Merge pull request #688 from matrix-org/matthew/preview_urls

URL previewing support

README.rst

@@ -104,7 +104,7 @@ Installing prerequisites on Ubuntu or Debian::
 
     sudo apt-get install build-essential python2.7-dev libffi-dev \
                          python-pip python-setuptools sqlite3 \
-                         libssl-dev python-virtualenv libjpeg-dev
+                         libssl-dev python-virtualenv libjpeg-dev libxslt1-dev
 
 Installing prerequisites on ArchLinux::
 
@@ -557,6 +557,23 @@ as the primary means of identity and E2E encryption is not complete. As such,
 we are running a single identity server (https://matrix.org) at the current
 time.
 
+
+URL Previews
+============
+
+Synapse 0.15.0 introduces an experimental new API for previewing URLs at
+/_matrix/media/r0/preview_url.  This is disabled by default.  To turn it on
+you must enable the `url_preview_enabled: True` config parameter and explicitly
+specify the IP ranges that Synapse is not allowed to spider for previewing in
+the `url_preview_ip_range_blacklist` configuration parameter.  This is critical
+from a security perspective to stop arbitrary Matrix users spidering 'internal'
+URLs on your network.  At the very least we recommend that your loopback and
+RFC1918 IP addresses are blacklisted.
+
+This also requires the optional lxml and netaddr python dependencies to be
+installed.
+
+
 Password reset
 ==============
 

UPGRADE.rst

@@ -30,6 +30,14 @@ running:
     python synapse/python_dependencies.py | xargs -n1 pip install
 
 
+Upgrading to v0.15.0
+====================
+
+If you want to use the new URL previewing API (/_matrix/media/r0/preview_url)
+then you have to explicitly enable it in the config and update your dependencies
+dependencies.  See README.rst for details.
+
+
 Upgrading to v0.11.0
 ====================
 

docs/url_previews.rst

@@ -0,0 +1,74 @@
+URL Previews
+============
+
+Design notes on a URL previewing service for Matrix:
+
+Options are:
+
+ 1. Have an AS which listens for URLs, downloads them, and inserts an event that describes their metadata.
+   * Pros:
+     * Decouples the implementation entirely from Synapse.
+     * Uses existing Matrix events & content repo to store the metadata.
+   * Cons:
+     * Which AS should provide this service for a room, and why should you trust it?
+     * Doesn't work well with E2E; you'd have to cut the AS into every room
+     * the AS would end up subscribing to every room anyway.
+
+ 2. Have a generic preview API (nothing to do with Matrix) that provides a previewing service:
+   * Pros:
+     * Simple and flexible; can be used by any clients at any point
+   * Cons:
+     * If each HS provides one of these independently, all the HSes in a room may needlessly DoS the target URI
+     * We need somewhere to store the URL metadata rather than just using Matrix itself
+     * We can't piggyback on matrix to distribute the metadata between HSes.
+
+ 3. Make the synapse of the sending user responsible for spidering the URL and inserting an event asynchronously which describes the metadata.
+   * Pros:
+     * Works transparently for all clients
+     * Piggy-backs nicely on using Matrix for distributing the metadata.
+     * No confusion as to which AS
+   * Cons:
+     * Doesn't work with E2E
+     * We might want to decouple the implementation of the spider from the HS, given spider behaviour can be quite complicated and evolve much more rapidly than the HS.  It's more like a bot than a core part of the server.
+
+ 4. Make the sending client use the preview API and insert the event itself when successful.
+   * Pros:
+      * Works well with E2E
+      * No custom server functionality
+      * Lets the client customise the preview that they send (like on FB)
+   * Cons:
+      * Entirely specific to the sending client, whereas it'd be nice if /any/ URL was correctly previewed if clients support it.
+
+ 5. Have the option of specifying a shared (centralised) previewing service used by a room, to avoid all the different HSes in the room DoSing the target.
+
+Best solution is probably a combination of both 2 and 4.
+ * Sending clients do their best to create and send a preview at the point of sending the message, perhaps delaying the message until the preview is computed?  (This also lets the user validate the preview before sending)
+ * Receiving clients have the option of going and creating their own preview if one doesn't arrive soon enough (or if the original sender didn't create one)
+
+This is a bit magical though in that the preview could come from two entirely different sources - the sending HS or your local one.  However, this can always be exposed to users: "Generate your own URL previews if none are available?"
+
+This is tantamount also to senders calculating their own thumbnails for sending in advance of the main content - we are trusting the sender not to lie about the content in the thumbnail.  Whereas currently thumbnails are calculated by the receiving homeserver to avoid this attack.
+
+However, this kind of phishing attack does exist whether we let senders pick their thumbnails or not, in that a malicious sender can send normal text messages around the attachment claiming it to be legitimate.  We could rely on (future) reputation/abuse management to punish users who phish (be it with bogus metadata or bogus descriptions).   Bogus metadata is particularly bad though, especially if it's avoidable.
+
+As a first cut, let's do #2 and have the receiver hit the API to calculate its own previews (as it does currently for image thumbnails).  We can then extend/optimise this to option 4 as a special extra if needed.
+
+API
+---
+
+GET /_matrix/media/r0/preview_url?url=http://wherever.com
+200 OK
+{
+    "og:type"        : "article"
+    "og:url"         : "https://twitter.com/matrixdotorg/status/684074366691356672"
+    "og:title"       : "Matrix on Twitter"
+    "og:image"       : "https://pbs.twimg.com/profile_images/500400952029888512/yI0qtFi7_400x400.png"
+    "og:description" : "“Synapse 0.12 is out! Lots of polishing, performance & bugfixes: /sync API, /r0 prefix, fulltext search, 3PID invites https://t.co/5alhXLLEGP”"
+    "og:site_name"   : "Twitter"
+}
+
+* Downloads the URL
+  * If HTML, just stores it in RAM and parses it for OG meta tags
+    * Download any media OG meta tags to the media repo, and refer to them in the OG via mxc:// URIs.
+  * If a media filetype we know we can thumbnail: store it on disk, and hand it to the thumbnailer. Generate OG meta tags from the thumbnailer contents.
+  * Otherwise, don't bother downloading further.

