Merge branch '38-websocket-proxying' into 'master'

Resolve "Websocket Proxying"

Closes #38

See merge request BeryJu.org/passbook!24

Dockerfile

@@ -6,7 +6,7 @@ COPY ./requirements.txt /app/
 
 WORKDIR /app/
 
-RUN apt-get update && apt-get install build-essential libssl-dev libffi-dev -y && \
+RUN apt-get update && apt-get install build-essential libssl-dev libffi-dev libpq-dev -y && \
     mkdir /app/static/ && \
     pip install -r requirements.txt && \
     pip install psycopg2 && \
@@ -23,7 +23,7 @@ COPY --from=build /app/static /app/static/
 
 WORKDIR /app/
 
-RUN apt-get update && apt-get install build-essential libssl-dev libffi-dev -y && \
+RUN apt-get update && apt-get install build-essential libssl-dev libffi-dev libpq-dev -y && \
     pip install -r requirements.txt && \
     pip install psycopg2 && \
     adduser --system --home /app/ passbook && \

helm/passbook/templates/passbook-web-deployment.yaml

@@ -29,7 +29,7 @@ spec:
           image: "docker.pkg.beryju.org/passbook:{{ .Values.image.tag }}"
           imagePullPolicy: IfNotPresent
           command: ["/bin/sh","-c"]
-          args: ["./manage.py migrate && ./manage.py web"]
+          args: ["./manage.py migrate && daphne -p 8000 passbook.core.asgi:application"]
           ports:
             - name: http
               containerPort: 8000

passbook/app_gw/middleware.py

@@ -10,10 +10,10 @@ from django.core.cache import cache
 from django.utils.http import urlencode
 from django.views.generic import RedirectView
 from revproxy.exceptions import InvalidUpstream
-from revproxy.response import get_django_response
-from revproxy.utils import encode_items, normalize_request_headers
 
 from passbook.app_gw.models import ApplicationGatewayProvider
+from passbook.app_gw.proxy.response import get_django_response
+from passbook.app_gw.proxy.utils import encode_items, normalize_request_headers
 from passbook.app_gw.rewrite import Rewriter
 from passbook.core.models import Application
 from passbook.core.policies import PolicyEngine

passbook/app_gw/proxy/exceptions.py

@@ -0,0 +1,8 @@
+"""Exception classes"""
+
+class ReverseProxyException(Exception):
+    """Base for revproxy exception"""
+
+
+class InvalidUpstream(ReverseProxyException):
+    """Invalid upstream set"""

passbook/app_gw/proxy/response.py

@@ -0,0 +1,63 @@
+"""response functions from django-revproxy"""
+import logging
+
+from django.http import HttpResponse, StreamingHttpResponse
+
+from passbook.app_gw.proxy.utils import (cookie_from_string,
+                                         set_response_headers, should_stream)
+
+#: Default number of bytes that are going to be read in a file lecture
+DEFAULT_AMT = 2 ** 16
+
+logger = logging.getLogger('revproxy.response')
+
+
+def get_django_response(proxy_response, strict_cookies=False):
+    """This method is used to create an appropriate response based on the
+    Content-Length of the proxy_response. If the content is bigger than
+    MIN_STREAMING_LENGTH, which is found on utils.py,
+    than django.http.StreamingHttpResponse will be created,
+    else a django.http.HTTPResponse will be created instead
+
+    :param proxy_response: An Instance of urllib3.response.HTTPResponse that
+                           will create an appropriate response
+    :param strict_cookies: Whether to only accept RFC-compliant cookies
+    :returns: Returns an appropriate response based on the proxy_response
+              content-length
+    """
+    status = proxy_response.status
+    headers = proxy_response.headers
+
+    logger.debug('Proxy response headers: %s', headers)
+
+    content_type = headers.get('Content-Type')
+
+    logger.debug('Content-Type: %s', content_type)
+
+    if should_stream(proxy_response):
+        logger.info('Content-Length is bigger than %s', DEFAULT_AMT)
+        response = StreamingHttpResponse(proxy_response.stream(DEFAULT_AMT),
+                                         status=status,
+                                         content_type=content_type)
+    else:
+        content = proxy_response.data or b''
+        response = HttpResponse(content, status=status,
+                                content_type=content_type)
+
+    logger.info('Normalizing response headers')
+    set_response_headers(response, headers)
+
+    logger.debug('Response headers: %s', getattr(response, '_headers'))
+
+    cookies = proxy_response.headers.getlist('set-cookie')
+    logger.info('Checking for invalid cookies')
+    for cookie_string in cookies:
+        cookie_dict = cookie_from_string(cookie_string,
+                                         strict_cookies=strict_cookies)
+        # if cookie is invalid cookie_dict will be None
+        if cookie_dict:
+            response.set_cookie(**cookie_dict)
+
+    logger.debug('Response cookies: %s', response.cookies)
+
+    return response

