current_state_delta_pos = events_max
+
+    def stream_positions(self):
+        result = super(UserDirectorySlaveStore, self).stream_positions()
+        result["current_state_deltas"] = self._current_state_delta_pos
+        return result
+
+    def process_replication_rows(self, stream_name, token, rows):
+        if stream_name == "current_state_deltas":
+            self._current_state_delta_pos = token
+            for row in rows:
+                self._curr_state_delta_stream_cache.entity_has_changed(
+                    row.room_id, token
+                )
+        return super(UserDirectorySlaveStore, self).process_replication_rows(
+            stream_name, token, rows
+        )
+
+
+class UserDirectoryServer(HomeServer):
+    def get_db_conn(self, run_new_connection=True):
+        # Any param beginning with cp_ is a parameter for adbapi, and should
+        # not be passed to the database engine.
+        db_params = {
+            k: v for k, v in self.db_config.get("args", {}).items()
+            if not k.startswith("cp_")
+        }
+        db_conn = self.database_engine.module.connect(**db_params)
+
+        if run_new_connection:
+            self.database_engine.on_new_connection(db_conn)
+        return db_conn
+
+    def setup(self):
+        logger.info("Setting up.")
+        self.datastore = UserDirectorySlaveStore(self.get_db_conn(), self)
+        logger.info("Finished setting up.")
+
+    def _listen_http(self, listener_config):
+        port = listener_config["port"]
+        bind_addresses = listener_config["bind_addresses"]
+        site_tag = listener_config.get("tag", port)
+        resources = {}
+        for res in listener_config["resources"]:
+            for name in res["names"]:
+                if name == "metrics":
+                    resources[METRICS_PREFIX] = MetricsResource(self)
+                elif name == "client":
+                    resource = JsonResource(self, canonical_json=False)
+                    user_directory.register_servlets(self, resource)
+                    resources.update({
+                        "/_matrix/client/r0": resource,
+                        "/_matrix/client/unstable": resource,
+                        "/_matrix/client/v2_alpha": resource,
+                        "/_matrix/client/api/v1": resource,
+                    })
+
+        root_resource = create_resource_tree(resources, Resource())
+
+        for address in bind_addresses:
+            reactor.listenTCP(
+                port,
+                SynapseSite(
+                    "synapse.access.http.%s" % (site_tag,),
+                    site_tag,
+                    listener_config,
+                    root_resource,
+                ),
+                interface=address
+            )
+
+        logger.info("Synapse user_dir now listening on port %d", port)
+
+    def start_listening(self, listeners):
+        for listener in listeners:
+            if listener["type"] == "http":
+                self._listen_http(listener)
+            elif listener["type"] == "manhole":
+                bind_addresses = listener["bind_addresses"]
+
+                for address in bind_addresses:
+                    reactor.listenTCP(
+                        listener["port"],
+                        manhole(
+                            username="matrix",
+                            password="rabbithole",
+                            globals={"hs": self},
+                        ),
+                        interface=address
+                    )
+            else:
+                logger.warn("Unrecognized listener type: %s", listener["type"])
+
+        self.get_tcp_replication().start_replication(self)
+
+    def build_tcp_replication(self):
+        return UserDirectoryReplicationHandler(self)
+
+
+class UserDirectoryReplicationHandler(ReplicationClientHandler):
+    def __init__(self, hs):
+        super(UserDirectoryReplicationHandler, self).__init__(hs.get_datastore())
+        self.user_directory = hs.get_user_directory_handler()
+
+    def on_rdata(self, stream_name, token, rows):
+        super(UserDirectoryReplicationHandler, self).on_rdata(
+            stream_name, token, rows
+        )
+        if stream_name == "current_state_deltas":
+            preserve_fn(self.user_directory.notify_new_event)()
+
+
+def start(config_options):
+    try:
+        config = HomeServerConfig.load_config(
+            "Synapse user directory", config_options
+        )
+    except ConfigError as e:
+        sys.stderr.write("\n" + e.message + "\n")
+        sys.exit(1)
+
+    assert config.worker_app == "synapse.app.user_dir"
+
+    setup_logging(config, use_worker_options=True)
+
+    events.USE_FROZEN_DICTS = config.use_frozen_dicts
+
+    database_engine = create_engine(config.database_config)
+
+    if config.update_user_directory:
+        sys.stderr.write(
+            "\nThe update_user_directory must be disabled in the main synapse process"
+            "\nbefore they can be run in a separate worker."
+            "\nPlease add ``update_user_directory: false`` to the main config"
+            "\n"
+        )
+        sys.exit(1)
+
+    # Force the pushers to start since they will be disabled in the main config
+    config.update_user_directory = True
+
+    tls_server_context_factory = context_factory.ServerContextFactory(config)
+
+    ps = UserDirectoryServer(
+        config.server_name,
+        db_config=config.database_config,
+        tls_server_context_factory=tls_server_context_factory,
+        config=config,
+        version_string="Synapse/" + get_version_string(synapse),
+        database_engine=database_engine,
+    )
+
+    ps.setup()
+    ps.start_listening(config.worker_listeners)
+
+    def start():
+        ps.get_datastore().start_profiling()
+        ps.get_state_handler().start_caching()
+
+    reactor.callWhenRunning(start)
+
+    _base.start_worker_reactor("synapse-user-dir", config)
+
+
+if __name__ == '__main__':
+    with LoggingContext("main"):
+        start(sys.argv[1:])

synapse/appservice/__init__.py

@@ -241,6 +241,16 @@ class ApplicationService(object):
     def is_exclusive_room(self, room_id):
         return self._is_exclusive(ApplicationService.NS_ROOMS, room_id)
 
+    def get_exlusive_user_regexes(self):
+        """Get the list of regexes used to determine if a user is exclusively
+        registered by the AS
+        """
+        return [
+            regex_obj["regex"]
+            for regex_obj in self.namespaces[ApplicationService.NS_USERS]
+            if regex_obj["exclusive"]
+        ]
+
     def is_rate_limited(self):
         return self.rate_limited
 

synapse/config/homeserver.py

@@ -33,6 +33,8 @@ from .jwt import JWTConfig
 from .password_auth_providers import PasswordAuthProviderConfig
 from .emailconfig import EmailConfig
 from .workers import WorkerConfig
+from .push import PushConfig
+from .spam_checker import SpamCheckerConfig
 
 
 class HomeServerConfig(TlsConfig, ServerConfig, DatabaseConfig, LoggingConfig,
@@ -40,7 +42,8 @@ class HomeServerConfig(TlsConfig, ServerConfig, DatabaseConfig, LoggingConfig,
                        VoipConfig, RegistrationConfig, MetricsConfig, ApiConfig,
                        AppServiceConfig, KeyConfig, SAML2Config, CasConfig,
                        JWTConfig, PasswordConfig, EmailConfig,
-                       WorkerConfig, PasswordAuthProviderConfig,):
+                       WorkerConfig, PasswordAuthProviderConfig, PushConfig,
+                       SpamCheckerConfig,):
     pass
 
 

synapse/config/password_auth_providers.py

@@ -15,13 +15,15 @@
 
 from ._base import Config, ConfigError
 
-import importlib
+from synapse.util.module_loader import load_module
 
 
 class PasswordAuthProviderConfig(Config):
     def read_config(self, config):
         self.password_providers = []
 
+        provider_config = None
+
         # We want to be backwards compatible with the old `ldap_config`
         # param.
         ldap_config = config.get("ldap_config", {})
@@ -38,19 +40,15 @@ class PasswordAuthProviderConfig(Config):
             if provider['module'] == "synapse.util.ldap_auth_provider.LdapAuthProvider":
                 from ldap_auth_provider import LdapAuthProvider
                 provider_class = LdapAuthProvider
+                try:
+                    provider_config = provider_class.parse_config(provider["config"])
+                except Exception as e:
+                    raise ConfigError(
+                        "Failed to parse config for %r: %r" % (provider['module'], e)
+                    )
             else:
-                # We need to import the module, and then pick the class out of
-                # that, so we split based on the last dot.
-                module, clz = provider['module'].rsplit(".", 1)
-                module = importlib.import_module(module)
-                provider_class = getattr(module, clz)
+                (provider_class, provider_config) = load_module(provider)
 
-            try:
-                provider_config = provider_class.parse_config(provider["config"])
-            except Exception as e:
-                raise ConfigError(
-                    "Failed to parse config for %r: %r" % (provider['module'], e)
-                )
             self.password_providers.append((provider_class, provider_config))
 
     def default_config(self, **kwargs):

synapse/config/push.py

@@ -0,0 +1,45 @@
+# -*- coding: utf-8 -*-
+# Copyright 2015, 2016 OpenMarket Ltd
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+from ._base import Config
+
+
+class PushConfig(Config):
+    def read_config(self, config):
+        self.push_redact_content = False
+
+        push_config = config.get("email", {})
+        self.push_redact_content = push_config.get("redact_content", False)
+
+    def default_config(self, config_dir_path, server_name, **kwargs):
+        return """
+        # Control how push messages are sent to google/apple to notifications.
+        # Normally every message said in a room with one or more people using
+        # mobile devices will be posted to a push server hosted by matrix.org
+        # which is registered with google and apple in order to allow push
+        # notifications to be sent to these mobile devices.
+        #
+        # Setting redact_content to true will make the push messages contain no
+        # message content which will provide increased privacy. This is a
+        # temporary solution pending improvements to Android and iPhone apps
+        # to get content from the app rather than the notification.
+        #
+        # For modern android devices the notification content will still appear
+        # because it is loaded by the app. iPhone, however will send a
+        # notification saying only that a message arrived and who it came from.
+        #
+        #push:
+        #   redact_content: false
+        """

synapse/config/server.py

@@ -1,5 +1,6 @@
 # -*- coding: utf-8 -*-
 # Copyright 2014-2016 OpenMarket Ltd
+# Copyright 2017 New Vector Ltd
 #
 # Licensed under the Apache License, Version 2.0 (the "License");
 # you may not use this file except in compliance with the License.
@@ -29,14 +30,25 @@ class ServerConfig(Config):
         self.user_agent_suffix = config.get("user_agent_suffix")
         self.use_frozen_dicts = config.get("use_frozen_dicts", False)
         self.public_baseurl = config.get("public_baseurl")
+        self.cpu_affinity = config.get("cpu_affinity")
 
         # 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)
 
+        # Whether to update the user directory or not. This should be set to
+        # false only if we are updating the user directory in a worker
+        self.update_user_directory = config.get("update_user_directory", True)
+
         self.filter_timeline_limit = config.get("filter_timeline_limit", -1)
 
+        # Whether we should block invites sent to users on this server
+        # (other than those sent by local server admins)
+        self.block_non_admin_invites = config.get(
+            "block_non_admin_invites", False,
+        )
+
         if self.public_baseurl is not None:
             if self.public_baseurl[-1] != '/':
                 self.public_baseurl += '/'
@@ -143,6 +155,27 @@ class ServerConfig(Config):
         # When running as a daemon, the file to store the pid in
         pid_file: %(pid_file)s
 
+        # CPU affinity mask. Setting this restricts the CPUs on which the
+        # process will be scheduled. It is represented as a bitmask, with the
+        # lowest order bit corresponding to the first logical CPU and the
+        # highest order bit corresponding to the last logical CPU. Not all CPUs
+        # may exist on a given system but a mask may specify more CPUs than are
+        # present.
+        #
+        # For example:
+        #    0x00000001  is processor #0,
+        #    0x00000003  is processors #0 and #1,
+        #    0xFFFFFFFF  is all processors (#0 through #31).
+        #
+        # Pinning a Python process to a single CPU is desirable, because Python
+        # is inherently single-threaded due to the GIL, and can suffer a
+        # 30-40%% slowdown due to cache blow-out and thread context switching
+        # if the scheduler happens to schedule the underlying threads across
+        # different cores. See
+        # https://www.mirantis.com/blog/improve-performance-python-programs-restricting-single-cpu/.
+        #
+        # cpu_affinity: 0xFFFFFFFF
+
         # Whether to serve a web client from the HTTP/HTTPS root resource.
         web_client: True
 
@@ -167,6 +200,10 @@ class ServerConfig(Config):
         # and sync operations. The default value is -1, means no upper limit.
         # filter_timeline_limit: 5000
 
+        # Whether room invites to users on this server should be blocked
+        # (except those sent by local server admins). The default is False.
+        # block_non_admin_invites: True
+
         # List of ports that Synapse should listen on, their purpose and their
         # configuration.
         listeners:

synapse/config/spam_checker.py

@@ -0,0 +1,35 @@
+# -*- coding: utf-8 -*-
+# Copyright 2017 New Vector Ltd
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+from synapse.util.module_loader import load_module
+
+from ._base import Config
+
+
+class SpamCheckerConfig(Config):
+    def read_config(self, config):
+        self.spam_checker = None
+
+        provider = config.get("spam_checker", None)
+        if provider is not None:
+            self.spam_checker = load_module(provider)
+
+    def default_config(self, **kwargs):
+        return """\
+        # spam_checker:
+        #     module: "my_custom_project.SuperSpamChecker"
+        #     config:
+        #         example_option: 'things'
+        """

synapse/config/workers.py

@@ -32,6 +32,9 @@ class WorkerConfig(Config):
         self.worker_replication_port = config.get("worker_replication_port", None)
         self.worker_name = config.get("worker_name", self.worker_app)
 
+        self.worker_main_http_uri = config.get("worker_main_http_uri", None)
+        self.worker_cpu_affinity = config.get("worker_cpu_affinity")
+
         if self.worker_listeners:
             for listener in self.worker_listeners:
                 bind_address = listener.pop("bind_address", None)

synapse/crypto/keyclient.py

@@ -13,14 +13,11 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 
-
+from synapse.util import logcontext
 from twisted.web.http import HTTPClient
 from twisted.internet.protocol import Factory
 from twisted.internet import defer, reactor
 from synapse.http.endpoint import matrix_federation_endpoint
-from synapse.util.logcontext import (
-    preserve_context_over_fn, preserve_context_over_deferred
-)
 import simplejson as json
 import logging
 
@@ -43,14 +40,10 @@ def fetch_server_key(server_name, ssl_context_factory, path=KEY_API_V1):
 
     for i in range(5):
         try:
-            protocol = yield preserve_context_over_fn(
-                endpoint.connect, factory
-            )
-            server_response, server_certificate = yield preserve_context_over_deferred(
-                protocol.remote_key
-            )
-            defer.returnValue((server_response, server_certificate))
-            return
+            with logcontext.PreserveLoggingContext():
+                protocol = yield endpoint.connect(factory)
+                server_response, server_certificate = yield protocol.remote_key
+                defer.returnValue((server_response, server_certificate))
         except SynapseKeyClientError as e:
             logger.exception("Error getting key for %r" % (server_name,))
             if e.status.startswith("4"):

synapse/crypto/keyring.py

@@ -1,5 +1,6 @@
 # -*- coding: utf-8 -*-
 # Copyright 2014-2016 OpenMarket Ltd
+# Copyright 2017 New Vector Ltd.
 #
 # Licensed under the Apache License, Version 2.0 (the "License");
 # you may not use this file except in compliance with the License.
@@ -15,10 +16,9 @@
 
 from synapse.crypto.keyclient import fetch_server_key
 from synapse.api.errors import SynapseError, Codes
-from synapse.util import unwrapFirstError
-from synapse.util.async import ObservableDeferred
+from synapse.util import unwrapFirstError, logcontext
 from synapse.util.logcontext import (
-    preserve_context_over_deferred, preserve_context_over_fn, PreserveLoggingContext,
+    PreserveLoggingContext,
     preserve_fn
 )
 from synapse.util.metrics import Measure
@@ -57,7 +57,8 @@ Attributes:
     json_object(dict): The JSON object to verify.
     deferred(twisted.internet.defer.Deferred):
         A deferred (server_name, key_id, verify_key) tuple that resolves when
-        a verify key has been fetched
+        a verify key has been fetched. The deferreds' callbacks are run with no
+        logcontext.
 """
 
 
@@ -74,23 +75,32 @@ class Keyring(object):
         self.perspective_servers = self.config.perspectives
         self.hs = hs
 
+        # map from server name to Deferred. Has an entry for each server with
+        # an ongoing key download; the Deferred completes once the download
+        # completes.
+        #
+        # These are regular, logcontext-agnostic Deferreds.
         self.key_downloads = {}
 
     def verify_json_for_server(self, server_name, json_object):
-        return self.verify_json_objects_for_server(
-            [(server_name, json_object)]
-        )[0]
+        return logcontext.make_deferred_yieldable(
+            self.verify_json_objects_for_server(
+                [(server_name, json_object)]
+            )[0]
+        )
 
     def verify_json_objects_for_server(self, server_and_json):
-        """Bulk verfies signatures of json objects, bulk fetching keys as
+        """Bulk verifies signatures of json objects, bulk fetching keys as
         necessary.
 
         Args:
             server_and_json (list): List of pairs of (server_name, json_object)
 
         Returns:
-            list of deferreds indicating success or failure to verify each
-            json object's signature for the given server_name.
+            List<Deferred>: for each input pair, a deferred indicating success
+                or failure to verify each json object's signature for the given
+                server_name. The deferreds run their callbacks in the sentinel
+                logcontext.
         """
         verify_requests = []
 
@@ -117,94 +127,72 @@ class Keyring(object):
 
             verify_requests.append(verify_request)
 
-        @defer.inlineCallbacks
-        def handle_key_deferred(verify_request):
-            server_name = verify_request.server_name
-            try:
-                _, key_id, verify_key = yield verify_request.deferred
-            except IOError as e:
-                logger.warn(
-                    "Got IOError when downloading keys for %s: %s %s",
-                    server_name, type(e).__name__, str(e.message),
-                )
-                raise SynapseError(
-                    502,
-                    "Error downloading keys for %s" % (server_name,),
-                    Codes.UNAUTHORIZED,
-                )
-            except Exception as e:
-                logger.exception(
-                    "Got Exception when downloading keys for %s: %s %s",
-                    server_name, type(e).__name__, str(e.message),
-                )
-                raise SynapseError(
-                    401,
-                    "No key for %s with id %s" % (server_name, key_ids),
-                    Codes.UNAUTHORIZED,
-                )
-
-            json_object = verify_request.json_object
-
-            logger.debug("Got key %s %s:%s for server %s, verifying" % (
-                key_id, verify_key.alg, verify_key.version, server_name,
-            ))
-            try:
-                verify_signed_json(json_object, server_name, verify_key)
-            except:
-                raise SynapseError(
-                    401,
-                    "Invalid signature for server %s with key %s:%s" % (
-                        server_name, verify_key.alg, verify_key.version
-                    ),
-                    Codes.UNAUTHORIZED,
-                )
-
-        server_to_deferred = {
-            server_name: defer.Deferred()
-            for server_name, _ in server_and_json
-        }
-
-        with PreserveLoggingContext():
-
-            # We want to wait for any previous lookups to complete before
-            # proceeding.
-            wait_on_deferred = self.wait_for_previous_lookups(
-                [server_name for server_name, _ in server_and_json],
-                server_to_deferred,
-            )
-
-            # Actually start fetching keys.
-            wait_on_deferred.addBoth(
-                lambda _: self.get_server_verify_keys(verify_requests)
-            )
-
-            # When we've finished fetching all the keys for a given server_name,
-            # resolve the deferred passed to `wait_for_previous_lookups` so that
-            # any lookups waiting will proceed.
-            server_to_request_ids = {}
-
-            def remove_deferreds(res, server_name, verify_request):
-                request_id = id(verify_request)
-                server_to_request_ids[server_name].discard(request_id)
-                if not server_to_request_ids[server_name]:
-                    d = server_to_deferred.pop(server_name, None)
-                    if d:
-                        d.callback(None)
-                return res
-
-            for verify_request in verify_requests:
-                server_name = verify_request.server_name
-                request_id = id(verify_request)
-                server_to_request_ids.setdefault(server_name, set()).add(request_id)
-                deferred.addBoth(remove_deferreds, server_name, verify_request)
+        preserve_fn(self._start_key_lookups)(verify_requests)
 
         # Pass those keys to handle_key_deferred so that the json object
         # signatures can be verified
+        handle = preserve_fn(_handle_key_deferred)
         return [
-            preserve_context_over_fn(handle_key_deferred, verify_request)
-            for verify_request in verify_requests
+            handle(rq) for rq in verify_requests
         ]
 
+    @defer.inlineCallbacks
+    def _start_key_lookups(self, verify_requests):
+        """Sets off the key fetches for each verify request
+
+        Once each fetch completes, verify_request.deferred will be resolved.
+
+        Args:
+            verify_requests (List[VerifyKeyRequest]):
+        """
+
+        # create a deferred for each server we're going to look up the keys
+        # for; we'll resolve them once we have completed our lookups.
+        # These will be passed into wait_for_previous_lookups to block
+        # any other lookups until we have finished.
+        # The deferreds are called with no logcontext.
+        server_to_deferred = {
+            rq.server_name: defer.Deferred()
+            for rq in verify_requests
+        }
+
+        # We want to wait for any previous lookups to complete before
+        # proceeding.
+        yield self.wait_for_previous_lookups(
+            [rq.server_name for rq in verify_requests],
+            server_to_deferred,
+        )
+
+        # Actually start fetching keys.
+        self._get_server_verify_keys(verify_requests)
+
+        # When we've finished fetching all the keys for a given server_name,
+        # resolve the deferred passed to `wait_for_previous_lookups` so that
+        # any lookups waiting will proceed.
+        #
+        # map from server name to a set of request ids
+        server_to_request_ids = {}
+
+        for verify_request in verify_requests:
+            server_name = verify_request.server_name
+            request_id = id(verify_request)
+            server_to_request_ids.setdefault(server_name, set()).add(request_id)
+
+        def remove_deferreds(res, verify_request):
+            server_name = verify_request.server_name
+            request_id = id(verify_request)
+            server_to_request_ids[server_name].discard(request_id)
+            if not server_to_request_ids[server_name]:
+                d = server_to_deferred.pop(server_name, None)
+                if d:
+                    d.callback(None)
+            return res
+
+        for verify_request in verify_requests:
+            verify_request.deferred.addBoth(
+                remove_deferreds, verify_request,
+            )
+
     @defer.inlineCallbacks
     def wait_for_previous_lookups(self, server_names, server_to_deferred):
         """Waits for any previous key lookups for the given servers to finish.
@@ -212,7 +200,13 @@ class Keyring(object):
         Args:
             server_names (list): list of server_names we want to lookup
             server_to_deferred (dict): server_name to deferred which gets
-                resolved once we've finished looking up keys for that server
+                resolved once we've finished looking up keys for that server.
+                The Deferreds should be regular twisted ones which call their
+                callbacks with no logcontext.
+
+        Returns: a Deferred which resolves once all key lookups for the given
+            servers have completed. Follows the synapse rules of logcontext
+            preservation.
         """
         while True:
             wait_on = [
@@ -226,17 +220,15 @@ class Keyring(object):
             else:
                 break
 
+        def rm(r, server_name_):
+            self.key_downloads.pop(server_name_, None)
+            return r
+
         for server_name, deferred in server_to_deferred.items():
-            d = ObservableDeferred(preserve_context_over_deferred(deferred))
-            self.key_downloads[server_name] = d
+            self.key_downloads[server_name] = deferred
+            deferred.addBoth(rm, server_name)
 
-            def rm(r, server_name):
-                self.key_downloads.pop(server_name, None)
-                return r
-
-            d.addBoth(rm, server_name)
-
-    def get_server_verify_keys(self, verify_requests):
+    def _get_server_verify_keys(self, verify_requests):
         """Tries to find at least one key for each verify request
 
         For each verify_request, verify_request.deferred is called back with
@@ -305,21 +297,23 @@ class Keyring(object):
                     if not missing_keys:
                         break
 
-                for verify_request in requests_missing_keys.values():
-                    verify_request.deferred.errback(SynapseError(
-                        401,
-                        "No key for %s with id %s" % (
-                            verify_request.server_name, verify_request.key_ids,
-                        ),
-                        Codes.UNAUTHORIZED,
-                    ))
+                with PreserveLoggingContext():
+                    for verify_request in requests_missing_keys:
+                        verify_request.deferred.errback(SynapseError(
+                            401,
+                            "No key for %s with id %s" % (
+                                verify_request.server_name, verify_request.key_ids,
+                            ),
+                            Codes.UNAUTHORIZED,
+                        ))
 
         def on_err(err):
-            for verify_request in verify_requests:
-                if not verify_request.deferred.called:
-                    verify_request.deferred.errback(err)
+            with PreserveLoggingContext():
+                for verify_request in verify_requests:
+                    if not verify_request.deferred.called:
+                        verify_request.deferred.errback(err)
 
-        do_iterations().addErrback(on_err)
+        preserve_fn(do_iterations)().addErrback(on_err)
 
     @defer.inlineCallbacks
     def get_keys_from_store(self, server_name_and_key_ids):
@@ -333,7 +327,7 @@ class Keyring(object):
             Deferred: resolves to dict[str, dict[str, VerifyKey]]: map from
                 server_name -> key_id -> VerifyKey
         """
-        res = yield preserve_context_over_deferred(defer.gatherResults(
+        res = yield logcontext.make_deferred_yieldable(defer.gatherResults(
             [
                 preserve_fn(self.store.get_server_verify_keys)(
                     server_name, key_ids
@@ -341,7 +335,7 @@ class Keyring(object):
                 for server_name, key_ids in server_name_and_key_ids
             ],
             consumeErrors=True,
-        )).addErrback(unwrapFirstError)
+        ).addErrback(unwrapFirstError))
 
         defer.returnValue(dict(res))
 
@@ -362,13 +356,13 @@ class Keyring(object):
                 )
                 defer.returnValue({})
 
-        results = yield preserve_context_over_deferred(defer.gatherResults(
+        results = yield logcontext.make_deferred_yieldable(defer.gatherResults(
             [
                 preserve_fn(get_key)(p_name, p_keys)
                 for p_name, p_keys in self.perspective_servers.items()
             ],
             consumeErrors=True,
-        )).addErrback(unwrapFirstError)
+        ).addErrback(unwrapFirstError))
 
         union_of_keys = {}
         for result in results:
@@ -402,13 +396,13 @@ class Keyring(object):
 
             defer.returnValue(keys)
 
-        results = yield preserve_context_over_deferred(defer.gatherResults(
+        results = yield logcontext.make_deferred_yieldable(defer.gatherResults(
             [
                 preserve_fn(get_key)(server_name, key_ids)
                 for server_name, key_ids in server_name_and_key_ids
             ],
             consumeErrors=True,
-        )).addErrback(unwrapFirstError)
+        ).addErrback(unwrapFirstError))
 
         merged = {}
         for result in results:
@@ -485,7 +479,7 @@ class Keyring(object):
             for server_name, response_keys in processed_response.items():
                 keys.setdefault(server_name, {}).update(response_keys)
 
-        yield preserve_context_over_deferred(defer.gatherResults(
+        yield logcontext.make_deferred_yieldable(defer.gatherResults(
             [
                 preserve_fn(self.store_keys)(
                     server_name=server_name,
@@ -495,7 +489,7 @@ class Keyring(object):
                 for server_name, response_keys in keys.items()
             ],
             consumeErrors=True
-        )).addErrback(unwrapFirstError)
+        ).addErrback(unwrapFirstError))
 
         defer.returnValue(keys)
 
@@ -543,7 +537,7 @@ class Keyring(object):
 
             keys.update(response_keys)
 
-        yield preserve_context_over_deferred(defer.gatherResults(
+        yield logcontext.make_deferred_yieldable(defer.gatherResults(
             [
                 preserve_fn(self.store_keys)(
                     server_name=key_server_name,
@@ -553,7 +547,7 @@ class Keyring(object):
                 for key_server_name, verify_keys in keys.items()
             ],
             consumeErrors=True
-        )).addErrback(unwrapFirstError)
+        ).addErrback(unwrapFirstError))
 
         defer.returnValue(keys)
 
@@ -619,7 +613,7 @@ class Keyring(object):
         response_keys.update(verify_keys)
         response_keys.update(old_verify_keys)
 