synapse/config/repository.py

@@ -16,6 +16,8 @@
 from ._base import Config
 from collections import namedtuple
 
+import sys
+
 ThumbnailRequirement = namedtuple(
     "ThumbnailRequirement", ["width", "height", "method", "media_type"]
 )
@@ -23,7 +25,7 @@ ThumbnailRequirement = namedtuple(
 
 def parse_thumbnail_requirements(thumbnail_sizes):
     """ Takes a list of dictionaries with "width", "height", and "method" keys
-    and creates a map from image media types to the thumbnail size, thumnailing
+    and creates a map from image media types to the thumbnail size, thumbnailing
     method, and thumbnail media type to precalculate
 
     Args:
@@ -53,12 +55,25 @@ class ContentRepositoryConfig(Config):
     def read_config(self, config):
         self.max_upload_size = self.parse_size(config["max_upload_size"])
         self.max_image_pixels = self.parse_size(config["max_image_pixels"])
+        self.max_spider_size = self.parse_size(config["max_spider_size"])
         self.media_store_path = self.ensure_directory(config["media_store_path"])
         self.uploads_path = self.ensure_directory(config["uploads_path"])
         self.dynamic_thumbnails = config["dynamic_thumbnails"]
         self.thumbnail_requirements = parse_thumbnail_requirements(
             config["thumbnail_sizes"]
         )
+        self.url_preview_enabled = config["url_preview_enabled"]
+        if self.url_preview_enabled:
+            try:
+                from netaddr import IPSet
+                if "url_preview_ip_range_blacklist" in config:
+                    self.url_preview_ip_range_blacklist = IPSet(
+                        config["url_preview_ip_range_blacklist"]
+                    )
+                if "url_preview_url_blacklist" in config:
+                    self.url_preview_url_blacklist = config["url_preview_url_blacklist"]
+            except ImportError:
+                sys.stderr.write("\nmissing netaddr dep - disabling preview_url API\n")
 
     def default_config(self, **kwargs):
         media_store = self.default_path("media_store")
@@ -80,7 +95,7 @@ class ContentRepositoryConfig(Config):
         # the resolution requested by the client. If true then whenever
         # a new resolution is requested by the client the server will
         # generate a new thumbnail. If false the server will pick a thumbnail
-        # from a precalcualted list.
+        # from a precalculated list.
         dynamic_thumbnails: false
 
         # List of thumbnail to precalculate when an image is uploaded.
@@ -100,4 +115,62 @@ class ContentRepositoryConfig(Config):
         - width: 800
           height: 600
           method: scale
+
+        # Is the preview URL API enabled?  If enabled, you *must* specify
+        # an explicit url_preview_ip_range_blacklist of IPs that the spider is
+        # denied from accessing.
+        url_preview_enabled: False
+
+        # List of IP address CIDR ranges that the URL preview spider is denied
+        # from accessing.  There are no defaults: you must explicitly
+        # specify a list for URL previewing to work.  You should specify any
+        # internal services in your network that you do not want synapse to try
+        # to connect to, otherwise anyone in any Matrix room could cause your
+        # synapse to issue arbitrary GET requests to your internal services,
+        # causing serious security issues.
+        #
+        # url_preview_ip_range_blacklist:
+        # - '127.0.0.0/8'
+        # - '10.0.0.0/8'
+        # - '172.16.0.0/12'
+        # - '192.168.0.0/16'
+
+        # Optional list of URL matches that the URL preview spider is
+        # denied from accessing.  You should use url_preview_ip_range_blacklist
+        # in preference to this, otherwise someone could define a public DNS
+        # entry that points to a private IP address and circumvent the blacklist.
+        # This is more useful if you know there is an entire shape of URL that
+        # you know that will never want synapse to try to spider.
+        #
+        # Each list entry is a dictionary of url component attributes as returned
+        # by urlparse.urlsplit as applied to the absolute form of the URL.  See
+        # https://docs.python.org/2/library/urlparse.html#urlparse.urlsplit
+        # The values of the dictionary are treated as an filename match pattern
+        # applied to that component of URLs, unless they start with a ^ in which
+        # case they are treated as a regular expression match.  If all the
+        # specified component matches for a given list item succeed, the URL is
+        # blacklisted.
+        #
+        # url_preview_url_blacklist:
+        # # blacklist any URL with a username in its URI
+        # - username: '*'
+        #
+        # # blacklist all *.google.com URLs
+        # - netloc: 'google.com'
+        # - netloc: '*.google.com'
+        #
+        # # blacklist all plain HTTP URLs
+        # - scheme: 'http'
+        #
+        # # blacklist http(s)://www.acme.com/foo
+        # - netloc: 'www.acme.com'
+        #   path: '/foo'
+        #
+        # # blacklist any URL with a literal IPv4 address
+        # - netloc: '^[0-9]+\.[0-9]+\.[0-9]+\.[0-9]+$'
+
+        # The largest allowed URL preview spidering size in bytes
+        max_spider_size: "10M"
+
+
         """ % locals()

synapse/http/client.py

@@ -15,17 +15,24 @@
 from OpenSSL import SSL
 from OpenSSL.SSL import VERIFY_NONE
 
