1
0

Merge commit 'e22b71810' into anoa/dinsic_release_1_31_0

This commit is contained in:
Andrew Morgan
2021-04-23 14:12:47 +01:00
15 changed files with 231 additions and 86 deletions

1
changelog.d/8957.feature Normal file
View File

@@ -0,0 +1 @@
Add rate limiters to cross-user key sharing requests.

1
changelog.d/9385.feature Normal file
View File

@@ -0,0 +1 @@
Add a configuration option, `user_directory.prefer_local_users`, which when enabled will make it more likely for users on the same server as you to appear above other users.

1
changelog.d/9432.misc Normal file
View File

@@ -0,0 +1 @@
Add documentation and type hints to `parse_duration`.

View File

@@ -2724,29 +2724,41 @@ spam_checker:
# User Directory configuration
#
# 'enabled' defines whether users can search the user directory. If
# false then empty responses are returned to all queries. Defaults to
# true.
#
# 'search_all_users' defines whether to search all users visible to your HS
# when searching the user directory, rather than limiting to users visible
# in public rooms. Defaults to false. If you set it True, you'll have to
# rebuild the user_directory search indexes, see
# https://github.com/matrix-org/synapse/blob/master/docs/user_directory.md
#
# 'prefer_local_users' defines whether to prioritise local users in
# search query results. If True, local users are more likely to appear above
# remote users when searching the user directory. Defaults to false.
#
#user_directory:
# enabled: true
# search_all_users: false
# prefer_local_users: false
#
# # If this is set, user search will be delegated to this ID server instead
# # of synapse performing the search itself.
# # This is an experimental API.
# defer_to_id_server: https://id.example.com
user_directory:
# Defines whether users can search the user directory. If false then
# empty responses are returned to all queries. Defaults to true.
#
# Uncomment to disable the user directory.
#
#enabled: false
# Defines whether to search all users visible to your HS when searching
# the user directory, rather than limiting to users visible in public
# rooms. Defaults to false.
#
# If you set it true, you'll have to rebuild the user_directory search
# indexes, see:
# https://github.com/matrix-org/synapse/blob/master/docs/user_directory.md
#
# Uncomment to return search results containing all known users, even if that
# user does not share a room with the requester.
#
#search_all_users: true
# If this is set, user search will be delegated to this ID server instead
# of Synapse performing the search itself.
# This is an experimental API.
#
#defer_to_id_server: https://id.example.com
# Defines whether to prefer local users in search query results.
# If True, local users are more likely to appear above remote users
# when searching the user directory. Defaults to false.
#
# Uncomment to prefer local over remote users in user directory search
# results.
#
#prefer_local_users: true
# User Consent configuration

View File

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

View File

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

View File

@@ -22,7 +22,7 @@ from collections import OrderedDict
from hashlib import sha256
from io import open as io_open
from textwrap import dedent
from typing import Any, Iterable, List, MutableMapping, Optional
from typing import Any, Iterable, List, MutableMapping, Optional, Union
import attr
import jinja2
@@ -148,7 +148,20 @@ class Config:
return int(value) * size
@staticmethod
def parse_duration(value):
def parse_duration(value: Union[str, int]) -> int:
"""Convert a duration as a string or integer to a number of milliseconds.
If an integer is provided it is treated as milliseconds and is unchanged.
String durations can have a suffix of 's', 'm', 'h', 'd', 'w', or 'y'.
No suffix is treated as milliseconds.
Args:
value: The duration to parse.
Returns:
The number of milliseconds in the duration.
"""
if isinstance(value, int):
return value
second = 1000

View File

@@ -105,6 +105,16 @@ class RatelimitConfig(Config):
defaults={"per_second": 0.01, "burst_count": 3},
)
# Ratelimit cross-user key requests:
# * For local requests this is keyed by the sending device.
# * For requests received over federation this is keyed by the origin.
#
# Note that this isn't exposed in the configuration as it is obscure.
self.rc_key_requests = RateLimitConfig(
config.get("rc_key_requests", {}),
defaults={"per_second": 20, "burst_count": 100},
)
self.rc_3pid_validation = RateLimitConfig(
config.get("rc_3pid_validation") or {},
defaults={"per_second": 0.003, "burst_count": 5},