-        yield preserve_context_over_deferred(defer.gatherResults(
+        yield logcontext.make_deferred_yieldable(defer.gatherResults(
             [
                 preserve_fn(self.store.store_server_keys_json)(
                     server_name=server_name,
@@ -632,7 +626,7 @@ class Keyring(object):
                 for key_id in updated_key_ids
             ],
             consumeErrors=True,
-        )).addErrback(unwrapFirstError)
+        ).addErrback(unwrapFirstError))
 
         results[server_name] = response_keys
 
@@ -710,7 +704,6 @@ class Keyring(object):
 
         defer.returnValue(verify_keys)
 
-    @defer.inlineCallbacks
     def store_keys(self, server_name, from_server, verify_keys):
         """Store a collection of verify keys for a given server
         Args:
@@ -721,7 +714,7 @@ class Keyring(object):
             A deferred that completes when the keys are stored.
         """
         # TODO(markjh): Store whether the keys have expired.
-        yield preserve_context_over_deferred(defer.gatherResults(
+        return logcontext.make_deferred_yieldable(defer.gatherResults(
             [
                 preserve_fn(self.store.store_server_verify_key)(
                     server_name, server_name, key.time_added, key
@@ -729,4 +722,48 @@ class Keyring(object):
                 for key_id, key in verify_keys.items()
             ],
             consumeErrors=True,
-        )).addErrback(unwrapFirstError)
+        ).addErrback(unwrapFirstError))
+
+
+@defer.inlineCallbacks
+def _handle_key_deferred(verify_request):
+    server_name = verify_request.server_name
+    try:
+        with PreserveLoggingContext():
+            _, key_id, verify_key = yield verify_request.deferred
+    except IOError as e:
+        logger.warn(
+            "Got IOError when downloading keys for %s: %s %s",
+            server_name, type(e).__name__, str(e.message),
+        )
+        raise SynapseError(
+            502,
+            "Error downloading keys for %s" % (server_name,),
+            Codes.UNAUTHORIZED,
+        )
+    except Exception as e:
+        logger.exception(
+            "Got Exception when downloading keys for %s: %s %s",
+            server_name, type(e).__name__, str(e.message),
+        )
+        raise SynapseError(
+            401,
+            "No key for %s with id %s" % (server_name, verify_request.key_ids),
+            Codes.UNAUTHORIZED,
+        )
+
+    json_object = verify_request.json_object
+
+    logger.debug("Got key %s %s:%s for server %s, verifying" % (
+        key_id, verify_key.alg, verify_key.version, server_name,
+    ))
+    try:
+        verify_signed_json(json_object, server_name, verify_key)
+    except:
+        raise SynapseError(
+            401,
+            "Invalid signature for server %s with key %s:%s" % (
+                server_name, verify_key.alg, verify_key.version
+            ),
+            Codes.UNAUTHORIZED,
+        )

synapse/events/spamcheck.py

@@ -0,0 +1,47 @@
+# -*- coding: utf-8 -*-
+# Copyright 2017 New Vector Ltd.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+
+class SpamChecker(object):
+    def __init__(self, hs):
+        self.spam_checker = None
+
+        module = None
+        config = None
+        try:
+            module, config = hs.config.spam_checker
+        except:
+            pass
+
+        if module is not None:
+            self.spam_checker = module(config=config)
+
+    def check_event_for_spam(self, event):
+        """Checks if a given event is considered "spammy" by this server.
+
+        If the server considers an event spammy, then it will be rejected if
+        sent by a local user. If it is sent by a user on another server, then
+        users receive a blank event.
+
+        Args:
+            event (synapse.events.EventBase): the event to be checked
+
+        Returns:
+            bool: True if the event is spammy.
+        """
+        if self.spam_checker is None:
+            return False
+
+        return self.spam_checker.check_event_for_spam(event)

synapse/federation/federation_base.py

@@ -12,28 +12,20 @@
 # 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.
-
-
-from twisted.internet import defer
-
-from synapse.events.utils import prune_event
-
-from synapse.crypto.event_signing import check_event_content_hash
-
-from synapse.api.errors import SynapseError
-
-from synapse.util import unwrapFirstError
-from synapse.util.logcontext import preserve_fn, preserve_context_over_deferred
-
 import logging
 
+from synapse.api.errors import SynapseError
+from synapse.crypto.event_signing import check_event_content_hash
+from synapse.events.utils import prune_event
+from synapse.util import unwrapFirstError, logcontext
+from twisted.internet import defer
 
 logger = logging.getLogger(__name__)
 
 
 class FederationBase(object):
     def __init__(self, hs):
-        pass
+        self.spam_checker = hs.get_spam_checker()
 
     @defer.inlineCallbacks
     def _check_sigs_and_hash_and_fetch(self, origin, pdus, outlier=False,
@@ -57,56 +49,52 @@ class FederationBase(object):
         """
         deferreds = self._check_sigs_and_hashes(pdus)
 
-        def callback(pdu):
-            return pdu
+        @defer.inlineCallbacks
+        def handle_check_result(pdu, deferred):
+            try:
+                res = yield logcontext.make_deferred_yieldable(deferred)
+            except SynapseError:
+                res = None
 
-        def errback(failure, pdu):
-            failure.trap(SynapseError)
-            return None
-
-        def try_local_db(res, pdu):
             if not res:
                 # Check local db.
-                return self.store.get_event(
+                res = yield self.store.get_event(
                     pdu.event_id,
                     allow_rejected=True,
                     allow_none=True,
                 )
-            return res
 
-        def try_remote(res, pdu):
             if not res and pdu.origin != origin:
-                return self.get_pdu(
-                    destinations=[pdu.origin],
-                    event_id=pdu.event_id,
-                    outlier=outlier,
-                    timeout=10000,
-                ).addErrback(lambda e: None)
-            return res
+                try:
+                    res = yield self.get_pdu(
+                        destinations=[pdu.origin],
+                        event_id=pdu.event_id,
+                        outlier=outlier,
+                        timeout=10000,
+                    )
+                except SynapseError:
+                    pass
 
-        def warn(res, pdu):
             if not res:
                 logger.warn(
                     "Failed to find copy of %s with valid signature",
                     pdu.event_id,
                 )
-            return res
 
-        for pdu, deferred in zip(pdus, deferreds):
-            deferred.addCallbacks(
-                callback, errback, errbackArgs=[pdu]
-            ).addCallback(
-                try_local_db, pdu
-            ).addCallback(
-                try_remote, pdu
-            ).addCallback(
-                warn, pdu
+            defer.returnValue(res)
+
+        handle = logcontext.preserve_fn(handle_check_result)
+        deferreds2 = [
+            handle(pdu, deferred)
+            for pdu, deferred in zip(pdus, deferreds)
+        ]
+
+        valid_pdus = yield logcontext.make_deferred_yieldable(
+            defer.gatherResults(
+                deferreds2,
+                consumeErrors=True,
             )
-
-        valid_pdus = yield preserve_context_over_deferred(defer.gatherResults(
-            deferreds,
-            consumeErrors=True
-        )).addErrback(unwrapFirstError)
+        ).addErrback(unwrapFirstError)
 
         if include_none:
             defer.returnValue(valid_pdus)
@@ -114,15 +102,24 @@ class FederationBase(object):
             defer.returnValue([p for p in valid_pdus if p])
 
     def _check_sigs_and_hash(self, pdu):
-        return self._check_sigs_and_hashes([pdu])[0]
+        return logcontext.make_deferred_yieldable(
+            self._check_sigs_and_hashes([pdu])[0],
+        )
 
     def _check_sigs_and_hashes(self, pdus):
-        """Throws a SynapseError if a PDU does not have the correct
-        signatures.
+        """Checks that each of the received events is correctly signed by the
+        sending server.
+
+        Args:
+            pdus (list[FrozenEvent]): the events to be checked
 
         Returns:
-            FrozenEvent: Either the given event or it redacted if it failed the
-            content hash check.
+            list[Deferred]: for each input event, a deferred which:
+              * returns the original event if the checks pass
+              * returns a redacted version of the event (if the signature
+                matched but the hash did not)
+              * throws a SynapseError if the signature check failed.
+            The deferreds run their callbacks in the sentinel logcontext.
         """
 
         redacted_pdus = [
@@ -130,26 +127,38 @@ class FederationBase(object):
             for pdu in pdus
         ]
 
-        deferreds = preserve_fn(self.keyring.verify_json_objects_for_server)([
+        deferreds = self.keyring.verify_json_objects_for_server([
             (p.origin, p.get_pdu_json())
             for p in redacted_pdus
         ])
 
+        ctx = logcontext.LoggingContext.current_context()
+
         def callback(_, pdu, redacted):
-            if not check_event_content_hash(pdu):
-                logger.warn(
-                    "Event content has been tampered, redacting %s: %s",
-                    pdu.event_id, pdu.get_pdu_json()
-                )
-                return redacted
-            return pdu
+            with logcontext.PreserveLoggingContext(ctx):
+                if not check_event_content_hash(pdu):
+                    logger.warn(
+                        "Event content has been tampered, redacting %s: %s",
+                        pdu.event_id, pdu.get_pdu_json()
+                    )
+                    return redacted
+
+                if self.spam_checker.check_event_for_spam(pdu):
+                    logger.warn(
+                        "Event contains spam, redacting %s: %s",
+                        pdu.event_id, pdu.get_pdu_json()
+                    )
+                    return redacted
+
+                return pdu
 
         def errback(failure, pdu):
             failure.trap(SynapseError)
-            logger.warn(
-                "Signature check failed for %s",
-                pdu.event_id,
-            )
+            with logcontext.PreserveLoggingContext(ctx):
+                logger.warn(
+                    "Signature check failed for %s",
+                    pdu.event_id,
+                )
             return failure
 
         for deferred, pdu, redacted in zip(deferreds, pdus, redacted_pdus):

synapse/federation/federation_client.py

@@ -22,7 +22,7 @@ from synapse.api.constants import Membership
 from synapse.api.errors import (
     CodeMessageException, HttpResponseException, SynapseError,
 )
-from synapse.util import unwrapFirstError
+from synapse.util import unwrapFirstError, logcontext
 from synapse.util.caches.expiringcache import ExpiringCache
 from synapse.util.logutils import log_function
 from synapse.util.logcontext import preserve_fn, preserve_context_over_deferred
@@ -189,10 +189,10 @@ class FederationClient(FederationBase):
         ]
 
         # FIXME: We should handle signature failures more gracefully.
-        pdus[:] = yield preserve_context_over_deferred(defer.gatherResults(
+        pdus[:] = yield logcontext.make_deferred_yieldable(defer.gatherResults(
             self._check_sigs_and_hashes(pdus),
             consumeErrors=True,
-        )).addErrback(unwrapFirstError)
+        ).addErrback(unwrapFirstError))
 
         defer.returnValue(pdus)
 
@@ -252,7 +252,7 @@ class FederationClient(FederationBase):
                     pdu = pdu_list[0]
 
                     # Check signatures are correct.
-                    signed_pdu = yield self._check_sigs_and_hashes([pdu])[0]
+                    signed_pdu = yield self._check_sigs_and_hash(pdu)
 
                     break
 

synapse/federation/transaction_queue.py

@@ -187,6 +187,7 @@ class TransactionQueue(object):
                             prev_id for prev_id, _ in event.prev_events
                         ],
                     )
+                    destinations = set(destinations)
 
                     if send_on_behalf_of is not None:
                         # If we are sending the event on behalf of another server

synapse/federation/transport/server.py

@@ -153,12 +153,10 @@ class Authenticator(object):
 class BaseFederationServlet(object):
     REQUIRE_AUTH = True
 
-    def __init__(self, handler, authenticator, ratelimiter, server_name,
-                 room_list_handler):
+    def __init__(self, handler, authenticator, ratelimiter, server_name):
         self.handler = handler
         self.authenticator = authenticator
         self.ratelimiter = ratelimiter
-        self.room_list_handler = room_list_handler
 
     def _wrap(self, func):
         authenticator = self.authenticator
@@ -590,7 +588,7 @@ class PublicRoomList(BaseFederationServlet):
         else:
             network_tuple = ThirdPartyInstanceID(None, None)
 
-        data = yield self.room_list_handler.get_local_public_room_list(
+        data = yield self.handler.get_local_public_room_list(
             limit, since_token,
             network_tuple=network_tuple
         )
@@ -611,7 +609,7 @@ class FederationVersionServlet(BaseFederationServlet):
         }))
 
 
-SERVLET_CLASSES = (
+FEDERATION_SERVLET_CLASSES = (
     FederationSendServlet,
     FederationPullServlet,
     FederationEventServlet,
@@ -634,17 +632,27 @@ SERVLET_CLASSES = (
     FederationThirdPartyInviteExchangeServlet,
     On3pidBindServlet,
     OpenIdUserInfo,
-    PublicRoomList,
     FederationVersionServlet,
 )
 
+ROOM_LIST_CLASSES = (
+    PublicRoomList,
+)
+
 
 def register_servlets(hs, resource, authenticator, ratelimiter):
-    for servletclass in SERVLET_CLASSES:
+    for servletclass in FEDERATION_SERVLET_CLASSES:
         servletclass(
             handler=hs.get_replication_layer(),
             authenticator=authenticator,
             ratelimiter=ratelimiter,
             server_name=hs.hostname,
-            room_list_handler=hs.get_room_list_handler(),
+        ).register(resource)
+
+    for servletclass in ROOM_LIST_CLASSES:
+        servletclass(
+            handler=hs.get_room_list_handler(),
+            authenticator=authenticator,
+            ratelimiter=ratelimiter,
+            server_name=hs.hostname,
         ).register(resource)

synapse/handlers/auth.py

@@ -21,6 +21,7 @@ from synapse.api.constants import LoginType
 from synapse.types import UserID
 from synapse.api.errors import AuthError, LoginError, Codes, StoreError, SynapseError
 from synapse.util.async import run_on_reactor
+from synapse.util.caches.expiringcache import ExpiringCache
 
 from twisted.web.client import PartialDownloadError
 
@@ -52,7 +53,15 @@ class AuthHandler(BaseHandler):
             LoginType.DUMMY: self._check_dummy_auth,
         }
         self.bcrypt_rounds = hs.config.bcrypt_rounds
-        self.sessions = {}
+
+        # This is not a cache per se, but a store of all current sessions that
+        # expire after N hours
+        self.sessions = ExpiringCache(
+            cache_name="register_sessions",
+            clock=hs.get_clock(),
+            expiry_ms=self.SESSION_EXPIRE_MS,
+            reset_expiry_on_get=True,
+        )
 
         account_handler = _AccountHandler(
             hs, check_user_exists=self.check_user_exists
@@ -617,16 +626,6 @@ class AuthHandler(BaseHandler):
         logger.debug("Saving session %s", session)
         session["last_used"] = self.hs.get_clock().time_msec()
         self.sessions[session["id"]] = session
-        self._prune_sessions()
-
-    def _prune_sessions(self):
-        for sid, sess in self.sessions.items():
-            last_used = 0
-            if 'last_used' in sess:
-                last_used = sess['last_used']
-            now = self.hs.get_clock().time_msec()
-            if last_used < now - AuthHandler.SESSION_EXPIRE_MS:
-                del self.sessions[sid]
 
     def hash(self, password):
         """Computes a secure hash of password.

synapse/handlers/device.py

@@ -106,7 +106,7 @@ class DeviceHandler(BaseHandler):
         device_map = yield self.store.get_devices_by_user(user_id)
 
         ips = yield self.store.get_last_client_ip_by_device(
-            devices=((user_id, device_id) for device_id in device_map.keys())
+            user_id, device_id=None
         )
 
         devices = device_map.values()
@@ -133,7 +133,7 @@ class DeviceHandler(BaseHandler):
         except errors.StoreError:
             raise errors.NotFoundError
         ips = yield self.store.get_last_client_ip_by_device(
-            devices=((user_id, device_id),)
+            user_id, device_id,
         )
         _update_device_from_client_ips(device, ips)
         defer.returnValue(device)
@@ -270,6 +270,8 @@ class DeviceHandler(BaseHandler):
             user_id (str)
             from_token (StreamToken)
         """
+        now_token = yield self.hs.get_event_sources().get_current_token()
+
         room_ids = yield self.store.get_rooms_for_user(user_id)
 
         # First we check if any devices have changed
@@ -280,11 +282,30 @@ class DeviceHandler(BaseHandler):
         # Then work out if any users have since joined
         rooms_changed = self.store.get_rooms_that_changed(room_ids, from_token.room_key)
 
+        member_events = yield self.store.get_membership_changes_for_user(
+            user_id, from_token.room_key, now_token.room_key
+        )
+        rooms_changed.update(event.room_id for event in member_events)
+
         stream_ordering = RoomStreamToken.parse_stream_token(
-            from_token.room_key).stream
+            from_token.room_key
+        ).stream
 
         possibly_changed = set(changed)
+        possibly_left = set()
         for room_id in rooms_changed:
+            current_state_ids = yield self.store.get_current_state_ids(room_id)
+
+            # The user may have left the room
+            # TODO: Check if they actually did or if we were just invited.
+            if room_id not in room_ids:
+                for key, event_id in current_state_ids.iteritems():
+                    etype, state_key = key
+                    if etype != EventTypes.Member:
+                        continue
+                    possibly_left.add(state_key)
+                continue
+
             # Fetch the current state at the time.
             try:
                 event_ids = yield self.store.get_forward_extremeties_for_room(
@@ -295,8 +316,6 @@ class DeviceHandler(BaseHandler):
                 # ordering: treat it the same as a new room
                 event_ids = []
 
-            current_state_ids = yield self.store.get_current_state_ids(room_id)
-
             # special-case for an empty prev state: include all members
             # in the changed list
             if not event_ids:
@@ -307,9 +326,25 @@ class DeviceHandler(BaseHandler):
                     possibly_changed.add(state_key)
                 continue
 
+            current_member_id = current_state_ids.get((EventTypes.Member, user_id))
+            if not current_member_id:
+                continue
+
             # mapping from event_id -> state_dict
             prev_state_ids = yield self.store.get_state_ids_for_events(event_ids)
 
+            # Check if we've joined the room? If so we just blindly add all the users to
+            # the "possibly changed" users.
+            for state_dict in prev_state_ids.itervalues():
+                member_event = state_dict.get((EventTypes.Member, user_id), None)
+                if not member_event or member_event != current_member_id:
+                    for key, event_id in current_state_ids.iteritems():
+                        etype, state_key = key
+                        if etype != EventTypes.Member:
+                            continue
+                        possibly_changed.add(state_key)
+                    break
+
             # If there has been any change in membership, include them in the
             # possibly changed list. We'll check if they are joined below,
             # and we're not toooo worried about spuriously adding users.
@@ -320,19 +355,30 @@ class DeviceHandler(BaseHandler):
 
                 # check if this member has changed since any of the extremities
                 # at the stream_ordering, and add them to the list if so.
-                for state_dict in prev_state_ids.values():
+                for state_dict in prev_state_ids.itervalues():
                     prev_event_id = state_dict.get(key, None)
                     if not prev_event_id or prev_event_id != event_id:
-                        possibly_changed.add(state_key)
+                        if state_key != user_id:
+                            possibly_changed.add(state_key)
                         break
 
-        users_who_share_room = yield self.store.get_users_who_share_room_with_user(
-            user_id
-        )
+        if possibly_changed or possibly_left:
+            users_who_share_room = yield self.store.get_users_who_share_room_with_user(
+                user_id
+            )
 
-        # Take the intersection of the users whose devices may have changed
-        # and those that actually still share a room with the user
-        defer.returnValue(users_who_share_room & possibly_changed)
+            # Take the intersection of the users whose devices may have changed
+            # and those that actually still share a room with the user
+            possibly_joined = possibly_changed & users_who_share_room
+            possibly_left = (possibly_changed | possibly_left) - users_who_share_room
+        else:
+            possibly_joined = []
+            possibly_left = []
+
+        defer.returnValue({
+            "changed": list(possibly_joined),
+            "left": list(possibly_left),
+        })
 
     @defer.inlineCallbacks
     def on_federation_query_user_devices(self, user_id):

synapse/handlers/federation.py

@@ -43,7 +43,6 @@ from synapse.events.utils import prune_event
 
 from synapse.util.retryutils import NotRetryingDestination
 
-from synapse.push.action_generator import ActionGenerator
 from synapse.util.distributor import user_joined_room
 
 from twisted.internet import defer
@@ -75,6 +74,9 @@ class FederationHandler(BaseHandler):
         self.state_handler = hs.get_state_handler()
         self.server_name = hs.hostname
         self.keyring = hs.get_keyring()
+        self.action_generator = hs.get_action_generator()
+        self.is_mine_id = hs.is_mine_id
+        self.pusher_pool = hs.get_pusherpool()
 
         self.replication_layer.set_handler(self)
 
@@ -832,7 +834,11 @@ class FederationHandler(BaseHandler):
 
     @defer.inlineCallbacks
     def on_event_auth(self, event_id):
-        auth = yield self.store.get_auth_chain([event_id])
+        event = yield self.store.get_event(event_id)
+        auth = yield self.store.get_auth_chain(
+            [auth_id for auth_id, _ in event.auth_events],
+            include_given=True
+        )
 
         for event in auth:
             event.signatures.update(
@@ -1047,9 +1053,7 @@ class FederationHandler(BaseHandler):
                 yield user_joined_room(self.distributor, user, event.room_id)
 
         state_ids = context.prev_state_ids.values()
-        auth_chain = yield self.store.get_auth_chain(set(
-            [event.event_id] + state_ids
-        ))
+        auth_chain = yield self.store.get_auth_chain(state_ids)
 
         state = yield self.store.get_events(context.prev_state_ids.values())
 
@@ -1066,6 +1070,27 @@ class FederationHandler(BaseHandler):
         """
         event = pdu
 
+        is_blocked = yield self.store.is_room_blocked(event.room_id)
+        if is_blocked:
+            raise SynapseError(403, "This room has been blocked on this server")
+
+        if self.hs.config.block_non_admin_invites:
+            raise SynapseError(403, "This server does not accept room invites")
+
+        membership = event.content.get("membership")
+        if event.type != EventTypes.Member or membership != Membership.INVITE:
+            raise SynapseError(400, "The event was not an m.room.member invite event")
+
+        sender_domain = get_domain_from_id(event.sender)
+        if sender_domain != origin:
+            raise SynapseError(400, "The invite event was not from the server sending it")
+
+        if event.state_key is None:
+            raise SynapseError(400, "The invite event did not have a state key")
+
+        if not self.is_mine_id(event.state_key):
+            raise SynapseError(400, "The invite event must be for this server")
+
         event.internal_metadata.outlier = True
         event.internal_metadata.invite_from_remote = True
 
@@ -1100,6 +1125,9 @@ class FederationHandler(BaseHandler):
             user_id,
             "leave"
         )
+        # Mark as outlier as we don't have any state for this event; we're not
+        # even in the room.
+        event.internal_metadata.outlier = True
         event = self._sign_event(event)
 
         # Try the host that we succesfully called /make_leave/ on first for
@@ -1271,7 +1299,7 @@ class FederationHandler(BaseHandler):
             for event in res:
                 # We sign these again because there was a bug where we
                 # incorrectly signed things the first time round
-                if self.hs.is_mine_id(event.event_id):
+                if self.is_mine_id(event.event_id):
                     event.signatures.update(
                         compute_event_signature(
                             event,
@@ -1344,7 +1372,7 @@ class FederationHandler(BaseHandler):
         )
 
         if event:
-            if self.hs.is_mine_id(event.event_id):
+            if self.is_mine_id(event.event_id):
                 # FIXME: This is a temporary work around where we occasionally
                 # return events slightly differently than when they were
                 # originally signed
@@ -1388,9 +1416,8 @@ class FederationHandler(BaseHandler):
             auth_events=auth_events,
         )
 
-        if not event.internal_metadata.is_outlier():
-            action_generator = ActionGenerator(self.hs)
-            yield action_generator.handle_push_actions_for_event(
+        if not event.internal_metadata.is_outlier() and not backfilled:
+            yield self.action_generator.handle_push_actions_for_event(
                 event, context
             )
 
@@ -1403,7 +1430,7 @@ class FederationHandler(BaseHandler):
         if not backfilled:
             # this intentionally does not yield: we don't care about the result
             # and don't need to wait for it.
-            preserve_fn(self.hs.get_pusherpool().on_new_notifications)(
+            preserve_fn(self.pusher_pool.on_new_notifications)(
                 event_stream_id, max_stream_id
             )
 
@@ -1582,7 +1609,7 @@ class FederationHandler(BaseHandler):
 
             context.rejected = RejectedReason.AUTH_ERROR
 
-        if event.type == EventTypes.GuestAccess:
+        if event.type == EventTypes.GuestAccess and not context.rejected:
             yield self.maybe_kick_guest_users(event)
 
         defer.returnValue(context)
@@ -1599,7 +1626,11 @@ class FederationHandler(BaseHandler):
                 pass
 
         # Now get the current auth_chain for the event.
-        local_auth_chain = yield self.store.get_auth_chain([event_id])
+        event = yield self.store.get_event(event_id)
+        local_auth_chain = yield self.store.get_auth_chain(
+            [auth_id for auth_id, _ in event.auth_events],
+            include_given=True
+        )
 
         # TODO: Check if we would now reject event_id. If so we need to tell
         # everyone.
@@ -1792,7 +1823,9 @@ class FederationHandler(BaseHandler):
                 auth_ids = yield self.auth.compute_auth_events(
                     event, context.prev_state_ids
                 )
-                local_auth_chain = yield self.store.get_auth_chain(auth_ids)
+                local_auth_chain = yield self.store.get_auth_chain(
+                    auth_ids, include_given=True
+                )
 
                 try:
                     # 2. Get remote difference.
@@ -2060,6 +2093,14 @@ class FederationHandler(BaseHandler):
     @defer.inlineCallbacks
     @log_function
     def on_exchange_third_party_invite_request(self, origin, room_id, event_dict):
+        """Handle an exchange_third_party_invite request from a remote server
+
+        The remote server will call this when it wants to turn a 3pid invite
+        into a normal m.room.member invite.
+
+        Returns:
+            Deferred: resolves (to None)
+        """
         builder = self.event_builder_factory.new(event_dict)
 
         message_handler = self.hs.get_handlers().message_handler
@@ -2078,9 +2119,12 @@ class FederationHandler(BaseHandler):
             raise e
         yield self._check_signature(event, context)
 
+        # XXX we send the invite here, but send_membership_event also sends it,
+        # so we end up making two requests. I think this is redundant.
         returned_invite = yield self.send_invite(origin, event)
         # TODO: Make sure the signatures actually are correct.
         event.signatures.update(returned_invite.signatures)
+
         member_handler = self.hs.get_handlers().room_member_handler
         yield member_handler.send_membership_event(None, event, context)
 

synapse/handlers/message.py

@@ -12,7 +12,6 @@
 # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 # See the License for the specific language governing permissions and
 # limitations under the License.
-
 from twisted.internet import defer
 
 from synapse.api.constants import EventTypes, Membership
@@ -20,7 +19,6 @@ from synapse.api.errors import AuthError, Codes, SynapseError
 from synapse.crypto.event_signing import add_hashes_and_signatures
 from synapse.events.utils import serialize_event
 from synapse.events.validator import EventValidator
-from synapse.push.action_generator import ActionGenerator
 from synapse.types import (
     UserID, RoomAlias, RoomStreamToken,
 )
@@ -35,6 +33,7 @@ from canonicaljson import encode_canonical_json
 
 import logging
 import random
+import ujson
 
 logger = logging.getLogger(__name__)
 
@@ -50,10 +49,16 @@ class MessageHandler(BaseHandler):
 
         self.pagination_lock = ReadWriteLock()
 
+        self.pusher_pool = hs.get_pusherpool()
+
         # We arbitrarily limit concurrent event creation for a room to 5.
         # This is to stop us from diverging history *too* much.
         self.limiter = Limiter(max_count=5)
 
+        self.action_generator = hs.get_action_generator()
+
+        self.spam_checker = hs.get_spam_checker()
+
     @defer.inlineCallbacks
     def purge_history(self, room_id, event_id):
         event = yield self.store.get_event(event_id)
@@ -317,6 +322,12 @@ class MessageHandler(BaseHandler):
             token_id=requester.access_token_id,
             txn_id=txn_id
         )
+
+        if self.spam_checker.check_event_for_spam(event):
+            raise SynapseError(
+                403, "Spam is not permitted here", Codes.FORBIDDEN
+            )
+
         yield self.send_nonmember_event(
             requester,
             event,
@@ -408,6 +419,51 @@ class MessageHandler(BaseHandler):
             [serialize_event(c, now) for c in room_state.values()]
         )
 
+    @defer.inlineCallbacks
+    def get_joined_members(self, requester, room_id):
+        """Get all the joined members in the room and their profile information.
+
+        If the user has left the room return the state events from when they left.
+
+        Args:
+            requester(Requester): The user requesting state events.
+            room_id(str): The room ID to get all state events from.
+        Returns:
+            A dict of user_id to profile info
+        """
+        user_id = requester.user.to_string()
+        if not requester.app_service:
+            # We check AS auth after fetching the room membership, as it
+            # requires us to pull out all joined members anyway.
+            membership, _ = yield self._check_in_room_or_world_readable(
+                room_id, user_id
+            )
+            if membership != Membership.JOIN:
+                raise NotImplementedError(
+                    "Getting joined members after leaving is not implemented"
+                )
+
+        users_with_profile = yield self.state.get_current_user_in_room(room_id)
+
+        # If this is an AS, double check that they are allowed to see the members.
+        # This can either be because the AS user is in the room or becuase there
+        # is a user in the room that the AS is "interested in"
+        if requester.app_service and user_id not in users_with_profile:
+            for uid in users_with_profile:
+                if requester.app_service.is_interested_in_user(uid):
+                    break
+            else:
+                # Loop fell through, AS has no interested users in room
+                raise AuthError(403, "Appservice not in room")
+
+        defer.returnValue({
+            user_id: {
+                "avatar_url": profile.avatar_url,
+                "display_name": profile.display_name,
+            }
+            for user_id, profile in users_with_profile.iteritems()
+        })
+
     @measure_func("_create_new_client_event")
     @defer.inlineCallbacks
     def _create_new_client_event(self, builder, requester=None, prev_event_ids=None):
@@ -497,6 +553,14 @@ class MessageHandler(BaseHandler):
             logger.warn("Denying new event %r because %s", event, err)
             raise err
 
+        # Ensure that we can round trip before trying to persist in db
+        try:
+            dump = ujson.dumps(event.content)
+            ujson.loads(dump)
+        except:
+            logger.exception("Failed to encode content: %r", event.content)
+            raise
+
         yield self.maybe_kick_guest_users(event, context)
 
         if event.type == EventTypes.CanonicalAlias:
@@ -590,8 +654,7 @@ class MessageHandler(BaseHandler):
                 "Changing the room create event is forbidden",
             )
 
-        action_generator = ActionGenerator(self.hs)
-        yield action_generator.handle_push_actions_for_event(
+        yield self.action_generator.handle_push_actions_for_event(
             event, context
         )
 
@@ -601,7 +664,7 @@ class MessageHandler(BaseHandler):
 
         # this intentionally does not yield: we don't care about the result
         # and don't need to wait for it.
-        preserve_fn(self.hs.get_pusherpool().on_new_notifications)(
+        preserve_fn(self.pusher_pool.on_new_notifications)(
             event_stream_id, max_stream_id
         )
 

synapse/handlers/room.py

@@ -61,7 +61,7 @@ class RoomCreationHandler(BaseHandler):
     }
 
     @defer.inlineCallbacks
-    def create_room(self, requester, config):
+    def create_room(self, requester, config, ratelimit=True):
         """ Creates a new room.
 