-from synapse.api.errors import CodeMessageException
+from synapse.api.errors import (
+    CodeMessageException, SynapseError, Codes,
+)
 from synapse.util.logcontext import preserve_context_over_fn
 import synapse.metrics
+from synapse.http.endpoint import SpiderEndpoint
 
 from canonicaljson import encode_canonical_json
 
-from twisted.internet import defer, reactor, ssl
+from twisted.internet import defer, reactor, ssl, protocol
+from twisted.internet.endpoints import SSL4ClientEndpoint, TCP4ClientEndpoint
 from twisted.web.client import (
-    Agent, readBody, FileBodyProducer, PartialDownloadError,
+    BrowserLikeRedirectAgent, ContentDecoderAgent, GzipDecoder, Agent,
+    readBody, FileBodyProducer, PartialDownloadError,
 )
+from twisted.web.http import PotentialDataLoss
 from twisted.web.http_headers import Headers
+from twisted.web._newclient import ResponseDone
 
 from StringIO import StringIO
 
@@ -238,6 +245,107 @@ class SimpleHttpClient(object):
         else:
             raise CodeMessageException(response.code, body)
 
+    # XXX: FIXME: This is horribly copy-pasted from matrixfederationclient.
+    # The two should be factored out.
+
+    @defer.inlineCallbacks
+    def get_file(self, url, output_stream, max_size=None):
+        """GETs a file from a given URL
+        Args:
+            url (str): The URL to GET
+            output_stream (file): File to write the response body to.
+        Returns:
+            A (int,dict,string,int) tuple of the file length, dict of the response
+            headers, absolute URI of the response and HTTP response code.
+        """
+
+        response = yield self.request(
+            "GET",
+            url.encode("ascii"),
+            headers=Headers({
+                b"User-Agent": [self.user_agent],
+            })
+        )
+
+        headers = dict(response.headers.getAllRawHeaders())
+
+        if 'Content-Length' in headers and headers['Content-Length'] > max_size:
+            logger.warn("Requested URL is too large > %r bytes" % (self.max_size,))
+            raise SynapseError(
+                502,
+                "Requested file is too large > %r bytes" % (self.max_size,),
+                Codes.TOO_LARGE,
+            )
+
+        if response.code > 299:
+            logger.warn("Got %d when downloading %s" % (response.code, url))
+            raise SynapseError(
+                502,
+                "Got error %d" % (response.code,),
+                Codes.UNKNOWN,
+            )
+
+        # TODO: if our Content-Type is HTML or something, just read the first
+        # N bytes into RAM rather than saving it all to disk only to read it
+        # straight back in again
+
+        try:
+            length = yield preserve_context_over_fn(
+                _readBodyToFile,
+                response, output_stream, max_size
+            )
+        except Exception as e:
+            logger.exception("Failed to download body")
+            raise SynapseError(
+                502,
+                ("Failed to download remote body: %s" % e),
+                Codes.UNKNOWN,
+            )
+
+        defer.returnValue((length, headers, response.request.absoluteURI, response.code))
+
+
+# XXX: FIXME: This is horribly copy-pasted from matrixfederationclient.
+# The two should be factored out.
+
+class _ReadBodyToFileProtocol(protocol.Protocol):
+    def __init__(self, stream, deferred, max_size):
+        self.stream = stream
+        self.deferred = deferred
+        self.length = 0
+        self.max_size = max_size
+
+    def dataReceived(self, data):
+        self.stream.write(data)
+        self.length += len(data)
+        if self.max_size is not None and self.length >= self.max_size:
+            self.deferred.errback(SynapseError(
+                502,
+                "Requested file is too large > %r bytes" % (self.max_size,),
+                Codes.TOO_LARGE,
+            ))
+            self.deferred = defer.Deferred()
+            self.transport.loseConnection()
+
+    def connectionLost(self, reason):
+        if reason.check(ResponseDone):
+            self.deferred.callback(self.length)
+        elif reason.check(PotentialDataLoss):
+            # stolen from https://github.com/twisted/treq/pull/49/files
+            # http://twistedmatrix.com/trac/ticket/4840
+            self.deferred.callback(self.length)
+        else:
+            self.deferred.errback(reason)
+
+
+# XXX: FIXME: This is horribly copy-pasted from matrixfederationclient.
+# The two should be factored out.
+
+def _readBodyToFile(response, stream, max_size):
+    d = defer.Deferred()
+    response.deliverBody(_ReadBodyToFileProtocol(stream, d, max_size))
+    return d
+
 
 class CaptchaServerHttpClient(SimpleHttpClient):
     """
@@ -269,6 +377,59 @@ class CaptchaServerHttpClient(SimpleHttpClient):
             defer.returnValue(e.response)
 