passbook/app_gw/proxy/utils.py

@@ -0,0 +1,227 @@
+"""Utils from django-revproxy, slightly adjusted"""
+import logging
+import re
+from wsgiref.util import is_hop_by_hop
+
+try:
+    from http.cookies import SimpleCookie
+    COOKIE_PREFIX = ''
+except ImportError:
+    from Cookie import SimpleCookie
+    COOKIE_PREFIX = 'Set-Cookie: '
+
+
+#: List containing string constant that are used to represent headers that can
+#: be ignored in the required_header function
+IGNORE_HEADERS = (
+    'HTTP_ACCEPT_ENCODING',  # We want content to be uncompressed so
+                             # we remove the Accept-Encoding from
+                             # original request
+    'HTTP_HOST',
+    'HTTP_REMOTE_USER',
+)
+
+
+# Default from HTTP RFC 2616
+#   See: http://www.w3.org/Protocols/rfc2616/rfc2616-sec3.html#sec3.7.1
+#: Variable that represent the default charset used
+DEFAULT_CHARSET = 'latin-1'
+
+#: List containing string constants that represents possible html content type
+HTML_CONTENT_TYPES = (
+    'text/html',
+    'application/xhtml+xml'
+)
+
+#: Variable used to represent a minimal content size required for response
+#: to be turned into stream
+MIN_STREAMING_LENGTH = 4 * 1024  # 4KB
+
+#: Regex used to find charset in a html content type
+_get_charset_re = re.compile(r';\s*charset=(?P<charset>[^\s;]+)', re.I)
+
+
+def is_html_content_type(content_type):
+    """Function used to verify if the parameter is a proper html content type
+
+    :param content_type: String variable that represent a content-type
+    :returns:  A boolean value stating if the content_type is a valid html
+               content type
+    """
+    for html_content_type in HTML_CONTENT_TYPES:
+        if content_type.startswith(html_content_type):
+            return True
+
+    return False
+
+
+def should_stream(proxy_response):
+    """Function to verify if the proxy_response must be converted into
+    a stream.This will be done by checking the proxy_response content-length
+    and verify if its length is bigger than one stipulated
+    by MIN_STREAMING_LENGTH.
+
+    :param proxy_response: An Instance of urllib3.response.HTTPResponse
+    :returns: A boolean stating if the proxy_response should
+              be treated as a stream
+    """
+    content_type = proxy_response.headers.get('Content-Type')
+
+    if is_html_content_type(content_type):
+        return False
+
+    try:
+        content_length = int(proxy_response.headers.get('Content-Length', 0))
+    except ValueError:
+        content_length = 0
+
+    if not content_length or content_length > MIN_STREAMING_LENGTH:
+        return True
+
+    return False
+
+
+def get_charset(content_type):
+    """Function used to retrieve the charset from a content-type.If there is no
+    charset in the content type then the charset defined on DEFAULT_CHARSET
+    will be returned
+
+    :param  content_type:   A string containing a Content-Type header
+    :returns:               A string containing the charset
+    """
+    if not content_type:
+        return DEFAULT_CHARSET
+
+    matched = _get_charset_re.search(content_type)
+    if matched:
+        # Extract the charset and strip its double quotes
+        return matched.group('charset').replace('"', '')
+    return DEFAULT_CHARSET
+
+
+def required_header(header):
+    """Function that verify if the header parameter is a essential header
+
+    :param header:  A string represented a header
+    :returns:       A boolean value that represent if the header is required
+    """
+    if header in IGNORE_HEADERS:
+        return False
+
+    if header.startswith('HTTP_') or header == 'CONTENT_TYPE':
+        return True
+
+    return False
+
+
+def set_response_headers(response, response_headers):
+    """Set response's header"""
+    for header, value in response_headers.items():
+        if is_hop_by_hop(header) or header.lower() == 'set-cookie':