View File

@@ -24,50 +24,55 @@ class UserDirectoryConfig(Config):
section = "userdirectory"
def read_config(self, config, **kwargs):
self.user_directory_search_enabled = True
self.user_directory_search_all_users = False
self.user_directory_defer_to_id_server = None
self.user_directory_search_prefer_local_users = False
user_directory_config = config.get("user_directory", None)
if user_directory_config:
self.user_directory_search_enabled = user_directory_config.get(
"enabled", True
)
self.user_directory_search_all_users = user_directory_config.get(
"search_all_users", False
)
self.user_directory_defer_to_id_server = user_directory_config.get(
"defer_to_id_server", None
)
self.user_directory_search_prefer_local_users = user_directory_config.get(
"prefer_local_users", False
)
user_directory_config = config.get("user_directory") or {}
self.user_directory_search_enabled = user_directory_config.get("enabled", True)
self.user_directory_search_all_users = user_directory_config.get(
"search_all_users", False
)
self.user_directory_defer_to_id_server = user_directory_config.get(
"defer_to_id_server", None
)
self.user_directory_search_prefer_local_users = user_directory_config.get(
"prefer_local_users", False
)
def generate_config_section(self, config_dir_path, server_name, **kwargs):
return """
# User Directory configuration
#
# 'enabled' defines whether users can search the user directory. If
# false then empty responses are returned to all queries. Defaults to
# true.
#
# 'search_all_users' defines whether to search all users visible to your HS
# when searching the user directory, rather than limiting to users visible
# in public rooms. Defaults to false. If you set it True, you'll have to
# rebuild the user_directory search indexes, see
# https://github.com/matrix-org/synapse/blob/master/docs/user_directory.md
#
# 'prefer_local_users' defines whether to prioritise local users in
# search query results. If True, local users are more likely to appear above
# remote users when searching the user directory. Defaults to false.
#
#user_directory:
# enabled: true
# search_all_users: false
# prefer_local_users: false
#
# # If this is set, user search will be delegated to this ID server instead
# # of synapse performing the search itself.
# # This is an experimental API.
# defer_to_id_server: https://id.example.com
user_directory:
# Defines whether users can search the user directory. If false then
# empty responses are returned to all queries. Defaults to true.
#
# Uncomment to disable the user directory.
#
#enabled: false
# Defines whether to search all users visible to your HS when searching
# the user directory, rather than limiting to users visible in public
# rooms. Defaults to false.
#
# If you set it true, you'll have to rebuild the user_directory search
# indexes, see:
# https://github.com/matrix-org/synapse/blob/master/docs/user_directory.md
#
# Uncomment to return search results containing all known users, even if that
# user does not share a room with the requester.
#
#search_all_users: true
# If this is set, user search will be delegated to this ID server instead
# of Synapse performing the search itself.
# This is an experimental API.
#
#defer_to_id_server: https://id.example.com
# Defines whether to prefer local users in search query results.
# If True, local users are more likely to appear above remote users
# when searching the user directory. Defaults to false.
#
# Uncomment to prefer local over remote users in user directory search
# results.
#
#prefer_local_users: true
"""

View File

@@ -34,7 +34,7 @@ from twisted.internet import defer
from twisted.internet.abstract import isIPAddress
from twisted.python import failure
from synapse.api.constants import EventTypes, Membership
from synapse.api.constants import EduTypes, EventTypes, Membership
from synapse.api.errors import (
AuthError,
Codes,
@@ -44,6 +44,7 @@ from synapse.api.errors import (
SynapseError,
UnsupportedRoomVersionError,
)
from synapse.api.ratelimiting import Ratelimiter
from synapse.api.room_versions import KNOWN_ROOM_VERSIONS
from synapse.config.api import DEFAULT_ROOM_STATE_TYPES
from synapse.events import EventBase
@@ -945,6 +946,13 @@ class FederationHandlerRegistry:
# EDU received.
self._edu_type_to_instance = {} # type: Dict[str, List[str]]
# A rate limiter for incoming room key requests per origin.
self._room_key_request_rate_limiter = Ratelimiter(
clock=self.clock,
rate_hz=self.config.rc_key_requests.per_second,
burst_count=self.config.rc_key_requests.burst_count,
)
def register_edu_handler(
self, edu_type: str, handler: Callable[[str, JsonDict], Awaitable[None]]
):
@@ -993,7 +1001,15 @@ class FederationHandlerRegistry:
self._edu_type_to_instance[edu_type] = instance_names
async def on_edu(self, edu_type: str, origin: str, content: dict):
if not self.config.use_presence and edu_type == "m.presence":
if not self.config.use_presence and edu_type == EduTypes.Presence:
return
# If the incoming room key requests from a particular origin are over
# the limit, drop them.
if (
edu_type == EduTypes.RoomKeyRequest
and not self._room_key_request_rate_limiter.can_do_action(origin)
):
return
# Check if we have a handler on this instance