 
+class SpiderEndpointFactory(object):
+    def __init__(self, hs):
+        self.blacklist = hs.config.url_preview_ip_range_blacklist
+        self.policyForHTTPS = hs.get_http_client_context_factory()
+
+    def endpointForURI(self, uri):
+        logger.info("Getting endpoint for %s", uri.toBytes())
+        if uri.scheme == "http":
+            return SpiderEndpoint(
+                reactor, uri.host, uri.port, self.blacklist,
+                endpoint=TCP4ClientEndpoint,
+                endpoint_kw_args={
+                    'timeout': 15
+                },
+            )
+        elif uri.scheme == "https":
+            tlsPolicy = self.policyForHTTPS.creatorForNetloc(uri.host, uri.port)
+            return SpiderEndpoint(
+                reactor, uri.host, uri.port, self.blacklist,
+                endpoint=SSL4ClientEndpoint,
+                endpoint_kw_args={
+                    'sslContextFactory': tlsPolicy,
+                    'timeout': 15
+                },
+            )
+        else:
+            logger.warn("Can't get endpoint for unrecognised scheme %s", uri.scheme)
+
+
+class SpiderHttpClient(SimpleHttpClient):
+    """
+    Separate HTTP client for spidering arbitrary URLs.
+    Special in that it follows retries and has a UA that looks
+    like a browser.
+
+    used by the preview_url endpoint in the content repo.
+    """
+    def __init__(self, hs):
+        SimpleHttpClient.__init__(self, hs)
+        # clobber the base class's agent and UA:
+        self.agent = ContentDecoderAgent(
+            BrowserLikeRedirectAgent(
+                Agent.usingEndpointFactory(
+                    reactor,
+                    SpiderEndpointFactory(hs)
+                )
+            ), [('gzip', GzipDecoder)]
+        )
+        # We could look like Chrome:
+        # self.user_agent = ("Mozilla/5.0 (%s) (KHTML, like Gecko)
+        #                   Chrome Safari" % hs.version_string)
+
+
 def encode_urlencode_args(args):
     return {k: encode_urlencode_arg(v) for k, v in args.items()}
 

synapse/http/endpoint.py

@@ -75,6 +75,37 @@ def matrix_federation_endpoint(reactor, destination, ssl_context_factory=None,
         return transport_endpoint(reactor, domain, port, **endpoint_kw_args)
 
 
+class SpiderEndpoint(object):
+    """An endpoint which refuses to connect to blacklisted IP addresses
+    Implements twisted.internet.interfaces.IStreamClientEndpoint.
+    """
+    def __init__(self, reactor, host, port, blacklist,
+                 endpoint=TCP4ClientEndpoint, endpoint_kw_args={}):
+        self.reactor = reactor
+        self.host = host
+        self.port = port
+        self.blacklist = blacklist
+        self.endpoint = endpoint
+        self.endpoint_kw_args = endpoint_kw_args
+
+    @defer.inlineCallbacks
+    def connect(self, protocolFactory):
+        address = yield self.reactor.resolve(self.host)
+
+        from netaddr import IPAddress
+        if IPAddress(address) in self.blacklist:
+            raise ConnectError(
+                "Refusing to spider blacklisted IP address %s" % address
+            )
+
+        logger.info("Connecting to %s:%s", address, self.port)
+        endpoint = self.endpoint(
+            self.reactor, address, self.port, **self.endpoint_kw_args
+        )
+        connection = yield endpoint.connect(protocolFactory)
+        defer.returnValue(connection)
+
+
 class SRVClientEndpoint(object):
     """An endpoint which looks up SRV records for a service.
     Cycles through the list of servers starting with each call to connect
@@ -120,7 +151,7 @@ class SRVClientEndpoint(object):
                 return self.default_server
             else:
                 raise ConnectError(
-                    "Not server available for %s", self.service_name
+                    "Not server available for %s" % self.service_name
                 )
 
         min_priority = self.servers[0].priority
@@ -174,7 +205,7 @@ def resolve_service(service_name, dns_client=client, cache=SERVER_CACHE, clock=t
                 and answers[0].type == dns.SRV
                 and answers[0].payload
                 and answers[0].payload.target == dns.Name('.')):
-            raise ConnectError("Service %s unavailable", service_name)
+            raise ConnectError("Service %s unavailable" % service_name)
 
         for answer in answers:
             if answer.type != dns.SRV or not answer.payload:

synapse/python_dependencies.py

@@ -41,7 +41,11 @@ REQUIREMENTS = {
 CONDITIONAL_REQUIREMENTS = {
     "web_client": {
         "matrix_angular_sdk>=0.6.8": ["syweb>=0.6.8"],
-    }
+    },
+    "preview_url": {
+        "lxml>=3.6.0": ["lxml"],
+        "netaddr>=0.7.18": ["netaddr"],
+    },
 }
 
 

synapse/rest/media/v1/base_resource.py

@@ -72,6 +72,7 @@ class BaseMediaResource(Resource):
         self.store = hs.get_datastore()
         self.max_upload_size = hs.config.max_upload_size
         self.max_image_pixels = hs.config.max_image_pixels
+        self.max_spider_size = hs.config.max_spider_size
         self.filepaths = filepaths
         self.version_string = hs.version_string
         self.downloads = {}

synapse/rest/media/v1/media_repository.py

@@ -17,6 +17,7 @@ from .upload_resource import UploadResource
 from .download_resource import DownloadResource
 from .thumbnail_resource import ThumbnailResource
 from .identicon_resource import IdenticonResource
+from .preview_url_resource import PreviewUrlResource
 from .filepath import MediaFilePaths
 