         Args:
@@ -75,7 +75,8 @@ class RoomCreationHandler(BaseHandler):
         """
         user_id = requester.user.to_string()
 
-        yield self.ratelimit(requester)
+        if ratelimit:
+            yield self.ratelimit(requester)
 
         if "room_alias_name" in config:
             for wchar in string.whitespace:
@@ -167,6 +168,7 @@ class RoomCreationHandler(BaseHandler):
             initial_state=initial_state,
             creation_content=creation_content,
             room_alias=room_alias,
+            power_level_content_override=config.get("power_level_content_override", {})
         )
 
         if "name" in config:
@@ -245,7 +247,8 @@ class RoomCreationHandler(BaseHandler):
             invite_list,
             initial_state,
             creation_content,
-            room_alias
+            room_alias,
+            power_level_content_override,
     ):
         def create(etype, content, **kwargs):
             e = {
@@ -291,7 +294,15 @@ class RoomCreationHandler(BaseHandler):
             ratelimit=False,
         )
 
-        if (EventTypes.PowerLevels, '') not in initial_state:
+        # We treat the power levels override specially as this needs to be one
+        # of the first events that get sent into a room.
+        pl_content = initial_state.pop((EventTypes.PowerLevels, ''), None)
+        if pl_content is not None:
+            yield send(
+                etype=EventTypes.PowerLevels,
+                content=pl_content,
+            )
+        else:
             power_level_content = {
                 "users": {
                     creator_id: 100,
@@ -316,6 +327,8 @@ class RoomCreationHandler(BaseHandler):
                 for invitee in invite_list:
                     power_level_content["users"][invitee] = 100
 
+            power_level_content.update(power_level_content_override)
+
             yield send(
                 etype=EventTypes.PowerLevels,
                 content=power_level_content,

synapse/handlers/room_member.py

@@ -191,6 +191,8 @@ class RoomMemberHandler(BaseHandler):
         if action in ["kick", "unban"]:
             effective_membership_state = "leave"
 
+        # if this is a join with a 3pid signature, we may need to turn a 3pid
+        # invite into a normal invite before we can handle the join.
         if third_party_signed is not None:
             replication = self.hs.get_replication_layer()
             yield replication.exchange_third_party_invite(
@@ -203,6 +205,21 @@ class RoomMemberHandler(BaseHandler):
         if not remote_room_hosts:
             remote_room_hosts = []
 
+        if effective_membership_state not in ("leave", "ban",):
+            is_blocked = yield self.store.is_room_blocked(room_id)
+            if is_blocked:
+                raise SynapseError(403, "This room has been blocked on this server")
+
+        if (effective_membership_state == "invite" and
+                self.hs.config.block_non_admin_invites):
+            is_requester_admin = yield self.auth.is_server_admin(
+                requester.user,
+            )
+            if not is_requester_admin:
+                raise SynapseError(
+                    403, "Invites have been disabled on this server",
+                )
+
         latest_event_ids = yield self.store.get_latest_event_ids_in_room(room_id)
         current_state_ids = yield self.state_handler.get_current_state_ids(
             room_id, latest_event_ids=latest_event_ids,
@@ -369,6 +386,11 @@ class RoomMemberHandler(BaseHandler):
                     # so don't really fit into the general auth process.
                     raise AuthError(403, "Guest access not allowed")
 
+        if event.membership not in (Membership.LEAVE, Membership.BAN):
+            is_blocked = yield self.store.is_room_blocked(room_id)
+            if is_blocked:
+                raise SynapseError(403, "This room has been blocked on this server")
+
         yield message_handler.handle_new_client_event(
             requester,
             event,
@@ -461,6 +483,16 @@ class RoomMemberHandler(BaseHandler):
             requester,
             txn_id
     ):
+        if self.hs.config.block_non_admin_invites:
+            is_requester_admin = yield self.auth.is_server_admin(
+                requester.user,
+            )
+            if not is_requester_admin:
+                raise SynapseError(
+                    403, "Invites have been disabled on this server",
+                    Codes.FORBIDDEN,
+                )
+
         invitee = yield self._lookup_3pid(
             id_server, medium, address
         )

synapse/handlers/sync.py

@@ -108,6 +108,16 @@ class InvitedSyncResult(collections.namedtuple("InvitedSyncResult", [
         return True
 
 
+class DeviceLists(collections.namedtuple("DeviceLists", [
+    "changed",   # list of user_ids whose devices may have changed
+    "left",      # list of user_ids whose devices we no longer track
+])):
+    __slots__ = []
+
+    def __nonzero__(self):
+        return bool(self.changed or self.left)
+
+
 class SyncResult(collections.namedtuple("SyncResult", [
     "next_batch",  # Token for the next sync
     "presence",  # List of presence events for the user.
@@ -117,6 +127,8 @@ class SyncResult(collections.namedtuple("SyncResult", [
     "archived",  # ArchivedSyncResult for each archived room.
     "to_device",  # List of direct messages for the device.
     "device_lists",  # List of user_ids whose devices have chanegd
+    "device_one_time_keys_count",  # Dict of algorithm to count for one time keys
+                                   # for this device
 ])):
     __slots__ = []
 
@@ -288,10 +300,20 @@ class SyncHandler(object):
 
             if recents:
                 recents = sync_config.filter_collection.filter_room_timeline(recents)
+
+                # We check if there are any state events, if there are then we pass
+                # all current state events to the filter_events function. This is to
+                # ensure that we always include current state in the timeline
+                current_state_ids = frozenset()
+                if any(e.is_state() for e in recents):
+                    current_state_ids = yield self.state.get_current_state_ids(room_id)
+                    current_state_ids = frozenset(current_state_ids.itervalues())
+
                 recents = yield filter_events_for_client(
                     self.store,
                     sync_config.user.to_string(),
                     recents,
+                    always_include_ids=current_state_ids,
                 )
             else:
                 recents = []
@@ -323,10 +345,20 @@ class SyncHandler(object):
                 loaded_recents = sync_config.filter_collection.filter_room_timeline(
                     events
                 )
+
+                # We check if there are any state events, if there are then we pass
+                # all current state events to the filter_events function. This is to
+                # ensure that we always include current state in the timeline
+                current_state_ids = frozenset()
+                if any(e.is_state() for e in loaded_recents):
+                    current_state_ids = yield self.state.get_current_state_ids(room_id)
+                    current_state_ids = frozenset(current_state_ids.itervalues())
+
                 loaded_recents = yield filter_events_for_client(
                     self.store,
                     sync_config.user.to_string(),
                     loaded_recents,
+                    always_include_ids=current_state_ids,
                 )
                 loaded_recents.extend(recents)
                 recents = loaded_recents
@@ -533,7 +565,8 @@ class SyncHandler(object):
         res = yield self._generate_sync_entry_for_rooms(
             sync_result_builder, account_data_by_room
         )
-        newly_joined_rooms, newly_joined_users = res
+        newly_joined_rooms, newly_joined_users, _, _ = res
+        _, _, newly_left_rooms, newly_left_users = res
 
         block_all_presence_data = (
             since_token is None and
@@ -547,9 +580,21 @@ class SyncHandler(object):
         yield self._generate_sync_entry_for_to_device(sync_result_builder)
 
         device_lists = yield self._generate_sync_entry_for_device_list(
-            sync_result_builder
+            sync_result_builder,
+            newly_joined_rooms=newly_joined_rooms,
+            newly_joined_users=newly_joined_users,
+            newly_left_rooms=newly_left_rooms,
+            newly_left_users=newly_left_users,
         )
 
+        device_id = sync_config.device_id
+        one_time_key_counts = {}
+        if device_id:
+            user_id = sync_config.user.to_string()
+            one_time_key_counts = yield self.store.count_e2e_one_time_keys(
+                user_id, device_id
+            )
+
         defer.returnValue(SyncResult(
             presence=sync_result_builder.presence,
             account_data=sync_result_builder.account_data,
@@ -558,30 +603,56 @@ class SyncHandler(object):
             archived=sync_result_builder.archived,
             to_device=sync_result_builder.to_device,
             device_lists=device_lists,
+            device_one_time_keys_count=one_time_key_counts,
             next_batch=sync_result_builder.now_token,
         ))
 
     @measure_func("_generate_sync_entry_for_device_list")
     @defer.inlineCallbacks
-    def _generate_sync_entry_for_device_list(self, sync_result_builder):
+    def _generate_sync_entry_for_device_list(self, sync_result_builder,
+                                             newly_joined_rooms, newly_joined_users,
+                                             newly_left_rooms, newly_left_users):
         user_id = sync_result_builder.sync_config.user.to_string()
         since_token = sync_result_builder.since_token
 
         if since_token and since_token.device_list_key:
-            room_ids = yield self.store.get_rooms_for_user(user_id)
-
-            user_ids_changed = set()
             changed = yield self.store.get_user_whose_devices_changed(
                 since_token.device_list_key
             )
-            for other_user_id in changed:
-                other_room_ids = yield self.store.get_rooms_for_user(other_user_id)
-                if room_ids.intersection(other_room_ids):
-                    user_ids_changed.add(other_user_id)
 
-            defer.returnValue(user_ids_changed)
+            # TODO: Be more clever than this, i.e. remove users who we already
+            # share a room with?
+            for room_id in newly_joined_rooms:
+                joined_users = yield self.state.get_current_user_in_room(room_id)
+                newly_joined_users.update(joined_users)
+
+            for room_id in newly_left_rooms:
+                left_users = yield self.state.get_current_user_in_room(room_id)
+                newly_left_users.update(left_users)
+
+            # TODO: Check that these users are actually new, i.e. either they
+            # weren't in the previous sync *or* they left and rejoined.
+            changed.update(newly_joined_users)
+
+            if not changed and not newly_left_users:
+                defer.returnValue(DeviceLists(
+                    changed=[],
+                    left=newly_left_users,
+                ))
+
+            users_who_share_room = yield self.store.get_users_who_share_room_with_user(
+                user_id
+            )
+
+            defer.returnValue(DeviceLists(
+                changed=users_who_share_room & changed,
+                left=set(newly_left_users) - users_who_share_room,
+            ))
         else:
-            defer.returnValue([])
+            defer.returnValue(DeviceLists(
+                changed=[],
+                left=[],
+            ))
 
     @defer.inlineCallbacks
     def _generate_sync_entry_for_to_device(self, sync_result_builder):
@@ -745,8 +816,8 @@ class SyncHandler(object):
             account_data_by_room(dict): Dictionary of per room account data
 
         Returns:
-            Deferred(tuple): Returns a 2-tuple of
-            `(newly_joined_rooms, newly_joined_users)`
+            Deferred(tuple): Returns a 4-tuple of
+            `(newly_joined_rooms, newly_joined_users, newly_left_rooms, newly_left_users)`
         """
         user_id = sync_result_builder.sync_config.user.to_string()
         block_all_room_ephemeral = (
@@ -777,7 +848,7 @@ class SyncHandler(object):
                     )
                     if not tags_by_room:
                         logger.debug("no-oping sync")
-                        defer.returnValue(([], []))
+                        defer.returnValue(([], [], [], []))
 
         ignored_account_data = yield self.store.get_global_account_data_by_type_for_user(
             "m.ignored_user_list", user_id=user_id,
@@ -790,7 +861,7 @@ class SyncHandler(object):
 
         if since_token:
             res = yield self._get_rooms_changed(sync_result_builder, ignored_users)
-            room_entries, invited, newly_joined_rooms = res
+            room_entries, invited, newly_joined_rooms, newly_left_rooms = res
 
             tags_by_room = yield self.store.get_updated_tags(
                 user_id, since_token.account_data_key,
@@ -798,6 +869,7 @@ class SyncHandler(object):
         else:
             res = yield self._get_all_rooms(sync_result_builder, ignored_users)
             room_entries, invited, newly_joined_rooms = res
+            newly_left_rooms = []
 
             tags_by_room = yield self.store.get_tags_for_user(user_id)
 
@@ -818,17 +890,30 @@ class SyncHandler(object):
 
         # Now we want to get any newly joined users
         newly_joined_users = set()
+        newly_left_users = set()
         if since_token:
             for joined_sync in sync_result_builder.joined:
                 it = itertools.chain(
-                    joined_sync.timeline.events, joined_sync.state.values()
+                    joined_sync.timeline.events, joined_sync.state.itervalues()
                 )
                 for event in it:
                     if event.type == EventTypes.Member:
                         if event.membership == Membership.JOIN:
                             newly_joined_users.add(event.state_key)
+                        else:
+                            prev_content = event.unsigned.get("prev_content", {})
+                            prev_membership = prev_content.get("membership", None)
+                            if prev_membership == Membership.JOIN:
+                                newly_left_users.add(event.state_key)
 
-        defer.returnValue((newly_joined_rooms, newly_joined_users))
+        newly_left_users -= newly_joined_users
+
+        defer.returnValue((
+            newly_joined_rooms,
+            newly_joined_users,
+            newly_left_rooms,
+            newly_left_users,
+        ))
 
     @defer.inlineCallbacks
     def _have_rooms_changed(self, sync_result_builder):
@@ -898,15 +983,28 @@ class SyncHandler(object):
             mem_change_events_by_room_id.setdefault(event.room_id, []).append(event)
 
         newly_joined_rooms = []
+        newly_left_rooms = []
         room_entries = []
         invited = []
-        for room_id, events in mem_change_events_by_room_id.items():
+        for room_id, events in mem_change_events_by_room_id.iteritems():
             non_joins = [e for e in events if e.membership != Membership.JOIN]
             has_join = len(non_joins) != len(events)
 
             # We want to figure out if we joined the room at some point since
             # the last sync (even if we have since left). This is to make sure
             # we do send down the room, and with full state, where necessary
+
+            old_state_ids = None
+            if room_id in joined_room_ids and non_joins:
+                # Always include if the user (re)joined the room, especially
+                # important so that device list changes are calculated correctly.
+                # If there are non join member events, but we are still in the room,
+                # then the user must have left and joined
+                newly_joined_rooms.append(room_id)
+
+                # User is in the room so we don't need to do the invite/leave checks
+                continue
+
             if room_id in joined_room_ids or has_join:
                 old_state_ids = yield self.get_state_at(room_id, since_token)
                 old_mem_ev_id = old_state_ids.get((EventTypes.Member, user_id), None)
@@ -918,12 +1016,33 @@ class SyncHandler(object):
                 if not old_mem_ev or old_mem_ev.membership != Membership.JOIN:
                     newly_joined_rooms.append(room_id)
 
-                if room_id in joined_room_ids:
-                    continue
+            # If user is in the room then we don't need to do the invite/leave checks
+            if room_id in joined_room_ids:
+                continue
 
             if not non_joins:
                 continue
 
+            # Check if we have left the room. This can either be because we were
+            # joined before *or* that we since joined and then left.
+            if events[-1].membership != Membership.JOIN:
+                if has_join:
+                    newly_left_rooms.append(room_id)
+                else:
+                    if not old_state_ids:
+                        old_state_ids = yield self.get_state_at(room_id, since_token)
+                        old_mem_ev_id = old_state_ids.get(
+                            (EventTypes.Member, user_id),
+                            None,
+                        )
+                        old_mem_ev = None
+                        if old_mem_ev_id:
+                            old_mem_ev = yield self.store.get_event(
+                                old_mem_ev_id, allow_none=True
+                            )
+                    if old_mem_ev and old_mem_ev.membership == Membership.JOIN:
+                        newly_left_rooms.append(room_id)
+
             # Only bother if we're still currently invited
             should_invite = non_joins[-1].membership == Membership.INVITE
             if should_invite:
@@ -1001,7 +1120,7 @@ class SyncHandler(object):
                     upto_token=since_token,
                 ))
 
-        defer.returnValue((room_entries, invited, newly_joined_rooms))
+        defer.returnValue((room_entries, invited, newly_joined_rooms, newly_left_rooms))
 
     @defer.inlineCallbacks
     def _get_all_rooms(self, sync_result_builder, ignored_users):
@@ -1249,6 +1368,7 @@ class SyncResultBuilder(object):
         self.invited = []
         self.archived = []
         self.device = []
+        self.to_device = []
 
 
 class RoomSyncResultBuilder(object):

synapse/handlers/typing.py

@@ -89,7 +89,7 @@ class TypingHandler(object):
             until = self._member_typing_until.get(member, None)
             if not until or until <= now:
                 logger.info("Timing out typing for: %s", member.user_id)
-                preserve_fn(self._stopped_typing)(member)
+                self._stopped_typing(member)
                 continue
 
             # Check if we need to resend a keep alive over federation for this
@@ -147,7 +147,7 @@ class TypingHandler(object):
             # No point sending another notification
             defer.returnValue(None)
 
-        yield self._push_update(
+        self._push_update(
             member=member,
             typing=True,
         )
@@ -171,7 +171,7 @@ class TypingHandler(object):
 
         member = RoomMember(room_id=room_id, user_id=target_user_id)
 
-        yield self._stopped_typing(member)
+        self._stopped_typing(member)
 
     @defer.inlineCallbacks
     def user_left_room(self, user, room_id):
@@ -180,7 +180,6 @@ class TypingHandler(object):
             member = RoomMember(room_id=room_id, user_id=user_id)
             yield self._stopped_typing(member)
 
-    @defer.inlineCallbacks
     def _stopped_typing(self, member):
         if member.user_id not in self._room_typing.get(member.room_id, set()):
             # No point
@@ -189,16 +188,15 @@ class TypingHandler(object):
         self._member_typing_until.pop(member, None)
         self._member_last_federation_poke.pop(member, None)
 
-        yield self._push_update(
+        self._push_update(
             member=member,
             typing=False,
         )
 
-    @defer.inlineCallbacks
     def _push_update(self, member, typing):
         if self.hs.is_mine_id(member.user_id):
             # Only send updates for changes to our own users.
-            yield self._push_remote(member, typing)
+            preserve_fn(self._push_remote)(member, typing)
 
