diff --git a/changelog.d/17213.feature b/changelog.d/17213.feature new file mode 100644 index 0000000000..ca60afa8f3 --- /dev/null +++ b/changelog.d/17213.feature @@ -0,0 +1 @@ +Support MSC3916 by adding unstable media endpoints to `_matrix/client` (#17213). \ No newline at end of file diff --git a/synapse/config/experimental.py b/synapse/config/experimental.py index cda7afc5c4..75fe6d7b24 100644 --- a/synapse/config/experimental.py +++ b/synapse/config/experimental.py @@ -439,3 +439,7 @@ class ExperimentalConfig(Config): self.msc4115_membership_on_events = experimental.get( "msc4115_membership_on_events", False ) + + self.msc3916_authenticated_media_enabled = experimental.get( + "msc3916_authenticated_media_enabled", False + ) diff --git a/synapse/media/thumbnailer.py b/synapse/media/thumbnailer.py index 5538020bec..cc3acf51e1 100644 --- a/synapse/media/thumbnailer.py +++ b/synapse/media/thumbnailer.py @@ -22,11 +22,27 @@ import logging from io import BytesIO from types import TracebackType -from typing import Optional, Tuple, Type +from typing import TYPE_CHECKING, List, Optional, Tuple, Type from PIL import Image +from synapse.api.errors import Codes, SynapseError, cs_error +from synapse.config.repository import THUMBNAIL_SUPPORTED_MEDIA_FORMAT_MAP +from synapse.http.server import respond_with_json +from synapse.http.site import SynapseRequest from synapse.logging.opentracing import trace +from synapse.media._base import ( + FileInfo, + ThumbnailInfo, + respond_404, + respond_with_file, + respond_with_responder, +) +from synapse.media.media_storage import MediaStorage + +if TYPE_CHECKING: + from synapse.media.media_repository import MediaRepository + from synapse.server import HomeServer logger = logging.getLogger(__name__) @@ -231,3 +247,471 @@ class Thumbnailer: def __del__(self) -> None: # Make sure we actually do close the image, rather than leak data. self.close() + + +class ThumbnailProvider: + def __init__( + self, + hs: "HomeServer", + media_repo: "MediaRepository", + media_storage: MediaStorage, + ): + self.hs = hs + self.media_repo = media_repo + self.media_storage = media_storage + self.store = hs.get_datastores().main + self.dynamic_thumbnails = hs.config.media.dynamic_thumbnails + + async def respond_local_thumbnail( + self, + request: SynapseRequest, + media_id: str, + width: int, + height: int, + method: str, + m_type: str, + max_timeout_ms: int, + ) -> None: + media_info = await self.media_repo.get_local_media_info( + request, media_id, max_timeout_ms + ) + if not media_info: + return + + thumbnail_infos = await self.store.get_local_media_thumbnails(media_id) + await self._select_and_respond_with_thumbnail( + request, + width, + height, + method, + m_type, + thumbnail_infos, + media_id, + media_id, + url_cache=bool(media_info.url_cache), + server_name=None, + ) + + async def select_or_generate_local_thumbnail( + self, + request: SynapseRequest, + media_id: str, + desired_width: int, + desired_height: int, + desired_method: str, + desired_type: str, + max_timeout_ms: int, + ) -> None: + media_info = await self.media_repo.get_local_media_info( + request, media_id, max_timeout_ms + ) + + if not media_info: + return + + thumbnail_infos = await self.store.get_local_media_thumbnails(media_id) + for info in thumbnail_infos: + t_w = info.width == desired_width + t_h = info.height == desired_height + t_method = info.method == desired_method + t_type = info.type == desired_type + + if t_w and t_h and t_method and t_type: + file_info = FileInfo( + server_name=None, + file_id=media_id, + url_cache=bool(media_info.url_cache), + thumbnail=info, + ) + + responder = await self.media_storage.fetch_media(file_info) + if responder: + await respond_with_responder( + request, responder, info.type, info.length + ) + return + + logger.debug("We don't have a thumbnail of that size. Generating") + + # Okay, so we generate one. + file_path = await self.media_repo.generate_local_exact_thumbnail( + media_id, + desired_width, + desired_height, + desired_method, + desired_type, + url_cache=bool(media_info.url_cache), + ) + + if file_path: + await respond_with_file(request, desired_type, file_path) + else: + logger.warning("Failed to generate thumbnail") + raise SynapseError(400, "Failed to generate thumbnail.") + + async def select_or_generate_remote_thumbnail( + self, + request: SynapseRequest, + server_name: str, + media_id: str, + desired_width: int, + desired_height: int, + desired_method: str, + desired_type: str, + max_timeout_ms: int, + ) -> None: + media_info = await self.media_repo.get_remote_media_info( + server_name, media_id, max_timeout_ms + ) + if not media_info: + respond_404(request) + return + + thumbnail_infos = await self.store.get_remote_media_thumbnails( + server_name, media_id + ) + + file_id = media_info.filesystem_id + + for info in thumbnail_infos: + t_w = info.width == desired_width + t_h = info.height == desired_height + t_method = info.method == desired_method + t_type = info.type == desired_type + + if t_w and t_h and t_method and t_type: + file_info = FileInfo( + server_name=server_name, + file_id=file_id, + thumbnail=info, + ) + + responder = await self.media_storage.fetch_media(file_info) + if responder: + await respond_with_responder( + request, responder, info.type, info.length + ) + return + + logger.debug("We don't have a thumbnail of that size. Generating") + + # Okay, so we generate one. + file_path = await self.media_repo.generate_remote_exact_thumbnail( + server_name, + file_id, + media_id, + desired_width, + desired_height, + desired_method, + desired_type, + ) + + if file_path: + await respond_with_file(request, desired_type, file_path) + else: + logger.warning("Failed to generate thumbnail") + raise SynapseError(400, "Failed to generate thumbnail.") + + async def respond_remote_thumbnail( + self, + request: SynapseRequest, + server_name: str, + media_id: str, + width: int, + height: int, + method: str, + m_type: str, + max_timeout_ms: int, + ) -> None: + # TODO: Don't download the whole remote file + # We should proxy the thumbnail from the remote server instead of + # downloading the remote file and generating our own thumbnails. + media_info = await self.media_repo.get_remote_media_info( + server_name, media_id, max_timeout_ms + ) + if not media_info: + return + + thumbnail_infos = await self.store.get_remote_media_thumbnails( + server_name, media_id + ) + await self._select_and_respond_with_thumbnail( + request, + width, + height, + method, + m_type, + thumbnail_infos, + media_id, + media_info.filesystem_id, + url_cache=False, + server_name=server_name, + ) + + async def _select_and_respond_with_thumbnail( + self, + request: SynapseRequest, + desired_width: int, + desired_height: int, + desired_method: str, + desired_type: str, + thumbnail_infos: List[ThumbnailInfo], + media_id: str, + file_id: str, + url_cache: bool, + server_name: Optional[str] = None, + ) -> None: + """ + Respond to a request with an appropriate thumbnail from the previously generated thumbnails. + + Args: + request: The incoming request. + desired_width: The desired width, the returned thumbnail may be larger than this. + desired_height: The desired height, the returned thumbnail may be larger than this. + desired_method: The desired method used to generate the thumbnail. + desired_type: The desired content-type of the thumbnail. + thumbnail_infos: A list of thumbnail info of candidate thumbnails. + file_id: The ID of the media that a thumbnail is being requested for. + url_cache: True if this is from a URL cache. + server_name: The server name, if this is a remote thumbnail. + """ + logger.debug( + "_select_and_respond_with_thumbnail: media_id=%s desired=%sx%s (%s) thumbnail_infos=%s", + media_id, + desired_width, + desired_height, + desired_method, + thumbnail_infos, + ) + + # If `dynamic_thumbnails` is enabled, we expect Synapse to go down a + # different code path to handle it. + assert not self.dynamic_thumbnails + + if thumbnail_infos: + file_info = self._select_thumbnail( + desired_width, + desired_height, + desired_method, + desired_type, + thumbnail_infos, + file_id, + url_cache, + server_name, + ) + if not file_info: + logger.info("Couldn't find a thumbnail matching the desired inputs") + respond_404(request) + return + + # The thumbnail property must exist. + assert file_info.thumbnail is not None + + responder = await self.media_storage.fetch_media(file_info) + if responder: + await respond_with_responder( + request, + responder, + file_info.thumbnail.type, + file_info.thumbnail.length, + ) + return + + # If we can't find the thumbnail we regenerate it. This can happen + # if e.g. we've deleted the thumbnails but still have the original + # image somewhere. + # + # Since we have an entry for the thumbnail in the DB we a) know we + # have have successfully generated the thumbnail in the past (so we + # don't need to worry about repeatedly failing to generate + # thumbnails), and b) have already calculated that appropriate + # width/height/method so we can just call the "generate exact" + # methods. + + # First let's check that we do actually have the original image + # still. This will throw a 404 if we don't. + # TODO: We should refetch the thumbnails for remote media. + await self.media_storage.ensure_media_is_in_local_cache( + FileInfo(server_name, file_id, url_cache=url_cache) + ) + + if server_name: + await self.media_repo.generate_remote_exact_thumbnail( + server_name, + file_id=file_id, + media_id=media_id, + t_width=file_info.thumbnail.width, + t_height=file_info.thumbnail.height, + t_method=file_info.thumbnail.method, + t_type=file_info.thumbnail.type, + ) + else: + await self.media_repo.generate_local_exact_thumbnail( + media_id=media_id, + t_width=file_info.thumbnail.width, + t_height=file_info.thumbnail.height, + t_method=file_info.thumbnail.method, + t_type=file_info.thumbnail.type, + url_cache=url_cache, + ) + + responder = await self.media_storage.fetch_media(file_info) + await respond_with_responder( + request, + responder, + file_info.thumbnail.type, + file_info.thumbnail.length, + ) + else: + # This might be because: + # 1. We can't create thumbnails for the given media (corrupted or + # unsupported file type), or + # 2. The thumbnailing process never ran or errored out initially + # when the media was first uploaded (these bugs should be + # reported and fixed). + # Note that we don't attempt to generate a thumbnail now because + # `dynamic_thumbnails` is disabled. + logger.info("Failed to find any generated thumbnails") + + assert request.path is not None + respond_with_json( + request, + 400, + cs_error( + "Cannot find any thumbnails for the requested media ('%s'). This might mean the media is not a supported_media_format=(%s) or that thumbnailing failed for some other reason. (Dynamic thumbnails are disabled on this server.)" + % ( + request.path.decode(), + ", ".join(THUMBNAIL_SUPPORTED_MEDIA_FORMAT_MAP.keys()), + ), + code=Codes.UNKNOWN, + ), + send_cors=True, + ) + + def _select_thumbnail( + self, + desired_width: int, + desired_height: int, + desired_method: str, + desired_type: str, + thumbnail_infos: List[ThumbnailInfo], + file_id: str, + url_cache: bool, + server_name: Optional[str], + ) -> Optional[FileInfo]: + """ + Choose an appropriate thumbnail from the previously generated thumbnails. + + Args: + desired_width: The desired