 from twisted.web.resource import Resource
@@ -78,3 +79,9 @@ class MediaRepositoryResource(Resource):
         self.putChild("download", DownloadResource(hs, filepaths))
         self.putChild("thumbnail", ThumbnailResource(hs, filepaths))
         self.putChild("identicon", IdenticonResource())
+        if hs.config.url_preview_enabled:
+            try:
+                self.putChild("preview_url", PreviewUrlResource(hs, filepaths))
+            except Exception as e:
+                logger.warn("Failed to mount preview_url")
+                logger.exception(e)

synapse/rest/media/v1/preview_url_resource.py

@@ -0,0 +1,462 @@
+# -*- 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.
+
+from .base_resource import BaseMediaResource
+
+from twisted.web.server import NOT_DONE_YET
+from twisted.internet import defer
+from urlparse import urlparse, urlsplit, urlunparse
+
+from synapse.api.errors import (
+    SynapseError, Codes,
+)
+from synapse.util.stringutils import random_string
+from synapse.util.caches.expiringcache import ExpiringCache
+from synapse.http.client import SpiderHttpClient
+from synapse.http.server import (
+    request_handler, respond_with_json_bytes
+)
+from synapse.util.async import ObservableDeferred
+from synapse.util.stringutils import is_ascii
+
+import os
+import re
+import fnmatch
+import cgi
+import ujson as json
+
+import logging
+logger = logging.getLogger(__name__)
+
+try:
+    from lxml import html
+except ImportError:
+    pass
+
+
+class PreviewUrlResource(BaseMediaResource):
+    isLeaf = True
+
+    def __init__(self, hs, filepaths):
+        try:
+            if html:
+                pass
+        except:
+            raise RuntimeError("Disabling PreviewUrlResource as lxml not available")
+
+        if not hasattr(hs.config, "url_preview_ip_range_blacklist"):
+            logger.warn(
+                "For security, you must specify an explicit target IP address "
+                "blacklist in url_preview_ip_range_blacklist for url previewing "
+                "to work"
+            )
+            raise RuntimeError(
+                "Disabling PreviewUrlResource as "
+                "url_preview_ip_range_blacklist not specified"
+            )
+
+        BaseMediaResource.__init__(self, hs, filepaths)
+        self.client = SpiderHttpClient(hs)
+        if hasattr(hs.config, "url_preview_url_blacklist"):
+            self.url_preview_url_blacklist = hs.config.url_preview_url_blacklist
+
+        # simple memory cache mapping urls to OG metadata
+        self.cache = ExpiringCache(
+            cache_name="url_previews",
+            clock=self.clock,
+            # don't spider URLs more often than once an hour
+            expiry_ms=60 * 60 * 1000,
+        )
+        self.cache.start()
+
+        self.downloads = {}
+
+    def render_GET(self, request):
+        self._async_render_GET(request)
+        return NOT_DONE_YET
+
+    @request_handler
+    @defer.inlineCallbacks
+    def _async_render_GET(self, request):
+
+        # XXX: if get_user_by_req fails, what should we do in an async render?
+        requester = yield self.auth.get_user_by_req(request)
+        url = request.args.get("url")[0]
+        if "ts" in request.args:
+            ts = int(request.args.get("ts")[0])
+        else:
+            ts = self.clock.time_msec()
+
+        # impose the URL pattern blacklist
+        if hasattr(self, "url_preview_url_blacklist"):
+            url_tuple = urlsplit(url)
+            for entry in self.url_preview_url_blacklist:
+                match = True
+                for attrib in entry:
+                    pattern = entry[attrib]
+                    value = getattr(url_tuple, attrib)
+                    logger.debug((
+                        "Matching attrib '%s' with value '%s' against"
+                        " pattern '%s'"
+                    ) % (attrib, value, pattern))
+
+                    if value is None:
+                        match = False
+                        continue
+
+                    if pattern.startswith('^'):
+                        if not re.match(pattern, getattr(url_tuple, attrib)):
+                            match = False
+                            continue
+                    else:
+                        if not fnmatch.fnmatch(getattr(url_tuple, attrib), pattern):
+                            match = False
+                            continue
+                if match:
+                    logger.warn(
+                        "URL %s blocked by url_blacklist entry %s", url, entry
+                    )
+                    raise SynapseError(
+                        403, "URL blocked by url pattern blacklist entry",
+                        Codes.UNKNOWN
+                    )
+
+        # first check the memory cache - good to handle all the clients on this
+        # HS thundering away to preview the same URL at the same time.
+        og = self.cache.get(url)
+        if og:
+            respond_with_json_bytes(request, 200, json.dumps(og), send_cors=True)
+            return
+
+        # then check the URL cache in the DB (which will also provide us with
+        # historical previews, if we have any)
+        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["response_code"] / 100 == 2
+        ):
+            respond_with_json_bytes(
+                request, 200, cache_result["og"].encode('utf-8'),
+                send_cors=True
+            )
+            return
+
+        # Ensure only one download for a given URL is active at a time
+        download = self.downloads.get(url)
+        if download is None:
+            download = self._download_url(url, requester.user)
+            download = ObservableDeferred(
+                download,
+                consumeErrors=True
+            )
+            self.downloads[url] = download
+
+            @download.addBoth
+            def callback(media_info):
+                del self.downloads[url]
+                return media_info
+        media_info = yield download.observe()
+
+        # FIXME: we should probably update our cache now anyway, so that
+        # even if the OG calculation raises, we don't keep hammering on the