         self._push_update_local(
             member=member,

synapse/handlers/user_directory.py

@@ -0,0 +1,641 @@
+# -*- coding: utf-8 -*-
+# Copyright 2017 Vector Creations Ltd
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+import logging
+from twisted.internet import defer
+
+from synapse.api.constants import EventTypes, JoinRules, Membership
+from synapse.storage.roommember import ProfileInfo
+from synapse.util.metrics import Measure
+from synapse.util.async import sleep
+
+
+logger = logging.getLogger(__name__)
+
+
+class UserDirectoyHandler(object):
+    """Handles querying of and keeping updated the user_directory.
+
+    N.B.: ASSUMES IT IS THE ONLY THING THAT MODIFIES THE USER DIRECTORY
+
+    The user directory is filled with users who this server can see are joined to a
+    world_readable or publically joinable room. We keep a database table up to date
+    by streaming changes of the current state and recalculating whether users should
+    be in the directory or not when necessary.
+
+    For each user in the directory we also store a room_id which is public and that the
+    user is joined to. This allows us to ignore history_visibility and join_rules changes
+    for that user in all other public rooms, as we know they'll still be in at least
+    one public room.
+    """
+
+    INITIAL_SLEEP_MS = 50
+    INITIAL_SLEEP_COUNT = 100
+    INITIAL_BATCH_SIZE = 100
+
+    def __init__(self, hs):
+        self.store = hs.get_datastore()
+        self.state = hs.get_state_handler()
+        self.server_name = hs.hostname
+        self.clock = hs.get_clock()
+        self.notifier = hs.get_notifier()
+        self.is_mine_id = hs.is_mine_id
+        self.update_user_directory = hs.config.update_user_directory
+
+        # When start up for the first time we need to populate the user_directory.
+        # This is a set of user_id's we've inserted already
+        self.initially_handled_users = set()
+        self.initially_handled_users_in_public = set()
+
+        self.initially_handled_users_share = set()
+        self.initially_handled_users_share_private_room = set()
+
+        # The current position in the current_state_delta stream
+        self.pos = None
+
+        # Guard to ensure we only process deltas one at a time
+        self._is_processing = False
+
+        if self.update_user_directory:
+            self.notifier.add_replication_callback(self.notify_new_event)
+
+            # We kick this off so that we don't have to wait for a change before
+            # we start populating the user directory
+            self.clock.call_later(0, self.notify_new_event)
+
+    def search_users(self, user_id, search_term, limit):
+        """Searches for users in directory
+
+        Returns:
+            dict of the form::
+
+                {
+                    "limited": <bool>,  # whether there were more results or not
+                    "results": [  # Ordered by best match first
+                        {
+                            "user_id": <user_id>,
+                            "display_name": <display_name>,
+                            "avatar_url": <avatar_url>
+                        }
+                    ]
+                }
+        """
+        return self.store.search_user_dir(user_id, search_term, limit)
+
+    @defer.inlineCallbacks
+    def notify_new_event(self):
+        """Called when there may be more deltas to process
+        """
+        if not self.update_user_directory:
+            return
+
+        if self._is_processing:
+            return
+
+        self._is_processing = True
+        try:
+            yield self._unsafe_process()
+        finally:
+            self._is_processing = False
+
+    @defer.inlineCallbacks
+    def _unsafe_process(self):
+        # If self.pos is None then means we haven't fetched it from DB
+        if self.pos is None:
+            self.pos = yield self.store.get_user_directory_stream_pos()
+
+        # If still None then we need to do the initial fill of directory
+        if self.pos is None:
+            yield self._do_initial_spam()
+            self.pos = yield self.store.get_user_directory_stream_pos()
+
+        # Loop round handling deltas until we're up to date
+        while True:
+            with Measure(self.clock, "user_dir_delta"):
+                deltas = yield self.store.get_current_state_deltas(self.pos)
+                if not deltas:
+                    return
+
+                logger.info("Handling %d state deltas", len(deltas))
+                yield self._handle_deltas(deltas)
+
+                self.pos = deltas[-1]["stream_id"]
+                yield self.store.update_user_directory_stream_pos(self.pos)
+
+    @defer.inlineCallbacks
+    def _do_initial_spam(self):
+        """Populates the user_directory from the current state of the DB, used
+        when synapse first starts with user_directory support
+        """
+        new_pos = yield self.store.get_max_stream_id_in_current_state_deltas()
+
+        # Delete any existing entries just in case there are any
+        yield self.store.delete_all_from_user_dir()
+
+        # We process by going through each existing room at a time.
+        room_ids = yield self.store.get_all_rooms()
+
+        logger.info("Doing initial update of user directory. %d rooms", len(room_ids))
+        num_processed_rooms = 1
+
+        for room_id in room_ids:
+            logger.info("Handling room %d/%d", num_processed_rooms, len(room_ids))
+            yield self._handle_intial_room(room_id)
+            num_processed_rooms += 1
+            yield sleep(self.INITIAL_SLEEP_MS / 1000.)
+
+        logger.info("Processed all rooms.")
+
+        self.initially_handled_users = None
+        self.initially_handled_users_in_public = None
+        self.initially_handled_users_share = None
+        self.initially_handled_users_share_private_room = None
+
+        yield self.store.update_user_directory_stream_pos(new_pos)
+
+    @defer.inlineCallbacks
+    def _handle_intial_room(self, room_id):
+        """Called when we initially fill out user_directory one room at a time
+        """
+        is_in_room = yield self.store.is_host_joined(room_id, self.server_name)
+        if not is_in_room:
+            return
+
+        is_public = yield self.store.is_room_world_readable_or_publicly_joinable(room_id)
+
+        users_with_profile = yield self.state.get_current_user_in_room(room_id)
+        user_ids = set(users_with_profile)
+        unhandled_users = user_ids - self.initially_handled_users
+
+        yield self.store.add_profiles_to_user_dir(
+            room_id, {
+                user_id: users_with_profile[user_id] for user_id in unhandled_users
+            }
+        )
+
+        self.initially_handled_users |= unhandled_users
+
+        if is_public:
+            yield self.store.add_users_to_public_room(
+                room_id,
+                user_ids=user_ids - self.initially_handled_users_in_public
+            )
+            self.initially_handled_users_in_public |= user_ids
+
+        # We now go and figure out the new users who share rooms with user entries
+        # We sleep aggressively here as otherwise it can starve resources.
+        # We also batch up inserts/updates, but try to avoid too many at once.
+        to_insert = set()
+        to_update = set()
+        count = 0
+        for user_id in user_ids:
+            if count % self.INITIAL_SLEEP_COUNT == 0:
+                yield sleep(self.INITIAL_SLEEP_MS / 1000.)
+
+            if not self.is_mine_id(user_id):
+                count += 1
+                continue
+
+            if self.store.get_if_app_services_interested_in_user(user_id):
+                count += 1
+                continue
+
+            for other_user_id in user_ids:
+                if user_id == other_user_id:
+                    continue
+
+                if count % self.INITIAL_SLEEP_COUNT == 0:
+                    yield sleep(self.INITIAL_SLEEP_MS / 1000.)
+                count += 1
+
+                user_set = (user_id, other_user_id)
+
+                if user_set in self.initially_handled_users_share_private_room:
+                    continue
+
+                if user_set in self.initially_handled_users_share:
+                    if is_public:
+                        continue
+                    to_update.add(user_set)
+                else:
+                    to_insert.add(user_set)
+
+                if is_public:
+                    self.initially_handled_users_share.add(user_set)
+                else:
+                    self.initially_handled_users_share_private_room.add(user_set)
+
+                if len(to_insert) > self.INITIAL_BATCH_SIZE:
+                    yield self.store.add_users_who_share_room(
+                        room_id, not is_public, to_insert,
+                    )
+                    to_insert.clear()
+
+                if len(to_update) > self.INITIAL_BATCH_SIZE:
+                    yield self.store.update_users_who_share_room(
+                        room_id, not is_public, to_update,
+                    )
+                    to_update.clear()
+
+        if to_insert:
+            yield self.store.add_users_who_share_room(
+                room_id, not is_public, to_insert,
+            )
+            to_insert.clear()
+
+        if to_update:
+            yield self.store.update_users_who_share_room(
+                room_id, not is_public, to_update,
+            )
+            to_update.clear()
+
+    @defer.inlineCallbacks
+    def _handle_deltas(self, deltas):
+        """Called with the state deltas to process
+        """
+        for delta in deltas:
+            typ = delta["type"]
+            state_key = delta["state_key"]
+            room_id = delta["room_id"]
+            event_id = delta["event_id"]
+            prev_event_id = delta["prev_event_id"]
+
+            logger.debug("Handling: %r %r, %s", typ, state_key, event_id)
+
+            # For join rule and visibility changes we need to check if the room
+            # may have become public or not and add/remove the users in said room
+            if typ in (EventTypes.RoomHistoryVisibility, EventTypes.JoinRules):
+                yield self._handle_room_publicity_change(
+                    room_id, prev_event_id, event_id, typ,
+                )
+            elif typ == EventTypes.Member:
+                change = yield self._get_key_change(
+                    prev_event_id, event_id,
+                    key_name="membership",
+                    public_value=Membership.JOIN,
+                )
+
+                if change is None:
+                    # Handle any profile changes
+                    yield self._handle_profile_change(
+                        state_key, room_id, prev_event_id, event_id,
+                    )
+                    continue
+
+                if not change:
+                    # Need to check if the server left the room entirely, if so
+                    # we might need to remove all the users in that room
+                    is_in_room = yield self.store.is_host_joined(
+                        room_id, self.server_name,
+                    )
+                    if not is_in_room:
+                        logger.info("Server left room: %r", room_id)
+                        # Fetch all the users that we marked as being in user
+                        # directory due to being in the room and then check if
+                        # need to remove those users or not
+                        user_ids = yield self.store.get_users_in_dir_due_to_room(room_id)
+                        for user_id in user_ids:
+                            yield self._handle_remove_user(room_id, user_id)
+                        return
+                    else:
+                        logger.debug("Server is still in room: %r", room_id)
+
+                if change:  # The user joined
+                    event = yield self.store.get_event(event_id, allow_none=True)
+                    profile = ProfileInfo(
+                        avatar_url=event.content.get("avatar_url"),
+                        display_name=event.content.get("displayname"),
+                    )
+
+                    yield self._handle_new_user(room_id, state_key, profile)
+                else:  # The user left
+                    yield self._handle_remove_user(room_id, state_key)
+            else:
+                logger.debug("Ignoring irrelevant type: %r", typ)
+
+    @defer.inlineCallbacks
+    def _handle_room_publicity_change(self, room_id, prev_event_id, event_id, typ):
+        """Handle a room having potentially changed from/to world_readable/publically
+        joinable.
+
+        Args:
+            room_id (str)
+            prev_event_id (str|None): The previous event before the state change
+            event_id (str|None): The new event after the state change
+            typ (str): Type of the event
+        """
+        logger.debug("Handling change for %s: %s", typ, room_id)
+
+        if typ == EventTypes.RoomHistoryVisibility:
+            change = yield self._get_key_change(
+                prev_event_id, event_id,
+                key_name="history_visibility",
+                public_value="world_readable",
+            )
+        elif typ == EventTypes.JoinRules:
+            change = yield self._get_key_change(
+                prev_event_id, event_id,
+                key_name="join_rule",
+                public_value=JoinRules.PUBLIC,
+            )
+        else:
+            raise Exception("Invalid event type")
+        # If change is None, no change. True => become world_readable/public,
+        # False => was world_readable/public
+        if change is None:
+            logger.debug("No change")
+            return
+
+        # There's been a change to or from being world readable.
+
+        is_public = yield self.store.is_room_world_readable_or_publicly_joinable(
+            room_id
+        )
+
+        logger.debug("Change: %r, is_public: %r", change, is_public)
+
+        if change and not is_public:
+            # If we became world readable but room isn't currently public then
+            # we ignore the change
+            return
+        elif not change and is_public:
+            # If we stopped being world readable but are still public,
+            # ignore the change
+            return
+
+        if change:
+            users_with_profile = yield self.state.get_current_user_in_room(room_id)
+            for user_id, profile in users_with_profile.iteritems():
+                yield self._handle_new_user(room_id, user_id, profile)
+        else:
+            users = yield self.store.get_users_in_public_due_to_room(room_id)
+            for user_id in users:
+                yield self._handle_remove_user(room_id, user_id)
+
+    @defer.inlineCallbacks
+    def _handle_new_user(self, room_id, user_id, profile):
+        """Called when we might need to add user to directory
+
+        Args:
+            room_id (str): room_id that user joined or started being public that
+            user_id (str)
+        """
+        logger.debug("Adding user to dir, %r", user_id)
+
+        row = yield self.store.get_user_in_directory(user_id)
+        if not row:
+            yield self.store.add_profiles_to_user_dir(room_id, {user_id: profile})
+
+        is_public = yield self.store.is_room_world_readable_or_publicly_joinable(
+            room_id
+        )
+
+        if is_public:
+            row = yield self.store.get_user_in_public_room(user_id)
+            if not row:
+                yield self.store.add_users_to_public_room(room_id, [user_id])
+        else:
+            logger.debug("Not adding user to public dir, %r", user_id)
+
+        # Now we update users who share rooms with users. We do this by getting
+        # all the current users in the room and seeing which aren't already
+        # marked in the database as sharing with `user_id`
+
+        users_with_profile = yield self.state.get_current_user_in_room(room_id)
+
+        to_insert = set()
+        to_update = set()
+
+        is_appservice = self.store.get_if_app_services_interested_in_user(user_id)
+
+        # First, if they're our user then we need to update for every user
+        if self.is_mine_id(user_id) and not is_appservice:
+            # Returns a map of other_user_id -> shared_private. We only need
+            # to update mappings if for users that either don't share a room
+            # already (aren't in the map) or, if the room is private, those that
+            # only share a public room.
+            user_ids_shared = yield self.store.get_users_who_share_room_from_dir(
+                user_id
+            )
+
+            for other_user_id in users_with_profile:
+                if user_id == other_user_id:
+                    continue
+
+                shared_is_private = user_ids_shared.get(other_user_id)
+                if shared_is_private is True:
+                    # We've already marked in the database they share a private room
+                    continue
+                elif shared_is_private is False:
+                    # They already share a public room, so only update if this is
+                    # a private room
+                    if not is_public:
+                        to_update.add((user_id, other_user_id))
+                elif shared_is_private is None:
+                    # This is the first time they both share a room
+                    to_insert.add((user_id, other_user_id))
+
+        # Next we need to update for every local user in the room
+        for other_user_id in users_with_profile:
+            if user_id == other_user_id:
+                continue
+
+            is_appservice = self.store.get_if_app_services_interested_in_user(
+                other_user_id
+            )
+            if self.is_mine_id(other_user_id) and not is_appservice:
+                shared_is_private = yield self.store.get_if_users_share_a_room(
+                    other_user_id, user_id,
+                )
+                if shared_is_private is True:
+                    # We've already marked in the database they share a private room
+                    continue
+                elif shared_is_private is False:
+                    # They already share a public room, so only update if this is
+                    # a private room
+                    if not is_public:
+                        to_update.add((other_user_id, user_id))
+                elif shared_is_private is None:
+                    # This is the first time they both share a room
+                    to_insert.add((other_user_id, user_id))
+
+        if to_insert:
+            yield self.store.add_users_who_share_room(
+                room_id, not is_public, to_insert,
+            )
+
+        if to_update:
+            yield self.store.update_users_who_share_room(
+                room_id, not is_public, to_update,
+            )
+
+    @defer.inlineCallbacks
+    def _handle_remove_user(self, room_id, user_id):
+        """Called when we might need to remove user to directory
+
+        Args:
+            room_id (str): room_id that user left or stopped being public that
+            user_id (str)
+        """
+        logger.debug("Maybe removing user %r", user_id)
+
+        row = yield self.store.get_user_in_directory(user_id)
+        update_user_dir = row and row["room_id"] == room_id
+
+        row = yield self.store.get_user_in_public_room(user_id)
+        update_user_in_public = row and row["room_id"] == room_id
+
+        if (update_user_in_public or update_user_dir):
+            # XXX: Make this faster?
+            rooms = yield self.store.get_rooms_for_user(user_id)
+            for j_room_id in rooms:
+                if (not update_user_in_public and not update_user_dir):
+                    break
+
+                is_in_room = yield self.store.is_host_joined(
+                    j_room_id, self.server_name,
+                )
+
+                if not is_in_room:
+                    continue
+
+                if update_user_dir:
+                    update_user_dir = False
+                    yield self.store.update_user_in_user_dir(user_id, j_room_id)
+
+                is_public = yield self.store.is_room_world_readable_or_publicly_joinable(
+                    j_room_id
+                )
+
+                if update_user_in_public and is_public:
+                    yield self.store.update_user_in_public_user_list(user_id, j_room_id)
+                    update_user_in_public = False
+
+        if update_user_dir:
+            yield self.store.remove_from_user_dir(user_id)
+        elif update_user_in_public:
+            yield self.store.remove_from_user_in_public_room(user_id)
+
+        # Now handle users_who_share_rooms.
+
+        # Get a list of user tuples that were in the DB due to this room and
+        # users (this includes tuples where the other user matches `user_id`)
+        user_tuples = yield self.store.get_users_in_share_dir_with_room_id(
+            user_id, room_id,
+        )
+
+        for user_id, other_user_id in user_tuples:
+            # For each user tuple get a list of rooms that they still share,
+            # trying to find a private room, and update the entry in the DB
+            rooms = yield self.store.get_rooms_in_common_for_users(user_id, other_user_id)
+
+            # If they dont share a room anymore, remove the mapping
+            if not rooms:
+                yield self.store.remove_user_who_share_room(
+                    user_id, other_user_id,
+                )
+                continue
+
+            found_public_share = None
+            for j_room_id in rooms:
+                is_public = yield self.store.is_room_world_readable_or_publicly_joinable(
+                    j_room_id
+                )
+
+                if is_public:
+                    found_public_share = j_room_id
+                else:
+                    found_public_share = None
+                    yield self.store.update_users_who_share_room(
+                        room_id, not is_public, [(user_id, other_user_id)],
+                    )
+                    break
+
+            if found_public_share:
+                yield self.store.update_users_who_share_room(
+                    room_id, not is_public, [(user_id, other_user_id)],
+                )
+
+    @defer.inlineCallbacks
+    def _handle_profile_change(self, user_id, room_id, prev_event_id, event_id):
+        """Check member event changes for any profile changes and update the
+        database if there are.
+        """
+        if not prev_event_id or not event_id:
+            return
+
+        prev_event = yield self.store.get_event(prev_event_id, allow_none=True)
+        event = yield self.store.get_event(event_id, allow_none=True)
+
+        if not prev_event or not event:
+            return
+
+        if event.membership != Membership.JOIN:
+            return
+
+        prev_name = prev_event.content.get("displayname")
+        new_name = event.content.get("displayname")
+
+        prev_avatar = prev_event.content.get("avatar_url")
+        new_avatar = event.content.get("avatar_url")
+
+        if prev_name != new_name or prev_avatar != new_avatar:
+            yield self.store.update_profile_in_user_dir(
+                user_id, new_name, new_avatar, room_id,
+            )
+
+    @defer.inlineCallbacks
+    def _get_key_change(self, prev_event_id, event_id, key_name, public_value):
+        """Given two events check if the `key_name` field in content changed
+        from not matching `public_value` to doing so.
+
+        For example, check if `history_visibility` (`key_name`) changed from
+        `shared` to `world_readable` (`public_value`).
+
+        Returns:
+            None if the field in the events either both match `public_value`
+            or if neither do, i.e. there has been no change.
+            True if it didnt match `public_value` but now does
+            False if it did match `public_value` but now doesn't
+        """
+        prev_event = None
+        event = None
+        if prev_event_id:
+            prev_event = yield self.store.get_event(prev_event_id, allow_none=True)
+
+        if event_id:
+            event = yield self.store.get_event(event_id, allow_none=True)
+
+        if not event and not prev_event:
+            logger.debug("Neither event exists: %r %r", prev_event_id, event_id)
+            defer.returnValue(None)
+
+        prev_value = None
+        value = None
+
+        if prev_event:
+            prev_value = prev_event.content.get(key_name)
+
+        if event:
+            value = event.content.get(key_name)
+
+        logger.debug("prev_value: %r -> value: %r", prev_value, value)
+
+        if value == public_value and prev_value != public_value:
+            defer.returnValue(True)
+        elif value != public_value and prev_value == public_value:
+            defer.returnValue(False)
+        else:
+            defer.returnValue(None)

synapse/http/endpoint.py

@@ -12,6 +12,7 @@
 # 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 socket
 
 from twisted.internet.endpoints import HostnameEndpoint, wrapClientTLS
 from twisted.internet import defer, reactor
@@ -30,7 +31,10 @@ logger = logging.getLogger(__name__)
 
 SERVER_CACHE = {}
 
-
+# our record of an individual server which can be tried to reach a destination.
+#
+# "host" is actually a dotted-quad or ipv6 address string. Except when there's
+# no SRV record, in which case it is the original hostname.
 _Server = collections.namedtuple(
     "_Server", "priority weight host port expires"
 )
@@ -219,9 +223,10 @@ class SRVClientEndpoint(object):
                 return self.default_server
             else:
                 raise ConnectError(
-                    "Not server available for %s" % self.service_name
+                    "No server available for %s" % self.service_name
                 )
 
+        # look for all servers with the same priority
         min_priority = self.servers[0].priority
         weight_indexes = list(
             (index, server.weight + 1)
@@ -231,11 +236,22 @@ class SRVClientEndpoint(object):
 
         total_weight = sum(weight for index, weight in weight_indexes)
         target_weight = random.randint(0, total_weight)
-
         for index, weight in weight_indexes:
             target_weight -= weight
             if target_weight <= 0:
                 server = self.servers[index]
+                # XXX: this looks totally dubious:
+                #
+                # (a) we never reuse a server until we have been through
+                #     all of the servers at the same priority, so if the
+                #     weights are A: 100, B:1, we always do ABABAB instead of
+                #     AAAA...AAAB (approximately).
+                #
+                # (b) After using all the servers at the lowest priority,
+                #     we move onto the next priority. We should only use the
+                #     second priority if servers at the top priority are
+                #     unreachable.
+                #
                 del self.servers[index]
                 self.used_servers.append(server)
                 return server
@@ -280,26 +296,21 @@ def resolve_service(service_name, dns_client=client, cache=SERVER_CACHE, clock=t
                 continue
 
             payload = answer.payload
-            host = str(payload.target)
-            srv_ttl = answer.ttl
 
-            try:
-                answers, _, _ = yield dns_client.lookupAddress(host)
-            except DNSNameError:
-                continue
+            hosts = yield _get_hosts_for_srv_record(
+                dns_client, str(payload.target)
+            )
 
-            for answer in answers:
-                if answer.type == dns.A and answer.payload:
-                    ip = answer.payload.dottedQuad()
-                    host_ttl = min(srv_ttl, answer.ttl)
+            for (ip, ttl) in hosts:
+                host_ttl = min(answer.ttl, ttl)
 
-                    servers.append(_Server(
-                        host=ip,
-                        port=int(payload.port),
-                        priority=int(payload.priority),
-                        weight=int(payload.weight),
-                        expires=int(clock.time()) + host_ttl,
-                    ))
+                servers.append(_Server(
+                    host=ip,
+                    port=int(payload.port),
+                    priority=int(payload.priority),
+                    weight=int(payload.weight),
+                    expires=int(clock.time()) + host_ttl,
+                ))
 
         servers.sort()
         cache[service_name] = list(servers)
@@ -317,3 +328,80 @@ def resolve_service(service_name, dns_client=client, cache=SERVER_CACHE, clock=t
             raise e
 
     defer.returnValue(servers)
+
+
+@defer.inlineCallbacks
+def _get_hosts_for_srv_record(dns_client, host):
+    """Look up each of the hosts in a SRV record
+
+    Args:
+        dns_client (twisted.names.dns.IResolver):
+        host (basestring): host to look up
+
+    Returns:
+        Deferred[list[(str, int)]]: a list of (host, ttl) pairs
+
+    """
+    ip4_servers = []
+    ip6_servers = []
+
+    def cb(res):
+        # lookupAddress and lookupIP6Address return a three-tuple
+        # giving the answer, authority, and additional sections of the
+        # response.
+        #
+        # we only care about the answers.
+
+        return res[0]
+
+    def eb(res, record_type):
+        if res.check(DNSNameError):
+            return []
+        logger.warn("Error looking up %s for %s: %s",
+                    record_type, host, res, res.value)
+        return res
+
+    # no logcontexts here, so we can safely fire these off and gatherResults
+    d1 = dns_client.lookupAddress(host).addCallbacks(cb, eb)
+    d2 = dns_client.lookupIPV6Address(host).addCallbacks(cb, eb)
+    results = yield defer.DeferredList(
+        [d1, d2], consumeErrors=True)
+
+    # if all of the lookups failed, raise an exception rather than blowing out
+    # the cache with an empty result.
+    if results and all(s == defer.FAILURE for (s, _) in results):
+        defer.returnValue(results[0][1])
+
+    for (success, result) in results:
+        if success == defer.FAILURE:
+            continue
+
+        for answer in result:
+            if not answer.payload:
+                continue
+
+            try:
+                if answer.type == dns.A:
+                    ip = answer.payload.dottedQuad()
+                    ip4_servers.append((ip, answer.ttl))
+                elif answer.type == dns.AAAA:
+                    ip = socket.inet_ntop(
+                        socket.AF_INET6, answer.payload.address,
+                    )
+                    ip6_servers.append((ip, answer.ttl))
+                else:
+                    # the most likely candidate here is a CNAME record.
+                    # rfc2782 says srvs may not point to aliases.
+                    logger.warn(
+                        "Ignoring unexpected DNS record type %s for %s",
+                        answer.type, host,
+                    )
+                    continue
+            except Exception as e:
+                logger.warn("Ignoring invalid DNS response for %s: %s",
+                            host, e)
+                continue
+
+    # keep the ipv4 results before the ipv6 results, mostly to match historical
+    # behaviour.
+    defer.returnValue(ip4_servers + ip6_servers)

synapse/http/matrixfederationclient.py

@@ -204,18 +204,15 @@ class MatrixFederationHttpClient(object):
                             raise
 
                         logger.warn(
-                            "{%s} Sending request failed to %s: %s %s: %s - %s",
+                            "{%s} Sending request failed to %s: %s %s: %s",
                             txn_id,
                             destination,
                             method,
                             url_bytes,
-                            type(e).__name__,
                             _flatten_response_never_received(e),
                         )
 
-                        log_result = "%s - %s" % (
-                            type(e).__name__, _flatten_response_never_received(e),
-                        )
+                        log_result = _flatten_response_never_received(e)
 
                         if retries_left and not timeout:
                             if long_retries:
@@ -578,12 +575,14 @@ class _JsonProducer(object):
 
 def _flatten_response_never_received(e):
     if hasattr(e, "reasons"):
-        return ", ".join(
+        reasons = ", ".join(
             _flatten_response_never_received(f.value)
             for f in e.reasons
         )
+
+        return "%s:[%s]" % (type(e).__name__, reasons)
     else:
-        return "%s: %s" % (type(e).__name__, e.message,)
+        return repr(e)
 
 
 def check_content_type_is_json(headers):

synapse/http/server.py

@@ -412,7 +412,7 @@ def set_cors_headers(request):
     )
     request.setHeader(
         "Access-Control-Allow-Headers",
-        "Origin, X-Requested-With, Content-Type, Accept"
+        "Origin, X-Requested-With, Content-Type, Accept, Authorization"
     )
 
 

synapse/http/site.py

@@ -20,7 +20,7 @@ import logging
 import re
 import time
 
-ACCESS_TOKEN_RE = re.compile(r'(\?.*access(_|%5[Ff])token=)[^&]*(.*)$')
+ACCESS_TOKEN_RE = re.compile(r'(\?\S*?access(_|%5[Ff])token=).*?(&|\s|$)')
 
 
 class SynapseRequest(Request):

synapse/notifier.py

@@ -251,7 +251,8 @@ class Notifier(object):
         """Notify any user streams that are interested in this room event"""
         # poke any interested application service.
         preserve_fn(self.appservice_handler.notify_interested_services)(
-            room_stream_id)
+            room_stream_id
+        )
 
         if self.federation_sender:
             preserve_fn(self.federation_sender.notify_new_events)(

synapse/push/action_generator.py

@@ -15,7 +15,7 @@
 
 from twisted.internet import defer
 
-from .bulk_push_rule_evaluator import evaluator_for_event
+from .bulk_push_rule_evaluator import BulkPushRuleEvaluator
 
 from synapse.util.metrics import Measure
 
@@ -24,11 +24,12 @@ import logging
 logger = logging.getLogger(__name__)
 
 
-class ActionGenerator:
+class ActionGenerator(object):
     def __init__(self, hs):
         self.hs = hs
         self.clock = hs.get_clock()
         self.store = hs.get_datastore()
+        self.bulk_evaluator = BulkPushRuleEvaluator(hs)
         # really we want to get all user ids and all profile tags too,
         # since we want the actions for each profile tag for every user and
         # also actions for a client with no profile tag for each user.
@@ -38,16 +39,11 @@ class ActionGenerator:
 
     @defer.inlineCallbacks
     def handle_push_actions_for_event(self, event, context):
-        with Measure(self.clock, "evaluator_for_event"):
-            bulk_evaluator = yield evaluator_for_event(
-                event, self.hs, self.store, context
-            )
-
         with Measure(self.clock, "action_for_event_by_user"):
-            actions_by_user = yield bulk_evaluator.action_for_event_by_user(
+            actions_by_user = yield self.bulk_evaluator.action_for_event_by_user(
                 event, context
             )
 
         context.push_actions = [
-            (uid, actions) for uid, actions in actions_by_user.items()
+            (uid, actions) for uid, actions in actions_by_user.iteritems()
         ]

synapse/push/bulk_push_rule_evaluator.py

@@ -19,65 +19,106 @@ from twisted.internet import defer
 
 from .push_rule_evaluator import PushRuleEvaluatorForEvent
 
-from synapse.api.constants import EventTypes
-from synapse.visibility import filter_events_for_clients_context
+from synapse.api.constants import EventTypes, Membership
+from synapse.metrics import get_metrics_for
+from synapse.util.caches import metrics as cache_metrics
+from synapse.util.caches.descriptors import cached
+from synapse.util.async import Linearizer
+
+from collections import namedtuple
 
 
 logger = logging.getLogger(__name__)
 
 
-@defer.inlineCallbacks
-def evaluator_for_event(event, hs, store, context):
-    rules_by_user = yield store.bulk_get_push_rules_for_room(
-        event, context
-    )
+rules_by_room = {}
 
-    # if this event is an invite event, we may need to run rules for the user
-    # who's been invited, otherwise they won't get told they've been invited
-    if event.type == 'm.room.member' and event.content['membership'] == 'invite':
-        invited_user = event.state_key
-        if invited_user and hs.is_mine_id(invited_user):
-            has_pusher = yield store.user_has_pusher(invited_user)
-            if has_pusher:
-                rules_by_user = dict(rules_by_user)
-                rules_by_user[invited_user] = yield store.get_push_rules_for_user(
-                    invited_user
-                )
+push_metrics = get_metrics_for(__name__)
 
-    defer.returnValue(BulkPushRuleEvaluator(
-        event.room_id, rules_by_user, store
-    ))
+push_rules_invalidation_counter = push_metrics.register_counter(
+    "push_rules_invalidation_counter"
+)
+push_rules_state_size_counter = push_metrics.register_counter(
+    "push_rules_state_size_counter"
+)
+
+# Measures whether we use the fast path of using state deltas, or if we have to
+# recalculate from scratch
+push_rules_delta_state_cache_metric = cache_metrics.register_cache(
+    "cache",
+    size_callback=lambda: 0,  # Meaningless size, as this isn't a cache that stores values
+    cache_name="push_rules_delta_state_cache_metric",
+)
 
 
-class BulkPushRuleEvaluator:
+class BulkPushRuleEvaluator(object):
+    """Calculates the outcome of push rules for an event for all users in the
+    room at once.
     """
-    Runs push rules for all users in a room.
-    This is faster than running PushRuleEvaluator for each user because it
-    fetches all the rules for all the users in one (batched) db query
-    rather than doing multiple queries per-user. It currently uses
-    the same logic to run the actual rules, but could be optimised further
-    (see https://matrix.org/jira/browse/SYN-562)
-    """
-    def __init__(self, room_id, rules_by_user, store):
-        self.room_id = room_id
-        self.rules_by_user = rules_by_user
-        self.store = store
+
+    def __init__(self, hs):
+        self.hs = hs
+        self.store = hs.get_datastore()
+
+        self.room_push_rule_cache_metrics = cache_metrics.register_cache(
+            "cache",
+            size_callback=lambda: 0,  # There's not good value for this
+            cache_name="room_push_rule_cache",
+        )
+
+    @defer.inlineCallbacks
+    def _get_rules_for_event(self, event, context):
+        """This gets the rules for all users in the room at the time of the event,
+        as well as the push rules for the invitee if the event is an invite.
+
+        Returns:
+            dict of user_id -> push_rules
+        """
+        room_id = event.room_id
+        rules_for_room = self._get_rules_for_room(room_id)
+
+        rules_by_user = yield rules_for_room.get_rules(event, context)
+
+        # if this event is an invite event, we may need to run rules for the user
+        # who's been invited, otherwise they won't get told they've been invited
+        if event.type == 'm.room.member' and event.content['membership'] == 'invite':
+            invited = event.state_key
+            if invited and self.hs.is_mine_id(invited):
+                has_pusher = yield self.store.user_has_pusher(invited)
+                if has_pusher:
+                    rules_by_user = dict(rules_by_user)
+                    rules_by_user[invited] = yield self.store.get_push_rules_for_user(
+                        invited
+                    )
+
+        defer.returnValue(rules_by_user)
+
+    @cached()
+    def _get_rules_for_room(self, room_id):
+        """Get the current RulesForRoom object for the given room id
+
+        Returns:
+            RulesForRoom
+        """
+        # It's important that RulesForRoom gets added to self._get_rules_for_room.cache
+        # before any lookup methods get called on it as otherwise there may be
+        # a race if invalidate_all gets called (which assumes its in the cache)
+        return RulesForRoom(
+            self.hs, room_id, self._get_rules_for_room.cache,
+            self.room_push_rule_cache_metrics,
+        )
 
     @defer.inlineCallbacks
     def action_for_event_by_user(self, event, context):
+        """Given an event and context, evaluate the push rules and return
+        the results
+
+        Returns:
+            dict of user_id -> action
+        """
+        rules_by_user = yield self._get_rules_for_event(event, context)
         actions_by_user = {}
 
-        # None of these users can be peeking since this list of users comes
-        # from the set of users in the room, so we know for sure they're all
-        # actually in the room.
-        user_tuples = [
-            (u, False) for u in self.rules_by_user.keys()
-        ]
-
-        filtered_by_user = yield filter_events_for_clients_context(
-            self.store, user_tuples, [event], {event.event_id: context}
-        )
-
         room_members = yield self.store.get_joined_users_from_context(
             event, context
         )
@@ -86,7 +127,15 @@ class BulkPushRuleEvaluator:
 
         condition_cache = {}
 
-        for uid, rules in self.rules_by_user.items():
+        for uid, rules in rules_by_user.iteritems():
+            if event.sender == uid:
+                continue
+
+            if not event.is_state():
+                is_ignored = yield self.store.is_ignored_by(event.sender, uid)
+                if is_ignored:
+                    continue
+
             display_name = None
             profile_info = room_members.get(uid)
             if profile_info:
@@ -98,13 +147,6 @@ class BulkPushRuleEvaluator:
                 if event.type == EventTypes.Member and event.state_key == uid:
                     display_name = event.content.get("displayname", None)
 