width, the returned thumbnail may be larger than this. + desired_height: The desired height, the returned thumbnail may be larger than this. + desired_method: The desired method used to generate the thumbnail. + desired_type: The desired content-type of the thumbnail. + thumbnail_infos: A list of thumbnail infos of candidate thumbnails. + file_id: The ID of the media that a thumbnail is being requested for. + url_cache: True if this is from a URL cache. + server_name: The server name, if this is a remote thumbnail. + + Returns: + The thumbnail which best matches the desired parameters. + """ + desired_method = desired_method.lower() + + # The chosen thumbnail. + thumbnail_info = None + + d_w = desired_width + d_h = desired_height + + if desired_method == "crop": + # Thumbnails that match equal or larger sizes of desired width/height. + crop_info_list: List[ + Tuple[int, int, int, bool, Optional[int], ThumbnailInfo] + ] = [] + # Other thumbnails. + crop_info_list2: List[ + Tuple[int, int, int, bool, Optional[int], ThumbnailInfo] + ] = [] + for info in thumbnail_infos: + # Skip thumbnails generated with different methods. + if info.method != "crop": + continue + + t_w = info.width + t_h = info.height + aspect_quality = abs(d_w * t_h - d_h * t_w) + min_quality = 0 if d_w <= t_w and d_h <= t_h else 1 + size_quality = abs((d_w - t_w) * (d_h - t_h)) + type_quality = desired_type != info.type + length_quality = info.length + if t_w >= d_w or t_h >= d_h: + crop_info_list.append( + ( + aspect_quality, + min_quality, + size_quality, + type_quality, + length_quality, + info, + ) + ) + else: + crop_info_list2.append( + ( + aspect_quality, + min_quality, + size_quality, + type_quality, + length_quality, + info, + ) + ) + # Pick the most appropriate thumbnail. Some values of `desired_width` and + # `desired_height` may result in a tie, in which case we avoid comparing on + # the thumbnail info and pick the thumbnail that appears earlier + # in the list of candidates. + if crop_info_list: + thumbnail_info = min(crop_info_list, key=lambda t: t[:-1])[-1] + elif crop_info_list2: + thumbnail_info = min(crop_info_list2, key=lambda t: t[:-1])[-1] + elif desired_method == "scale": + # Thumbnails that match equal or larger sizes of desired width/height. + info_list: List[Tuple[int, bool, int, ThumbnailInfo]] = [] + # Other thumbnails. + info_list2: List[Tuple[int, bool, int, ThumbnailInfo]] = [] + + for info in thumbnail_infos: + # Skip thumbnails generated with different methods. + if info.method != "scale": + continue + + t_w = info.width + t_h = info.height + size_quality = abs((d_w - t_w) * (d_h - t_h)) + type_quality = desired_type != info.type + length_quality = info.length + if t_w >= d_w or t_h >= d_h: + info_list.append((size_quality, type_quality, length_quality, info)) + else: + info_list2.append( + (size_quality, type_quality, length_quality, info) + ) + # Pick the most appropriate thumbnail. Some values of `desired_width` and + # `desired_height` may result in a tie, in which case we avoid comparing on + # the thumbnail info and pick the thumbnail that appears earlier + # in the list of candidates. + if info_list: + thumbnail_info = min(info_list, key=lambda t: t[:-1])[-1] + elif info_list2: + thumbnail_info = min(info_list2, key=lambda t: t[:-1])[-1] + + if thumbnail_info: + return FileInfo( + file_id=file_id, + url_cache=url_cache, + server_name=server_name, + thumbnail=thumbnail_info, + ) + + # No matching thumbnail was found. + return None diff --git a/synapse/rest/client/media.py b/synapse/rest/client/media.py new file mode 100644 index 0000000000..172d240783 --- /dev/null +++ b/synapse/rest/client/media.py @@ -0,0 +1,205 @@ +# +# This file is licensed under the Affero General Public License (AGPL) version 3. +# +# Copyright 2020 The Matrix.org Foundation C.I.C. +# Copyright 2015, 2016 OpenMarket Ltd +# Copyright (C) 2024 New Vector, Ltd +# +# This program is free software: you can redistribute it and/or modify +# it under the terms of the GNU Affero General Public License as +# published by the Free Software Foundation, either version 3 of the +# License, or (at your option) any later version. +# +# See the GNU Affero General Public License for more details: +# . +# +# Originally licensed under the Apache License, Version 2.0: +# . +# +# [This file includes modifications made by New Vector Limited] +# +# + +import logging +import re + +from synapse.http.server import ( + HttpServer, + respond_with_json, + respond_with_json_bytes, + set_corp_headers, + set_cors_headers, +) +from synapse.http.servlet import RestServlet, parse_integer, parse_string +from synapse.http.site import SynapseRequest +from synapse.media._base import ( + DEFAULT_MAX_TIMEOUT_MS, + MAXIMUM_ALLOWED_MAX_TIMEOUT_MS, + respond_404, +) +from synapse.media.media_repository import MediaRepository +from synapse.media.media_storage import MediaStorage +from synapse.media.thumbnailer import ThumbnailProvider +from synapse.server import HomeServer +from synapse.util.stringutils import parse_and_validate_server_name + +logger = logging.getLogger(__name__) + + +class UnstablePreviewURLServlet(RestServlet): + """ + Same as `GET /_matrix/media/r0/preview_url`, this endpoint provides a generic preview API + for URLs which outputs Open Graph (https://ogp.me/) responses (with some Matrix + specific additions). + + This does have trade-offs compared to other designs: + + * Pros: + * Simple and flexible; can be used by any clients at any point + * Cons: + * If each homeserver provides one of these independently, all the homeservers in a + room may needlessly DoS the target URI + * The URL metadata must be stored somewhere, rather than just using Matrix + itself to store the media. + * Matrix cannot be used to distribute the metadata between homeservers. + """ + + PATTERNS = [ + re.compile(r"^/_matrix/client/unstable/org.matrix.msc3916/media/preview_url$") + ] + + def __init__( + self, + hs: "HomeServer", + media_repo: "MediaRepository", + media_storage: MediaStorage, + ): + super().__init__() + + self.auth = hs.get_auth() + self.clock = hs.get_clock() + self.media_repo = media_repo + self.media_storage = media_storage + assert self.media_repo.url_previewer is not None + self.url_previewer = self.media_repo.url_previewer + + async def on_GET(self, request: SynapseRequest) -> None: + requester = await self.auth.get_user_by_req(request) + url = parse_string(request, "url", required=True) + ts = parse_integer(request, "ts") + if ts is None: + ts = self.clock.time_msec() + + og = await self.url_previewer.preview(url, requester.user, ts) + respond_with_json_bytes(request, 200, og, send_cors=True) + + +class UnstableMediaConfigResource(RestServlet): + PATTERNS = [ + re.compile(r"^/_matrix/client/unstable/org.matrix.msc3916/media/config$") + ] + + def __init__(self, hs: "HomeServer"): + super().__init__() + config = hs.config + self.clock = hs.get_clock() + self.auth = hs.get_auth() + self.limits_dict = {"m.upload.size": config.media.max_upload_size} + + async def on_GET(self, request: SynapseRequest) -> None: + await self.auth.get_user_by_req(request) + respond_with_json(request, 200, self.limits_dict, send_cors=True) + + +class UnstableThumbnailResource(RestServlet): + PATTERNS = [ + re.compile( + "/_matrix/client/unstable/org.matrix.msc3916/media/thumbnail/(?P[^/]*)/(?P[^/]*)$" + ) + ] + + def __init__( + self, + hs: "HomeServer", + media_repo: "MediaRepository", + media_storage: MediaStorage, + ): + super().__init__() + + self.store = hs.get_datastores().main + self.media_repo = media_repo + self.media_storage = media_storage + self.dynamic_thumbnails = hs.config.media.dynamic_thumbnails + self._is_mine_server_name = hs.is_mine_server_name + self._server_name = hs.hostname + self.prevent_media_downloads_from = hs.config.media.prevent_media_downloads_from + self.thumbnailer = ThumbnailProvider(hs, media_repo, media_storage) + self.auth = hs.get_auth() + + async def on_GET( + self, request: SynapseRequest, server_name: str, media_id: str + ) -> None: + # Validate the server name, raising if invalid + parse_and_validate_server_name(server_name) + await self.auth.get_user_by_req(request) + + set_cors_headers(request) + set_corp_headers(request) + width = parse_integer(request, "width", required=True) + height = parse_integer(request, "height", required=True) + method = parse_string(request, "method", "scale") + # TODO Parse the Accept header to get an prioritised list of thumbnail types. + m_type = "image/png" + max_timeout_ms = parse_integer( + request, "timeout_ms", default=DEFAULT_MAX_TIMEOUT_MS + ) + max_timeout_ms = min(max_timeout_ms, MAXIMUM_ALLOWED_MAX_TIMEOUT_MS) + + if self._is_mine_server_name(server_name): + if self.dynamic_thumbnails: + await self.thumbnailer.select_or_generate_local_thumbnail( + request, media_id, width, height, method, m_type, max_timeout_ms + ) + else: + await self.thumbnailer.respond_local_thumbnail( + request, media_id, width, height, method, m_type, max_timeout_ms + ) + self.media_repo.mark_recently_accessed(None, media_id) + else: + # Don't let users download media from configured domains, even if it + # is already downloaded. This is Trust & Safety tooling to make some + # media inaccessible to local users. + # See `prevent_media_downloads_from` config docs for more info. + if server_name in self.prevent_media_downloads_from: + respond_404(request) + return + + remote_resp_function = ( + self.thumbnailer.select_or_generate_remote_thumbnail + if self.dynamic_thumbnails + else self.thumbnailer.respond_remote_thumbnail + ) + await remote_resp_function( + request, + server_name, + media_id, + width, + height, + method, + m_type, + max_timeout_ms, + ) + self.media_repo.mark_recently_accessed(server_name, media_id) + + +def register_servlets(hs: "HomeServer", http_server: HttpServer) -> None: + if hs.config.experimental.msc3916_authenticated_media_enabled: + media_repo = hs.get_media_repository() + if hs.config.media.url_preview_enabled: + UnstablePreviewURLServlet( + hs, media_repo, media_repo.media_storage + ).register(http_server) + UnstableMediaConfigResource(hs).register(http_server) + UnstableThumbnailResource(hs, media_repo, media_repo.media_storage).register( + http_server + ) diff --git a/synapse/rest/media/thumbnail_resource.py b/synapse/rest/media/thumbnail_resource.py index 7cb335c7c3..fe8fbb06e4 100644 --- a/synapse/rest/media/thumbnail_resource.py +++ b/synapse/rest/media/thumbnail_resource.py @@ -22,23 +22,18 @@ import logging import re -from typing import TYPE_CHECKING, List, Optional, Tuple +from typing import TYPE_CHECKING -from synapse.api.errors import Codes, SynapseError, cs_error -from synapse.config.repository import THUMBNAIL_SUPPORTED_MEDIA_FORMAT_MAP -from synapse.http.server import respond_with_json, set_corp_headers, set_cors_headers +from synapse.http.server import set_corp_headers, set_cors_headers from synapse.http.servlet import RestServlet, parse_integer, parse_string from synapse.http.site import SynapseRequest from synapse.media._base import ( DEFAULT_MAX_TIMEOUT_MS, MAXIMUM_ALLOWED_MAX_TIMEOUT_MS, - FileInfo, - ThumbnailInfo, respond_404, - respond_with_file, - respond_with_responder, ) from synapse.media.media_storage import MediaStorage +from synapse.media.thumbnailer import ThumbnailProvider from synapse.util.stringutils import parse_and_validate_server_name