+        # remote server.  For now, leave it uncached to aid debugging OG
+        # calculation problems
+
+        logger.debug("got media_info of '%s'" % media_info)
+
+        if self._is_media(media_info['media_type']):
+            dims = yield self._generate_local_thumbnails(
+                media_info['filesystem_id'], media_info
+            )
+
+            og = {
+                "og:description": media_info['download_name'],
+                "og:image": "mxc://%s/%s" % (
+                    self.server_name, media_info['filesystem_id']
+                ),
+                "og:image:type": media_info['media_type'],
+                "matrix:image:size": media_info['media_length'],
+            }
+
+            if dims:
+                og["og:image:width"] = dims['width']
+                og["og:image:height"] = dims['height']
+            else:
+                logger.warn("Couldn't get dims for %s" % url)
+
+            # define our OG response for this media
+        elif self._is_html(media_info['media_type']):
+            # TODO: somehow stop a big HTML tree from exploding synapse's RAM
+
+            try:
+                tree = html.parse(media_info['filename'])
+                og = yield self._calc_og(tree, media_info, requester)
+            except UnicodeDecodeError:
+                # XXX: evil evil bodge
+                # Empirically, sites like google.com mix Latin-1 and utf-8
+                # encodings in the same page.  The rogue Latin-1 characters
+                # cause lxml to choke with a UnicodeDecodeError, so if we
+                # see this we go and do a manual decode of the HTML before
+                # handing it to lxml as utf-8 encoding, counter-intuitively,
+                # which seems to make it happier...
+                file = open(media_info['filename'])
+                body = file.read()
+                file.close()
+                tree = html.fromstring(body.decode('utf-8', 'ignore'))
+                og = yield self._calc_og(tree, media_info, requester)
+
+        else:
+            logger.warn("Failed to find any OG data in %s", url)
+            og = {}
+
+        logger.debug("Calculated OG for %s as %s" % (url, og))
+
+        # store OG in ephemeral in-memory cache
+        self.cache[url] = og
+
+        # store OG in history-aware DB cache
+        yield self.store.store_url_cache(
+            url,
+            media_info["response_code"],
+            media_info["etag"],
+            media_info["expires"],
+            json.dumps(og),
+            media_info["filesystem_id"],
+            media_info["created_ts"],
+        )
+
+        respond_with_json_bytes(request, 200, json.dumps(og), send_cors=True)
+
+
+    @defer.inlineCallbacks
+    def _calc_og(self, tree, media_info, requester):
+        # suck our tree into lxml and define our OG response.
+
+        # if we see any image URLs in the OG response, then spider them
+        # (although the client could choose to do this by asking for previews of those
+        # URLs to avoid DoSing the server)
+
+        # "og:type"         : "video",
+        # "og:url"          : "https://www.youtube.com/watch?v=LXDBoHyjmtw",
+        # "og:site_name"    : "YouTube",
+        # "og:video:type"   : "application/x-shockwave-flash",
+        # "og:description"  : "Fun stuff happening here",
+        # "og:title"        : "RemoteJam - Matrix team hack for Disrupt Europe Hackathon",
+        # "og:image"        : "https://i.ytimg.com/vi/LXDBoHyjmtw/maxresdefault.jpg",
+        # "og:video:url"    : "http://www.youtube.com/v/LXDBoHyjmtw?version=3&autohide=1",
+        # "og:video:width"  : "1280"
+        # "og:video:height" : "720",
+        # "og:video:secure_url": "https://www.youtube.com/v/LXDBoHyjmtw?version=3",
+
+        og = {}
+        for tag in tree.xpath("//*/meta[starts-with(@property, 'og:')]"):
+            og[tag.attrib['property']] = tag.attrib['content']
+
+        # TODO: grab article: meta tags too, e.g.:
+
+        # "article:publisher" : "https://www.facebook.com/thethudonline" />
+        # "article:author" content="https://www.facebook.com/thethudonline" />
+        # "article:tag" content="baby" />
+        # "article:section" content="Breaking News" />
+        # "article:published_time" content="2016-03-31T19:58:24+00:00" />
+        # "article:modified_time" content="2016-04-01T18:31:53+00:00" />
+
+        if 'og:title' not in og:
+            # do some basic spidering of the HTML
+            title = tree.xpath("(//title)[1] | (//h1)[1] | (//h2)[1] | (//h3)[1]")
+            og['og:title'] = title[0].text.strip() if title else None
+
+        if 'og:image' not in og:
+            # TODO: extract a favicon failing all else
+            meta_image = tree.xpath(
+                "//*/meta[translate(@itemprop, 'IMAGE', 'image')='image']/@content"
+            )
+            if meta_image:
+                og['og:image'] = self._rebase_url(meta_image[0], media_info['uri'])
+            else:
+                # TODO: consider inlined CSS styles as well as width & height attribs
+                images = tree.xpath("//img[@src][number(@width)>10][number(@height)>10]")
+                images = sorted(images, key=lambda i: (
+                    -1 * int(i.attrib['width']) * int(i.attrib['height'])
+                ))
+                if not images:
+                    images = tree.xpath("//img[@src]")
+                if images:
+                    og['og:image'] = images[0].attrib['src']
+
+        # pre-cache the image for posterity
+        # FIXME: it might be cleaner to use the same flow as the main /preview_url request
+        # itself and benefit from the same caching etc.  But for now we just rely on the
+        # caching on the master request to speed things up.
+        if 'og:image' in og and og['og:image']:
+            image_info = yield self._download_url(
+                self._rebase_url(og['og:image'], media_info['uri']), requester.user
+            )
+
+            if self._is_media(image_info['media_type']):
+                # TODO: make sure we don't choke on white-on-transparent images
+                dims = yield self._generate_local_thumbnails(