-            filtered = filtered_by_user[uid]
-            if len(filtered) == 0:
-                continue
-
-            if filtered[0].sender == uid:
-                continue
-
             for rule in rules:
                 if 'enabled' in rule and not rule['enabled']:
                     continue
@@ -138,3 +180,264 @@ def _condition_checker(evaluator, conditions, uid, display_name, cache):
             return False
 
     return True
+
+
+class RulesForRoom(object):
+    """Caches push rules for users in a room.
+
+    This efficiently handles users joining/leaving the room by not invalidating
+    the entire cache for the room.
+    """
+
+    def __init__(self, hs, room_id, rules_for_room_cache, room_push_rule_cache_metrics):
+        """
+        Args:
+            hs (HomeServer)
+            room_id (str)
+            rules_for_room_cache(Cache): The cache object that caches these
+                RoomsForUser objects.
+            room_push_rule_cache_metrics (CacheMetric)
+        """
+        self.room_id = room_id
+        self.is_mine_id = hs.is_mine_id
+        self.store = hs.get_datastore()
+        self.room_push_rule_cache_metrics = room_push_rule_cache_metrics
+
+        self.linearizer = Linearizer(name="rules_for_room")
+
+        self.member_map = {}  # event_id -> (user_id, state)
+        self.rules_by_user = {}  # user_id -> rules
+
+        # The last state group we updated the caches for. If the state_group of
+        # a new event comes along, we know that we can just return the cached
+        # result.
+        # On invalidation of the rules themselves (if the user changes them),
+        # we invalidate everything and set state_group to `object()`
+        self.state_group = object()
+
+        # A sequence number to keep track of when we're allowed to update the
+        # cache. We bump the sequence number when we invalidate the cache. If
+        # the sequence number changes while we're calculating stuff we should
+        # not update the cache with it.
+        self.sequence = 0
+
+        # A cache of user_ids that we *know* aren't interesting, e.g. user_ids
+        # owned by AS's, or remote users, etc. (I.e. users we will never need to
+        # calculate push for)
+        # These never need to be invalidated as we will never set up push for
+        # them.
+        self.uninteresting_user_set = set()
+
+        # We need to be clever on the invalidating caches callbacks, as
+        # otherwise the invalidation callback holds a reference to the object,
+        # potentially causing it to leak.
+        # To get around this we pass a function that on invalidations looks ups
+        # the RoomsForUser entry in the cache, rather than keeping a reference
+        # to self around in the callback.
+        self.invalidate_all_cb = _Invalidation(rules_for_room_cache, room_id)
+
+    @defer.inlineCallbacks
+    def get_rules(self, event, context):
+        """Given an event context return the rules for all users who are
+        currently in the room.
+        """
+        state_group = context.state_group
+
+        if state_group and self.state_group == state_group:
+            logger.debug("Using cached rules for %r", self.room_id)
+            self.room_push_rule_cache_metrics.inc_hits()
+            defer.returnValue(self.rules_by_user)
+
+        with (yield self.linearizer.queue(())):
+            if state_group and self.state_group == state_group:
+                logger.debug("Using cached rules for %r", self.room_id)
+                self.room_push_rule_cache_metrics.inc_hits()
+                defer.returnValue(self.rules_by_user)
+
+            self.room_push_rule_cache_metrics.inc_misses()
+
+            ret_rules_by_user = {}
+            missing_member_event_ids = {}
+            if state_group and self.state_group == context.prev_group:
+                # If we have a simple delta then we can reuse most of the previous
+                # results.
+                ret_rules_by_user = self.rules_by_user
+                current_state_ids = context.delta_ids
+
+                push_rules_delta_state_cache_metric.inc_hits()
+            else:
+                current_state_ids = context.current_state_ids
+                push_rules_delta_state_cache_metric.inc_misses()
+
+            push_rules_state_size_counter.inc_by(len(current_state_ids))
+
+            logger.debug(
+                "Looking for member changes in %r %r", state_group, current_state_ids
+            )
+
+            # Loop through to see which member events we've seen and have rules
+            # for and which we need to fetch
+            for key in current_state_ids:
+                typ, user_id = key
+                if typ != EventTypes.Member:
+                    continue
+
+                if user_id in self.uninteresting_user_set:
+                    continue
+
+                if not self.is_mine_id(user_id):
+                    self.uninteresting_user_set.add(user_id)
+                    continue
+
+                if self.store.get_if_app_services_interested_in_user(user_id):
+                    self.uninteresting_user_set.add(user_id)
+                    continue
+
+                event_id = current_state_ids[key]
+
+                res = self.member_map.get(event_id, None)
+                if res:
+                    user_id, state = res
+                    if state == Membership.JOIN:
+                        rules = self.rules_by_user.get(user_id, None)
+                        if rules:
+                            ret_rules_by_user[user_id] = rules
+                    continue
+
+                # If a user has left a room we remove their push rule. If they
+                # joined then we readd it later in _update_rules_with_member_event_ids
+                ret_rules_by_user.pop(user_id, None)
+                missing_member_event_ids[user_id] = event_id
+
+            if missing_member_event_ids:
+                # If we have some memebr events we haven't seen, look them up
+                # and fetch push rules for them if appropriate.
+                logger.debug("Found new member events %r", missing_member_event_ids)
+                yield self._update_rules_with_member_event_ids(
+                    ret_rules_by_user, missing_member_event_ids, state_group, event
+                )
+            else:
+                # The push rules didn't change but lets update the cache anyway
+                self.update_cache(
+                    self.sequence,
+                    members={},  # There were no membership changes
+                    rules_by_user=ret_rules_by_user,
+                    state_group=state_group
+                )
+
+        if logger.isEnabledFor(logging.DEBUG):
+            logger.debug(
+                "Returning push rules for %r %r",
+                self.room_id, ret_rules_by_user.keys(),
+            )
+        defer.returnValue(ret_rules_by_user)
+
+    @defer.inlineCallbacks
+    def _update_rules_with_member_event_ids(self, ret_rules_by_user, member_event_ids,
+                                            state_group, event):
+        """Update the partially filled rules_by_user dict by fetching rules for
+        any newly joined users in the `member_event_ids` list.
+
+        Args:
+            ret_rules_by_user (dict): Partiallly filled dict of push rules. Gets
+                updated with any new rules.
+            member_event_ids (list): List of event ids for membership events that
+                have happened since the last time we filled rules_by_user
+            state_group: The state group we are currently computing push rules
+                for. Used when updating the cache.
+        """
+        sequence = self.sequence
+
+        rows = yield self.store._simple_select_many_batch(
+            table="room_memberships",
+            column="event_id",
+            iterable=member_event_ids.values(),
+            retcols=('user_id', 'membership', 'event_id'),
+            keyvalues={},
+            batch_size=500,
+            desc="_get_rules_for_member_event_ids",
+        )
+
+        members = {
+            row["event_id"]: (row["user_id"], row["membership"])
+            for row in rows
+        }
+
+        # If the event is a join event then it will be in current state evnts
+        # map but not in the DB, so we have to explicitly insert it.
+        if event.type == EventTypes.Member:
+            for event_id in member_event_ids.itervalues():
+                if event_id == event.event_id:
+                    members[event_id] = (event.state_key, event.membership)
+
+        if logger.isEnabledFor(logging.DEBUG):
+            logger.debug("Found members %r: %r", self.room_id, members.values())
+
+        interested_in_user_ids = set(
+            user_id for user_id, membership in members.itervalues()
+            if membership == Membership.JOIN
+        )
+
+        logger.debug("Joined: %r", interested_in_user_ids)
+
+        if_users_with_pushers = yield self.store.get_if_users_have_pushers(
+            interested_in_user_ids,
+            on_invalidate=self.invalidate_all_cb,
+        )
+
+        user_ids = set(
+            uid for uid, have_pusher in if_users_with_pushers.iteritems() if have_pusher
+        )
+
+        logger.debug("With pushers: %r", user_ids)
+
+        users_with_receipts = yield self.store.get_users_with_read_receipts_in_room(
+            self.room_id, on_invalidate=self.invalidate_all_cb,
+        )
+
+        logger.debug("With receipts: %r", users_with_receipts)
+
+        # any users with pushers must be ours: they have pushers
+        for uid in users_with_receipts:
+            if uid in interested_in_user_ids:
+                user_ids.add(uid)
+
+        rules_by_user = yield self.store.bulk_get_push_rules(
+            user_ids, on_invalidate=self.invalidate_all_cb,
+        )
+
+        ret_rules_by_user.update(
+            item for item in rules_by_user.iteritems() if item[0] is not None
+        )
+
+        self.update_cache(sequence, members, ret_rules_by_user, state_group)
+
+    def invalidate_all(self):
+        # Note: Don't hand this function directly to an invalidation callback
+        # as it keeps a reference to self and will stop this instance from being
+        # GC'd if it gets dropped from the rules_to_user cache. Instead use
+        # `self.invalidate_all_cb`
+        logger.debug("Invalidating RulesForRoom for %r", self.room_id)
+        self.sequence += 1
+        self.state_group = object()
+        self.member_map = {}
+        self.rules_by_user = {}
+        push_rules_invalidation_counter.inc()
+
+    def update_cache(self, sequence, members, rules_by_user, state_group):
+        if sequence == self.sequence:
+            self.member_map.update(members)
+            self.rules_by_user = rules_by_user
+            self.state_group = state_group
+
+
+class _Invalidation(namedtuple("_Invalidation", ("cache", "room_id"))):
+    # We rely on _CacheContext implementing __eq__ and __hash__ sensibly,
+    # which namedtuple does for us (i.e. two _CacheContext are the same if
+    # their caches and keys match). This is important in particular to
+    # dedupe when we add callbacks to lru cache nodes, otherwise the number
+    # of callbacks would grow.
+    def __call__(self):
+        rules = self.cache.get(self.room_id, None, update_metrics=False)
+        if rules:
+            rules.invalidate_all()

synapse/push/emailpusher.py

@@ -21,7 +21,6 @@ import logging
 from synapse.util.metrics import Measure
 from synapse.util.logcontext import LoggingContext
 
-from mailer import Mailer
 
 logger = logging.getLogger(__name__)
 
@@ -56,8 +55,10 @@ class EmailPusher(object):
     This shares quite a bit of code with httpusher: it would be good to
     factor out the common parts
     """
-    def __init__(self, hs, pusherdict):
+    def __init__(self, hs, pusherdict, mailer):
         self.hs = hs
+        self.mailer = mailer
+
         self.store = self.hs.get_datastore()
         self.clock = self.hs.get_clock()
         self.pusher_id = pusherdict['id']
@@ -73,16 +74,6 @@ class EmailPusher(object):
 
         self.processing = False
 
-        if self.hs.config.email_enable_notifs:
-            if 'data' in pusherdict and 'brand' in pusherdict['data']:
-                app_name = pusherdict['data']['brand']
-            else:
-                app_name = self.hs.config.email_app_name
-
-            self.mailer = Mailer(self.hs, app_name)
-        else:
-            self.mailer = None
-
     @defer.inlineCallbacks
     def on_started(self):
         if self.mailer is not None:

synapse/push/httppusher.py

@@ -244,6 +244,26 @@ class HttpPusher(object):
 
     @defer.inlineCallbacks
     def _build_notification_dict(self, event, tweaks, badge):
+        if self.data.get('format') == 'event_id_only':
+            d = {
+                'notification': {
+                    'event_id': event.event_id,
+                    'room_id': event.room_id,
+                    'counts': {
+                        'unread': badge,
+                    },
+                    'devices': [
+                        {
+                            'app_id': self.app_id,
+                            'pushkey': self.pushkey,
+                            'pushkey_ts': long(self.pushkey_ts / 1000),
+                            'data': self.data_minus_url,
+                        }
+                    ]
+                }
+            }
+            defer.returnValue(d)
+
         ctx = yield push_tools.get_context_for_event(
             self.store, self.state_handler, event, self.user_id
         )
@@ -275,7 +295,7 @@ class HttpPusher(object):
         if event.type == 'm.room.member':
             d['notification']['membership'] = event.content['membership']
             d['notification']['user_is_target'] = event.state_key == self.user_id
-        if 'content' in event:
+        if not self.hs.config.push_redact_content and 'content' in event:
             d['notification']['content'] = event.content
 
         # We no longer send aliases separately, instead, we send the human

synapse/push/mailer.py

@@ -78,23 +78,17 @@ ALLOWED_ATTRS = {
 
 
 class Mailer(object):
-    def __init__(self, hs, app_name):
+    def __init__(self, hs, app_name, notif_template_html, notif_template_text):
         self.hs = hs
+        self.notif_template_html = notif_template_html
+        self.notif_template_text = notif_template_text
+
         self.store = self.hs.get_datastore()
         self.macaroon_gen = self.hs.get_macaroon_generator()
         self.state_handler = self.hs.get_state_handler()
-        loader = jinja2.FileSystemLoader(self.hs.config.email_template_dir)
         self.app_name = app_name
+
         logger.info("Created Mailer for app_name %s" % app_name)
-        env = jinja2.Environment(loader=loader)
-        env.filters["format_ts"] = format_ts_filter
-        env.filters["mxc_to_http"] = self.mxc_to_http_filter
-        self.notif_template_html = env.get_template(
-            self.hs.config.email_notif_template_html
-        )
-        self.notif_template_text = env.get_template(
-            self.hs.config.email_notif_template_text
-        )
 
     @defer.inlineCallbacks
     def send_notification_mail(self, app_id, user_id, email_address,
@@ -481,28 +475,6 @@ class Mailer(object):
             urllib.urlencode(params),
         )
 
-    def mxc_to_http_filter(self, value, width, height, resize_method="crop"):
-        if value[0:6] != "mxc://":
-            return ""
-
-        serverAndMediaId = value[6:]
-        fragment = None
-        if '#' in serverAndMediaId:
-            (serverAndMediaId, fragment) = serverAndMediaId.split('#', 1)
-            fragment = "#" + fragment
-
-        params = {
-            "width": width,
-            "height": height,
-            "method": resize_method,
-        }
-        return "%s_matrix/media/v1/thumbnail/%s?%s%s" % (
-            self.hs.config.public_baseurl,
-            serverAndMediaId,
-            urllib.urlencode(params),
-            fragment or "",
-        )
-
 
 def safe_markup(raw_html):
     return jinja2.Markup(bleach.linkify(bleach.clean(
@@ -543,3 +515,52 @@ def string_ordinal_total(s):
 
 def format_ts_filter(value, format):
     return time.strftime(format, time.localtime(value / 1000))
+
+
+def load_jinja2_templates(config):
+    """Load the jinja2 email templates from disk
+
+    Returns:
+        (notif_template_html, notif_template_text)
+    """
+    logger.info("loading jinja2")
+
+    loader = jinja2.FileSystemLoader(config.email_template_dir)
+    env = jinja2.Environment(loader=loader)
+    env.filters["format_ts"] = format_ts_filter
+    env.filters["mxc_to_http"] = _create_mxc_to_http_filter(config)
+
+    notif_template_html = env.get_template(
+        config.email_notif_template_html
+    )
+    notif_template_text = env.get_template(
+        config.email_notif_template_text
+    )
+
+    return notif_template_html, notif_template_text
+
+
+def _create_mxc_to_http_filter(config):
+    def mxc_to_http_filter(value, width, height, resize_method="crop"):
+        if value[0:6] != "mxc://":
+            return ""
+
+        serverAndMediaId = value[6:]
+        fragment = None
+        if '#' in serverAndMediaId:
+            (serverAndMediaId, fragment) = serverAndMediaId.split('#', 1)
+            fragment = "#" + fragment
+
+        params = {
+            "width": width,
+            "height": height,
+            "method": resize_method,
+        }
+        return "%s_matrix/media/v1/thumbnail/%s?%s%s" % (
+            config.public_baseurl,
+            serverAndMediaId,
+            urllib.urlencode(params),
+            fragment or "",
+        )
+
+    return mxc_to_http_filter

synapse/push/push_rule_evaluator.py

@@ -200,7 +200,9 @@ def _glob_to_re(glob, word_boundary):
         return re.compile(r, flags=re.IGNORECASE)
 
 
-def _flatten_dict(d, prefix=[], result={}):
+def _flatten_dict(d, prefix=[], result=None):
+    if result is None:
+        result = {}
     for key, value in d.items():
         if isinstance(value, basestring):
             result[".".join(prefix + [key])] = value.lower()

synapse/push/pusher.py

@@ -26,22 +26,54 @@ logger = logging.getLogger(__name__)
 # process works fine)
 try:
     from synapse.push.emailpusher import EmailPusher
+    from synapse.push.mailer import Mailer, load_jinja2_templates
 except:
     pass
 
 
-def create_pusher(hs, pusherdict):
-    logger.info("trying to create_pusher for %r", pusherdict)
+class PusherFactory(object):
+    def __init__(self, hs):
+        self.hs = hs
 
-    PUSHER_TYPES = {
-        "http": HttpPusher,
-    }
+        self.pusher_types = {
+            "http": HttpPusher,
+        }
 
-    logger.info("email enable notifs: %r", hs.config.email_enable_notifs)
-    if hs.config.email_enable_notifs:
-        PUSHER_TYPES["email"] = EmailPusher
-        logger.info("defined email pusher type")
+        logger.info("email enable notifs: %r", hs.config.email_enable_notifs)
+        if hs.config.email_enable_notifs:
+            self.mailers = {}  # app_name -> Mailer
 
-    if pusherdict['kind'] in PUSHER_TYPES:
-        logger.info("found pusher")
-        return PUSHER_TYPES[pusherdict['kind']](hs, pusherdict)
+            templates = load_jinja2_templates(hs.config)
+            self.notif_template_html, self.notif_template_text = templates
+
+            self.pusher_types["email"] = self._create_email_pusher
+
+            logger.info("defined email pusher type")
+
+    def create_pusher(self, pusherdict):
+        logger.info("trying to create_pusher for %r", pusherdict)
+
+        if pusherdict['kind'] in self.pusher_types:
+            logger.info("found pusher")
+            return self.pusher_types[pusherdict['kind']](self.hs, pusherdict)
+
+    def _create_email_pusher(self, _hs, pusherdict):
+        app_name = self._app_name_from_pusherdict(pusherdict)
+        mailer = self.mailers.get(app_name)
+        if not mailer:
+            mailer = Mailer(
+                hs=self.hs,
+                app_name=app_name,
+                notif_template_html=self.notif_template_html,
+                notif_template_text=self.notif_template_text,
+            )
+            self.mailers[app_name] = mailer
+        return EmailPusher(self.hs, pusherdict, mailer)
+
+    def _app_name_from_pusherdict(self, pusherdict):
+        if 'data' in pusherdict and 'brand' in pusherdict['data']:
+            app_name = pusherdict['data']['brand']
+        else:
+            app_name = self.hs.config.email_app_name
+
+        return app_name

synapse/push/pusherpool.py

@@ -16,7 +16,7 @@
 
 from twisted.internet import defer
 
-import pusher
+from .pusher import PusherFactory
 from synapse.util.logcontext import preserve_fn, preserve_context_over_deferred
 from synapse.util.async import run_on_reactor
 
@@ -28,6 +28,7 @@ logger = logging.getLogger(__name__)
 class PusherPool:
     def __init__(self, _hs):
         self.hs = _hs
+        self.pusher_factory = PusherFactory(_hs)
         self.start_pushers = _hs.config.start_pushers
         self.store = self.hs.get_datastore()
         self.clock = self.hs.get_clock()
@@ -48,7 +49,7 @@ class PusherPool:
         # will then get pulled out of the database,
         # recreated, added and started: this means we have only one
         # code path adding pushers.
-        pusher.create_pusher(self.hs, {
+        self.pusher_factory.create_pusher({
             "id": None,
             "user_name": user_id,
             "kind": kind,
@@ -186,7 +187,7 @@ class PusherPool:
         logger.info("Starting %d pushers", len(pushers))
         for pusherdict in pushers:
             try:
-                p = pusher.create_pusher(self.hs, pusherdict)
+                p = self.pusher_factory.create_pusher(pusherdict)
             except:
                 logger.exception("Couldn't start a pusher: caught Exception")
                 continue

synapse/python_dependencies.py

@@ -31,7 +31,7 @@ REQUIREMENTS = {
     "pyyaml": ["yaml"],
     "pyasn1": ["pyasn1"],
     "daemonize": ["daemonize"],
-    "py-bcrypt": ["bcrypt"],
+    "bcrypt": ["bcrypt"],
     "pillow": ["PIL"],
     "pydenticon": ["pydenticon"],
     "ujson": ["ujson"],
@@ -40,6 +40,7 @@ REQUIREMENTS = {
     "pymacaroons-pynacl": ["pymacaroons"],
     "msgpack-python>=0.3.0": ["msgpack"],
     "phonenumbers>=8.2.0": ["phonenumbers"],
+    "affinity": ["affinity"],
 }
 CONDITIONAL_REQUIREMENTS = {
     "web_client": {

synapse/replication/slave/storage/appservice.py

@@ -16,6 +16,7 @@
 from ._base import BaseSlavedStore
 from synapse.storage import DataStore
 from synapse.config.appservice import load_appservices
+from synapse.storage.appservice import _make_exclusive_regex
 
 
 class SlavedApplicationServiceStore(BaseSlavedStore):
@@ -25,6 +26,7 @@ class SlavedApplicationServiceStore(BaseSlavedStore):
             hs.config.server_name,
             hs.config.app_service_config_files
         )
+        self.exclusive_user_regex = _make_exclusive_regex(self.services_cache)
 
     get_app_service_by_token = DataStore.get_app_service_by_token.__func__
     get_app_service_by_user_id = DataStore.get_app_service_by_user_id.__func__
@@ -38,3 +40,6 @@ class SlavedApplicationServiceStore(BaseSlavedStore):
     get_appservice_state = DataStore.get_appservice_state.__func__
     set_appservice_last_pos = DataStore.set_appservice_last_pos.__func__
     set_appservice_state = DataStore.set_appservice_state.__func__
+    get_if_app_services_interested_in_user = (
+        DataStore.get_if_app_services_interested_in_user.__func__
+    )

synapse/replication/slave/storage/client_ips.py

@@ -0,0 +1,47 @@
+# -*- coding: utf-8 -*-
+# Copyright 2017 Vector Creations Ltd
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+from ._base import BaseSlavedStore
+from synapse.storage.client_ips import LAST_SEEN_GRANULARITY
+from synapse.util.caches import CACHE_SIZE_FACTOR
+from synapse.util.caches.descriptors import Cache
+
+
+class SlavedClientIpStore(BaseSlavedStore):
+    def __init__(self, db_conn, hs):
+        super(SlavedClientIpStore, self).__init__(db_conn, hs)
+
+        self.client_ip_last_seen = Cache(
+            name="client_ip_last_seen",
+            keylen=4,
+            max_entries=50000 * CACHE_SIZE_FACTOR,
+        )
+
+    def insert_client_ip(self, user_id, access_token, ip, user_agent, device_id):
+        now = int(self._clock.time_msec())
+        key = (user_id, access_token, ip)
+
+        try:
+            last_seen = self.client_ip_last_seen.get(key)
+        except KeyError:
+            last_seen = None
+
+        # Rate-limited inserts
+        if last_seen is not None and (now - last_seen) < LAST_SEEN_GRANULARITY:
+            return
+
+        self.hs.get_tcp_replication().send_user_ip(
+            user_id, access_token, ip, user_agent, device_id, now
+        )

synapse/replication/slave/storage/devices.py

@@ -16,6 +16,7 @@
 from ._base import BaseSlavedStore
 from ._slaved_id_tracker import SlavedIdTracker
 from synapse.storage import DataStore
+from synapse.storage.end_to_end_keys import EndToEndKeyStore
 from synapse.util.caches.stream_change_cache import StreamChangeCache
 
 
@@ -45,6 +46,7 @@ class SlavedDeviceStore(BaseSlavedStore):
     _mark_as_sent_devices_by_remote_txn = (
         DataStore._mark_as_sent_devices_by_remote_txn.__func__
     )
+    count_e2e_one_time_keys = EndToEndKeyStore.__dict__["count_e2e_one_time_keys"]
 
     def stream_positions(self):
         result = super(SlavedDeviceStore, self).stream_positions()

synapse/replication/slave/storage/events.py

@@ -108,6 +108,8 @@ class SlavedEventStore(BaseSlavedStore):
     get_current_state_ids = (
         StateStore.__dict__["get_current_state_ids"]
     )
+    get_state_group_delta = StateStore.__dict__["get_state_group_delta"]
+    _get_joined_hosts_cache = RoomMemberStore.__dict__["_get_joined_hosts_cache"]
     has_room_changed_since = DataStore.has_room_changed_since.__func__
 
     get_unread_push_actions_for_user_in_range_for_http = (
@@ -151,8 +153,7 @@ class SlavedEventStore(BaseSlavedStore):
     get_room_events_stream_for_rooms = (
         DataStore.get_room_events_stream_for_rooms.__func__
     )
-    is_host_joined = DataStore.is_host_joined.__func__
-    _is_host_joined = RoomMemberStore.__dict__["_is_host_joined"]
+    is_host_joined = RoomMemberStore.__dict__["is_host_joined"]
     get_stream_token_for_event = DataStore.get_stream_token_for_event.__func__
 
     _set_before_and_after = staticmethod(DataStore._set_before_and_after)

synapse/replication/tcp/client.py

@@ -20,6 +20,7 @@ from twisted.internet.protocol import ReconnectingClientFactory
 
 from .commands import (
     FederationAckCommand, UserSyncCommand, RemovePusherCommand, InvalidateCacheCommand,
+    UserIpCommand,
 )
 from .protocol import ClientReplicationStreamProtocol
 
@@ -178,6 +179,12 @@ class ReplicationClientHandler(object):
         cmd = InvalidateCacheCommand(cache_func.__name__, keys)
         self.send_command(cmd)
 
+    def send_user_ip(self, user_id, access_token, ip, user_agent, device_id, last_seen):
+        """Tell the master that the user made a request.
+        """
+        cmd = UserIpCommand(user_id, access_token, ip, user_agent, device_id, last_seen)
+        self.send_command(cmd)
+
     def await_sync(self, data):
         """Returns a deferred that is resolved when we receive a SYNC command
         with given data.

synapse/replication/tcp/commands.py

@@ -304,6 +304,40 @@ class InvalidateCacheCommand(Command):
         return " ".join((self.cache_func, json.dumps(self.keys)))
 
 
+class UserIpCommand(Command):
+    """Sent periodically when a worker sees activity from a client.
+
+    Format::
+
+        USER_IP <user_id>, <access_token>, <ip>, <device_id>, <last_seen>, <user_agent>
+    """
+    NAME = "USER_IP"
+
+    def __init__(self, user_id, access_token, ip, user_agent, device_id, last_seen):
+        self.user_id = user_id
+        self.access_token = access_token
+        self.ip = ip
+        self.user_agent = user_agent
+        self.device_id = device_id
+        self.last_seen = last_seen
+
+    @classmethod
+    def from_line(cls, line):
+        user_id, jsn = line.split(" ", 1)
+
+        access_token, ip, user_agent, device_id, last_seen = json.loads(jsn)
+
+        return cls(
+            user_id, access_token, ip, user_agent, device_id, last_seen
+        )
+
+    def to_line(self):
+        return self.user_id + " " + json.dumps((
+            self.access_token, self.ip, self.user_agent, self.device_id,
+            self.last_seen,
+        ))
+
+
 # Map of command name to command type.
 COMMAND_MAP = {
     cmd.NAME: cmd
@@ -320,6 +354,7 @@ COMMAND_MAP = {
         SyncCommand,
         RemovePusherCommand,
         InvalidateCacheCommand,
+        UserIpCommand,
     )
 }
 