if TYPE_CHECKING: @@ -66,10 +61,11 @@ class ThumbnailResource(RestServlet): self.store = hs.get_datastores().main self.media_repo = media_repo self.media_storage = media_storage - self.dynamic_thumbnails = hs.config.media.dynamic_thumbnails self._is_mine_server_name = hs.is_mine_server_name self._server_name = hs.hostname self.prevent_media_downloads_from = hs.config.media.prevent_media_downloads_from + self.dynamic_thumbnails = hs.config.media.dynamic_thumbnails + self.thumbnail_provider = ThumbnailProvider(hs, media_repo, media_storage) async def on_GET( self, request: SynapseRequest, server_name: str, media_id: str @@ -91,11 +87,11 @@ class ThumbnailResource(RestServlet): if self._is_mine_server_name(server_name): if self.dynamic_thumbnails: - await self._select_or_generate_local_thumbnail( + await self.thumbnail_provider.select_or_generate_local_thumbnail( request, media_id, width, height, method, m_type, max_timeout_ms ) else: - await self._respond_local_thumbnail( + await self.thumbnail_provider.respond_local_thumbnail( request, media_id, width, height, method, m_type, max_timeout_ms ) self.media_repo.mark_recently_accessed(None, media_id) @@ -109,9 +105,9 @@ class ThumbnailResource(RestServlet): return remote_resp_function = ( - self._select_or_generate_remote_thumbnail + self.thumbnail_provider.select_or_generate_remote_thumbnail if self.dynamic_thumbnails - else self._respond_remote_thumbnail + else self.thumbnail_provider.respond_remote_thumbnail ) await remote_resp_function( request, @@ -124,457 +120,3 @@ class ThumbnailResource(RestServlet): max_timeout_ms, ) self.media_repo.mark_recently_accessed(server_name, media_id) - - async def _respond_local_thumbnail( - self, - request: SynapseRequest, - media_id: str, - width: int, - height: int, - method: str, - m_type: str, - max_timeout_ms: int, - ) -> None: - media_info = await self.media_repo.get_local_media_info( - request, media_id, max_timeout_ms - ) - if not media_info: - return - - thumbnail_infos = await self.store.get_local_media_thumbnails(media_id) - await self._select_and_respond_with_thumbnail( - request, - width, - height, - method, - m_type, - thumbnail_infos, - media_id, - media_id, - url_cache=bool(media_info.url_cache), - server_name=None, - ) - - async def _select_or_generate_local_thumbnail( - self, - request: SynapseRequest, - media_id: str, - desired_width: int, - desired_height: int, - desired_method: str, - desired_type: str, - max_timeout_ms: int, - ) -> None: - media_info = await self.media_repo.get_local_media_info( - request, media_id, max_timeout_ms - ) - - if not media_info: - return - - thumbnail_infos = await self.store.get_local_media_thumbnails(media_id) - for info in thumbnail_infos: - t_w = info.width == desired_width - t_h = info.height == desired_height - t_method = info.method == desired_method - t_type = info.type == desired_type - - if t_w and t_h and t_method and t_type: - file_info = FileInfo( - server_name=None, - file_id=media_id, - url_cache=bool(media_info.url_cache), - thumbnail=info, - ) - - responder = await self.media_storage.fetch_media(file_info) - if responder: - await respond_with_responder( - request, responder, info.type, info.length - ) - return - - logger.debug("We don't have a thumbnail of that size. Generating") - - # Okay, so we generate one. - file_path = await self.media_repo.generate_local_exact_thumbnail( - media_id, - desired_width, - desired_height, - desired_method, - desired_type, - url_cache=bool(media_info.url_cache), - ) - - if file_path: - await respond_with_file(request, desired_type, file_path) - else: - logger.warning("Failed to generate thumbnail") - raise SynapseError(400, "Failed to generate thumbnail.") - - async def _select_or_generate_remote_thumbnail( - self, - request: SynapseRequest, - server_name: str, - media_id: str, - desired_width: int, - desired_height: int, - desired_method: str, - desired_type: str, - max_timeout_ms: int, - ) -> None: - media_info = await self.media_repo.get_remote_media_info( - server_name, media_id, max_timeout_ms - ) - if not media_info: - respond_404(request) - return - - thumbnail_infos = await self.store.get_remote_media_thumbnails( - server_name, media_id - ) - - file_id = media_info.filesystem_id - - for info in thumbnail_infos: - t_w = info.width == desired_width - t_h = info.height == desired_height - t_method = info.method == desired_method - t_type = info.type == desired_type - - if t_w and t_h and t_method and t_type: - file_info = FileInfo( - server_name=server_name, - file_id=file_id, - thumbnail=info, - ) - - responder = await self.media_storage.fetch_media(file_info) - if responder: - await respond_with_responder( - request, responder, info.type, info.length - ) - return - - logger.debug("We don't have a thumbnail of that size. Generating") - - # Okay, so we generate one. - file_path = await self.media_repo.generate_remote_exact_thumbnail( - server_name, - file_id, - media_id, - desired_width, - desired_height, - desired_method, - desired_type, - ) - - if file_path: - await respond_with_file(request, desired_type, file_path) - else: - logger.warning("Failed to generate thumbnail") - raise SynapseError(400, "Failed to generate thumbnail.") - - async def _respond_remote_thumbnail( - self, - request: SynapseRequest, - server_name: str, - media_id: str, - width: int, - height: int, - method: str, - m_type: str, - max_timeout_ms: int, - ) -> None: - # TODO: Don't download the whole remote file - # We should proxy the thumbnail from the remote server instead of - # downloading the remote file and generating our own thumbnails. - media_info = await self.media_repo.get_remote_media_info( - server_name, media_id, max_timeout_ms - ) - if not media_info: - return - - thumbnail_infos = await self.store.get_remote_media_thumbnails( - server_name, media_id - ) - await self._select_and_respond_with_thumbnail( - request, - width, - height, - method, - m_type, - thumbnail_infos, - media_id, - media_info.filesystem_id, - url_cache=False, - server_name=server_name, - ) - - async def _select_and_respond_with_thumbnail( - self, - request: SynapseRequest, - desired_width: int, - desired_height: int, - desired_method: str, - desired_type: str, - thumbnail_infos: List[ThumbnailInfo], - media_id: str, - file_id: str, - url_cache: bool, - server_name: Optional[str] = None, - ) -> None: - """ - Respond to a request with an appropriate thumbnail from the previously generated thumbnails. - - Args: - request: The incoming request. - desired_width: The desired width, the returned thumbnail may be larger than this. - desired_height: The desired height, the returned thumbnail may be larger than this. - desired_method: The desired method used to generate the thumbnail. - desired_type: The desired content-type of the thumbnail. - thumbnail_infos: A list of thumbnail info of candidate thumbnails. - file_id: The ID of the media that a thumbnail is being requested for. - url_cache: True if this is from a URL cache. - server_name: The server name, if this is a remote thumbnail. - """ - logger.debug( - "_select_and_respond_with_thumbnail: media_id=%s desired=%sx%s (%s) thumbnail_infos=%s", - media_id, - desired_width, - desired_height, - desired_method, - thumbnail_infos, - ) - - # If `dynamic_thumbnails` is enabled, we expect Synapse to go down a - # different code path to handle it. - assert not self.dynamic_thumbnails - - if thumbnail_infos: - file_info = self._select_thumbnail( - desired_width, - desired_height, - desired_method, - desired_type, - thumbnail_infos, - file_id, - url_cache, - server_name, - ) - if not file_info: - logger.info("Couldn't find a thumbnail matching the desired inputs") - respond_404(request) - return - - # The thumbnail property must exist. - assert file_info.thumbnail is not None - - responder = await self.media_storage.fetch_media(file_info) - if responder: - await respond_with_responder( - request, - responder, - file_info.thumbnail.type, - file_info.thumbnail.length, - ) - return - - # If we can't find the thumbnail we regenerate it. This can happen - # if e.g. we've deleted the thumbnails but still have the original - # image somewhere. - # - # Since we have an entry for the thumbnail in the DB we a) know we - # have have successfully generated the thumbnail in the past (so we - # don't need to worry about repeatedly failing to generate - # thumbnails), and b) have already calculated that appropriate - # width/height/method so we can just call the "generate exact" - # methods. - - # First let's check that we do actually have the original image - # still. This will throw a 404 if we don't. - # TODO: We should refetch the thumbnails for remote media. - await self.media_storage.ensure_media_is_in_local_cache( - FileInfo(server_name, file_id, url_cache=url_cache) - ) - - if server_name: - await self.media_repo.generate_remote_exact_thumbnail( - server_name, - file_id=file_id, - media_id=media_id, - t_width=file_info.thumbnail.width, - t_height=file_info.thumbnail.height, - t_method=file_info.thumbnail.method, - t_type=file_info.thumbnail.type, - ) - else: - await self.media_repo.generate_local_exact_thumbnail( - media_id=media_id, - t_width=file_info.thumbnail.width, - t_height=file_info.thumbnail.height, - t_method=file_info.thumbnail.method, - t_type=file_info.thumbnail.type, - url_cache=url_cache, - ) - - responder = await self.media_storage.fetch_media(file_info) - await respond_with_responder( - request, - responder, - file_info.thumbnail.type, - file_info.thumbnail.length, - ) - else: - # This might be because: - # 1. We can't create thumbnails for the given media (corrupted or - # unsupported file type), or - # 2. The thumbnailing process never ran or errored out initially - # when the media was first uploaded (these bugs should be - # reported and fixed). - # Note that we don't attempt to generate a thumbnail now because - # `dynamic_thumbnails` is disabled. - logger.info("Failed to find any generated thumbnails") - - assert request.path is not None - respond_with_json( - request, - 400, - cs_error( - "Cannot find any thumbnails for the requested media ('%s'). This might mean the media is not a supported_media_format=(%s) or that thumbnailing failed for some other reason. (Dynamic thumbnails are disabled on this server.)" - % ( - request.path.decode(), - ", ".join(THUMBNAIL_SUPPORTED_MEDIA_FORMAT_MAP.keys()), - ), - code=Codes.UNKNOWN, - ), - send_cors=True, - ) - - def _select_thumbnail( - self, - desired_width: int, - desired_height: int, - desired_method: str, - desired_type: str, - thumbnail_infos: List[ThumbnailInfo], - file_id: str, - url_cache: bool, - server_name: Optional[str], - ) -> Optional[FileInfo]: - """ - Choose an appropriate thumbnail from the previously generated thumbnails. - - Args: - desired_width: The desired width, the returned thumbnail may be larger than this. - desired_height: The desired height, the returned thumbnail may be larger than this. - desired_method: The desired method used to generate the thumbnail. - desired_type: The desired content-type of the thumbnail. - thumbnail_infos: A list of thumbnail infos of candidate thumbnails. - file_id: The ID of the media that a thumbnail is being requested for. - url_cache: True if this is from a URL cache. - server_name: The server name, if this is a remote thumbnail. - - Returns: - The thumbnail which best matches