+                    image_info['filesystem_id'], image_info
+                )
+                if dims:
+                    og["og:image:width"] = dims['width']
+                    og["og:image:height"] = dims['height']
+                else:
+                    logger.warn("Couldn't get dims for %s" % og["og:image"])
+
+                og["og:image"] = "mxc://%s/%s" % (
+                    self.server_name, image_info['filesystem_id']
+                )
+                og["og:image:type"] = image_info['media_type']
+                og["matrix:image:size"] = image_info['media_length']
+            else:
+                del og["og:image"]
+
+        if 'og:description' not in og:
+            meta_description = tree.xpath(
+                "//*/meta"
+                "[translate(@name, 'DESCRIPTION', 'description')='description']"
+                "/@content")
+            if meta_description:
+                og['og:description'] = meta_description[0]
+            else:
+                # grab any text nodes which are inside the <body/> tag...
+                # unless they are within an HTML5 semantic markup tag...
+                # <header/>, <nav/>, <aside/>, <footer/>
+                # ...or if they are within a <script/> or <style/> tag.
+                # This is a very very very coarse approximation to a plain text
+                # render of the page.
+                text_nodes = tree.xpath("//text()[not(ancestor::header | ancestor::nav | "
+                                        "ancestor::aside | ancestor::footer | "
+                                        "ancestor::script | ancestor::style)]" +
+                                        "[ancestor::body]")
+                text = ''
+                for text_node in text_nodes:
+                    if len(text) < 500:
+                        text += text_node + ' '
+                    else:
+                        break
+                text = re.sub(r'[\t ]+', ' ', text)
+                text = re.sub(r'[\t \r\n]*[\r\n]+', '\n', text)
+                text = text.strip()[:500]
+                og['og:description'] = text if text else None
+
+        # TODO: delete the url downloads to stop diskfilling,
+        # as we only ever cared about its OG
+        defer.returnValue(og)
+
+    def _rebase_url(self, url, base):
+        base = list(urlparse(base))
+        url = list(urlparse(url))
+        if not url[0]:  # fix up schema
+            url[0] = base[0] or "http"
+        if not url[1]:  # fix up hostname
+            url[1] = base[1]
+            if not url[2].startswith('/'):
+                url[2] = re.sub(r'/[^/]+$', '/', base[2]) + url[2]
+        return urlunparse(url)
+
+    @defer.inlineCallbacks
+    def _download_url(self, url, user):
+        # TODO: we should probably honour robots.txt... except in practice
+        # 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)
+
+        fname = self.filepaths.local_media_filepath(file_id)
+        self._makedirs(fname)
+
+        try:
+            with open(fname, "wb") as f:
+                logger.debug("Trying to get url '%s'" % url)
+                length, headers, uri, code = yield self.client.get_file(
+                    url, output_stream=f, max_size=self.max_spider_size,
+                )
+                # FIXME: pass through 404s and other error messages nicely
+
+            media_type = headers["Content-Type"][0]
+            time_now_ms = self.clock.time_msec()
+
+            content_disposition = headers.get("Content-Disposition", None)
+            if content_disposition:
+                _, params = cgi.parse_header(content_disposition[0],)
+                download_name = None
+
+                # First check if there is a valid UTF-8 filename
+                download_name_utf8 = params.get("filename*", None)
+                if download_name_utf8:
+                    if download_name_utf8.lower().startswith("utf-8''"):
+                        download_name = download_name_utf8[7:]
+
+                # If there isn't check for an ascii name.
+                if not download_name:
+                    download_name_ascii = params.get("filename", None)
+                    if download_name_ascii and is_ascii(download_name_ascii):
+                        download_name = download_name_ascii
+
+                if download_name:
+                    download_name = urlparse.unquote(download_name)
+                    try:
+                        download_name = download_name.decode("utf-8")
+                    except UnicodeDecodeError:
+                        download_name = None
+            else:
+                download_name = None
+
+            yield self.store.store_local_media(
+                media_id=file_id,
+                media_type=media_type,
+                time_now_ms=self.clock.time_msec(),
+                upload_name=download_name,
+                media_length=length,
+                user_id=user,
+            )
+
+        except Exception as e:
+            os.remove(fname)
+            raise SynapseError(
+                500, ("Failed to download content: %s" % e),
+                Codes.UNKNOWN
+            )
+
+        defer.returnValue({
+            "media_type": media_type,
+            "media_length": length,
+            "download_name": download_name,
+            "created_ts": time_now_ms,
+            "filesystem_id": file_id,
+            "filename": fname,
+            "uri": uri,
+            "response_code": code,
+            # FIXME: we should calculate a proper expiration based on the
+            # Cache-Control and Expire headers.  But for now, assume 1 hour.
+            "expires": 60 * 60 * 1000,
+            "etag": headers["ETag"][0] if "ETag" in headers else None,
+        })
+
+    def _is_media(self, content_type):
+        if content_type.lower().startswith("image/"):
+            return True
+
+    def _is_html(self, content_type):
+        content_type = content_type.lower()
+        if (
+            content_type.startswith("text/html") or
+            content_type.startswith("application/xhtml")
+        ):
+            return True