@@ -342,5 +377,6 @@ VALID_CLIENT_COMMANDS = (
     FederationAckCommand.NAME,
     RemovePusherCommand.NAME,
     InvalidateCacheCommand.NAME,
+    UserIpCommand.NAME,
     ErrorCommand.NAME,
 )

synapse/replication/tcp/protocol.py

@@ -244,7 +244,7 @@ class BaseReplicationStreamProtocol(LineOnlyReceiver):
                 becoming full.
         """
         if self.state == ConnectionStates.CLOSED:
-            logger.info("[%s] Not sending, connection closed", self.id())
+            logger.debug("[%s] Not sending, connection closed", self.id())
             return
 
         if do_buffer and self.state != ConnectionStates.ESTABLISHED:
@@ -264,7 +264,7 @@ class BaseReplicationStreamProtocol(LineOnlyReceiver):
     def _queue_command(self, cmd):
         """Queue the command until the connection is ready to write to again.
         """
-        logger.info("[%s] Queing as conn %r, cmd: %r", self.id(), self.state, cmd)
+        logger.debug("[%s] Queing as conn %r, cmd: %r", self.id(), self.state, cmd)
         self.pending_commands.append(cmd)
 
         if len(self.pending_commands) > self.max_line_buffer:
@@ -406,6 +406,12 @@ class ServerReplicationStreamProtocol(BaseReplicationStreamProtocol):
     def on_INVALIDATE_CACHE(self, cmd):
         self.streamer.on_invalidate_cache(cmd.cache_func, cmd.keys)
 
+    def on_USER_IP(self, cmd):
+        self.streamer.on_user_ip(
+            cmd.user_id, cmd.access_token, cmd.ip, cmd.user_agent, cmd.device_id,
+            cmd.last_seen,
+        )
+
     @defer.inlineCallbacks
     def subscribe_to_stream(self, stream_name, token):
         """Subscribe the remote to a streams.

synapse/replication/tcp/resource.py

@@ -35,6 +35,7 @@ user_sync_counter = metrics.register_counter("user_sync")
 federation_ack_counter = metrics.register_counter("federation_ack")
 remove_pusher_counter = metrics.register_counter("remove_pusher")
 invalidate_cache_counter = metrics.register_counter("invalidate_cache")
+user_ip_cache_counter = metrics.register_counter("user_ip_cache")
 
 logger = logging.getLogger(__name__)
 
@@ -67,6 +68,7 @@ class ReplicationStreamer(object):
         self.store = hs.get_datastore()
         self.presence_handler = hs.get_presence_handler()
         self.clock = hs.get_clock()
+        self.notifier = hs.get_notifier()
 
         # Current connections.
         self.connections = []
@@ -99,7 +101,7 @@ class ReplicationStreamer(object):
         if not hs.config.send_federation:
             self.federation_sender = hs.get_federation_sender()
 
-        hs.get_notifier().add_replication_callback(self.on_notifier_poke)
+        self.notifier.add_replication_callback(self.on_notifier_poke)
 
         # Keeps track of whether we are currently checking for updates
         self.is_looping = False
@@ -237,6 +239,15 @@ class ReplicationStreamer(object):
         invalidate_cache_counter.inc()
         getattr(self.store, cache_func).invalidate(tuple(keys))
 
+    @measure_func("repl.on_user_ip")
+    def on_user_ip(self, user_id, access_token, ip, user_agent, device_id, last_seen):
+        """The client saw a user request
+        """
+        user_ip_cache_counter.inc()
+        self.store.insert_client_ip(
+            user_id, access_token, ip, user_agent, device_id, last_seen,
+        )
+
     def send_sync_to_all_connections(self, data):
         """Sends a SYNC command to all clients.
 

synapse/replication/tcp/streams.py

@@ -112,6 +112,12 @@ AccountDataStreamRow = namedtuple("AccountDataStream", (
     "data_type",  # str
     "data",  # dict
 ))
+CurrentStateDeltaStreamRow = namedtuple("CurrentStateDeltaStream", (
+    "room_id",  # str
+    "type",  # str
+    "state_key",  # str
+    "event_id",  # str, optional
+))
 
 
 class Stream(object):
@@ -443,6 +449,21 @@ class AccountDataStream(Stream):
         defer.returnValue(results)
 
 
+class CurrentStateDeltaStream(Stream):
+    """Current state for a room was changed
+    """
+    NAME = "current_state_deltas"
+    ROW_TYPE = CurrentStateDeltaStreamRow
+
+    def __init__(self, hs):
+        store = hs.get_datastore()
+
+        self.current_token = store.get_max_current_state_delta_stream_id
+        self.update_function = store.get_all_updated_current_state_deltas
+
+        super(CurrentStateDeltaStream, self).__init__(hs)
+
+
 STREAMS_MAP = {
     stream.NAME: stream
     for stream in (
@@ -460,5 +481,6 @@ STREAMS_MAP = {
         FederationStream,
         TagAccountDataStream,
         AccountDataStream,
+        CurrentStateDeltaStream,
     )
 }

synapse/rest/__init__.py

@@ -51,6 +51,7 @@ from synapse.rest.client.v2_alpha import (
     devices,
     thirdparty,
     sendtodevice,
+    user_directory,
 )
 
 from synapse.http.server import JsonResource
@@ -100,3 +101,4 @@ class ClientRestResource(JsonResource):
         devices.register_servlets(hs, client_resource)
         thirdparty.register_servlets(hs, client_resource)
         sendtodevice.register_servlets(hs, client_resource)
+        user_directory.register_servlets(hs, client_resource)

synapse/rest/client/v1/admin.py

@@ -15,8 +15,9 @@
 
 from twisted.internet import defer
 
+from synapse.api.constants import Membership
 from synapse.api.errors import AuthError, SynapseError
-from synapse.types import UserID
+from synapse.types import UserID, create_requester
 from synapse.http.servlet import parse_json_object_from_request
 
 from .base import ClientV1RestServlet, client_path_patterns
@@ -157,9 +158,145 @@ class DeactivateAccountRestServlet(ClientV1RestServlet):
         defer.returnValue((200, {}))
 
 
+class ShutdownRoomRestServlet(ClientV1RestServlet):
+    """Shuts down a room by removing all local users from the room and blocking
+    all future invites and joins to the room. Any local aliases will be repointed
+    to a new room created by `new_room_user_id` and kicked users will be auto
+    joined to the new room.
+    """
+    PATTERNS = client_path_patterns("/admin/shutdown_room/(?P<room_id>[^/]+)")
+
+    DEFAULT_MESSAGE = (
+        "Sharing illegal content on this server is not permitted and rooms in"
+        " violation will be blocked."
+    )
+
+    def __init__(self, hs):
+        super(ShutdownRoomRestServlet, self).__init__(hs)
+        self.store = hs.get_datastore()
+        self.handlers = hs.get_handlers()
+        self.state = hs.get_state_handler()
+
+    @defer.inlineCallbacks
+    def on_POST(self, request, room_id):
+        requester = yield self.auth.get_user_by_req(request)
+        is_admin = yield self.auth.is_server_admin(requester.user)
+        if not is_admin:
+            raise AuthError(403, "You are not a server admin")
+
+        content = parse_json_object_from_request(request)
+
+        new_room_user_id = content.get("new_room_user_id")
+        if not new_room_user_id:
+            raise SynapseError(400, "Please provide field `new_room_user_id`")
+
+        room_creator_requester = create_requester(new_room_user_id)
+
+        message = content.get("message", self.DEFAULT_MESSAGE)
+        room_name = content.get("room_name", "Content Violation Notification")
+
+        info = yield self.handlers.room_creation_handler.create_room(
+            room_creator_requester,
+            config={
+                "preset": "public_chat",
+                "name": room_name,
+                "power_level_content_override": {
+                    "users_default": -10,
+                },
+            },
+            ratelimit=False,
+        )
+        new_room_id = info["room_id"]
+
+        msg_handler = self.handlers.message_handler
+        yield msg_handler.create_and_send_nonmember_event(
+            room_creator_requester,
+            {
+                "type": "m.room.message",
+                "content": {"body": message, "msgtype": "m.text"},
+                "room_id": new_room_id,
+                "sender": new_room_user_id,
+            },
+            ratelimit=False,
+        )
+
+        requester_user_id = requester.user.to_string()
+
+        logger.info("Shutting down room %r", room_id)
+
+        yield self.store.block_room(room_id, requester_user_id)
+
+        users = yield self.state.get_current_user_in_room(room_id)
+        kicked_users = []
+        for user_id in users:
+            if not self.hs.is_mine_id(user_id):
+                continue
+
+            logger.info("Kicking %r from %r...", user_id, room_id)
+
+            target_requester = create_requester(user_id)
+            yield self.handlers.room_member_handler.update_membership(
+                requester=target_requester,
+                target=target_requester.user,
+                room_id=room_id,
+                action=Membership.LEAVE,
+                content={},
+                ratelimit=False
+            )
+
+            yield self.handlers.room_member_handler.forget(target_requester.user, room_id)
+
+            yield self.handlers.room_member_handler.update_membership(
+                requester=target_requester,
+                target=target_requester.user,
+                room_id=new_room_id,
+                action=Membership.JOIN,
+                content={},
+                ratelimit=False
+            )
+
+            kicked_users.append(user_id)
+
+        aliases_for_room = yield self.store.get_aliases_for_room(room_id)
+
+        yield self.store.update_aliases_for_room(
+            room_id, new_room_id, requester_user_id
+        )
+
+        defer.returnValue((200, {
+            "kicked_users": kicked_users,
+            "local_aliases": aliases_for_room,
+            "new_room_id": new_room_id,
+        }))
+
+
+class QuarantineMediaInRoom(ClientV1RestServlet):
+    """Quarantines all media in a room so that no one can download it via
+    this server.
+    """
+    PATTERNS = client_path_patterns("/admin/quarantine_media/(?P<room_id>[^/]+)")
+
+    def __init__(self, hs):
+        super(QuarantineMediaInRoom, self).__init__(hs)
+        self.store = hs.get_datastore()
+
+    @defer.inlineCallbacks
+    def on_POST(self, request, room_id):
+        requester = yield self.auth.get_user_by_req(request)
+        is_admin = yield self.auth.is_server_admin(requester.user)
+        if not is_admin:
+            raise AuthError(403, "You are not a server admin")
+
+        num_quarantined = yield self.store.quarantine_media_ids_in_room(
+            room_id, requester.user.to_string(),
+        )
+
+        defer.returnValue((200, {"num_quarantined": num_quarantined}))
+
+
 class ResetPasswordRestServlet(ClientV1RestServlet):
     """Post request to allow an administrator reset password for a user.
-    This need a user have a administrator access in Synapse.
+    This needs user to have administrator access in Synapse.
         Example:
             http://localhost:8008/_matrix/client/api/v1/admin/reset_password/
             @user:to_reset_password?access_token=admin_access_token
@@ -182,7 +319,7 @@ class ResetPasswordRestServlet(ClientV1RestServlet):
     @defer.inlineCallbacks
     def on_POST(self, request, target_user_id):
         """Post request to allow an administrator reset password for a user.
-        This need a user have a administrator access in Synapse.
+        This needs user to have administrator access in Synapse.
         """
         UserID.from_string(target_user_id)
         requester = yield self.auth.get_user_by_req(request)
@@ -206,7 +343,7 @@ class ResetPasswordRestServlet(ClientV1RestServlet):
 
 class GetUsersPaginatedRestServlet(ClientV1RestServlet):
     """Get request to get specific number of users from Synapse.
-    This need a user have a administrator access in Synapse.
+    This needs user to have administrator access in Synapse.
         Example:
             http://localhost:8008/_matrix/client/api/v1/admin/users_paginate/
             @admin:user?access_token=admin_access_token&start=0&limit=10
@@ -225,7 +362,7 @@ class GetUsersPaginatedRestServlet(ClientV1RestServlet):
     @defer.inlineCallbacks
     def on_GET(self, request, target_user_id):
         """Get request to get specific number of users from Synapse.
-        This need a user have a administrator access in Synapse.
+        This needs user to have administrator access in Synapse.
         """
         target_user = UserID.from_string(target_user_id)
         requester = yield self.auth.get_user_by_req(request)
@@ -258,7 +395,7 @@ class GetUsersPaginatedRestServlet(ClientV1RestServlet):
     @defer.inlineCallbacks
     def on_POST(self, request, target_user_id):
         """Post request to get specific number of users from Synapse..
-        This need a user have a administrator access in Synapse.
+        This needs user to have administrator access in Synapse.
         Example:
             http://localhost:8008/_matrix/client/api/v1/admin/users_paginate/
             @admin:user?access_token=admin_access_token
@@ -296,7 +433,7 @@ class GetUsersPaginatedRestServlet(ClientV1RestServlet):
 class SearchUsersRestServlet(ClientV1RestServlet):
     """Get request to search user table for specific users according to
     search term.
-    This need a user have a administrator access in Synapse.
+    This needs user to have administrator access in Synapse.
         Example:
             http://localhost:8008/_matrix/client/api/v1/admin/search_users/
             @admin:user?access_token=admin_access_token&term=alice
@@ -316,7 +453,7 @@ class SearchUsersRestServlet(ClientV1RestServlet):
     def on_GET(self, request, target_user_id):
         """Get request to search user table for specific users according to
         search term.
-        This need a user have a administrator access in Synapse.
+        This needs user to have a administrator access in Synapse.
         """
         target_user = UserID.from_string(target_user_id)
         requester = yield self.auth.get_user_by_req(request)
@@ -353,3 +490,5 @@ def register_servlets(hs, http_server):
     ResetPasswordRestServlet(hs).register(http_server)
     GetUsersPaginatedRestServlet(hs).register(http_server)
     SearchUsersRestServlet(hs).register(http_server)
+    ShutdownRoomRestServlet(hs).register(http_server)
+    QuarantineMediaInRoom(hs).register(http_server)

synapse/rest/client/v1/pusher.py

@@ -73,6 +73,7 @@ class PushersSetRestServlet(ClientV1RestServlet):
     def __init__(self, hs):
         super(PushersSetRestServlet, self).__init__(hs)
         self.notifier = hs.get_notifier()
+        self.pusher_pool = self.hs.get_pusherpool()
 
     @defer.inlineCallbacks
     def on_POST(self, request):
@@ -81,12 +82,10 @@ class PushersSetRestServlet(ClientV1RestServlet):
 
         content = parse_json_object_from_request(request)
 
-        pusher_pool = self.hs.get_pusherpool()
-
         if ('pushkey' in content and 'app_id' in content
                 and 'kind' in content and
                 content['kind'] is None):
-            yield pusher_pool.remove_pusher(
+            yield self.pusher_pool.remove_pusher(
                 content['app_id'], content['pushkey'], user_id=user.to_string()
             )
             defer.returnValue((200, {}))
@@ -109,14 +108,14 @@ class PushersSetRestServlet(ClientV1RestServlet):
             append = content['append']
 
         if not append:
-            yield pusher_pool.remove_pushers_by_app_id_and_pushkey_not_user(
+            yield self.pusher_pool.remove_pushers_by_app_id_and_pushkey_not_user(
                 app_id=content['app_id'],
                 pushkey=content['pushkey'],
                 not_user_id=user.to_string()
             )
 
         try:
-            yield pusher_pool.add_pusher(
+            yield self.pusher_pool.add_pusher(
                 user_id=user.to_string(),
                 access_token=requester.access_token_id,
                 kind=content['kind'],
@@ -152,6 +151,7 @@ class PushersRemoveRestServlet(RestServlet):
         self.hs = hs
         self.notifier = hs.get_notifier()
         self.auth = hs.get_v1auth()
+        self.pusher_pool = self.hs.get_pusherpool()
 
     @defer.inlineCallbacks
     def on_GET(self, request):
@@ -161,10 +161,8 @@ class PushersRemoveRestServlet(RestServlet):
         app_id = parse_string(request, "app_id", required=True)
         pushkey = parse_string(request, "pushkey", required=True)
 
-        pusher_pool = self.hs.get_pusherpool()
-
         try:
-            yield pusher_pool.remove_pusher(
+            yield self.pusher_pool.remove_pusher(
                 app_id=app_id,
                 pushkey=pushkey,
                 user_id=user.to_string(),

synapse/rest/client/v1/room.py

@@ -398,22 +398,18 @@ class JoinedRoomMemberListRestServlet(ClientV1RestServlet):
 
     def __init__(self, hs):
         super(JoinedRoomMemberListRestServlet, self).__init__(hs)
-        self.state = hs.get_state_handler()
+        self.message_handler = hs.get_handlers().message_handler
 
     @defer.inlineCallbacks
     def on_GET(self, request, room_id):
-        yield self.auth.get_user_by_req(request)
+        requester = yield self.auth.get_user_by_req(request)
 
-        users_with_profile = yield self.state.get_current_user_in_room(room_id)
+        users_with_profile = yield self.message_handler.get_joined_members(
+            requester, room_id,
+        )
 
         defer.returnValue((200, {
-            "joined": {
-                user_id: {
-                    "avatar_url": profile.avatar_url,
-                    "display_name": profile.display_name,
-                }
-                for user_id, profile in users_with_profile.iteritems()
-            }
+            "joined": users_with_profile,
         }))
 
 

synapse/rest/client/v2_alpha/keys.py

@@ -188,13 +188,11 @@ class KeyChangesServlet(RestServlet):
 
         user_id = requester.user.to_string()
 
-        changed = yield self.device_handler.get_user_ids_changed(
+        results = yield self.device_handler.get_user_ids_changed(
             user_id, from_token,
         )
 
-        defer.returnValue((200, {
-            "changed": list(changed),
-        }))
+        defer.returnValue((200, results))
 
 
 class OneTimeKeyServlet(RestServlet):

synapse/rest/client/v2_alpha/sync.py

@@ -110,7 +110,7 @@ class SyncRestServlet(RestServlet):
         filter_id = parse_string(request, "filter", default=None)
         full_state = parse_boolean(request, "full_state", default=False)
 
-        logger.info(
+        logger.debug(
             "/sync: user=%r, timeout=%r, since=%r,"
             " set_presence=%r, filter_id=%r, device_id=%r" % (
                 user, timeout, since, set_presence, filter_id, device_id
@@ -164,27 +164,35 @@ class SyncRestServlet(RestServlet):
             )
 
         time_now = self.clock.time_msec()
-
-        joined = self.encode_joined(
-            sync_result.joined, time_now, requester.access_token_id, filter.event_fields
+        response_content = self.encode_response(
+            time_now, sync_result, requester.access_token_id, filter
         )
 
-        invited = self.encode_invited(
-            sync_result.invited, time_now, requester.access_token_id
+        defer.returnValue((200, response_content))
+
+    @staticmethod
+    def encode_response(time_now, sync_result, access_token_id, filter):
+        joined = SyncRestServlet.encode_joined(
+            sync_result.joined, time_now, access_token_id, filter.event_fields
         )
 
-        archived = self.encode_archived(
-            sync_result.archived, time_now, requester.access_token_id,
+        invited = SyncRestServlet.encode_invited(
+            sync_result.invited, time_now, access_token_id,
+        )
+
+        archived = SyncRestServlet.encode_archived(
+            sync_result.archived, time_now, access_token_id,
             filter.event_fields,
         )
 
-        response_content = {
+        return {
             "account_data": {"events": sync_result.account_data},
             "to_device": {"events": sync_result.to_device},
             "device_lists": {
-                "changed": list(sync_result.device_lists),
+                "changed": list(sync_result.device_lists.changed),
+                "left": list(sync_result.device_lists.left),
             },
-            "presence": self.encode_presence(
+            "presence": SyncRestServlet.encode_presence(
                 sync_result.presence, time_now
             ),
             "rooms": {
@@ -192,12 +200,12 @@ class SyncRestServlet(RestServlet):
                 "invite": invited,
                 "leave": archived,
             },
+            "device_one_time_keys_count": sync_result.device_one_time_keys_count,
             "next_batch": sync_result.next_batch.to_string(),
         }
 
-        defer.returnValue((200, response_content))
-
-    def encode_presence(self, events, time_now):
+    @staticmethod
+    def encode_presence(events, time_now):
         return {
             "events": [
                 {
@@ -211,7 +219,8 @@ class SyncRestServlet(RestServlet):
             ]
         }
 
-    def encode_joined(self, rooms, time_now, token_id, event_fields):
+    @staticmethod
+    def encode_joined(rooms, time_now, token_id, event_fields):
         """
         Encode the joined rooms in a sync result
 
@@ -230,13 +239,14 @@ class SyncRestServlet(RestServlet):
         """
         joined = {}
         for room in rooms:
-            joined[room.room_id] = self.encode_room(
+            joined[room.room_id] = SyncRestServlet.encode_room(
                 room, time_now, token_id, only_fields=event_fields
             )
 
         return joined
 
-    def encode_invited(self, rooms, time_now, token_id):
+    @staticmethod
+    def encode_invited(rooms, time_now, token_id):
         """
         Encode the invited rooms in a sync result
 
@@ -269,7 +279,8 @@ class SyncRestServlet(RestServlet):
 
         return invited
 
-    def encode_archived(self, rooms, time_now, token_id, event_fields):
+    @staticmethod
+    def encode_archived(rooms, time_now, token_id, event_fields):
         """
         Encode the archived rooms in a sync result
 
@@ -288,7 +299,7 @@ class SyncRestServlet(RestServlet):
         """
         joined = {}
         for room in rooms:
-            joined[room.room_id] = self.encode_room(
+            joined[room.room_id] = SyncRestServlet.encode_room(
                 room, time_now, token_id, joined=False, only_fields=event_fields
             )
 

synapse/rest/client/v2_alpha/user_directory.py

@@ -0,0 +1,79 @@
+# -*- coding: utf-8 -*-
+# Copyright 2017 Vector Creations Ltd
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+import logging
+
+from twisted.internet import defer
+
+from synapse.api.errors import SynapseError
+from synapse.http.servlet import RestServlet, parse_json_object_from_request
+from ._base import client_v2_patterns
+
+logger = logging.getLogger(__name__)
+
+
+class UserDirectorySearchRestServlet(RestServlet):
+    PATTERNS = client_v2_patterns("/user_directory/search$")
+
+    def __init__(self, hs):
+        """
+        Args:
+            hs (synapse.server.HomeServer): server
+        """
+        super(UserDirectorySearchRestServlet, self).__init__()
+        self.hs = hs
+        self.auth = hs.get_auth()
+        self.user_directory_handler = hs.get_user_directory_handler()
+
+    @defer.inlineCallbacks
+    def on_POST(self, request):
+        """Searches for users in directory
+
+        Returns:
+            dict of the form::
+
+                {
+                    "limited": <bool>,  # whether there were more results or not
+                    "results": [  # Ordered by best match first
+                        {
+                            "user_id": <user_id>,
+                            "display_name": <display_name>,
+                            "avatar_url": <avatar_url>
+                        }
+                    ]
+                }
+        """
+        requester = yield self.auth.get_user_by_req(request, allow_guest=False)
+        user_id = requester.user.to_string()
+
+        body = parse_json_object_from_request(request)
+
+        limit = body.get("limit", 10)
+        limit = min(limit, 50)
+
+        try:
+            search_term = body["search_term"]
+        except:
+            raise SynapseError(400, "`search_term` is required field")
+
+        results = yield self.user_directory_handler.search_users(
+            user_id, search_term, limit,
+        )
+
+        defer.returnValue((200, results))
+
+
+def register_servlets(hs, http_server):
+    UserDirectorySearchRestServlet(hs).register(http_server)

synapse/rest/media/v1/download_resource.py

@@ -66,14 +66,19 @@ class DownloadResource(Resource):
     @defer.inlineCallbacks
     def _respond_local_file(self, request, media_id, name):
         media_info = yield self.store.get_local_media(media_id)
-        if not media_info:
+        if not media_info or media_info["quarantined_by"]:
             respond_404(request)
             return
 
         media_type = media_info["media_type"]
         media_length = media_info["media_length"]
         upload_name = name if name else media_info["upload_name"]
-        file_path = self.filepaths.local_media_filepath(media_id)
+        if media_info["url_cache"]:
+            # TODO: Check the file still exists, if it doesn't we can redownload
+            # it from the url `media_info["url_cache"]`
+            file_path = self.filepaths.url_cache_filepath(media_id)
+        else:
+            file_path = self.filepaths.local_media_filepath(media_id)
 
         yield respond_with_file(
             request, media_type, file_path, media_length,

synapse/rest/media/v1/filepath.py

@@ -14,6 +14,9 @@
 # limitations under the License.
 
 import os
+import re
+
+NEW_FORMAT_ID_RE = re.compile(r"^\d\d\d\d-\d\d-\d\d")
 
 
 class MediaFilePaths(object):
@@ -71,3 +74,107 @@ class MediaFilePaths(object):
             self.base_path, "remote_thumbnail", server_name,
             file_id[0:2], file_id[2:4], file_id[4:],
         )
+
+    def url_cache_filepath(self, media_id):
+        if NEW_FORMAT_ID_RE.match(media_id):
+            # Media id is of the form <DATE><RANDOM_STRING>
+            # E.g.: 2017-09-28-fsdRDt24DS234dsf
+            return os.path.join(
+                self.base_path, "url_cache",
+                media_id[:10], media_id[11:]
+            )
+        else:
+            return os.path.join(
+                self.base_path, "url_cache",
+                media_id[0:2], media_id[2:4], media_id[4:],
+            )
+
+    def url_cache_filepath_dirs_to_delete(self, media_id):
+        "The dirs to try and remove if we delete the media_id file"
+        if NEW_FORMAT_ID_RE.match(media_id):
+            return [
+                os.path.join(
+                    self.base_path, "url_cache",
+                    media_id[:10],
+                ),
+            ]
+        else:
+            return [
+                os.path.join(
+                    self.base_path, "url_cache",
+                    media_id[0:2], media_id[2:4],
+                ),
+                os.path.join(
+                    self.base_path, "url_cache",
+                    media_id[0:2],
+                ),
+            ]
+
+    def url_cache_thumbnail(self, media_id, width, height, content_type,
+                            method):
+        # Media id is of the form <DATE><RANDOM_STRING>
+        # E.g.: 2017-09-28-fsdRDt24DS234dsf
+
+        top_level_type, sub_type = content_type.split("/")
+        file_name = "%i-%i-%s-%s-%s" % (
+            width, height, top_level_type, sub_type, method
+        )
+
+        if NEW_FORMAT_ID_RE.match(media_id):
+            return os.path.join(
+                self.base_path, "url_cache_thumbnails",
+                media_id[:10], media_id[11:],
+                file_name
+            )
+        else:
+            return os.path.join(
+                self.base_path, "url_cache_thumbnails",
+                media_id[0:2], media_id[2:4], media_id[4:],
+                file_name
+            )
+
+    def url_cache_thumbnail_directory(self, media_id):
+        # Media id is of the form <DATE><RANDOM_STRING>
+        # E.g.: 2017-09-28-fsdRDt24DS234dsf
+
+        if NEW_FORMAT_ID_RE.match(media_id):
+            return os.path.join(
+                self.base_path, "url_cache_thumbnails",
+                media_id[:10], media_id[11:],
+            )
+        else:
+            return os.path.join(
+                self.base_path, "url_cache_thumbnails",
+                media_id[0:2], media_id[2:4], media_id[4:],
+            )
+
+    def url_cache_thumbnail_dirs_to_delete(self, media_id):
+        "The dirs to try and remove if we delete the media_id thumbnails"
+        # Media id is of the form <DATE><RANDOM_STRING>
+        # E.g.: 2017-09-28-fsdRDt24DS234dsf
+        if NEW_FORMAT_ID_RE.match(media_id):
+            return [
+                os.path.join(
+                    self.base_path, "url_cache_thumbnails",
+                    media_id[:10], media_id[11:],
+                ),
+                os.path.join(
+                    self.base_path, "url_cache_thumbnails",
+                    media_id[:10],
+                ),
+            ]
+        else:
+            return [
+                os.path.join(
+                    self.base_path, "url_cache_thumbnails",
+                    media_id[0:2], media_id[2:4], media_id[4:],
+                ),
+                os.path.join(
+                    self.base_path, "url_cache_thumbnails",
+                    media_id[0:2], media_id[2:4],
+                ),
+                os.path.join(
+                    self.base_path, "url_cache_thumbnails",
+                    media_id[0:2],
+                ),
+            ]

synapse/rest/media/v1/media_repository.py

@@ -135,6 +135,8 @@ class MediaRepository(object):
             media_info = yield self._download_remote_file(
                 server_name, media_id
             )
+        elif media_info["quarantined_by"]:
+            raise NotFoundError()
         else:
             self.recently_accessed_remotes.add((server_name, media_id))
             yield self.store.update_cached_last_access_time(
@@ -184,6 +186,7 @@ class MediaRepository(object):
                     raise
                 except NotRetryingDestination:
                     logger.warn("Not retrying destination %r", server_name)
+                    raise SynapseError(502, "Failed to fetch remote media")
                 except Exception:
                     logger.exception("Failed to fetch remote media %s/%s",
                                      server_name, media_id)
@@ -323,13 +326,17 @@ class MediaRepository(object):
             defer.returnValue(t_path)
 