the desired parameters. - """ - desired_method = desired_method.lower() - - # The chosen thumbnail. - thumbnail_info = None - - d_w = desired_width - d_h = desired_height - - if desired_method == "crop": - # Thumbnails that match equal or larger sizes of desired width/height. - crop_info_list: List[ - Tuple[int, int, int, bool, Optional[int], ThumbnailInfo] - ] = [] - # Other thumbnails. - crop_info_list2: List[ - Tuple[int, int, int, bool, Optional[int], ThumbnailInfo] - ] = [] - for info in thumbnail_infos: - # Skip thumbnails generated with different methods. - if info.method != "crop": - continue - - t_w = info.width - t_h = info.height - aspect_quality = abs(d_w * t_h - d_h * t_w) - min_quality = 0 if d_w <= t_w and d_h <= t_h else 1 - size_quality = abs((d_w - t_w) * (d_h - t_h)) - type_quality = desired_type != info.type - length_quality = info.length - if t_w >= d_w or t_h >= d_h: - crop_info_list.append( - ( - aspect_quality, - min_quality, - size_quality, - type_quality, - length_quality, - info, - ) - ) - else: - crop_info_list2.append( - ( - aspect_quality, - min_quality, - size_quality, - type_quality, - length_quality, - info, - ) - ) - # Pick the most appropriate thumbnail. Some values of `desired_width` and - # `desired_height` may result in a tie, in which case we avoid comparing on - # the thumbnail info and pick the thumbnail that appears earlier - # in the list of candidates. - if crop_info_list: - thumbnail_info = min(crop_info_list, key=lambda t: t[:-1])[-1] - elif crop_info_list2: - thumbnail_info = min(crop_info_list2, key=lambda t: t[:-1])[-1] - elif desired_method == "scale": - # Thumbnails that match equal or larger sizes of desired width/height. - info_list: List[Tuple[int, bool, int, ThumbnailInfo]] = [] - # Other thumbnails. - info_list2: List[Tuple[int, bool, int, ThumbnailInfo]] = [] - - for info in thumbnail_infos: - # Skip thumbnails generated with different methods. - if info.method != "scale": - continue - - t_w = info.width - t_h = info.height - size_quality = abs((d_w - t_w) * (d_h - t_h)) - type_quality = desired_type != info.type - length_quality = info.length - if t_w >= d_w or t_h >= d_h: - info_list.append((size_quality, type_quality, length_quality, info)) - else: - info_list2.append( - (size_quality, type_quality, length_quality, info) - ) - # Pick the most appropriate thumbnail. Some values of `desired_width` and - # `desired_height` may result in a tie, in which case we avoid comparing on - # the thumbnail info and pick the thumbnail that appears earlier - # in the list of candidates. - if info_list: - thumbnail_info = min(info_list, key=lambda t: t[:-1])[-1] - elif info_list2: - thumbnail_info = min(info_list2, key=lambda t: t[:-1])[-1] - - if thumbnail_info: - return FileInfo( - file_id=file_id, - url_cache=url_cache, - server_name=server_name, - thumbnail=thumbnail_info, - ) - - # No matching thumbnail was found. - return None diff --git a/tests/media/test_media_storage.py b/tests/media/test_media_storage.py index cae67e11c8..1bd51ceba2 100644 --- a/tests/media/test_media_storage.py +++ b/tests/media/test_media_storage.py @@ -18,6 +18,7 @@ # [This file includes modifications made by New Vector Limited] # # +import itertools import os import shutil import tempfile @@ -46,11 +47,11 @@ from synapse.media._base import FileInfo, ThumbnailInfo from synapse.media.filepath import MediaFilePaths from synapse.media.media_storage import MediaStorage, ReadableFileWrapper from synapse.media.storage_provider import FileStorageProviderBackend +from synapse.media.thumbnailer import ThumbnailProvider from synapse.module_api import ModuleApi from synapse.module_api.callbacks.spamchecker_callbacks import load_legacy_spam_checkers from synapse.rest import admin -from synapse.rest.client import login -from synapse.rest.media.thumbnail_resource import ThumbnailResource +from synapse.rest.client import login, media from synapse.server import HomeServer from synapse.types import JsonDict, RoomAlias from synapse.util import Clock @@ -153,68 +154,54 @@ class _TestImage: is_inline: bool = True -@parameterized_class( - ("test_image",), - [ - # small png - ( - _TestImage( - SMALL_PNG, - b"image/png", - b".png", - unhexlify( - b"89504e470d0a1a0a0000000d4948445200000020000000200806" - b"000000737a7af40000001a49444154789cedc101010000008220" - b"ffaf6e484001000000ef0610200001194334ee0000000049454e" - b"44ae426082" - ), - unhexlify( - b"89504e470d0a1a0a0000000d4948445200000001000000010806" - b"0000001f15c4890000000d49444154789c636060606000000005" - b"0001a5f645400000000049454e44ae426082" - ), - ), - ), - # small png with transparency. - ( - _TestImage( - unhexlify( - b"89504e470d0a1a0a0000000d49484452000000010000000101000" - b"00000376ef9240000000274524e5300010194fdae0000000a4944" - b"4154789c636800000082008177cd72b60000000049454e44ae426" - b"082" - ), - b"image/png", - b".png", - # Note that we don't check the output since it varies across - # different versions of Pillow. - ), - ), - # small lossless webp - ( - _TestImage( - unhexlify( - b"524946461a000000574542505650384c0d0000002f0000001007" - b"1011118888fe0700" - ), - b"image/webp", - b".webp", - ), - ), - # an empty file - ( - _TestImage( - b"", - b"image/gif", - b".gif", - expected_found=False, - unable_to_thumbnail=True, - ), - ), - # An SVG. - ( - _TestImage( - b""" +small_png = _TestImage( + SMALL_PNG, + b"image/png", + b".png", + unhexlify( + b"89504e470d0a1a0a0000000d4948445200000020000000200806" + b"000000737a7af40000001a49444154789cedc101010000008220" + b"ffaf6e484001000000ef0610200001194334ee0000000049454e" + b"44ae426082" + ), + unhexlify( + b"89504e470d0a1a0a0000000d4948445200000001000000010806" + b"0000001f15c4890000000d49444154789c636060606000000005" + b"0001a5f645400000000049454e44ae426082" + ), +) + +small_png_with_transparency = _TestImage( + unhexlify( + b"89504e470d0a1a0a0000000d49484452000000010000000101000" + b"00000376ef9240000000274524e5300010194fdae0000000a4944" + b"4154789c636800000082008177cd72b60000000049454e44ae426" + b"082" + ), + b"image/png", + b".png", + # Note that we don't check the output since it varies across + # different versions of Pillow. +) + +small_lossless_webp = _TestImage( + unhexlify( + b"524946461a000000574542505650384c0d0000002f0000001007" b"1011118888fe0700" + ), + b"image/webp", + b".webp", +) + +empty_file = _TestImage( + b"", + b"image/gif", + b".gif", + expected_found=False, + unable_to_thumbnail=True, +) + +SVG = _TestImage( + b""" @@ -223,19 +210,32 @@ class _TestImage: """, - b"image/svg", - b".svg", - expected_found=False, - unable_to_thumbnail=True, - is_inline=False, - ), - ), - ], + b"image/svg", + b".svg", + expected_found=False, + unable_to_thumbnail=True, + is_inline=False, ) +test_images = [ + small_png, + small_png_with_transparency, + small_lossless_webp, + empty_file, + SVG, +] +urls = [ + "_matrix/media/r0/thumbnail", + "_matrix/client/unstable/org.matrix.msc3916/media/thumbnail", +] + + +@parameterized_class(("test_image", "url"), itertools.product(test_images, urls)) class MediaRepoTests(unittest.HomeserverTestCase): + servlets = [media.register_servlets] test_image: ClassVar[_TestImage] hijack_auth = True user_id = "@test:user" + url: ClassVar[str] def make_homeserver(self, reactor: MemoryReactor, clock: Clock) -> HomeServer: self.fetches: List[ @@ -298,6 +298,7 @@ class MediaRepoTests(unittest.HomeserverTestCase): "config": {"directory": self.storage_path}, } config["media_storage_providers"] = [provider_config] + config["experimental_features"] = {"msc3916_authenticated_media_enabled": True} hs = self.setup_test_homeserver(config=config, federation_http_client=client) @@ -502,7 +503,7 @@ class MediaRepoTests(unittest.HomeserverTestCase): params = "?width=32&height=32&method=scale" channel = self.make_request( "GET", - f"/_matrix/media/v3/thumbnail/{self.media_id}{params}", + f"/{self.url}/{self.media_id}{params}", shorthand=False, await_result=False, ) @@ -530,7 +531,7 @@ class MediaRepoTests(unittest.HomeserverTestCase): channel = self.make_request( "GET", - f"/_matrix/media/v3/thumbnail/{self.media_id}{params}", + f"/{self.url}/{self.media_id}{params}", shorthand=False, await_result=False, ) @@ -566,12 +567,11 @@ class MediaRepoTests(unittest.HomeserverTestCase): params = "?width=32&height=32&method=" + method channel = self.make_request( "GET", - f"/_matrix/media/r0/thumbnail/{self.media_id}{params}", + f"/{self.url}/{self.media_id}{params}", shorthand=False, await_result=False, ) self.pump() - headers = { b"Content-Length": [b"%d" % (len(self.test_image.data))], b"Content-Type": [self.test_image.content_type], @@ -580,7 +580,6 @@ class MediaRepoTests(unittest.HomeserverTestCase): (self.test_image.data, (len(self.test_image.data), headers)) ) self.pump() - if expected_found: self.assertEqual(channel.code, 200) @@ -603,7 +602,7 @@ class MediaRepoTests(unittest.HomeserverTestCase): channel.json_body, { "errcode": "M_UNKNOWN", - "error": "Cannot find any thumbnails for the requested media ('/_matrix/media/r0/thumbnail/example.com/12345'). This might mean the media is not a supported_media_format=(image/jpeg, image/jpg, image/webp, image/gif, image/png) or that thumbnailing failed for some other reason. (Dynamic thumbnails are disabled on this server.)", + "error": f"Cannot find any thumbnails for the requested media ('/{self.url}/example.com/12345'). This might mean the media is not a supported_media_format=(image/jpeg, image/jpg, image/webp, image/gif, image/png) or that thumbnailing failed for some other reason. (Dynamic thumbnails are disabled on this server.)", }, ) else: @@ -613,7 +612,7 @@ class MediaRepoTests(unittest.HomeserverTestCase): channel.json_body, { "errcode": "M_NOT_FOUND", - "error": "Not found '/_matrix/media/r0/thumbnail/example.com/12345'", + "error": f"Not found '/{self.url}/example.com/12345'", }, ) @@ -625,12 +624,12 @@ class MediaRepoTests(unittest.HomeserverTestCase): content_type = self.test_image.content_type.decode() media_repo = self.hs.get_media_repository() - thumbnail_resouce = ThumbnailResource( + thumbnail_provider = ThumbnailProvider( self.hs, media_repo, media_repo.media_storage ) self.assertIsNotNone( - thumbnail_resouce._select_thumbnail( + thumbnail_provider._select_thumbnail( desired_width=desired_size, desired_height=desired_size, desired_method=method, diff --git a/tests/rest/client/test_media.py b/tests/rest/client/test_media.py new file mode 100644 index 0000000000..600cbf8963 --- /dev/null +++ b/tests/rest/client/test_media.py @@ -0,0 +1,1609 @@ +# +# This file is licensed under the Affero General Public License (AGPL) version 3. +# +# Copyright 2022 The Matrix.org Foundation C.I.C. +# Copyright (C) 2024 New Vector, Ltd +# +# This program is free software: you can redistribute it and/or modify +# it under the terms of the GNU Affero General Public License as +# published by the Free Software Foundation, either version 3 of the +# License, or (at your option) any later version. +# +# See the GNU Affero General Public License for more details: +# . +# +# Originally licensed under the Apache License, Version 2.0: +# . +# +# [This file includes modifications made by New Vector Limited] +# +# +import base64 +import json +import os +import re +from typing import Any, Dict, Optional, Sequence, Tuple, Type +from urllib.parse import quote, urlencode + +from twisted.internet._resolver import