View File

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

View File

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

View File

@@ -18,7 +18,7 @@ from typing import TYPE_CHECKING, Optional, Tuple
from twisted.internet import defer
from synapse.api.constants import EventTypes, Membership
from synapse.api.constants import EduTypes, EventTypes, Membership
from synapse.api.errors import SynapseError
from synapse.events.validator import EventValidator
from synapse.handlers.presence import format_user_presence_state
@@ -412,7 +412,7 @@ class InitialSyncHandler(BaseHandler):
return [
{
"type": EventTypes.Presence,
"type": EduTypes.Presence,
"content": format_user_presence_state(s, time_now),
}
for s in states

View File

@@ -56,10 +56,8 @@ class SendToDeviceRestServlet(servlet.RestServlet):
content = parse_json_object_from_request(request)
assert_params_in_dict(content, ("messages",))
sender_user_id = requester.user.to_string()
await self.device_message_handler.send_device_message(
sender_user_id, message_type, content["messages"]
requester, message_type, content["messages"]
)
response = (200, {}) # type: Tuple[int, dict]

View File

@@ -231,9 +231,11 @@ class MediaRepoTests(unittest.HomeserverTestCase):
def prepare(self, reactor, clock, hs):
self.media_repo = hs.get_media_repository_resource()
self.download_resource = self.media_repo.children[b"download"]
self.thumbnail_resource = self.media_repo.children[b"thumbnail"]
media_resource = hs.get_media_repository_resource()
self.download_resource = media_resource.children[b"download"]
self.thumbnail_resource = media_resource.children[b"thumbnail"]
self.store = hs.get_datastore()
self.media_repo = hs.get_media_repository()
self.media_id = "example.com/12345"
@@ -357,6 +359,67 @@ class MediaRepoTests(unittest.HomeserverTestCase):
"""
self._test_thumbnail("scale", None, False)
def test_thumbnail_repeated_thumbnail(self):
"""Test that fetching the same thumbnail works, and deleting the on disk
thumbnail regenerates it.
"""
self._test_thumbnail(
"scale", self.test_image.expected_scaled, self.test_image.expected_found
)
if not self.test_image.expected_found:
return
# Fetching again should work, without re-requesting the image from the
# remote.
params = "?width=32&height=32&method=scale"
channel = make_request(
self.reactor,
FakeSite(self.thumbnail_resource),
"GET",
self.media_id + params,
shorthand=False,
await_result=False,
)
self.pump()
self.assertEqual(channel.code, 200)
if self.test_image.expected_scaled:
self.assertEqual(
channel.result["body"],
self.test_image.expected_scaled,
channel.result["body"],
)
# Deleting the thumbnail on disk then re-requesting it should work as
# Synapse should regenerate missing thumbnails.
origin, media_id = self.media_id.split("/")
info = self.get_success(self.store.get_cached_remote_media(origin, media_id))
file_id = info["filesystem_id"]
thumbnail_dir = self.media_repo.filepaths.remote_media_thumbnail_dir(
origin, file_id
)
shutil.rmtree(thumbnail_dir, ignore_errors=True)
channel = make_request(
self.reactor,
FakeSite(self.thumbnail_resource),
"GET",
self.media_id + params,
shorthand=False,
await_result=False,
)
self.pump()
self.assertEqual(channel.code, 200)
if self.test_image.expected_scaled:
self.assertEqual(
channel.result["body"],
self.test_image.expected_scaled,
channel.result["body"],
)
def _test_thumbnail(self, method, expected_body, expected_found):
params = "?width=32&height=32&method=" + method
channel = make_request(