     @defer.inlineCallbacks
-    def _generate_local_thumbnails(self, media_id, media_info):
+    def _generate_local_thumbnails(self, media_id, media_info, url_cache=False):
         media_type = media_info["media_type"]
         requirements = self._get_thumbnail_requirements(media_type)
         if not requirements:
             return
 
-        input_path = self.filepaths.local_media_filepath(media_id)
+        if url_cache:
+            input_path = self.filepaths.url_cache_filepath(media_id)
+        else:
+            input_path = self.filepaths.local_media_filepath(media_id)
+
         thumbnailer = Thumbnailer(input_path)
         m_width = thumbnailer.width
         m_height = thumbnailer.height
@@ -357,9 +364,14 @@ class MediaRepository(object):
 
             for t_width, t_height, t_type in scales:
                 t_method = "scale"
-                t_path = self.filepaths.local_media_thumbnail(
-                    media_id, t_width, t_height, t_type, t_method
-                )
+                if url_cache:
+                    t_path = self.filepaths.url_cache_thumbnail(
+                        media_id, t_width, t_height, t_type, t_method
+                    )
+                else:
+                    t_path = self.filepaths.local_media_thumbnail(
+                        media_id, t_width, t_height, t_type, t_method
+                    )
                 self._makedirs(t_path)
                 t_len = thumbnailer.scale(t_path, t_width, t_height, t_type)
 
@@ -374,9 +386,14 @@ class MediaRepository(object):
                     # thumbnail.
                     continue
                 t_method = "crop"
-                t_path = self.filepaths.local_media_thumbnail(
-                    media_id, t_width, t_height, t_type, t_method
-                )
+                if url_cache:
+                    t_path = self.filepaths.url_cache_thumbnail(
+                        media_id, t_width, t_height, t_type, t_method
+                    )
+                else:
+                    t_path = self.filepaths.local_media_thumbnail(
+                        media_id, t_width, t_height, t_type, t_method
+                    )
                 self._makedirs(t_path)
                 t_len = thumbnailer.crop(t_path, t_width, t_height, t_type)
                 local_thumbnails.append((

synapse/rest/media/v1/preview_url_resource.py

@@ -36,6 +36,9 @@ import cgi
 import ujson as json
 import urlparse
 import itertools
+import datetime
+import errno
+import shutil
 
 import logging
 logger = logging.getLogger(__name__)
@@ -70,6 +73,10 @@ class PreviewUrlResource(Resource):
 
         self.downloads = {}
 
+        self._cleaner_loop = self.clock.looping_call(
+            self._expire_url_cache_data, 10 * 1000
+        )
+
     def render_GET(self, request):
         self._async_render_GET(request)
         return NOT_DONE_YET
@@ -130,7 +137,7 @@ class PreviewUrlResource(Resource):
         cache_result = yield self.store.get_url_cache(url, ts)
         if (
             cache_result and
-            cache_result["download_ts"] + cache_result["expires"] > ts and
+            cache_result["expires_ts"] > ts and
             cache_result["response_code"] / 100 == 2
         ):
             respond_with_json_bytes(
@@ -164,7 +171,7 @@ class PreviewUrlResource(Resource):
 
         if _is_media(media_info['media_type']):
             dims = yield self.media_repo._generate_local_thumbnails(
-                media_info['filesystem_id'], media_info
+                media_info['filesystem_id'], media_info, url_cache=True,
             )
 
             og = {
@@ -210,7 +217,7 @@ class PreviewUrlResource(Resource):
                 if _is_media(image_info['media_type']):
                     # TODO: make sure we don't choke on white-on-transparent images
                     dims = yield self.media_repo._generate_local_thumbnails(
-                        image_info['filesystem_id'], image_info
+                        image_info['filesystem_id'], image_info, url_cache=True,
                     )
                     if dims:
                         og["og:image:width"] = dims['width']
@@ -239,7 +246,7 @@ class PreviewUrlResource(Resource):
             url,
             media_info["response_code"],
             media_info["etag"],
-            media_info["expires"],
+            media_info["expires"] + media_info["created_ts"],
             json.dumps(og),
             media_info["filesystem_id"],
             media_info["created_ts"],
@@ -253,10 +260,9 @@ class PreviewUrlResource(Resource):
         # we're most likely being explicitly triggered by a human rather than a
         # bot, so are we really a robot?
 
-        # XXX: horrible duplication with base_resource's _download_remote_file()
-        file_id = random_string(24)
+        file_id = datetime.date.today().isoformat() + '_' + random_string(16)
 
-        fname = self.filepaths.local_media_filepath(file_id)
+        fname = self.filepaths.url_cache_filepath(file_id)
         self.media_repo._makedirs(fname)
 
         try:
@@ -303,6 +309,7 @@ class PreviewUrlResource(Resource):
                 upload_name=download_name,
                 media_length=length,
                 user_id=user,
+                url_cache=url,
             )
 
         except Exception as e:
@@ -327,6 +334,88 @@ class PreviewUrlResource(Resource):
             "etag": headers["ETag"][0] if "ETag" in headers else None,
         })
 
+    @defer.inlineCallbacks
+    def _expire_url_cache_data(self):
+        """Clean up expired url cache content, media and thumbnails.
+        """
+        now = self.clock.time_msec()
+
+        # First we delete expired url cache entries
+        media_ids = yield self.store.get_expired_url_cache(now)
+
+        removed_media = []
+        for media_id in media_ids:
+            fname = self.filepaths.url_cache_filepath(media_id)
+            try:
+                os.remove(fname)
+            except OSError as e:
+                # If the path doesn't exist, meh
+                if e.errno != errno.ENOENT:
+                    logger.warn("Failed to remove media: %r: %s", media_id, e)
+                    continue
+
+            removed_media.append(media_id)
+
+            try:
+                dirs = self.filepaths.url_cache_filepath_dirs_to_delete(media_id)
+                for dir in dirs:
+                    os.rmdir(dir)
+            except:
+                pass
+
+        yield self.store.delete_url_cache(removed_media)
+
+        if removed_media:
+            logger.info("Deleted %d entries from url cache", len(removed_media))
+
+        # Now we delete old images associated with the url cache.
+        # These may be cached for a bit on the client (i.e., they
+        # may have a room open with a preview url thing open).
+        # So we wait a couple of days before deleting, just in case.
+        expire_before = now - 2 * 24 * 60 * 60 * 1000
+        media_ids = yield self.store.get_url_cache_media_before(expire_before)
+
+        removed_media = []
+        for media_id in media_ids:
+            fname = self.filepaths.url_cache_filepath(media_id)
+            try:
+                os.remove(fname)
+            except OSError as e:
+                # If the path doesn't exist, meh
+                if e.errno != errno.ENOENT:
+                    logger.warn("Failed to remove media: %r: %s", media_id, e)
+                    continue
+
+            try:
+                dirs = self.filepaths.url_cache_filepath_dirs_to_delete(media_id)
+                for dir in dirs:
+                    os.rmdir(dir)
+            except:
+                pass
+
+            thumbnail_dir = self.filepaths.url_cache_thumbnail_directory(media_id)
+            try:
+                shutil.rmtree(thumbnail_dir)
+            except OSError as e:
+                # If the path doesn't exist, meh
+                if e.errno != errno.ENOENT:
+                    logger.warn("Failed to remove media: %r: %s", media_id, e)
+                    continue
+
+            removed_media.append(media_id)
+
+            try:
+                dirs = self.filepaths.url_cache_thumbnail_dirs_to_delete(media_id)
+                for dir in dirs:
+                    os.rmdir(dir)
+            except:
+                pass
+
+        yield self.store.delete_url_cache_media(removed_media)
+
+        if removed_media:
+            logger.info("Deleted %d media from url cache", len(removed_media))
+
 
 def decode_and_calc_og(body, media_uri, request_encoding=None):
     from lxml import etree
@@ -434,6 +523,8 @@ def _calc_og(tree, media_uri):
                 for el in _iterate_over_text(tree.find("body"), *TAGS_TO_REMOVE)
             )
             og['og:description'] = summarize_paragraphs(text_nodes)
+    else:
+        og['og:description'] = summarize_paragraphs([og['og:description']])
 
     # TODO: delete the url downloads to stop diskfilling,
     # as we only ever cared about its OG

synapse/rest/media/v1/thumbnail_resource.py

@@ -81,7 +81,7 @@ class ThumbnailResource(Resource):
                                  method, m_type):
         media_info = yield self.store.get_local_media(media_id)
 
-        if not media_info:
+        if not media_info or media_info["quarantined_by"]:
             respond_404(request)
             return
 
@@ -101,9 +101,16 @@ class ThumbnailResource(Resource):
             t_type = thumbnail_info["thumbnail_type"]
             t_method = thumbnail_info["thumbnail_method"]
 
-            file_path = self.filepaths.local_media_thumbnail(
-                media_id, t_width, t_height, t_type, t_method,
-            )
+            if media_info["url_cache"]:
+                # TODO: Check the file still exists, if it doesn't we can redownload
+                # it from the url `media_info["url_cache"]`
+                file_path = self.filepaths.url_cache_thumbnail(
+                    media_id, t_width, t_height, t_type, t_method,
+                )
+            else:
+                file_path = self.filepaths.local_media_thumbnail(
+                    media_id, t_width, t_height, t_type, t_method,
+                )
             yield respond_with_file(request, t_type, file_path)
 
         else:
@@ -117,7 +124,7 @@ class ThumbnailResource(Resource):
                                             desired_type):
         media_info = yield self.store.get_local_media(media_id)
 
-        if not media_info:
+        if not media_info or media_info["quarantined_by"]:
             respond_404(request)
             return
 
@@ -134,9 +141,18 @@ class ThumbnailResource(Resource):
             t_type = info["thumbnail_type"] == desired_type
 
             if t_w and t_h and t_method and t_type:
-                file_path = self.filepaths.local_media_thumbnail(
-                    media_id, desired_width, desired_height, desired_type, desired_method,
-                )
+                if media_info["url_cache"]:
+                    # TODO: Check the file still exists, if it doesn't we can redownload
+                    # it from the url `media_info["url_cache"]`
+                    file_path = self.filepaths.url_cache_thumbnail(
+                        media_id, desired_width, desired_height, desired_type,
+                        desired_method,
+                    )
+                else:
+                    file_path = self.filepaths.local_media_thumbnail(
+                        media_id, desired_width, desired_height, desired_type,
+                        desired_method,
+                    )
                 yield respond_with_file(request, desired_type, file_path)
                 return
 

synapse/server.py

@@ -31,6 +31,7 @@ from synapse.appservice.api import ApplicationServiceApi
 from synapse.appservice.scheduler import ApplicationServiceScheduler
 from synapse.crypto.keyring import Keyring
 from synapse.events.builder import EventBuilderFactory
+from synapse.events.spamcheck import SpamChecker
 from synapse.federation import initialize_http_replication
 from synapse.federation.send_queue import FederationRemoteSendQueue
 from synapse.federation.transport.client import TransportLayerClient
@@ -49,9 +50,11 @@ from synapse.handlers.events import EventHandler, EventStreamHandler
 from synapse.handlers.initial_sync import InitialSyncHandler
 from synapse.handlers.receipts import ReceiptsHandler
 from synapse.handlers.read_marker import ReadMarkerHandler
+from synapse.handlers.user_directory import UserDirectoyHandler
 from synapse.http.client import SimpleHttpClient, InsecureInterceptableContextFactory
 from synapse.http.matrixfederationclient import MatrixFederationHttpClient
 from synapse.notifier import Notifier
+from synapse.push.action_generator import ActionGenerator
 from synapse.push.pusherpool import PusherPool
 from synapse.rest.media.v1.media_repository import MediaRepository
 from synapse.state import StateHandler
@@ -135,6 +138,9 @@ class HomeServer(object):
         'macaroon_generator',
         'tcp_replication',
         'read_marker_handler',
+        'action_generator',
+        'user_directory_handler',
+        'spam_checker',
     ]
 
     def __init__(self, hostname, **kwargs):
@@ -299,6 +305,15 @@ class HomeServer(object):
     def build_tcp_replication(self):
         raise NotImplementedError()
 
+    def build_action_generator(self):
+        return ActionGenerator(self)
+
+    def build_user_directory_handler(self):
+        return UserDirectoyHandler(self)
+
+    def build_spam_checker(self):
+        return SpamChecker(self)
+
     def remove_pusher(self, app_id, push_key, user_id):
         return self.get_pusherpool().remove_pusher(app_id, push_key, user_id)
 

synapse/state.py

@@ -24,13 +24,13 @@ from synapse.api.constants import EventTypes
 from synapse.api.errors import AuthError
 from synapse.events.snapshot import EventContext
 from synapse.util.async import Linearizer
+from synapse.util.caches import CACHE_SIZE_FACTOR
 
 from collections import namedtuple
 from frozendict import frozendict
 
 import logging
 import hashlib
-import os
 
 logger = logging.getLogger(__name__)
 
@@ -38,9 +38,6 @@ logger = logging.getLogger(__name__)
 KeyStateTuple = namedtuple("KeyStateTuple", ("context", "type", "state_key"))
 
 
-CACHE_SIZE_FACTOR = float(os.environ.get("SYNAPSE_CACHE_FACTOR", 0.1))
-
-
 SIZE_OF_CACHE = int(100000 * CACHE_SIZE_FACTOR)
 EVICTION_TIMEOUT_SECONDS = 60 * 60
 
@@ -170,9 +167,7 @@ class StateHandler(object):
             latest_event_ids = yield self.store.get_latest_event_ids_in_room(room_id)
         logger.debug("calling resolve_state_groups from get_current_user_in_room")
         entry = yield self.resolve_state_groups(room_id, latest_event_ids)
-        joined_users = yield self.store.get_joined_users_from_state(
-            room_id, entry.state_id, entry.state
-        )
+        joined_users = yield self.store.get_joined_users_from_state(room_id, entry)
         defer.returnValue(joined_users)
 
     @defer.inlineCallbacks
@@ -181,9 +176,7 @@ class StateHandler(object):
             latest_event_ids = yield self.store.get_latest_event_ids_in_room(room_id)
         logger.debug("calling resolve_state_groups from get_current_hosts_in_room")
         entry = yield self.resolve_state_groups(room_id, latest_event_ids)
-        joined_hosts = yield self.store.get_joined_hosts(
-            room_id, entry.state_id, entry.state
-        )
+        joined_hosts = yield self.store.get_joined_hosts(room_id, entry)
         defer.returnValue(joined_hosts)
 
     @defer.inlineCallbacks
@@ -195,12 +188,12 @@ class StateHandler(object):
         Returns:
             synapse.events.snapshot.EventContext:
         """
-        context = EventContext()
 
         if event.internal_metadata.is_outlier():
             # If this is an outlier, then we know it shouldn't have any current
             # state. Certainly store.get_current_state won't return any, and
             # persisting the event won't store the state group.
+            context = EventContext()
             if old_state:
                 context.prev_state_ids = {
                     (s.type, s.state_key): s.event_id for s in old_state
@@ -219,6 +212,7 @@ class StateHandler(object):
             defer.returnValue(context)
 
         if old_state:
+            context = EventContext()
             context.prev_state_ids = {
                 (s.type, s.state_key): s.event_id for s in old_state
             }
@@ -239,19 +233,13 @@ class StateHandler(object):
             defer.returnValue(context)
 
         logger.debug("calling resolve_state_groups from compute_event_context")
-        if event.is_state():
-            entry = yield self.resolve_state_groups(
-                event.room_id, [e for e, _ in event.prev_events],
-                event_type=event.type,
-                state_key=event.state_key,
-            )
-        else:
-            entry = yield self.resolve_state_groups(
-                event.room_id, [e for e, _ in event.prev_events],
-            )
+        entry = yield self.resolve_state_groups(
+            event.room_id, [e for e, _ in event.prev_events],
+        )
 
         curr_state = entry.state
 
+        context = EventContext()
         context.prev_state_ids = curr_state
         if event.is_state():
             context.state_group = self.store.get_next_state_group()
@@ -264,10 +252,14 @@ class StateHandler(object):
             context.current_state_ids = dict(context.prev_state_ids)
             context.current_state_ids[key] = event.event_id
 
-            context.prev_group = entry.prev_group
-            context.delta_ids = entry.delta_ids
-            if context.delta_ids is not None:
-                context.delta_ids = dict(context.delta_ids)
+            if entry.state_group:
+                context.prev_group = entry.state_group
+                context.delta_ids = {
+                    key: event.event_id
+                }
+            elif entry.prev_group:
+                context.prev_group = entry.prev_group
+                context.delta_ids = dict(entry.delta_ids)
                 context.delta_ids[key] = event.event_id
         else:
             if entry.state_group is None:
@@ -284,7 +276,7 @@ class StateHandler(object):
 
     @defer.inlineCallbacks
     @log_function
-    def resolve_state_groups(self, room_id, event_ids, event_type=None, state_key=""):
+    def resolve_state_groups(self, room_id, event_ids):
         """ Given a list of event_ids this method fetches the state at each
         event, resolves conflicts between them and returns them.
 
@@ -309,11 +301,13 @@ class StateHandler(object):
         if len(group_names) == 1:
             name, state_list = state_groups_ids.items().pop()
 
+            prev_group, delta_ids = yield self.store.get_state_group_delta(name)
+
             defer.returnValue(_StateCacheEntry(
                 state=state_list,
                 state_group=name,
-                prev_group=name,
-                delta_ids={},
+                prev_group=prev_group,
+                delta_ids=delta_ids,
             ))
 
         with (yield self.resolve_linearizer.queue(group_names)):
@@ -357,20 +351,18 @@ class StateHandler(object):
                 if new_state_event_ids == frozenset(e_id for e_id in events):
                     state_group = sg
                     break
-            if state_group is None:
-                # Worker instances don't have access to this method, but we want
-                # to set the state_group on the main instance to increase cache
-                # hits.
-                if hasattr(self.store, "get_next_state_group"):
-                    state_group = self.store.get_next_state_group()
+
+            # TODO: We want to create a state group for this set of events, to
+            # increase cache hits, but we need to make sure that it doesn't
+            # end up as a prev_group without being added to the database
 
             prev_group = None
             delta_ids = None
-            for old_group, old_ids in state_groups_ids.items():
-                if not set(new_state.iterkeys()) - set(old_ids.iterkeys()):
+            for old_group, old_ids in state_groups_ids.iteritems():
+                if not set(new_state) - set(old_ids):
                     n_delta_ids = {
                         k: v
-                        for k, v in new_state.items()
+                        for k, v in new_state.iteritems()
                         if old_ids.get(k) != v
                     }
                     if not delta_ids or len(n_delta_ids) < len(delta_ids):

synapse/storage/__init__.py

@@ -49,6 +49,7 @@ from .tags import TagsStore
 from .account_data import AccountDataStore
 from .openid import OpenIdStore
 from .client_ips import ClientIpStore
+from .user_directory import UserDirectoryStore
 
 from .util.id_generators import IdGenerator, StreamIdGenerator, ChainedIdGenerator
 from .engines import PostgresEngine
@@ -86,6 +87,7 @@ class DataStore(RoomMemberStore, RoomStore,
                 ClientIpStore,
                 DeviceStore,
                 DeviceInboxStore,
+                UserDirectoryStore,
                 ):
 
     def __init__(self, db_conn, hs):
@@ -221,17 +223,30 @@ class DataStore(RoomMemberStore, RoomStore,
             "DeviceListFederationStreamChangeCache", device_list_max,
         )
 
+        curr_state_delta_prefill, min_curr_state_delta_id = self._get_cache_dict(
+            db_conn, "current_state_delta_stream",
+            entity_column="room_id",
+            stream_column="stream_id",
+            max_value=events_max,  # As we share the stream id with events token
+            limit=1000,
+        )
+        self._curr_state_delta_stream_cache = StreamChangeCache(
+            "_curr_state_delta_stream_cache", min_curr_state_delta_id,
+            prefilled_cache=curr_state_delta_prefill,
+        )
+
         cur = LoggingTransaction(
             db_conn.cursor(),
             name="_find_stream_orderings_for_times_txn",
             database_engine=self.database_engine,
-            after_callbacks=[]
+            after_callbacks=[],
+            final_callbacks=[],
         )
         self._find_stream_orderings_for_times_txn(cur)
         cur.close()
 
         self.find_stream_orderings_looping_call = self._clock.looping_call(
-            self._find_stream_orderings_for_times, 60 * 60 * 1000
+            self._find_stream_orderings_for_times, 10 * 60 * 1000
         )
 
         self._stream_order_on_start = self.get_room_max_stream_ordering()
@@ -272,31 +287,23 @@ class DataStore(RoomMemberStore, RoomStore,
         Counts the number of users who used this homeserver in the last 24 hours.
         """
         def _count_users(txn):
-            txn.execute(
-                "SELECT COUNT(DISTINCT user_id) AS users"
-                " FROM user_ips"
-                " WHERE last_seen > ?",
-                # This is close enough to a day for our purposes.
-                (int(self._clock.time_msec()) - (1000 * 60 * 60 * 24),)
-            )
-            rows = self.cursor_to_dict(txn)
-            if rows:
-                return rows[0]["users"]
-            return 0
+            yesterday = int(self._clock.time_msec()) - (1000 * 60 * 60 * 24),
+
+            sql = """
+                SELECT COALESCE(count(*), 0) FROM (
+                    SELECT user_id FROM user_ips
+                    WHERE last_seen > ?
+                    GROUP BY user_id
+                ) u
+            """
+
+            txn.execute(sql, (yesterday,))
+            count, = txn.fetchone()
+            return count
 
         ret = yield self.runInteraction("count_users", _count_users)
         defer.returnValue(ret)
 
-    def get_user_ip_and_agents(self, user):
-        return self._simple_select_list(
-            table="user_ips",
-            keyvalues={"user_id": user.to_string()},
-            retcols=[
-                "access_token", "ip", "user_agent", "last_seen"
-            ],
-            desc="get_user_ip_and_agents",
-        )
-
     def get_users(self):
         """Function to reterive a list of users in users table.
 

synapse/storage/_base.py

@@ -16,6 +16,7 @@ import logging
 
 from synapse.api.errors import StoreError
 from synapse.util.logcontext import LoggingContext, PreserveLoggingContext
+from synapse.util.caches import CACHE_SIZE_FACTOR
 from synapse.util.caches.dictionary_cache import DictionaryCache
 from synapse.util.caches.descriptors import Cache
 from synapse.storage.engines import PostgresEngine
@@ -27,10 +28,6 @@ from twisted.internet import defer
 import sys
 import time
 import threading
-import os
-
-
-CACHE_SIZE_FACTOR = float(os.environ.get("SYNAPSE_CACHE_FACTOR", 0.1))
 
 
 logger = logging.getLogger(__name__)
@@ -52,13 +49,17 @@ class LoggingTransaction(object):
     """An object that almost-transparently proxies for the 'txn' object
     passed to the constructor. Adds logging and metrics to the .execute()
     method."""
-    __slots__ = ["txn", "name", "database_engine", "after_callbacks"]
+    __slots__ = [
+        "txn", "name", "database_engine", "after_callbacks", "final_callbacks",
+    ]
 
-    def __init__(self, txn, name, database_engine, after_callbacks):
+    def __init__(self, txn, name, database_engine, after_callbacks,
+                 final_callbacks):
         object.__setattr__(self, "txn", txn)
         object.__setattr__(self, "name", name)
         object.__setattr__(self, "database_engine", database_engine)
         object.__setattr__(self, "after_callbacks", after_callbacks)
+        object.__setattr__(self, "final_callbacks", final_callbacks)
 
     def call_after(self, callback, *args, **kwargs):
         """Call the given callback on the main twisted thread after the
@@ -67,6 +68,9 @@ class LoggingTransaction(object):
         """
         self.after_callbacks.append((callback, args, kwargs))
 
+    def call_finally(self, callback, *args, **kwargs):
+        self.final_callbacks.append((callback, args, kwargs))
+
     def __getattr__(self, name):
         return getattr(self.txn, name)
 
@@ -217,8 +221,8 @@ class SQLBaseStore(object):
 
         self._clock.looping_call(loop, 10000)
 
-    def _new_transaction(self, conn, desc, after_callbacks, logging_context,
-                         func, *args, **kwargs):
+    def _new_transaction(self, conn, desc, after_callbacks, final_callbacks,
+                         logging_context, func, *args, **kwargs):
         start = time.time() * 1000
         txn_id = self._TXN_ID
 
@@ -237,7 +241,8 @@ class SQLBaseStore(object):
                 try:
                     txn = conn.cursor()
                     txn = LoggingTransaction(
-                        txn, name, self.database_engine, after_callbacks
+                        txn, name, self.database_engine, after_callbacks,
+                        final_callbacks,
                     )
                     r = func(txn, *args, **kwargs)
                     conn.commit()
@@ -298,6 +303,7 @@ class SQLBaseStore(object):
         start_time = time.time() * 1000
 
         after_callbacks = []
+        final_callbacks = []
 
         def inner_func(conn, *args, **kwargs):
             with LoggingContext("runInteraction") as context:
@@ -309,7 +315,7 @@ class SQLBaseStore(object):
 
                 current_context.copy_to(context)
                 return self._new_transaction(
-                    conn, desc, after_callbacks, current_context,
+                    conn, desc, after_callbacks, final_callbacks, current_context,
                     func, *args, **kwargs
                 )
 
@@ -318,9 +324,13 @@ class SQLBaseStore(object):
                 result = yield self._db_pool.runWithConnection(
                     inner_func, *args, **kwargs
                 )
-        finally:
+
             for after_callback, after_args, after_kwargs in after_callbacks:
                 after_callback(*after_args, **after_kwargs)
+        finally:
+            for after_callback, after_args, after_kwargs in final_callbacks:
+                after_callback(*after_args, **after_kwargs)
+
         defer.returnValue(result)
 
     @defer.inlineCallbacks
@@ -425,6 +435,11 @@ class SQLBaseStore(object):
 
         txn.execute(sql, vals)
 
+    def _simple_insert_many(self, table, values, desc):
+        return self.runInteraction(
+            desc, self._simple_insert_many_txn, table, values
+        )
+
     @staticmethod
     def _simple_insert_many_txn(txn, table, values):
         if not values:
@@ -936,7 +951,7 @@ class SQLBaseStore(object):
             # __exit__ called after the transaction finishes.
             ctx = self._cache_id_gen.get_next()
             stream_id = ctx.__enter__()
-            txn.call_after(ctx.__exit__, None, None, None)
+            txn.call_finally(ctx.__exit__, None, None, None)
             txn.call_after(self.hs.get_notifier().on_new_replication_data)
 
             self._simple_insert_txn(

synapse/storage/account_data.py

@@ -308,3 +308,16 @@ class AccountDataStore(SQLBaseStore):
             " WHERE stream_id < ?"
         )
         txn.execute(update_max_id_sql, (next_id, next_id))
+
+    @cachedInlineCallbacks(num_args=2, cache_context=True, max_entries=5000)
+    def is_ignored_by(self, ignored_user_id, ignorer_user_id, cache_context):
+        ignored_account_data = yield self.get_global_account_data_by_type_for_user(
+            "m.ignored_user_list", ignorer_user_id,
+            on_invalidate=cache_context.invalidate,
+        )
+        if not ignored_account_data:
+            defer.returnValue(False)
+
+        defer.returnValue(
+            ignored_user_id in ignored_account_data.get("ignored_users", {})
+        )

synapse/storage/appservice.py

@@ -13,6 +13,7 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 import logging
+import re
 import simplejson as json
 from twisted.internet import defer
 
@@ -26,6 +27,25 @@ from ._base import SQLBaseStore
 logger = logging.getLogger(__name__)
 