HostResolution +from twisted.internet.address import IPv4Address, IPv6Address +from twisted.internet.error import DNSLookupError +from twisted.internet.interfaces import IAddress, IResolutionReceiver +from twisted.test.proto_helpers import AccumulatingProtocol, MemoryReactor +from twisted.web.resource import Resource + +from synapse.config.oembed import OEmbedEndpointConfig +from synapse.media._base import FileInfo +from synapse.media.url_previewer import IMAGE_CACHE_EXPIRY_MS +from synapse.rest import admin +from synapse.rest.client import login, media +from synapse.server import HomeServer +from synapse.types import JsonDict +from synapse.util import Clock +from synapse.util.stringutils import parse_and_validate_mxc_uri + +from tests import unittest +from tests.server import FakeTransport, ThreadedMemoryReactorClock +from tests.test_utils import SMALL_PNG +from tests.unittest import override_config + +try: + import lxml +except ImportError: + lxml = None # type: ignore[assignment] + + +class UnstableMediaDomainBlockingTests(unittest.HomeserverTestCase): + remote_media_id = "doesnotmatter" + remote_server_name = "evil.com" + servlets = [ + media.register_servlets, + admin.register_servlets, + login.register_servlets, + ] + + def make_homeserver( + self, reactor: ThreadedMemoryReactorClock, clock: Clock + ) -> HomeServer: + config = self.default_config() + + self.storage_path = self.mktemp() + self.media_store_path = self.mktemp() + os.mkdir(self.storage_path) + os.mkdir(self.media_store_path) + config["media_store_path"] = self.media_store_path + + provider_config = { + "module": "synapse.media.storage_provider.FileStorageProviderBackend", + "store_local": True, + "store_synchronous": False, + "store_remote": True, + "config": {"directory": self.storage_path}, + } + + config["media_storage_providers"] = [provider_config] + + return self.setup_test_homeserver(config=config) + + def prepare(self, reactor: MemoryReactor, clock: Clock, hs: HomeServer) -> None: + self.store = hs.get_datastores().main + + # Inject a piece of media. We'll use this to ensure we're returning a sane + # response when we're not supposed to block it, distinguishing a media block + # from a regular 404. + file_id = "abcdefg12345" + file_info = FileInfo(server_name=self.remote_server_name, file_id=file_id) + with hs.get_media_repository().media_storage.store_into_file(file_info) as ( + f, + fname, + finish, + ): + f.write(SMALL_PNG) + self.get_success(finish()) + + self.get_success( + self.store.store_cached_remote_media( + origin=self.remote_server_name, + media_id=self.remote_media_id, + media_type="image/png", + media_length=1, + time_now_ms=clock.time_msec(), + upload_name="test.png", + filesystem_id=file_id, + ) + ) + self.register_user("user", "password") + self.tok = self.login("user", "password") + + @override_config( + { + # Disable downloads from the domain we'll be trying to download from. + # Should result in a 404. + "prevent_media_downloads_from": ["evil.com"], + "dynamic_thumbnails": True, + "experimental_features": {"msc3916_authenticated_media_enabled": True}, + } + ) + def test_cannot_download_blocked_media_thumbnail(self) -> None: + """ + Same test as test_cannot_download_blocked_media but for thumbnails. + """ + response = self.make_request( + "GET", + f"/_matrix/client/unstable/org.matrix.msc3916/media/thumbnail/evil.com/{self.remote_media_id}?width=100&height=100", + shorthand=False, + content={"width": 100, "height": 100}, + access_token=self.tok, + ) + self.assertEqual(response.code, 404) + + @override_config( + { + # Disable downloads from a domain we won't be requesting downloads from. + # This proves we haven't broken anything. + "prevent_media_downloads_from": ["not-listed.com"], + "dynamic_thumbnails": True, + "experimental_features": {"msc3916_authenticated_media_enabled": True}, + } + ) + def test_remote_media_thumbnail_normally_unblocked(self) -> None: + """ + Same test as test_remote_media_normally_unblocked but for thumbnails. + """ + response = self.make_request( + "GET", + f"/_matrix/client/unstable/org.matrix.msc3916/media/thumbnail/evil.com/{self.remote_media_id}?width=100&height=100", + shorthand=False, + access_token=self.tok, + ) + self.assertEqual(response.code, 200) + + +class UnstableURLPreviewTests(unittest.HomeserverTestCase): + if not lxml: + skip = "url preview feature requires lxml" + + servlets = [media.register_servlets] + hijack_auth = True + user_id = "@test:user" + end_content = ( + b"" + b'' + b'' + b"" + ) + + def make_homeserver(self, reactor: MemoryReactor, clock: Clock) -> HomeServer: + config = self.default_config() + config["experimental_features"] = {"msc3916_authenticated_media_enabled": True} + config["url_preview_enabled"] = True + config["max_spider_size"] = 9999999 + config["url_preview_ip_range_blacklist"] = ( + "192.168.1.1", + "1.0.0.0/8", + "3fff:ffff:ffff:ffff:ffff:ffff:ffff:ffff", + "2001:800::/21", + ) + config["url_preview_ip_range_whitelist"] = ("1.1.1.1",) + config["url_preview_accept_language"] = [ + "en-UK", + "en-US;q=0.9", + "fr;q=0.8", + "*;q=0.7", + ] + + self.storage_path = self.mktemp() + self.media_store_path = self.mktemp() + os.mkdir(self.storage_path) + os.mkdir(self.media_store_path) + config["media_store_path"] = self.media_store_path + + provider_config = { + "module": "synapse.media.storage_provider.FileStorageProviderBackend", + "store_local": True, + "store_synchronous": False, + "store_remote": True, + "config": {"directory": self.storage_path}, + } + + config["media_storage_providers"] = [provider_config] + + hs = self.setup_test_homeserver(config=config) + + # After the hs is created, modify the parsed oEmbed config (to avoid + # messing with files). + # + # Note that HTTP URLs are used to avoid having to deal with TLS in tests. + hs.config.oembed.oembed_patterns = [ + OEmbedEndpointConfig( + api_endpoint="http://publish.twitter.com/oembed", + url_patterns=[ + re.compile(r"http://twitter\.com/.+/status/.+"), + ], + formats=None, + ), + OEmbedEndpointConfig( + api_endpoint="http://www.hulu.com/api/oembed.{format}", + url_patterns=[ + re.compile(r"http://www\.hulu\.com/watch/.+"), + ], + formats=["json"], + ), + ] + + return hs + + def prepare(self, reactor: MemoryReactor, clock: Clock, hs: HomeServer) -> None: + self.media_repo = hs.get_media_repository() + assert self.media_repo.url_previewer is not None + self.url_previewer = self.media_repo.url_previewer + + self.lookups: Dict[str, Any] = {} + + class Resolver: + def resolveHostName( + _self, + resolutionReceiver: IResolutionReceiver, + hostName: str, + portNumber: int = 0, + addressTypes: Optional[Sequence[Type[IAddress]]] = None, + transportSemantics: str = "TCP", + ) -> IResolutionReceiver: + resolution = HostResolution(hostName) + resolutionReceiver.resolutionBegan(resolution) + if hostName not in self.lookups: + raise DNSLookupError("OH NO") + + for i in self.lookups[hostName]: + resolutionReceiver.addressResolved(i[0]("TCP", i[1], portNumber)) + resolutionReceiver.resolutionComplete() + return resolutionReceiver + + self.reactor.nameResolver = Resolver() # type: ignore[assignment] + + def create_resource_dict(self) -> Dict[str, Resource]: + """Create a resource tree for the test server + + A resource tree is a mapping from path to twisted.web.resource. + + The default implementation creates a JsonResource and calls each function in + `servlets` to register servlets against it. + """ + resources = super().create_resource_dict() + resources["/_matrix/media"] = self.hs.get_media_repository_resource() + return resources + + def _assert_small_png(self, json_body: JsonDict) -> None: + """Assert properties from the SMALL_PNG test image.""" + self.assertTrue(json_body["og:image"].startswith("mxc://")) + self.assertEqual(json_body["og:image:height"], 1) + self.assertEqual(json_body["og:image:width"], 1) + self.assertEqual(json_body["og:image:type"], "image/png") + self.assertEqual(json_body["matrix:image:size"], 67) + + def test_cache_returns_correct_type(self) -> None: + self.lookups["matrix.org"] = [(IPv4Address, "10.1.2.3")] + + channel = self.make_request( + "GET", + "/_matrix/client/unstable/org.matrix.msc3916/media/preview_url?url=http://matrix.org", + shorthand=False, + await_result=False, + ) + self.pump() + + client = self.reactor.tcpClients[0][2].buildProtocol(None) + server = AccumulatingProtocol() + server.makeConnection(FakeTransport(client, self.reactor)) + client.makeConnection(FakeTransport(server, self.reactor)) + client.dataReceived( + b"HTTP/1.0 200 OK\r\nContent-Length: %d\r\nContent-Type: text/html\r\n\r\n" + % (len(self.end_content),) + + self.end_content + ) + + self.pump() + self.assertEqual(channel.code, 200) + self.assertEqual( + channel.json_body, {"og:title": "~matrix~", "og:description": "hi"} + ) + + # Check the cache returns the correct response + channel = self.make_request( + "GET", + "/_matrix/client/unstable/org.matrix.msc3916/media/preview_url?url=http://matrix.org", + shorthand=False, + ) + + # Check the cache response has the same content + self.assertEqual(channel.code, 200) + self.assertEqual( + channel.json_body, {"og:title": "~matrix~", "og:description": "hi"} + ) + + # Clear the in-memory cache + self.assertIn("http://matrix.org", self.url_previewer._cache) + self.url_previewer._cache.pop("http://matrix.org") + self.assertNotIn("http://matrix.org", self.url_previewer._cache) + + # Check the database cache returns the correct response + channel = self.make_request( + "GET", + "/_matrix/client/unstable/org.matrix.msc3916/media/preview_url?url=http://matrix.org", + shorthand=False, + ) + + # Check the cache response has the same content + self.assertEqual(channel.code, 200) + self.assertEqual( + channel.json_body, {"og:title": "~matrix~", "og:description": "hi"} + ) + + def test_non_ascii_preview_httpequiv(self) -> None: + self.lookups["matrix.org"] = [(IPv4Address, "10.1.2.3")] + + end_content = ( + b"" + b'' + b'' + b'' + b"" + ) + + channel = self.make_request( + "GET", + "/_matrix/client/unstable/org.matrix.msc3916/media/preview_url?url=http://matrix.org", + shorthand=False, + await_result=False, + ) + self.pump() + + client = self.reactor.tcpClients[0][2].buildProtocol(None) + server = AccumulatingProtocol() + server.makeConnection(FakeTransport(client, self.reactor)) + client.makeConnection(FakeTransport(server, self.reactor)) + client.dataReceived( + ( + b"HTTP/1.0 200 OK\r\nContent-Length: %d\r\n" + b'Content-Type: text/html; charset="utf8"\r\n\r\n' + ) + % (len(end_content),) + + end_content + ) + + self.pump() + self.assertEqual(channel.code, 200) + self.assertEqual(channel.json_body["og:title"], "\u0434\u043a\u0430") + + def test_video_rejected(self) -> None: + self.lookups["matrix.org"] = [(IPv4Address, "10.1.2.3")] + + end_content = b"anything" + + channel = self.make_request( + "GET", + "/_matrix/client/unstable/org.matrix.msc3916/media/preview_url?url=http://matrix.org", + shorthand=False, + await_result=False, + ) + self.pump() + + client = self.reactor.tcpClients[0][2].buildProtocol(None) + server = AccumulatingProtocol() + server.makeConnection(FakeTransport(client, self.reactor)) + client.makeConnection(FakeTransport(server, self.reactor)) + client.dataReceived( + ( + b"HTTP/1.0 200 OK\r\nContent-Length: %d\r\n" + b"Content-Type: video/mp4\r\n\r\n" + ) + % (len(end_content)) + + end_content + ) + + self.pump() + self.assertEqual(channel.code, 502) + self.assertEqual( + channel.json_body, + { + "errcode": "M_UNKNOWN", + "error": "Requested file's content type not allowed for this operation: video/mp4", + }, + ) + + def test_audio_rejected(self) -> None: + self.lookups["matrix.org"] = [(IPv4Address, "10.1.2.3")] + + end_content = b"anything" + + channel = self.make_request( + "GET", + "/_matrix/client/unstable/org.matrix.msc3916/media/preview_url?url=http://matrix.org", + shorthand=False, + await_result=False, + ) + self.pump() + + client = self.reactor.tcpClients[0][2].buildProtocol(None) + server = AccumulatingProtocol() + server.makeConnection(FakeTransport(client, self.reactor)) + client.makeConnection(FakeTransport(server, self.reactor)) + client.dataReceived( + ( + b"HTTP/1.0 200 OK\r\nContent-Length: %d\r\n" + b"Content-Type: audio/aac\r\n\r\n" + ) + % (len(end_content)) + + end_content + ) + + self.pump() + self.assertEqual(channel.code, 502) + self.assertEqual( + channel.json_body, + { + "errcode": "M_UNKNOWN", + "error": "Requested file's content type not allowed for this operation: audio/aac", + }, + ) + + def test_non_ascii_preview_content_type(self) -> None: + self.lookups["matrix.org"] = [(IPv4Address, "10.1.2.3")] + + end_content = ( + b"" + b'' + b'' + b"" + ) + + channel = self.make_request( + "GET", + "/_matrix/client/unstable/org.matrix.msc3916/media/preview_url?url=http://matrix.org", + shorthand=False, + await_result=False, + ) + self.pump() + + client = self.reactor.tcpClients[0][2].buildProtocol(None) + server = AccumulatingProtocol() + server.makeConnection(FakeTransport(client, self.reactor)) + client.makeConnection(FakeTransport(server, self.reactor)) + client.dataReceived( + ( + b"HTTP/1.0 200 OK\r\nContent-Length: %d\r\n" + b'Content-Type: text/html; charset="windows-1251"\r\n\r\n' + ) + % (len(end_content),) + + end_content + ) + + self.pump() + self.assertEqual(channel.code, 200) + self.assertEqual(channel.json_body["og:title"], "\u0434\u043a\u0430") + + def test_overlong_title(self) -> None: + self.lookups["matrix.org"] = [(IPv4Address, "10.1.2.3")] + + end_content = ( + b"" + b"" + b"x" * 2000 + b"" + b'' + b"" + ) + + channel = self.make_request( + "GET", + "/_matrix/client/unstable/org.matrix.msc3916/media/preview_url?url=http://matrix.org", + shorthand=False, + await_result=False, + ) + self.pump() + + client = self.reactor.tcpClients[0][2].buildProtocol(None) + server = AccumulatingProtocol() + server.makeConnection(FakeTransport(client, self.reactor)) + client.makeConnection(FakeTransport(server, self.reactor)) + client.dataReceived( + ( + b"HTTP/1.0 200 OK\r\nContent-Length: %d\r\n" + b'Content-Type: text/html; charset="windows-1251"\r\n\r\n' + ) + % (len(end_content),) + + end_content + ) + + self.pump() + self.assertEqual(channel.code, 200) + res = channel.json_body + # We should only see the `og:description` field, as `title` is too long and should be stripped out + self.assertCountEqual(["og:description"], res.keys()) + + def test_ipaddr(self) -> None: + """ + IP addresses can be previewed directly. + """ + self.lookups["example.com"] = [(IPv4Address, "10.1.2.3")] + + channel = self.make_request( + "GET", + "/_matrix/client/unstable/org.matrix.msc3916/media/preview_url?url=http://example.com", + shorthand=False, + await_result=False, + ) + self.pump() + + client = self.reactor.tcpClients[0][2].buildProtocol(None) + server = AccumulatingProtocol() + server.makeConnection(FakeTransport(client, self.reactor)) + client.makeConnection(FakeTransport(server, self.reactor)) + client.dataReceived( + b"HTTP/1.0 200 OK\r\nContent-Length: %d\r\nContent-Type: text/html\r\n\r\n" + % (len(self.end_content),) + + self.end_content + ) + + self.pump() + self.assertEqual(channel.code, 200) + self.assertEqual( + channel.json_body, {"og:title": "~matrix~", "og:description": "hi"} + ) + + def test_blocked_ip_specific(self) -> None: + """ + Blocked IP addresses, found via DNS, are not spidered. + """ + self.lookups["example.com"] = [(IPv4Address, "192.168.1.1")] + + channel = self.make_request( + "GET", + "/_matrix/client/unstable/org.matrix.msc3916/media/preview_url?url=http://example.com", + shorthand=False, + ) + + # No requests made. + self.assertEqual(len(self.reactor.tcpClients), 0) + self.assertEqual(channel.code, 502) + self.assertEqual( + channel.json_body, + { + "errcode": "M_UNKNOWN", + "error": "DNS resolution failure during URL preview generation", + }, + ) + + def test_blocked_ip_range(self) -> None: + """ + Blocked IP ranges, IPs found over DNS, are not spidered. + """ + self.lookups["example.com"] = [(IPv4Address, "1.1.1.2")] + + channel = self.make_request( + "GET", + "/_matrix/client/unstable/org.matrix.msc3916/media/preview_url?url=http://example.com", + shorthand=False, + ) + + self.assertEqual(channel.code, 502) + self.assertEqual( + channel.json_body, + { + "errcode": "M_UNKNOWN", + "error": "DNS resolution failure during URL preview generation", + }, + ) + + def test_blocked_ip_specific_direct(self) -> None: + """ + Blocked IP addresses, accessed directly, are not spidered. + """ + channel = self.make_request( + "GET", + "/_matrix/client/unstable/org.matrix.msc3916/media/preview_url?url=http://192.168.1.1", + shorthand=False, + ) + + # No requests made. + self.assertEqual(len(self.reactor.tcpClients), 0) + self.assertEqual( + channel.json_body, + {"errcode": "M_UNKNOWN", "error": "IP address blocked"}, + ) + self.assertEqual(channel.code, 403) + + def test_blocked_ip_range_direct(self) -> None: + """ + Blocked IP ranges, accessed directly, are not spidered. + """ + channel = self.make_request( + "GET", + "/_matrix/client/unstable/org.matrix.msc3916/media/preview_url?url=http://1.1.1.2", + shorthand=False, + ) + + self.assertEqual(channel.code, 403) + self.assertEqual( + channel.json_body, + {"errcode": "M_UNKNOWN", "error": "IP address blocked"}, + ) + + def test_blocked_ip_range_whitelisted_ip(self) -> None: + """ + Blocked but then subsequently whitelisted IP addresses can be + spidered. + """ + self.lookups["example.com"] = [(IPv4Address, "1.1.1.1")] + + channel = self.make_request( + "GET", + "/_matrix/client/unstable/org.matrix.msc3916/media/preview_url?url=http://example.com", + shorthand=False, + await_result=False, + ) + self.pump() + + client = self.reactor.tcpClients[0][2].buildProtocol(None) + + server = AccumulatingProtocol() + server.makeConnection(FakeTransport(client, self.reactor)) + client.makeConnection(FakeTransport(server, self.reactor)) + + client.dataReceived( + b"HTTP/1.0 200 OK\r\nContent-Length: %d\r\nContent-Type: text/html\r\n\r\n" + % (len(self.end_content),) + + self.end_content + ) + + self.pump() + self.assertEqual(channel.code, 200) + self.assertEqual( + channel.json_body, {"og:title": "~matrix~", "og:description": "hi"} + ) + + def test_blocked_ip_with_external_ip(self) -> None: + """ + If a hostname resolves a blocked IP, even if there's a non-blocked one, + it will be rejected. + """ + # Hardcode the URL resolving to the IP we want. + self.lookups["example.com"] = [ + (IPv4Address, "1.1.1.2"), + (IPv4Address, "10.1.2.3"), + ] + + channel = self.make_request( + "GET", + "/_matrix/client/unstable/org.matrix.msc3916/media/preview_url?url=http://example.com", + shorthand=False, + ) + self.assertEqual(channel.code, 502) + self.assertEqual( + channel.json_body, + { + "errcode": "M_UNKNOWN", + "error": "DNS resolution failure during URL preview generation", + }, + ) + + def test_blocked_ipv6_specific(self) -> None: + """ + Blocked IP addresses, found via DNS, are not spidered. + """ + self.lookups["example.com"] = [ + (IPv6Address, "3fff:ffff:ffff:ffff:ffff:ffff:ffff:ffff") + ] + + channel = self.make_request( + "GET", + "/_matrix/client/unstable/org.matrix.msc3916/media/preview_url?url=http://example.com", + shorthand=False, + ) + + # No requests made. + self.assertEqual(len(self.reactor.tcpClients), 0) + self.assertEqual(channel.code, 502) + self.assertEqual( + channel.json_body, + { + "errcode": "M_UNKNOWN", + "error": "DNS resolution failure during URL preview generation", + }, + ) + + def test_blocked_ipv6_range(self) -> None: + """ + Blocked IP ranges, IPs found over DNS, are not spidered. + """ + self.lookups["example.com"] = [(IPv6Address, "2001:800::1")] + + channel = self.make_request( + "GET", + "/_matrix/client/unstable/org.matrix.msc3916/media/preview_url?url=http://example.com", + shorthand=False, + ) + + self.assertEqual(channel.code, 502) + self.assertEqual( + channel.json_body, + { + "errcode": "M_UNKNOWN", + "error": "DNS resolution failure during URL preview generation", + }, + ) + + def test_OPTIONS(self) -> None: + """ + OPTIONS returns the OPTIONS. + """ + channel = self.make_request( + "OPTIONS", + "/_matrix/client/unstable/org.matrix.msc3916/media/preview_url?url=http://example.com", + shorthand=False, + ) + self.assertEqual(channel.code, 204) + + def test_accept_language_config_option(self) -> None: + """ + Accept-Language header is sent to the remote server + """ + self.lookups["example.com"] = [(IPv4Address, "10.1.2.3")] + + # Build and make a request to the server + channel = self.make_request( + "GET", + "/_matrix/client/unstable/org.matrix.msc3916/media/preview_url?url=http://example.com", + shorthand=False, + await_result=False, + ) + self.pump() + + # Extract Synapse's tcp client + client = self.reactor.tcpClients[0][2].buildProtocol(None) + + # Build a fake remote server to reply with + server = AccumulatingProtocol() + + # Connect the two together + server.makeConnection(FakeTransport(client, self.reactor)) + client.makeConnection(FakeTransport(server, self.reactor)) + + # Tell Synapse that it has received some data from the remote server + client.dataReceived( + b"HTTP/1.0 200 OK\r\nContent-Length: %d\r\nContent-Type: text/html\r\n\r\n" + % (len(self.end_content),) + + self.end_content + ) + + # Move the reactor along until we get a response on our original channel + self.pump() + self.assertEqual(channel.code, 200) + self.assertEqual( + channel.json_body, {"og:title": "~matrix~", "og:description": "hi"} + ) + + # Check that the server received the Accept-Language header as part + # of the request from Synapse + self.assertIn( + ( + b"Accept-Language: en-UK\r\n" + b"Accept-Language: en-US;q=0.9\r\n" + b"Accept-Language: fr;q=0.8\r\n" + b"Accept-Language: *;q=0.7" + ), + server.data, + ) + + def test_image(self) -> None: + """An image should be precached if mentioned in the HTML.""" + self.lookups["matrix.org"] = [(IPv4Address, "10.1.2.3")] + self.lookups["cdn.matrix.org"] = [(IPv4Address, "10.1.2.4")] + + result = ( + b"""""" + ) + + channel = self.make_request( + "GET", + "/_matrix/client/unstable/org.matrix.msc3916/media/preview_url?url=http://matrix.org", + shorthand=False, + await_result=False, + ) + self.pump() + + # Respond with the HTML. + client = self.reactor.tcpClients[0][2].buildProtocol(None) + server = AccumulatingProtocol() + server.makeConnection(FakeTransport(client, self.reactor)) + client.makeConnection(FakeTransport(server, self.reactor)) + client.dataReceived( + ( + b"HTTP/1.0 200 OK\r\nContent-Length: %d\r\n" + b'Content-Type: text/html; charset="utf8"\r\n\r\n' + ) + % (len(result),) + + result + ) + self.pump() + + # Respond with the photo. + client = self.reactor.tcpClients[1][2].buildProtocol(None) + server = AccumulatingProtocol() + server.makeConnection(FakeTransport(client, self.reactor)) + client.makeConnection(FakeTransport(server, self.reactor)) + client.dataReceived( + ( + b"HTTP/1.0 200 OK\r\nContent-Length: %d\r\n" + b"Content-Type: image/png\r\n\r\n" + ) + % (len(SMALL_PNG),) + + SMALL_PNG + ) + self.pump() + + # The image should be in the result. + self.assertEqual(channel.code, 200) + self._assert_small_png(channel.json_body) + + def test_nonexistent_image(self) -> None: + """If the preview image doesn't exist, ensure some data is returned.""" + self.lookups["matrix.org"] = [(IPv4Address, "10.1.2.3")] + + result = ( + b"""""" + ) + + channel = self.make_request( + "GET", + "/_matrix/client/unstable/org.matrix.msc3916/media/preview_url?url=http://matrix.org", + shorthand=False, + await_result=False, + ) + self.pump() + + client = self.reactor.tcpClients[0][2].buildProtocol(None) + server = AccumulatingProtocol() + server.makeConnection(FakeTransport(client, self.reactor)) + client.makeConnection(FakeTransport(server, self.reactor)) + client.dataReceived( + ( + b"HTTP/1.0 200 OK\r\nContent-Length: %d\r\n" + b'Content-Type: text/html; charset="utf8"\r\n\r\n' + ) + % (len(result),) + + result + ) + + self.pump() + + # There should not be a second connection. + self.assertEqual(len(self.reactor.tcpClients), 1) + + # The image should not be in the result. + self.assertEqual(channel.code, 200) + self.assertNotIn("og:image", channel.json_body) + + @unittest.override_config( + {"url_preview_url_blacklist": [{"netloc": "cdn.matrix.org"}]} + ) + def test_image_blocked(self) -> None: + """If the preview image doesn't exist, ensure some data is returned.""" + self.lookups["matrix.org"] = [(IPv4Address, "10.1.2.3")] + self.lookups["cdn.matrix.org"] = [(IPv4Address, "10.1.2.4")] + + result = ( + b"""""" + ) + + channel = self.make_request( + "GET", + "/_matrix/client/unstable/org.matrix.msc3916/media/preview_url?url=http://matrix.org", + shorthand=False, + await_result=False, + ) + self.pump() + + client = self.reactor.tcpClients[0][2].buildProtocol(None) + server = AccumulatingProtocol() + server.makeConnection(FakeTransport(client, self.reactor)) + client.makeConnection(FakeTransport(server, self.reactor)) + client.dataReceived( + ( + b"HTTP/1.0 200 OK\r\nContent-Length: %d\r\n" + b'Content-Type: text/html; charset="utf8"\r\n\r\n' + ) + % (len(result),) + + result + ) + self.pump() + + # There should not be a second connection. + self.assertEqual(len(self.reactor.tcpClients), 1) + + # The image should not be in the result. + self.assertEqual(channel.code, 200) + self.assertNotIn("og:image", channel.json_body) + + def test_oembed_failure(self) -> None: + """If the autodiscovered oEmbed URL fails, ensure some data is returned.""" + self.lookups["matrix.org"] = [(IPv4Address, "10.1.2.3")] + + result = b""" + oEmbed Autodiscovery Fail + + """ + + channel = self.make_request( + "GET", + "/_matrix/client/unstable/org.matrix.msc3916/media/preview_url?url=http://matrix.org", + shorthand=False, + await_result=False, + ) + self.pump() + + client = self.reactor.tcpClients[0][2].buildProtocol(None) + server = AccumulatingProtocol() + server.makeConnection(FakeTransport(client, self.reactor)) + client.makeConnection(FakeTransport(server, self.reactor)) + client.dataReceived( + ( + b"HTTP/1.0 200 OK\r\nContent-Length: %d\r\n" + b'Content-Type: text/html; charset="utf8"\r\n\r\n' + ) + % (len(result),) + + result + ) + + self.pump() + self.assertEqual(channel.code, 200) + + # The image should not be in the result. + self.assertEqual(channel.json_body["og:title"], "oEmbed Autodiscovery Fail") + + def test_data_url(self) -> None: + """ + Requesting to preview a data URL is not supported. + """ + self.lookups["matrix.org"] = [(IPv4Address, "10.1.2.3")] + + data = base64.b64encode(SMALL_PNG).decode() + + query_params = urlencode( + { + "url": f'' + } + ) + + channel = self.make_request( + "GET", + f"/_matrix/client/unstable/org.matrix.msc3916/media/preview_url?{query_params}", + shorthand=False, + ) + self.pump() + + self.assertEqual(channel.code, 500) + + def test_inline_data_url(self) -> None: + """ + An inline image (as a data URL) should be parsed properly. + """ + self.lookups["matrix.org"] = [(IPv4Address, "10.1.2.3")] + + data = base64.b64encode(SMALL_PNG) + + end_content = ( + b"" b'' b"" + ) % (data,) + + channel = self.make_request( + "GET", + "/_matrix/client/unstable/org.matrix.msc3916/media/preview_url?url=http://matrix.org", + shorthand=False, + await_result=False, + ) + self.pump() + + client = self.reactor.tcpClients[0][2].buildProtocol(None) + server = AccumulatingProtocol() + server.makeConnection(FakeTransport(client, self.reactor)) + client.makeConnection(FakeTransport(server, self.reactor)) + client.dataReceived( + ( + b"HTTP/1.0 200 OK\r\nContent-Length: %d\r\n" + b'Content-Type: text/html; charset="utf8"\r\n\r\n' + ) + % (len(end_content),) + + end_content + ) + + self.pump() + self.assertEqual(channel.code, 200) + self._assert_small_png(channel.json_body) + + def test_oembed_photo(self) -> None: + """Test an oEmbed endpoint which returns a 'photo' type which redirects the preview to a new URL.""" + self.lookups["publish.twitter.com"] = [(IPv4Address, "10.1.2.3")] + self.lookups["cdn.twitter.com"] = [(IPv4Address, "10.1.2.3")] + + result = { + "version": "1.0", + "type": "photo", + "url": "http://cdn.twitter.com/matrixdotorg", + } + oembed_content = json.dumps(result).encode("utf-8") + + channel = self.make_request( + "GET", + "/_matrix/client/unstable/org.matrix.msc3916/media/preview_url?url=http://twitter.com/matrixdotorg/status/12345", + shorthand=False, + await_result=False, + ) + self.pump() + + client = self.reactor.tcpClients[0][2].buildProtocol(None) + server = AccumulatingProtocol() + server.makeConnection(FakeTransport(client, self.reactor)) + client.makeConnection(FakeTransport(server, self.reactor)) + client.dataReceived( + ( + b"HTTP/1.0 200 OK\r\nContent-Length: %d\r\n" + b'Content-Type: application/json; charset="utf8"\r\n\r\n' + ) + % (len(oembed_content),) + + oembed_content + ) + + self.pump() + + # Ensure a second request is made to the photo URL. + client = self.reactor.tcpClients[1][2].buildProtocol(None) + server = AccumulatingProtocol() + server.makeConnection(FakeTransport(client, self.reactor)) + client.makeConnection(FakeTransport(server, self.reactor)) + client.dataReceived( + ( + b"HTTP/1.0 200 OK\r\nContent-Length: %d\r\n" + b"Content-Type: image/png\r\n\r\n" + ) + % (len(SMALL_PNG),) + + SMALL_PNG + ) + + self.pump() + + # Ensure the URL is what was requested. + self.assertIn(b"/matrixdotorg", server.data) + + self.assertEqual(channel.code, 200) + body = channel.json_body + self.assertEqual(body["og:url"], "http://twitter.com/matrixdotorg/status/12345") + self._assert_small_png(body) + + def test_oembed_rich(self) -> None: + """Test an oEmbed endpoint which returns HTML content via the 'rich' type.""" + self.lookups["publish.twitter.com"] = [(IPv4Address, "10.1.2.3")] + + result = { + "version": "1.0", + "type": "rich", + # Note that this provides the author, not the title. + "author_name": "Alice", + "html": "
Content Preview
", + } + end_content = json.dumps(result).encode("utf-8") + + channel = self.make_request( + "GET", + "/_matrix/client/unstable/org.matrix.msc3916/media/preview_url?url=http://twitter.com/matrixdotorg/status/12345", + shorthand=False, + await_result=False, + ) + self.pump() + + client = self.reactor.tcpClients[0][2].buildProtocol(None) + server = AccumulatingProtocol() + server.makeConnection(FakeTransport(client, self.reactor)) + client.makeConnection(FakeTransport(server, self.reactor)) + client.dataReceived( + ( + b"HTTP/1.0 200 OK\r\nContent-Length: %d\r\n" + b'Content-Type: application/json; charset="utf8"\r\n\r\n' + ) + % (len(end_content),) + + end_content + ) + + self.pump() + + # Double check that the proper host is being connected to. (Note that + # twitter.com can't be resolved so this is already implicitly checked.) + self.assertIn(b"\r\nHost: publish.twitter.com\r\n", server.data) + + self.assertEqual(channel.code, 200) + body = channel.json_body + self.assertEqual( + body, + { + "og:url": "http://twitter.com/matrixdotorg/status/12345", + "og:title": "Alice", + "og:description": "Content Preview", + }, + ) + + def test_oembed_format(self) -> None: + """Test an oEmbed endpoint which requires the format in the URL.""" + self.lookups["www.hulu.com"] = [(IPv4Address, "10.1.2.3")] + + result = { + "version": "1.0", + "type": "rich", + "html": "
Content Preview
", + } + end_content = json.dumps(result).encode("utf-8") + + channel = self.make_request( + "GET", + "/_matrix/client/unstable/org.matrix.msc3916/media/preview_url?url=http://www.hulu.com/watch/12345", + shorthand=False, + await_result=False, + ) + self.pump() + + client = self.reactor.tcpClients[0][2].buildProtocol(None) + server = AccumulatingProtocol() + server.makeConnection(FakeTransport(client, self.reactor)) + client.makeConnection(FakeTransport(server, self.reactor)) + client.dataReceived( + ( + b"HTTP/1.0 200 OK\r\nContent-Length: %d\r\n" + b'Content-Type: application/json; charset="utf8"\r\n\r\n' + ) + % (len(end_content),) + + end_content + ) + + self.pump() + + # The {format} should have been turned into json. + self.assertIn(b"/api/oembed.json", server.data) + # A URL parameter of format=json should be provided. + self.assertIn(b"format=json", server.data) + + self.assertEqual(channel.code, 200) + body = channel.json_body + self.assertEqual( + body, + { + "og:url": "http://www.hulu.com/watch/12345", + "og:description": "Content Preview", + }, + ) + + @unittest.override_config( + {"url_preview_url_blacklist": [{"netloc": "publish.twitter.com"}]} + ) + def test_oembed_blocked(self) -> None: + """The oEmbed URL should not be downloaded if the oEmbed URL is blocked.""" + self.lookups["twitter.com"] = [(IPv4Address, "10.1.2.3")] + + channel = self.make_request( + "GET", + "/_matrix/client/unstable/org.matrix.msc3916/media/preview_url?url=http://twitter.com/matrixdotorg/status/12345", + shorthand=False, + await_result=False, + ) + self.pump() + self.assertEqual(channel.code, 403, channel.result) + + def test_oembed_autodiscovery(self) -> None: + """ + Autodiscovery works by finding the link in the HTML response and then requesting an oEmbed URL. + 1. Request a preview of a URL which is not known to the oEmbed code. + 2. It returns HTML including a link to an oEmbed preview. + 3. The oEmbed preview is requested and returns a URL for an image. + 4. The image is requested for thumbnailing. + """ + # This is a little cheesy in that we use the www subdomain (which isn't the + # list of oEmbed patterns) to get "raw" HTML response. + self.lookups["www.twitter.com"] = [(IPv4Address, "10.1.2.3")] + self.lookups["publish.twitter.com"] = [(IPv4Address, "10.1.2.3")] + self.lookups["cdn.twitter.com"] = [(IPv4Address, "10.1.2.3")] + + result = b""" + + """ + + channel = self.make_request( + "GET", + "/_matrix/client/unstable/org.matrix.msc3916/media/preview_url?url=http://www.twitter.com/matrixdotorg/status/12345", + shorthand=False, + await_result=False, + ) + self.pump() + + client = self.reactor.tcpClients[0][2].buildProtocol(None) + server = AccumulatingProtocol() + server.makeConnection(FakeTransport(client, self.reactor)) + client.makeConnection(FakeTransport(server, self.reactor)) + client.dataReceived( + ( + b"HTTP/1.0 