diff options
Diffstat (limited to 'synapse/http')
| -rw-r--r-- | synapse/http/__init__.py | 22 | ||||
| -rw-r--r-- | synapse/http/client.py | 24 | ||||
| -rw-r--r-- | synapse/http/endpoint.py | 116 | ||||
| -rw-r--r-- | synapse/http/matrixfederationclient.py | 47 | ||||
| -rw-r--r-- | synapse/http/server.py | 19 | ||||
| -rw-r--r-- | synapse/http/site.py | 6 |
6 files changed, 92 insertions, 142 deletions
diff --git a/synapse/http/__init__.py b/synapse/http/__init__.py index bfebb0f644..054372e179 100644 --- a/synapse/http/__init__.py +++ b/synapse/http/__init__.py @@ -1,5 +1,6 @@ # -*- coding: utf-8 -*- # Copyright 2014-2016 OpenMarket Ltd +# Copyright 2018 New Vector Ltd # # Licensed under the Apache License, Version 2.0 (the "License"); # you may not use this file except in compliance with the License. @@ -12,3 +13,24 @@ # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. # See the License for the specific language governing permissions and # limitations under the License. +from twisted.internet.defer import CancelledError +from twisted.python import failure + +from synapse.api.errors import SynapseError + + +class RequestTimedOutError(SynapseError): + """Exception representing timeout of an outbound request""" + def __init__(self): + super(RequestTimedOutError, self).__init__(504, "Timed out") + + +def cancelled_to_request_timed_out_error(value, timeout): + """Turns CancelledErrors into RequestTimedOutErrors. + + For use with async.add_timeout_to_deferred + """ + if isinstance(value, failure.Failure): + value.trap(CancelledError) + raise RequestTimedOutError() + return value diff --git a/synapse/http/client.py b/synapse/http/client.py index f3e4973c2e..70a19d9b74 100644 --- a/synapse/http/client.py +++ b/synapse/http/client.py @@ -1,5 +1,6 @@ # -*- coding: utf-8 -*- # Copyright 2014-2016 OpenMarket Ltd +# Copyright 2018 New Vector Ltd # # Licensed under the Apache License, Version 2.0 (the "License"); # you may not use this file except in compliance with the License. @@ -18,9 +19,10 @@ from OpenSSL.SSL import VERIFY_NONE from synapse.api.errors import ( CodeMessageException, MatrixCodeMessageException, SynapseError, Codes, ) +from synapse.http import cancelled_to_request_timed_out_error +from synapse.util.async import add_timeout_to_deferred from synapse.util.caches import CACHE_SIZE_FACTOR from synapse.util.logcontext import make_deferred_yieldable -from synapse.util import logcontext import synapse.metrics from synapse.http.endpoint import SpiderEndpoint @@ -38,7 +40,7 @@ from twisted.web.http import PotentialDataLoss from twisted.web.http_headers import Headers from twisted.web._newclient import ResponseDone -from StringIO import StringIO +from six import StringIO import simplejson as json import logging @@ -95,21 +97,17 @@ class SimpleHttpClient(object): # counters to it outgoing_requests_counter.inc(method) - def send_request(): + logger.info("Sending request %s %s", method, uri) + + try: request_deferred = self.agent.request( method, uri, *args, **kwargs ) - - return self.clock.time_bound_deferred( + add_timeout_to_deferred( request_deferred, - time_out=60, + 60, cancelled_to_request_timed_out_error, ) - - logger.info("Sending request %s %s", method, uri) - - try: - with logcontext.PreserveLoggingContext(): - response = yield send_request() + response = yield make_deferred_yieldable(request_deferred) incoming_responses_counter.inc(method, response.code) logger.info( @@ -509,7 +507,7 @@ class SpiderHttpClient(SimpleHttpClient): reactor, SpiderEndpointFactory(hs) ) - ), [('gzip', GzipDecoder)] + ), [(b'gzip', GzipDecoder)] ) # We could look like Chrome: # self.user_agent = ("Mozilla/5.0 (%s) (KHTML, like Gecko) diff --git a/synapse/http/endpoint.py b/synapse/http/endpoint.py index 87639b9151..87a482650d 100644 --- a/synapse/http/endpoint.py +++ b/synapse/http/endpoint.py @@ -12,8 +12,6 @@ # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. # See