 
+def _make_exclusive_regex(services_cache):
+    # We precompie a regex constructed from all the regexes that the AS's
+    # have registered for exclusive users.
+    exclusive_user_regexes = [
+        regex.pattern
+        for service in services_cache
+        for regex in service.get_exlusive_user_regexes()
+    ]
+    if exclusive_user_regexes:
+        exclusive_user_regex = "|".join("(" + r + ")" for r in exclusive_user_regexes)
+        exclusive_user_regex = re.compile(exclusive_user_regex)
+    else:
+        # We handle this case specially otherwise the constructed regex
+        # will always match
+        exclusive_user_regex = None
+
+    return exclusive_user_regex
+
+
 class ApplicationServiceStore(SQLBaseStore):
 
     def __init__(self, hs):
@@ -35,17 +55,18 @@ class ApplicationServiceStore(SQLBaseStore):
             hs.hostname,
             hs.config.app_service_config_files
         )
+        self.exclusive_user_regex = _make_exclusive_regex(self.services_cache)
 
     def get_app_services(self):
         return self.services_cache
 
     def get_if_app_services_interested_in_user(self, user_id):
-        """Check if the user is one associated with an app service
+        """Check if the user is one associated with an app service (exclusively)
         """
-        for service in self.services_cache:
-            if service.is_interested_in_user(user_id):
-                return True
-        return False
+        if self.exclusive_user_regex:
+            return bool(self.exclusive_user_regex.match(user_id))
+        else:
+            return False
 
     def get_app_service_by_user_id(self, user_id):
         """Retrieve an application service from their user ID.

synapse/storage/client_ips.py

@@ -15,11 +15,14 @@
 
 import logging
 
-from twisted.internet import defer
+from twisted.internet import defer, reactor
 
 from ._base import Cache
 from . import background_updates
 
+from synapse.util.caches import CACHE_SIZE_FACTOR
+
+
 logger = logging.getLogger(__name__)
 
 # Number of msec of granularity to store the user IP 'last seen' time. Smaller
@@ -33,7 +36,7 @@ class ClientIpStore(background_updates.BackgroundUpdateStore):
         self.client_ip_last_seen = Cache(
             name="client_ip_last_seen",
             keylen=4,
-            max_entries=5000,
+            max_entries=50000 * CACHE_SIZE_FACTOR,
         )
 
         super(ClientIpStore, self).__init__(hs)
@@ -45,10 +48,19 @@ class ClientIpStore(background_updates.BackgroundUpdateStore):
             columns=["user_id", "device_id", "last_seen"],
         )
 
-    @defer.inlineCallbacks
-    def insert_client_ip(self, user, access_token, ip, user_agent, device_id):
-        now = int(self._clock.time_msec())
-        key = (user.to_string(), access_token, ip)
+        # (user_id, access_token, ip) -> (user_agent, device_id, last_seen)
+        self._batch_row_update = {}
+
+        self._client_ip_looper = self._clock.looping_call(
+            self._update_client_ips_batch, 5 * 1000
+        )
+        reactor.addSystemEventTrigger("before", "shutdown", self._update_client_ips_batch)
+
+    def insert_client_ip(self, user_id, access_token, ip, user_agent, device_id,
+                         now=None):
+        if not now:
+            now = int(self._clock.time_msec())
+        key = (user_id, access_token, ip)
 
         try:
             last_seen = self.client_ip_last_seen.get(key)
@@ -57,34 +69,48 @@ class ClientIpStore(background_updates.BackgroundUpdateStore):
 
         # Rate-limited inserts
         if last_seen is not None and (now - last_seen) < LAST_SEEN_GRANULARITY:
-            defer.returnValue(None)
+            return
 
         self.client_ip_last_seen.prefill(key, now)
 
-        # It's safe not to lock here: a) no unique constraint,
-        # b) LAST_SEEN_GRANULARITY makes concurrent updates incredibly unlikely
-        yield self._simple_upsert(
-            "user_ips",
-            keyvalues={
-                "user_id": user.to_string(),
-                "access_token": access_token,
-                "ip": ip,
-                "user_agent": user_agent,
-                "device_id": device_id,
-            },
-            values={
-                "last_seen": now,
-            },
-            desc="insert_client_ip",
-            lock=False,
+        self._batch_row_update[key] = (user_agent, device_id, now)
+
+    def _update_client_ips_batch(self):
+        to_update = self._batch_row_update
+        self._batch_row_update = {}
+        return self.runInteraction(
+            "_update_client_ips_batch", self._update_client_ips_batch_txn, to_update
         )
 
+    def _update_client_ips_batch_txn(self, txn, to_update):
+        self.database_engine.lock_table(txn, "user_ips")
+
+        for entry in to_update.iteritems():
+            (user_id, access_token, ip), (user_agent, device_id, last_seen) = entry
+
+            self._simple_upsert_txn(
+                txn,
+                table="user_ips",
+                keyvalues={
+                    "user_id": user_id,
+                    "access_token": access_token,
+                    "ip": ip,
+                    "user_agent": user_agent,
+                    "device_id": device_id,
+                },
+                values={
+                    "last_seen": last_seen,
+                },
+                lock=False,
+            )
+
     @defer.inlineCallbacks
-    def get_last_client_ip_by_device(self, devices):
+    def get_last_client_ip_by_device(self, user_id, device_id):
         """For each device_id listed, give the user_ip it was last seen on
 
         Args:
-            devices (iterable[(str, str)]):  list of (user_id, device_id) pairs
+            user_id (str)
+            device_id (str): If None fetches all devices for the user
 
         Returns:
             defer.Deferred: resolves to a dict, where the keys
@@ -95,6 +121,7 @@ class ClientIpStore(background_updates.BackgroundUpdateStore):
         res = yield self.runInteraction(
             "get_last_client_ip_by_device",
             self._get_last_client_ip_by_device_txn,
+            user_id, device_id,
             retcols=(
                 "user_id",
                 "access_token",
@@ -103,23 +130,34 @@ class ClientIpStore(background_updates.BackgroundUpdateStore):
                 "device_id",
                 "last_seen",
             ),
-            devices=devices
         )
 
         ret = {(d["user_id"], d["device_id"]): d for d in res}
+        for key in self._batch_row_update:
+            uid, access_token, ip = key
+            if uid == user_id:
+                user_agent, did, last_seen = self._batch_row_update[key]
+                if not device_id or did == device_id:
+                    ret[(user_id, device_id)] = {
+                        "user_id": user_id,
+                        "access_token": access_token,
+                        "ip": ip,
+                        "user_agent": user_agent,
+                        "device_id": did,
+                        "last_seen": last_seen,
+                    }
         defer.returnValue(ret)
 
     @classmethod
-    def _get_last_client_ip_by_device_txn(cls, txn, devices, retcols):
+    def _get_last_client_ip_by_device_txn(cls, txn, user_id, device_id, retcols):
         where_clauses = []
         bindings = []
-        for (user_id, device_id) in devices:
-            if device_id is None:
-                where_clauses.append("(user_id = ? AND device_id IS NULL)")
-                bindings.extend((user_id, ))
-            else:
-                where_clauses.append("(user_id = ? AND device_id = ?)")
-                bindings.extend((user_id, device_id))
+        if device_id is None:
+            where_clauses.append("user_id = ?")
+            bindings.extend((user_id, ))
+        else:
+            where_clauses.append("(user_id = ? AND device_id = ?)")
+            bindings.extend((user_id, device_id))
 
         if not where_clauses:
             return []
@@ -147,3 +185,37 @@ class ClientIpStore(background_updates.BackgroundUpdateStore):
 
         txn.execute(sql, bindings)
         return cls.cursor_to_dict(txn)
+
+    @defer.inlineCallbacks
+    def get_user_ip_and_agents(self, user):
+        user_id = user.to_string()
+        results = {}
+
+        for key in self._batch_row_update:
+            uid, access_token, ip = key
+            if uid == user_id:
+                user_agent, _, last_seen = self._batch_row_update[key]
+                results[(access_token, ip)] = (user_agent, last_seen)
+
+        rows = yield self._simple_select_list(
+            table="user_ips",
+            keyvalues={"user_id": user_id},
+            retcols=[
+                "access_token", "ip", "user_agent", "last_seen"
+            ],
+            desc="get_user_ip_and_agents",
+        )
+
+        results.update(
+            ((row["access_token"], row["ip"]), (row["user_agent"], row["last_seen"]))
+            for row in rows
+        )
+        defer.returnValue(list(
+            {
+                "access_token": access_token,
+                "ip": ip,
+                "user_agent": user_agent,
+                "last_seen": last_seen,
+            }
+            for (access_token, ip), (user_agent, last_seen) in results.iteritems()
+        ))

synapse/storage/devices.py

@@ -368,7 +368,7 @@ class DeviceStore(SQLBaseStore):
 
         prev_sent_id_sql = """
             SELECT coalesce(max(stream_id), 0) as stream_id
-            FROM device_lists_outbound_pokes
+            FROM device_lists_outbound_last_success
             WHERE destination = ? AND user_id = ? AND stream_id <= ?
         """
 
@@ -510,32 +510,43 @@ class DeviceStore(SQLBaseStore):
         )
 
     def _mark_as_sent_devices_by_remote_txn(self, txn, destination, stream_id):
-        # First we DELETE all rows such that only the latest row for each
-        # (destination, user_id is left. We do this by selecting first and
-        # deleting.
+        # We update the device_lists_outbound_last_success with the successfully
+        # poked users. We do the join to see which users need to be inserted and
+        # which updated.
         sql = """
-            SELECT user_id, coalesce(max(stream_id), 0) FROM device_lists_outbound_pokes
-            WHERE destination = ? AND stream_id <= ?
+            SELECT user_id, coalesce(max(o.stream_id), 0), (max(s.stream_id) IS NOT NULL)
+            FROM device_lists_outbound_pokes as o
+            LEFT JOIN device_lists_outbound_last_success as s
+                USING (destination, user_id)
+            WHERE destination = ? AND o.stream_id <= ?
             GROUP BY user_id
-            HAVING count(*) > 1
         """
         txn.execute(sql, (destination, stream_id,))
         rows = txn.fetchall()
 
         sql = """
-            DELETE FROM device_lists_outbound_pokes
-            WHERE destination = ? AND user_id = ? AND stream_id < ?
+            UPDATE device_lists_outbound_last_success
+            SET stream_id = ?
+            WHERE destination = ? AND user_id = ?
         """
         txn.executemany(
-            sql, ((destination, row[0], row[1],) for row in rows)
+            sql, ((row[1], destination, row[0],) for row in rows if row[2])
         )
 
-        # Mark everything that is left as sent
         sql = """
-            UPDATE device_lists_outbound_pokes SET sent = ?
+            INSERT INTO device_lists_outbound_last_success
+            (destination, user_id, stream_id) VALUES (?, ?, ?)
+        """
+        txn.executemany(
+            sql, ((destination, row[0], row[1],) for row in rows if not row[2])
+        )
+
+        # Delete all sent outbound pokes
+        sql = """
+            DELETE FROM device_lists_outbound_pokes
             WHERE destination = ? AND stream_id <= ?
         """
-        txn.execute(sql, (True, destination, stream_id,))
+        txn.execute(sql, (destination, stream_id,))
 
     @defer.inlineCallbacks
     def get_user_whose_devices_changed(self, from_key):
@@ -670,6 +681,14 @@ class DeviceStore(SQLBaseStore):
                 )
             )
 
+            # Since we've deleted unsent deltas, we need to remove the entry
+            # of last successful sent so that the prev_ids are correctly set.
+            sql = """
+                DELETE FROM device_lists_outbound_last_success
+                WHERE destination = ? AND user_id = ?
+            """
+            txn.executemany(sql, ((row[0], row[1]) for row in rows))
+
             logger.info("Pruned %d device list outbound pokes", txn.rowcount)
 
         return self.runInteraction(

synapse/storage/directory.py

@@ -170,3 +170,17 @@ class DirectoryStore(SQLBaseStore):
             "room_alias",
             desc="get_aliases_for_room",
         )
+
+    def update_aliases_for_room(self, old_room_id, new_room_id, creator):
+        def _update_aliases_for_room_txn(txn):
+            sql = "UPDATE room_aliases SET room_id = ?, creator = ? WHERE room_id = ?"
+            txn.execute(sql, (new_room_id, creator, old_room_id,))
+            self._invalidate_cache_and_stream(
+                txn, self.get_aliases_for_room, (old_room_id,)
+            )
+            self._invalidate_cache_and_stream(
+                txn, self.get_aliases_for_room, (new_room_id,)
+            )
+        return self.runInteraction(
+            "_update_aliases_for_room_txn", _update_aliases_for_room_txn
+        )

synapse/storage/end_to_end_keys.py

@@ -185,8 +185,8 @@ class EndToEndKeyStore(SQLBaseStore):
                     for algorithm, key_id, json_bytes in new_keys
                 ],
             )
-            txn.call_after(
-                self.count_e2e_one_time_keys.invalidate, (user_id, device_id,)
+            self._invalidate_cache_and_stream(
+                txn, self.count_e2e_one_time_keys, (user_id, device_id,)
             )
         yield self.runInteraction(
             "add_e2e_one_time_keys_insert", _add_e2e_one_time_keys
@@ -237,24 +237,29 @@ class EndToEndKeyStore(SQLBaseStore):
             )
             for user_id, device_id, algorithm, key_id in delete:
                 txn.execute(sql, (user_id, device_id, algorithm, key_id))
-                txn.call_after(
-                    self.count_e2e_one_time_keys.invalidate, (user_id, device_id,)
+                self._invalidate_cache_and_stream(
+                    txn, self.count_e2e_one_time_keys, (user_id, device_id,)
                 )
             return result
         return self.runInteraction(
             "claim_e2e_one_time_keys", _claim_e2e_one_time_keys
         )
 
-    @defer.inlineCallbacks
     def delete_e2e_keys_by_device(self, user_id, device_id):
-        yield self._simple_delete(
-            table="e2e_device_keys_json",
-            keyvalues={"user_id": user_id, "device_id": device_id},
-            desc="delete_e2e_device_keys_by_device"
+        def delete_e2e_keys_by_device_txn(txn):
+            self._simple_delete_txn(
+                txn,
+                table="e2e_device_keys_json",
+                keyvalues={"user_id": user_id, "device_id": device_id},
+            )
+            self._simple_delete_txn(
+                txn,
+                table="e2e_one_time_keys_json",
+                keyvalues={"user_id": user_id, "device_id": device_id},
+            )
+            self._invalidate_cache_and_stream(
+                txn, self.count_e2e_one_time_keys, (user_id, device_id,)
+            )
+        return self.runInteraction(
+            "delete_e2e_keys_by_device", delete_e2e_keys_by_device_txn
         )
-        yield self._simple_delete(
-            table="e2e_one_time_keys_json",
-            keyvalues={"user_id": user_id, "device_id": device_id},
-            desc="delete_e2e_one_time_keys_by_device"
-        )
-        self.count_e2e_one_time_keys.invalidate((user_id, device_id,))

synapse/storage/event_federation.py

@@ -37,25 +37,55 @@ class EventFederationStore(SQLBaseStore):
     and backfilling from another server respectively.
     """
 
+    EVENT_AUTH_STATE_ONLY = "event_auth_state_only"
+
     def __init__(self, hs):
         super(EventFederationStore, self).__init__(hs)
 
+        self.register_background_update_handler(
+            self.EVENT_AUTH_STATE_ONLY,
+            self._background_delete_non_state_event_auth,
+        )
+
         hs.get_clock().looping_call(
             self._delete_old_forward_extrem_cache, 60 * 60 * 1000
         )
 
-    def get_auth_chain(self, event_ids):
-        return self.get_auth_chain_ids(event_ids).addCallback(self._get_events)
+    def get_auth_chain(self, event_ids, include_given=False):
+        """Get auth events for given event_ids. The events *must* be state events.
 
-    def get_auth_chain_ids(self, event_ids):
+        Args:
+            event_ids (list): state events
+            include_given (bool): include the given events in result
+
+        Returns:
+            list of events
+        """
+        return self.get_auth_chain_ids(
+            event_ids, include_given=include_given,
+        ).addCallback(self._get_events)
+
+    def get_auth_chain_ids(self, event_ids, include_given=False):
+        """Get auth events for given event_ids. The events *must* be state events.
+
+        Args:
+            event_ids (list): state events
+            include_given (bool): include the given events in result
+
+        Returns:
+            list of event_ids
+        """
         return self.runInteraction(
             "get_auth_chain_ids",
             self._get_auth_chain_ids_txn,
-            event_ids
+            event_ids, include_given
         )
 
-    def _get_auth_chain_ids_txn(self, txn, event_ids):
-        results = set()
+    def _get_auth_chain_ids_txn(self, txn, event_ids, include_given):
+        if include_given:
+            results = set(event_ids)
+        else:
+            results = set()
 
         base_sql = (
             "SELECT auth_id FROM event_auth WHERE event_id IN (%s)"
@@ -504,3 +534,52 @@ class EventFederationStore(SQLBaseStore):
 
         txn.execute(query, (room_id,))
         txn.call_after(self.get_latest_event_ids_in_room.invalidate, (room_id,))
+
+    @defer.inlineCallbacks
+    def _background_delete_non_state_event_auth(self, progress, batch_size):
+        def delete_event_auth(txn):
+            target_min_stream_id = progress.get("target_min_stream_id_inclusive")
+            max_stream_id = progress.get("max_stream_id_exclusive")
+
+            if not target_min_stream_id or not max_stream_id:
+                txn.execute("SELECT COALESCE(MIN(stream_ordering), 0) FROM events")
+                rows = txn.fetchall()
+                target_min_stream_id = rows[0][0]
+
+                txn.execute("SELECT COALESCE(MAX(stream_ordering), 0) FROM events")
+                rows = txn.fetchall()
+                max_stream_id = rows[0][0]
+
+            min_stream_id = max_stream_id - batch_size
+
+            sql = """
+                DELETE FROM event_auth
+                WHERE event_id IN (
+                    SELECT event_id FROM events
+                    LEFT JOIN state_events USING (room_id, event_id)
+                    WHERE ? <= stream_ordering AND stream_ordering < ?
+                        AND state_key IS null
+                )
+            """
+
+            txn.execute(sql, (min_stream_id, max_stream_id,))
+
+            new_progress = {
+                "target_min_stream_id_inclusive": target_min_stream_id,
+                "max_stream_id_exclusive": min_stream_id,
+            }
+
+            self._background_update_progress_txn(
+                txn, self.EVENT_AUTH_STATE_ONLY, new_progress
+            )
+
+            return min_stream_id >= target_min_stream_id
+
+        result = yield self.runInteraction(
+            self.EVENT_AUTH_STATE_ONLY, delete_event_auth
+        )
+
+        if not result:
+            yield self._end_background_update(self.EVENT_AUTH_STATE_ONLY)
+
+        defer.returnValue(batch_size)

synapse/storage/events.py

@@ -38,7 +38,6 @@ from functools import wraps
 import synapse.metrics
 
 import logging
-import math
 import ujson as json
 
 # these are only included to make the type annotations work
@@ -404,6 +403,11 @@ class EventsStore(SQLBaseStore):
                         (room_id, ), new_state
                     )
 
+                for room_id, latest_event_ids in new_forward_extremeties.iteritems():
+                    self.get_latest_event_ids_in_room.prefill(
+                        (room_id,), list(latest_event_ids)
+                    )
+
     @defer.inlineCallbacks
     def _calculate_new_extremeties(self, room_id, event_contexts, latest_event_ids):
         """Calculates the new forward extremeties for a room given events to
@@ -648,9 +652,10 @@ class EventsStore(SQLBaseStore):
                 list of the event ids which are the forward extremities.
 
         """
-        self._update_current_state_txn(txn, current_state_for_room)
-
         max_stream_order = events_and_contexts[-1][0].internal_metadata.stream_ordering
+
+        self._update_current_state_txn(txn, current_state_for_room, max_stream_order)
+
         self._update_forward_extremities_txn(
             txn,
             new_forward_extremities=new_forward_extremeties,
@@ -713,7 +718,7 @@ class EventsStore(SQLBaseStore):
             backfilled=backfilled,
         )
 
-    def _update_current_state_txn(self, txn, state_delta_by_room):
+    def _update_current_state_txn(self, txn, state_delta_by_room, max_stream_order):
         for room_id, current_state_tuple in state_delta_by_room.iteritems():
                 to_delete, to_insert, _ = current_state_tuple
                 txn.executemany(
@@ -735,6 +740,29 @@ class EventsStore(SQLBaseStore):
                     ],
                 )
 
+                state_deltas = {key: None for key in to_delete}
+                state_deltas.update(to_insert)
+
+                self._simple_insert_many_txn(
+                    txn,
+                    table="current_state_delta_stream",
+                    values=[
+                        {
+                            "stream_id": max_stream_order,
+                            "room_id": room_id,
+                            "type": key[0],
+                            "state_key": key[1],
+                            "event_id": ev_id,
+                            "prev_event_id": to_delete.get(key, None),
+                        }
+                        for key, ev_id in state_deltas.iteritems()
+                    ]
+                )
+
+                self._curr_state_delta_stream_cache.entity_has_changed(
+                    room_id, max_stream_order,
+                )
+
                 # Invalidate the various caches
 
                 # Figure out the changes of membership to invalidate the
@@ -743,11 +771,7 @@ class EventsStore(SQLBaseStore):
                 # and which we have added, then we invlidate the caches for all
                 # those users.
                 members_changed = set(
-                    state_key for ev_type, state_key in to_delete.iterkeys()
-                    if ev_type == EventTypes.Member
-                )
-                members_changed.update(
-                    state_key for ev_type, state_key in to_insert.iterkeys()
+                    state_key for ev_type, state_key in state_deltas
                     if ev_type == EventTypes.Member
                 )
 
@@ -756,6 +780,11 @@ class EventsStore(SQLBaseStore):
                         txn, self.get_rooms_for_user, (member,)
                     )
 
+                for host in set(get_domain_from_id(u) for u in members_changed):
+                    self._invalidate_cache_and_stream(
+                        txn, self.is_host_joined, (room_id, host)
+                    )
+
                 self._invalidate_cache_and_stream(
                     txn, self.get_users_in_room, (room_id,)
                 )
@@ -1120,6 +1149,7 @@ class EventsStore(SQLBaseStore):
                 }
                 for event, _ in events_and_contexts
                 for auth_id, _ in event.auth_events
+                if event.is_state()
             ],
         )
 
@@ -1419,7 +1449,7 @@ class EventsStore(SQLBaseStore):
                 ]
 
                 rows = self._new_transaction(
-                    conn, "do_fetch", [], None, self._fetch_event_rows, event_ids
+                    conn, "do_fetch", [], [], None, self._fetch_event_rows, event_ids
                 )
 
                 row_dict = {
@@ -1599,68 +1629,54 @@ class EventsStore(SQLBaseStore):
         call to this function, it will return None.
         """
         def _count_messages(txn):
-            now = self.hs.get_clock().time()
-
-            txn.execute(
-                "SELECT reported_stream_token, reported_time FROM stats_reporting"
-            )
-            last_reported = self.cursor_to_dict(txn)
-
-            txn.execute(
-                "SELECT stream_ordering"
-                " FROM events"
-                " ORDER BY stream_ordering DESC"
-                " LIMIT 1"
-            )
-            now_reporting = self.cursor_to_dict(txn)
-            if not now_reporting:
-                logger.info("Calculating daily messages skipped; no now_reporting")
-                return None
-            now_reporting = now_reporting[0]["stream_ordering"]
-
-            txn.execute("DELETE FROM stats_reporting")
-            txn.execute(
-                "INSERT INTO stats_reporting"
-                " (reported_stream_token, reported_time)"
-                " VALUES (?, ?)",
-                (now_reporting, now,)
-            )
-
-            if not last_reported:
-                logger.info("Calculating daily messages skipped; no last_reported")
-                return None
-
-            # Close enough to correct for our purposes.
-            yesterday = (now - 24 * 60 * 60)
-            since_yesterday_seconds = yesterday - last_reported[0]["reported_time"]
-            any_since_yesterday = math.fabs(since_yesterday_seconds) > 60 * 60
-            if any_since_yesterday:
-                logger.info(
-                    "Calculating daily messages skipped; since_yesterday_seconds: %d" %
-                    (since_yesterday_seconds,)
-                )
-                return None
-
-            txn.execute(
-                "SELECT COUNT(*) as messages"
-                " FROM events NATURAL JOIN event_json"
-                " WHERE json like '%m.room.message%'"
-                " AND stream_ordering > ?"
-                " AND stream_ordering <= ?",
-                (
-                    last_reported[0]["reported_stream_token"],
-                    now_reporting,
-                )
-            )
-            rows = self.cursor_to_dict(txn)
-            if not rows:
-                logger.info("Calculating daily messages skipped; messages count missing")
-                return None
-            return rows[0]["messages"]
+            sql = """
+                SELECT COALESCE(COUNT(*), 0) FROM events
+                WHERE type = 'm.room.message'
+                AND stream_ordering > ?
+            """
+            txn.execute(sql, (self.stream_ordering_day_ago,))
+            count, = txn.fetchone()
+            return count
 
         ret = yield self.runInteraction("count_messages", _count_messages)
         defer.returnValue(ret)
 
+    @defer.inlineCallbacks
+    def count_daily_sent_messages(self):
+        def _count_messages(txn):
+            # This is good enough as if you have silly characters in your own
+            # hostname then thats your own fault.
+            like_clause = "%:" + self.hs.hostname
+
+            sql = """
+                SELECT COALESCE(COUNT(*), 0) FROM events
+                WHERE type = 'm.room.message'
+                    AND sender LIKE ?
+                AND stream_ordering > ?
+            """
+
+            txn.execute(sql, (like_clause, self.stream_ordering_day_ago,))
+            count, = txn.fetchone()
+            return count
+
+        ret = yield self.runInteraction("count_daily_sent_messages", _count_messages)
+        defer.returnValue(ret)
+
+    @defer.inlineCallbacks
+    def count_daily_active_rooms(self):
+        def _count(txn):
+            sql = """
+                SELECT COALESCE(COUNT(DISTINCT room_id), 0) FROM events
+                WHERE type = 'm.room.message'
+                AND stream_ordering > ?
+            """
+            txn.execute(sql, (self.stream_ordering_day_ago,))
+            count, = txn.fetchone()
+            return count
+
+        ret = yield self.runInteraction("count_daily_active_rooms", _count)
+        defer.returnValue(ret)
+
     @defer.inlineCallbacks
     def _background_reindex_fields_sender(self, progress, batch_size):
         target_min_stream_id = progress["target_min_stream_id_inclusive"]
@@ -2258,6 +2274,24 @@ class EventsStore(SQLBaseStore):
 
         defer.returnValue((int(res["topological_ordering"]), int(res["stream_ordering"])))
 
+    def get_max_current_state_delta_stream_id(self):
+        return self._stream_id_gen.get_current_token()
+
+    def get_all_updated_current_state_deltas(self, from_token, to_token, limit):
+        def get_all_updated_current_state_deltas_txn(txn):
+            sql = """
+                SELECT stream_id, room_id, type, state_key, event_id
+                FROM current_state_delta_stream
+                WHERE ? < stream_id AND stream_id <= ?
+                ORDER BY stream_id ASC LIMIT ?
+            """
+            txn.execute(sql, (from_token, to_token, limit))
+            return txn.fetchall()
+        return self.runInteraction(
+            "get_all_updated_current_state_deltas",
+            get_all_updated_current_state_deltas_txn,
+        )
+
 
 AllNewEventsResult = namedtuple("AllNewEventsResult", [
     "new_forward_events", "new_backfill_events",

synapse/storage/filtering.py

@@ -19,6 +19,7 @@ from ._base import SQLBaseStore
 from synapse.api.errors import SynapseError, Codes
 from synapse.util.caches.descriptors import cachedInlineCallbacks
 
+from canonicaljson import encode_canonical_json
 import simplejson as json
 
 
@@ -46,11 +47,20 @@ class FilteringStore(SQLBaseStore):
         defer.returnValue(json.loads(str(def_json).decode("utf-8")))
 
     def add_user_filter(self, user_localpart, user_filter):
-        def_json = json.dumps(user_filter).encode("utf-8")
+        def_json = encode_canonical_json(user_filter)
 
         # Need an atomic transaction to SELECT the maximal ID so far then
         # INSERT a new one
         def _do_txn(txn):
+            sql = (
+                "SELECT filter_id FROM user_filters "
+                "WHERE user_id = ? AND filter_json = ?"
+            )
+            txn.execute(sql, (user_localpart, def_json))
+            filter_id_response = txn.fetchone()
+            if filter_id_response is not None:
+                return filter_id_response[0]
+
             sql = (
                 "SELECT MAX(filter_id) FROM user_filters "
                 "WHERE user_id = ?"