200 OK\r\nContent-Length: %d\r\n" + b'Content-Type: text/html; charset="utf8"\r\n\r\n' + ) + % (len(result),) + + result + ) + self.pump() + + # The oEmbed response. + result2 = { + "version": "1.0", + "type": "photo", + "url": "http://cdn.twitter.com/matrixdotorg", + } + oembed_content = json.dumps(result2).encode("utf-8") + + # Ensure a second request is made to the oEmbed URL. + client = self.reactor.tcpClients[1][2].buildProtocol(None) + server = AccumulatingProtocol() + server.makeConnection(FakeTransport(client, self.reactor)) + client.makeConnection(FakeTransport(server, self.reactor)) + client.dataReceived( + ( + b"HTTP/1.0 200 OK\r\nContent-Length: %d\r\n" + b'Content-Type: application/json; charset="utf8"\r\n\r\n' + ) + % (len(oembed_content),) + + oembed_content + ) + self.pump() + + # Ensure the URL is what was requested. + self.assertIn(b"/oembed?", server.data) + + # Ensure a third request is made to the photo URL. + client = self.reactor.tcpClients[2][2].buildProtocol(None) + server = AccumulatingProtocol() + server.makeConnection(FakeTransport(client, self.reactor)) + client.makeConnection(FakeTransport(server, self.reactor)) + client.dataReceived( + ( + b"HTTP/1.0 200 OK\r\nContent-Length: %d\r\n" + b"Content-Type: image/png\r\n\r\n" + ) + % (len(SMALL_PNG),) + + SMALL_PNG + ) + self.pump() + + # Ensure the URL is what was requested. + self.assertIn(b"/matrixdotorg", server.data) + + self.assertEqual(channel.code, 200) + body = channel.json_body + self.assertEqual( + body["og:url"], "http://www.twitter.com/matrixdotorg/status/12345" + ) + self._assert_small_png(body) + + @unittest.override_config( + {"url_preview_url_blacklist": [{"netloc": "publish.twitter.com"}]} + ) + def test_oembed_autodiscovery_blocked(self) -> None: + """ + If the discovered oEmbed URL is blocked, it should be discarded. + """ + # This is a little cheesy in that we use the www subdomain (which isn't the + # list of oEmbed patterns) to get "raw" HTML response. + self.lookups["www.twitter.com"] = [(IPv4Address, "10.1.2.3")] + self.lookups["publish.twitter.com"] = [(IPv4Address, "10.1.2.4")] + + result = b""" + Test + + """ + + channel = self.make_request( + "GET", + "/_matrix/client/unstable/org.matrix.msc3916/media/preview_url?url=http://www.twitter.com/matrixdotorg/status/12345", + shorthand=False, + await_result=False, + ) + self.pump() + + client = self.reactor.tcpClients[0][2].buildProtocol(None) + server = AccumulatingProtocol() + server.makeConnection(FakeTransport(client, self.reactor)) + client.makeConnection(FakeTransport(server, self.reactor)) + client.dataReceived( + ( + b"HTTP/1.0 200 OK\r\nContent-Length: %d\r\n" + b'Content-Type: text/html; charset="utf8"\r\n\r\n' + ) + % (len(result),) + + result + ) + + self.pump() + + # Ensure there's no additional connections. + self.assertEqual(len(self.reactor.tcpClients), 1) + + # Ensure the URL is what was requested. + self.assertIn(b"\r\nHost: www.twitter.com\r\n", server.data) + + self.assertEqual(channel.code, 200) + body = channel.json_body + self.assertEqual(body["og:title"], "Test") + self.assertNotIn("og:image", body) + + def _download_image(self) -> Tuple[str, str]: + """Downloads an image into the URL cache. + Returns: + A (host, media_id) tuple representing the MXC URI of the image. + """ + self.lookups["cdn.twitter.com"] = [(IPv4Address, "10.1.2.3")] + + channel = self.make_request( + "GET", + "/_matrix/client/unstable/org.matrix.msc3916/media/preview_url?url=http://cdn.twitter.com/matrixdotorg", + shorthand=False, + await_result=False, + ) + self.pump() + + client = self.reactor.tcpClients[0][2].buildProtocol(None) + server = AccumulatingProtocol() + server.makeConnection(FakeTransport(client, self.reactor)) + client.makeConnection(FakeTransport(server, self.reactor)) + client.dataReceived( + b"HTTP/1.0 200 OK\r\nContent-Length: %d\r\nContent-Type: image/png\r\n\r\n" + % (len(SMALL_PNG),) + + SMALL_PNG + ) + + self.pump() + self.assertEqual(channel.code, 200) + body = channel.json_body + mxc_uri = body["og:image"] + host, _port, media_id = parse_and_validate_mxc_uri(mxc_uri) + self.assertIsNone(_port) + return host, media_id + + def test_storage_providers_exclude_files(self) -> None: + """Test that files are not stored in or fetched from storage providers.""" + host, media_id = self._download_image() + + rel_file_path = self.media_repo.filepaths.url_cache_filepath_rel(media_id) + media_store_path = os.path.join(self.media_store_path, rel_file_path) + storage_provider_path = os.path.join(self.storage_path, rel_file_path) + + # Check storage + self.assertTrue(os.path.isfile(media_store_path)) + self.assertFalse( + os.path.isfile(storage_provider_path), + "URL cache file was unexpectedly stored in a storage provider", + ) + + # Check fetching + channel = self.make_request( + "GET", + f"/_matrix/media/v3/download/{host}/{media_id}", + shorthand=False, + await_result=False, + ) + self.pump() + self.assertEqual(channel.code, 200) + + # Move cached file into the storage provider + os.makedirs(os.path.dirname(storage_provider_path), exist_ok=True) + os.rename(media_store_path, storage_provider_path) + + channel = self.make_request( + "GET", + f"/_matrix/media/v3/download/{host}/{media_id}", + shorthand=False, + await_result=False, + ) + self.pump() + self.assertEqual( + channel.code, + 404, + "URL cache file was unexpectedly retrieved from a storage provider", + ) + + def test_storage_providers_exclude_thumbnails(self) -> None: + """Test that thumbnails are not stored in or fetched from storage providers.""" + host, media_id = self._download_image() + + rel_thumbnail_path = ( + self.media_repo.filepaths.url_cache_thumbnail_directory_rel(media_id) + ) + media_store_thumbnail_path = os.path.join( + self.media_store_path, rel_thumbnail_path + ) + storage_provider_thumbnail_path = os.path.join( + self.storage_path, rel_thumbnail_path + ) + + # Check storage + self.assertTrue(os.path.isdir(media_store_thumbnail_path)) + self.assertFalse( + os.path.isdir(storage_provider_thumbnail_path), + "URL cache thumbnails were unexpectedly stored in a storage provider", + ) + + # Check fetching + channel = self.make_request( + "GET", + f"/_matrix/client/unstable/org.matrix.msc3916/media/thumbnail/{host}/{media_id}?width=32&height=32&method=scale", + shorthand=False, + await_result=False, + ) + self.pump() + self.assertEqual(channel.code, 200) + + # Remove the original, otherwise thumbnails will regenerate + rel_file_path = self.media_repo.filepaths.url_cache_filepath_rel(media_id) + media_store_path = os.path.join(self.media_store_path, rel_file_path) + os.remove(media_store_path) + + # Move cached thumbnails into the storage provider + os.makedirs(os.path.dirname(storage_provider_thumbnail_path), exist_ok=True) + os.rename(media_store_thumbnail_path, storage_provider_thumbnail_path) + + channel = self.make_request( + "GET", + f"/_matrix/client/unstable/org.matrix.msc3916/media/thumbnail/{host}/{media_id}?width=32&height=32&method=scale", + shorthand=False, + await_result=False, + ) + self.pump() + self.assertEqual( + channel.code, + 404, + "URL cache thumbnail was unexpectedly retrieved from a storage provider", + ) + + def test_cache_expiry(self) -> None: + """Test that URL cache files and thumbnails are cleaned up properly on expiry.""" + _host, media_id = self._download_image() + + file_path = self.media_repo.filepaths.url_cache_filepath(media_id) + file_dirs = self.media_repo.filepaths.url_cache_filepath_dirs_to_delete( + media_id + ) + thumbnail_dir = self.media_repo.filepaths.url_cache_thumbnail_directory( + media_id + ) + thumbnail_dirs = self.media_repo.filepaths.url_cache_thumbnail_dirs_to_delete( + media_id + ) + + self.assertTrue(os.path.isfile(file_path)) + self.assertTrue(os.path.isdir(thumbnail_dir)) + + self.reactor.advance(IMAGE_CACHE_EXPIRY_MS * 1000 + 1) + self.get_success(self.url_previewer._expire_url_cache_data()) + + for path in [file_path] + file_dirs + [thumbnail_dir] + thumbnail_dirs: + self.assertFalse( + os.path.exists(path), + f"{os.path.relpath(path, self.media_store_path)} was not deleted", + ) + + @unittest.override_config({"url_preview_url_blacklist": [{"port": "*"}]}) + def test_blocked_port(self) -> None: + """Tests that blocking URLs with a port makes previewing such URLs + fail with a 403 error and doesn't impact other previews. + """ + self.lookups["matrix.org"] = [(IPv4Address, "10.1.2.3")] + + bad_url = quote("http://matrix.org:8888/foo") + good_url = quote("http://matrix.org/foo") + + channel = self.make_request( + "GET", + "/_matrix/client/unstable/org.matrix.msc3916/media/preview_url?url=" + + bad_url, + shorthand=False, + await_result=False, + ) + self.pump() + self.assertEqual(channel.code, 403, channel.result) + + channel = self.make_request( + "GET", + "/_matrix/client/unstable/org.matrix.msc3916/media/preview_url?url=" + + good_url, + shorthand=False, + await_result=False, + ) + self.pump() + + client = self.reactor.tcpClients[0][2].buildProtocol(None) + server = AccumulatingProtocol() + server.makeConnection(FakeTransport(client, self.reactor)) + client.makeConnection(FakeTransport(server, self.reactor)) + client.dataReceived( + b"HTTP/1.0 200 OK\r\nContent-Length: %d\r\nContent-Type: text/html\r\n\r\n" + % (len(self.end_content),) + + self.end_content + ) + + self.pump() + self.assertEqual(channel.code, 200) + + @unittest.override_config( + {"url_preview_url_blacklist": [{"netloc": "example.com"}]} + ) + def test_blocked_url(self) -> None: + """Tests that blocking URLs with a host makes previewing such URLs + fail with a 403 error. + """ + self.lookups["example.com"] = [(IPv4Address, "10.1.2.3")] + + bad_url = quote("http://example.com/foo") + + channel = self.make_request( + "GET", + "/_matrix/client/unstable/org.matrix.msc3916/media/preview_url?url=" + + bad_url, + shorthand=False, + await_result=False, + ) + self.pump() + self.assertEqual(channel.code, 403, channel.result) + + +class UnstableMediaConfigTest(unittest.HomeserverTestCase): + servlets = [ + media.register_servlets, + admin.register_servlets, + login.register_servlets, + ] + + def make_homeserver( + self, reactor: ThreadedMemoryReactorClock, clock: Clock + ) -> HomeServer: + config = self.default_config() + config["experimental_features"] = {"msc3916_authenticated_media_enabled": True} + + self.storage_path = self.mktemp() + self.media_store_path = self.mktemp() + os.mkdir(self.storage_path) + os.mkdir(self.media_store_path) + config["media_store_path"] = self.media_store_path + + provider_config = { + "module": "synapse.media.storage_provider.FileStorageProviderBackend", + "store_local": True, + "store_synchronous": False, + "store_remote": True, + "config": {"directory": self.storage_path}, + } + + config["media_storage_providers"] = [provider_config] + + return self.setup_test_homeserver(config=config) + + def prepare(self, reactor: MemoryReactor, clock: Clock, hs: HomeServer) -> None: + self.register_user("user", "password") + self.tok = self.login("user", "password") + + def test_media_config(self) -> None: + channel = self.make_request( + "GET", + "/_matrix/client/unstable/org.matrix.msc3916/media/config", + shorthand=False, + access_token=self.tok, + ) + self.assertEqual(channel.code, 200) + self.assertEqual( + channel.json_body["m.upload.size"], self.hs.config.media.max_upload_size + )