the License for the specific language governing permissions and # limitations under the License. -import socket - from twisted.internet.endpoints import HostnameEndpoint, wrapClientTLS from twisted.internet import defer, reactor from twisted.internet.error import ConnectError @@ -33,7 +31,7 @@ SERVER_CACHE = {} # our record of an individual server which can be tried to reach a destination. # -# "host" is actually a dotted-quad or ipv6 address string. Except when there's +# "host" is the hostname acquired from the SRV record. Except when there's # no SRV record, in which case it is the original hostname. _Server = collections.namedtuple( "_Server", "priority weight host port expires" @@ -117,10 +115,15 @@ class _WrappedConnection(object): if time.time() - self.last_request >= 2.5 * 60: self.abort() # Abort the underlying TLS connection. The abort() method calls - # loseConnection() on the underlying TLS connection which tries to + # loseConnection() on the TLS connection which tries to # shutdown the connection cleanly. We call abortConnection() - # since that will promptly close the underlying TCP connection. - self.transport.abortConnection() + # since that will promptly close the TLS connection. + # + # In Twisted >18.4; the TLS connection will be None if it has closed + # which will make abortConnection() throw. Check that the TLS connection + # is not None before trying to close it. + if self.transport.getHandle() is not None: + self.transport.abortConnection() def request(self, request): self.last_request = time.time() @@ -288,7 +291,7 @@ def resolve_service(service_name, dns_client=client, cache=SERVER_CACHE, clock=t if (len(answers) == 1 and answers[0].type == dns.SRV and answers[0].payload - and answers[0].payload.target == dns.Name('.')): + and answers[0].payload.target == dns.Name(b'.')): raise ConnectError("Service %s unavailable" % service_name) for answer in answers: @@ -297,20 +300,13 @@ def resolve_service(service_name, dns_client=client, cache=SERVER_CACHE, clock=t payload = answer.payload - hosts = yield _get_hosts_for_srv_record( - dns_client, str(payload.target) - ) - - for (ip, ttl) in hosts: - host_ttl = min(answer.ttl, ttl) - - servers.append(_Server( - host=ip, - port=int(payload.port), - priority=int(payload.priority), - weight=int(payload.weight), - expires=int(clock.time()) + host_ttl, - )) + servers.append(_Server( + host=str(payload.target), + port=int(payload.port), + priority=int(payload.priority), + weight=int(payload.weight), + expires=int(clock.time()) + answer.ttl, + )) servers.sort() cache[service_name] = list(servers) @@ -328,81 +324,3 @@ def resolve_service(service_name, dns_client=client, cache=SERVER_CACHE, clock=t raise e defer.returnValue(servers) - - -@defer.inlineCallbacks -def _get_hosts_for_srv_record(dns_client, host): - """Look up each of the hosts in a SRV record - - Args: - dns_client (twisted.names.dns.IResolver): - host (basestring): host to look up - - Returns: - Deferred[list[(str, int)]]: a list of (host, ttl) pairs - - """ - ip4_servers = [] - ip6_servers = [] - - def cb(res): - # lookupAddress and lookupIP6Address return a three-tuple - # giving the answer, authority, and additional sections of the - # response. - # - # we only care about the answers. - - return res[0] - - def eb(res, record_type): - if res.check(DNSNameError): - return [] - logger.warn("Error looking up %s for %s: %s", record_type, host, res) - return res - - # no logcontexts here, so we can safely fire these off and gatherResults - d1 = dns_client.lookupAddress(host).addCallbacks( - cb, eb, errbackArgs=("A", )) - d2 = dns_client.lookupIPV6Address(host).addCallbacks( - cb, eb, errbackArgs=("AAAA", )) - results = yield defer.DeferredList( - [d1, d2], consumeErrors=True) - - # if all of the lookups failed, raise an exception rather than blowing out - # the cache with an empty result. - if results and all(s == defer.FAILURE for (s, _) in results): - defer.returnValue(results[0][1]) - - for (success, result) in results: - if success == defer.FAILURE: - continue - - for answer in result: - if not answer.payload: - continue - - try: - if answer.type == dns.A: - ip = answer.payload.dottedQuad() - ip4_servers.append((ip, answer.ttl)) - elif answer.type == dns.AAAA: - ip = socket.inet_ntop( - socket.AF_INET6, answer.payload.address, - ) - ip6_servers.append((ip, answer.ttl)) - else: - # the most likely candidate here is a CNAME record. - # rfc2782 says srvs may not point to aliases. - logger.warn( - "Ignoring unexpected DNS record type %s for %s", - answer.type, host, - ) - continue - except Exception as e: - logger.warn("Ignoring invalid DNS response for %s: %s", - host, e) - continue - - # keep the ipv4 results before the ipv6 results, mostly to match historical - # behaviour. - defer.returnValue(ip4_servers + ip6_servers) diff --git a/synapse/http/matrixfederationclient.py b/synapse/http/matrixfederationclient.py index 9145405cb0..4b2b85464d 100644 --- a/synapse/http/matrixfederationclient.py +++ b/synapse/http/matrixfederationclient.py @@ -1,5 +1,6 @@ # -*- coding: utf-8 -*- # Copyright 2014-2016 OpenMarket Ltd +# Copyright 2018 New Vector Ltd # # Licensed under the Apache License, Version 2.0 (the "License"); # you may not use this file except in compliance with the License. @@ -12,17 +13,19 @@ # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. # See the License for the specific language governing permissions and # limitations under the License. -import synapse.util.retryutils from twisted.internet import defer, reactor, protocol from twisted.internet.error import DNSLookupError from twisted.web.client import readBody, HTTPConnectionPool, Agent from twisted.web.http_headers import Headers from twisted.web._newclient import ResponseDone +from synapse.http import cancelled_to_request_timed_out_error from synapse.http.endpoint import matrix_federation_endpoint -from synapse.util.async import sleep -from synapse.util import logcontext import synapse.metrics +from synapse.util.async import sleep, add_timeout_to_deferred +from synapse.util import logcontext +from synapse.util.logcontext import make_deferred_yieldable +import synapse.util.retryutils from canonicaljson import encode_canonical_json @@ -38,8 +41,7 @@ import logging import random import sys import urllib -import urlparse - +from six.moves.urllib import parse as urlparse logger = logging.getLogger(__name__) outbound_logger = logging.getLogger("synapse.http.outbound") @@ -184,21 +186,20 @@ class MatrixFederationHttpClient(object): producer = body_callback(method, http_url_bytes, headers_dict) try: - def send_request(): - request_deferred = self.agent.request( - method, - url_bytes, - Headers(headers_dict), - producer - ) - - return self.clock.time_bound_deferred( - request_deferred, - time_out=timeout / 1000. if timeout else 60, - ) - - with logcontext.PreserveLoggingContext(): - response = yield send_request() + request_deferred = self.agent.request( + method, + url_bytes, + Headers(headers_dict), + producer + ) + add_timeout_to_deferred( + request_deferred, + timeout / 1000. if timeout else 60, + cancelled_to_request_timed_out_error, + ) + response = yield make_deferred_yieldable( + request_deferred, + ) log_result = "%d %s" % (response.code, response.phrase,) break @@ -286,7 +287,8 @@ class MatrixFederationHttpClient(object): headers_dict[b"Authorization"] = auth_headers @defer.inlineCallbacks - def put_json(self, destination, path, data={}, json_data_callback=None, + def put_json(self, destination, path, args={}, data={}, + json_data_callback=None, long_retries=False, timeout=None, ignore_backoff=False, backoff_on_404=False): @@ -296,6 +298,7 @@ class MatrixFederationHttpClient(object): destination (str): The remote server to send the HTTP request to. path (str): The HTTP path. + args (dict): query params data (dict): A dict containing the data that will be used as the request body. This will be encoded as JSON. json_data_callback (callable): A callable returning the dict to @@ -342,6 +345,7 @@ class MatrixFederationHttpClient(object): path, body_callback=body_callback, headers_dict={"Content-Type": ["application/json"]}, + query_bytes=encode_query_args(args), long_retries=long_retries, timeout=timeout, ignore_backoff=ignore_backoff, @@ -373,6 +377,7 @@ class MatrixFederationHttpClient(object): giving up. None indicates no timeout. ignore_backoff (bool): true to ignore the historical backoff data and try the request anyway. + args (dict): query params Returns: Deferred: Succeeds when we get a 2xx HTTP response. The result will be the decoded JSON body. diff --git a/synapse/http/server.py b/synapse/http/server.py index 4b567215c8..55b9ad5251 100644 --- a/synapse/http/server.py +++ b/synapse/http/server.py @@ -37,7 +37,7 @@ from twisted.web.util import redirectTo import collections import logging import urllib -import ujson +import simplejson logger = logging.getLogger(__name__) @@ -113,6 +113,11 @@ response_db_sched_duration = metrics.register_counter( "response_db_sched_duration_seconds", labels=["method", "servlet", "tag"] ) +# size in bytes of the response written +response_size = metrics.register_counter( + "response_size", labels=["method", "servlet", "tag"] +) + _next_request_id = 0 @@ -324,7 +329,7 @@ class JsonResource(HttpServer, resource.Resource): register_paths, so will return (possibly via Deferred) either None, or a tuple of (http code, response body). """ - if request.method == "OPTIONS": + if request.method == b"OPTIONS": return _options_handler, {} # Loop through all the registered callbacks to check if the method @@ -426,6 +431,8 @@ class RequestMetrics(object): context.db_sched_duration_ms / 1000., request.method, self.name, tag ) + response_size.inc_by(request.sentLength, request.method, self.name, tag) + class RootRedirect(resource.Resource): """Redirects the root '/' path to another path.""" @@ -461,8 +468,7 @@ def respond_with_json(request, code, json_object, send_cors=False, if canonical_json or synapse.events.USE_FROZEN_DICTS: json_bytes = encode_canonical_json(json_object) else: - # ujson doesn't like frozen_dicts. - json_bytes = ujson.dumps(json_object, ensure_ascii=False) + json_bytes = simplejson.dumps(json_object) return respond_with_json_bytes( request, code, json_bytes, @@ -489,6 +495,7 @@ def respond_with_json_bytes(request, code, json_bytes, send_cors=False, request.setHeader(b"Content-Type", b"application/json") request.setHeader(b"Server", version_string) request.setHeader(b"Content-Length", b"%d" % (len(json_bytes),)) + request.setHeader(b"Cache-Control", b"no-cache, no-store, must-revalidate") if send_cors: set_cors_headers(request) @@ -536,9 +543,9 @@ def finish_request(request): def _request_user_agent_is_curl(request): user_agents = request.requestHeaders.getRawHeaders( - "User-Agent", default=[] + b"User-Agent", default=[] ) for user_agent in user_agents: - if "curl" in user_agent: + if b"curl" in user_agent: return True return False diff --git a/synapse/http/site.py b/synapse/http/site.py index e422c8dfae..c8b46e1af2 100644 --- a/synapse/http/site.py +++ b/synapse/http/site.py @@ -20,7 +20,7 @@ import logging import re import time -ACCESS_TOKEN_RE = re.compile(r'(\?.*access(_|%5[Ff])token=)[^&]*(.*)$') +ACCESS_TOKEN_RE = re.compile(br'(\?.*access(_|%5[Ff])token=)[^&]*(.*)$') class SynapseRequest(Request): @@ -43,12 +43,12 @@ class SynapseRequest(Request): def get_redacted_uri(self): return ACCESS_TOKEN_RE.sub( - r'\1<redacted>\3', + br'\1<redacted>\3', self.uri ) def get_user_agent(self): - return self.requestHeaders.getRawHeaders("User-Agent", [None])[-1] + return self.requestHeaders.getRawHeaders(b"User-Agent", [None])[-1] def started_processing(self): self.site.access_logger.info( |