synapse/rest/media/v1/thumbnail_resource.py

@@ -72,6 +72,11 @@ class ThumbnailResource(BaseMediaResource):
             self._respond_404(request)
             return
 
+        if media_info["media_type"] == "image/svg+xml":
+            file_path = self.filepaths.local_media_filepath(media_id)
+            yield self._respond_with_file(request, media_info["media_type"], file_path)
+            return
+
         thumbnail_infos = yield self.store.get_local_media_thumbnails(media_id)
 
         if thumbnail_infos:
@@ -103,6 +108,11 @@ class ThumbnailResource(BaseMediaResource):
             self._respond_404(request)
             return
 
+        if media_info["media_type"] == "image/svg+xml":
+            file_path = self.filepaths.local_media_filepath(media_id)
+            yield self._respond_with_file(request, media_info["media_type"], file_path)
+            return
+
         thumbnail_infos = yield self.store.get_local_media_thumbnails(media_id)
         for info in thumbnail_infos:
             t_w = info["thumbnail_width"] == desired_width
@@ -138,6 +148,11 @@ class ThumbnailResource(BaseMediaResource):
                                              desired_method, desired_type):
         media_info = yield self._get_remote_media(server_name, media_id)
 
+        if media_info["media_type"] == "image/svg+xml":
+            file_path = self.filepaths.remote_media_filepath(server_name, media_id)
+            yield self._respond_with_file(request, media_info["media_type"], file_path)
+            return
+
         thumbnail_infos = yield self.store.get_remote_media_thumbnails(
             server_name, media_id,
         )
@@ -181,6 +196,11 @@ class ThumbnailResource(BaseMediaResource):
         # We should proxy the thumbnail from the remote server instead.
         media_info = yield self._get_remote_media(server_name, media_id)
 
+        if media_info["media_type"] == "image/svg+xml":
+            file_path = self.filepaths.remote_media_filepath(server_name, media_id)
+            yield self._respond_with_file(request, media_info["media_type"], file_path)
+            return
+
         thumbnail_infos = yield self.store.get_remote_media_thumbnails(
             server_name, media_id,
         )
@@ -208,6 +228,8 @@ class ThumbnailResource(BaseMediaResource):
     @defer.inlineCallbacks
     def _respond_default_thumbnail(self, request, media_info, width, height,
                                    method, m_type):
+        # XXX: how is this meant to work? store.get_default_thumbnails
+        # appears to always return [] so won't this always 404?
         media_type = media_info["media_type"]
         top_level_type = media_type.split("/")[0]
         sub_type = media_type.split("/")[-1].split(";")[0]

synapse/storage/media_repository.py

@@ -25,7 +25,7 @@ class MediaRepositoryStore(SQLBaseStore):
     def get_local_media(self, media_id):
         """Get the metadata for a local piece of media
         Returns:
-            None if the meia_id doesn't exist.
+            None if the media_id doesn't exist.
         """
         return self._simple_select_one(
             "local_media_repository",
@@ -50,6 +50,61 @@ class MediaRepositoryStore(SQLBaseStore):
             desc="store_local_media",
         )
 
+    def get_url_cache(self, url, ts):
+        """Get the media_id and ts for a cached URL as of the given timestamp
+        Returns:
+            None if the URL isn't cached.
+        """
+        def get_url_cache_txn(txn):
+            # get the most recently cached result (relative to the given ts)
+            sql = (
+                "SELECT response_code, etag, expires, og, media_id, download_ts"
+                " FROM local_media_repository_url_cache"
+                " WHERE url = ? AND download_ts <= ?"
+                " ORDER BY download_ts DESC LIMIT 1"
+            )
+            txn.execute(sql, (url, ts))
+            row = txn.fetchone()
+
+            if not row:
+                # ...or if we've requested a timestamp older than the oldest
+                # copy in the cache, return the oldest copy (if any)
+                sql = (
+                    "SELECT response_code, etag, expires, og, media_id, download_ts"
+                    " FROM local_media_repository_url_cache"
+                    " WHERE url = ? AND download_ts > ?"
+                    " ORDER BY download_ts ASC LIMIT 1"
+                )
+                txn.execute(sql, (url, ts))
+                row = txn.fetchone()
+
+            if not row:
+                return None
+
+            return dict(zip((
+                'response_code', 'etag', 'expires', 'og', 'media_id', 'download_ts'
+            ), row))
+
+        return self.runInteraction(
+            "get_url_cache", get_url_cache_txn
+        )
+
+    def store_url_cache(self, url, response_code, etag, expires, og, media_id,
+                        download_ts):
+        return self._simple_insert(
+            "local_media_repository_url_cache",
+            {
+                "url": url,
+                "response_code": response_code,
+                "etag": etag,
+                "expires": expires,
+                "og": og,
+                "media_id": media_id,
+                "download_ts": download_ts,
+            },
+            desc="store_url_cache",
+        )
+
     def get_local_media_thumbnails(self, media_id):
         return self._simple_select_list(
             "local_media_repository_thumbnails",

synapse/storage/schema/delta/31/local_media_repository_url_cache.sql

@@ -0,0 +1,27 @@
+/* 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.
+ */
+
+CREATE TABLE local_media_repository_url_cache(
+    url TEXT,              -- the URL being cached
+    response_code INTEGER, -- the HTTP response code of this download attempt
+    etag TEXT,             -- the etag header of this response
+    expires INTEGER,       -- the number of ms this response was valid for
+    og TEXT,               -- cache of the OG metadata of this URL as JSON
+    media_id TEXT,         -- the media_id, if any, of the URL's content in the repo
+    download_ts BIGINT     -- the timestamp of this download attempt
+);
+
+CREATE INDEX local_media_repository_url_cache_by_url_download_ts
+    ON local_media_repository_url_cache(url, download_ts);