+            continue
+
+        response[header.title()] = value
+
+    logger.debug('Response headers: %s', getattr(response, '_headers'))
+
+
+def normalize_request_headers(request):
+    """Function used to transform header, replacing 'HTTP\\_' to ''
+    and replace '_' to '-'
+
+    :param request:  A HttpRequest that will be transformed
+    :returns:        A dictionary with the normalized headers
+    """
+    norm_headers = {}
+    for header, value in request.META.items():
+        if required_header(header):
+            norm_header = header.replace('HTTP_', '').title().replace('_', '-')
+            norm_headers[norm_header] = value
+
+    return norm_headers
+
+
+def encode_items(items):
+    """Function that encode all elements in the list of items passed as
+    a parameter
+
+    :param items:  A list of tuple
+    :returns:      A list of tuple with all items encoded in 'utf-8'
+    """
+    encoded = []
+    for key, values in items:
+        for value in values:
+            encoded.append((key.encode('utf-8'), value.encode('utf-8')))
+    return encoded
+
+
+logger = logging.getLogger('revproxy.cookies')
+
+
+def cookie_from_string(cookie_string, strict_cookies=False):
+    """Parser for HTTP header set-cookie
+    The return from this function will be used as parameters for
+    django's response.set_cookie method. Because set_cookie doesn't
+    have parameter comment, this cookie attribute will be ignored.
+
+    :param  cookie_string: A string representing a valid cookie
+    :param  strict_cookies: Whether to only accept RFC-compliant cookies
+    :returns: A dictionary containing the cookie_string attributes
+    """
+
+    if strict_cookies:
+
+        cookies = SimpleCookie(COOKIE_PREFIX + cookie_string)
+        if not cookies.keys():
+            return None
+        cookie_name, = cookies.keys()
+        cookie_dict = {k: v for k, v in cookies[cookie_name].items()
+                       if v and k != 'comment'}
+        cookie_dict['key'] = cookie_name
+        cookie_dict['value'] = cookies[cookie_name].value
+        return cookie_dict
+    valid_attrs = ('path', 'domain', 'comment', 'expires',
+                   'max_age', 'httponly', 'secure')
+
+    cookie_dict = {}
+
+    cookie_parts = cookie_string.split(';')
+    try:
+        cookie_dict['key'], cookie_dict['value'] = \
+            cookie_parts[0].split('=', 1)
+        cookie_dict['value'] = cookie_dict['value'].replace('"', '')
+        # print('aaaaaaaaaaaaaaaaaaaaaaaaaaaa')
+        # print(cookie_parts[0].split('=', 1))
+    except ValueError:
+        logger.warning('Invalid cookie: `%s`', cookie_string)
+        return None
+
+    if cookie_dict['value'].startswith('='):
+        logger.warning('Invalid cookie: `%s`', cookie_string)
+        return None
+
+    for part in cookie_parts[1:]:
+        if '=' in part:
+            attr, value = part.split('=', 1)
+            value = value.strip()
+        else:
+            attr = part
+            value = ''
+
+        attr = attr.strip().lower()
+        if not attr:
+            continue
+
+        if attr in valid_attrs:
+            if attr in ('httponly', 'secure'):
+                cookie_dict[attr] = True
+            elif attr in 'comment':
+                # ignoring comment attr as explained in the
+                # function docstring
+                continue
+            else:
+                cookie_dict[attr] = value
+        else:
+            logger.warning('Unknown cookie attribute %s', attr)
+
+        return cookie_dict

passbook/app_gw/requirements.txt

@@ -1,2 +1,7 @@
 django-revproxy
 urllib3[secure]
+channels
+service_identity
+websocket-client
+daphne<2.3.0
+asgiref~=2.3

passbook/app_gw/settings.py

@@ -1,5 +1,5 @@
 """Application Security Gateway settings"""
-
-# INSTALLED_APPS = [
-#     'revproxy'
-# ]
+INSTALLED_APPS = [
+    'channels'
+]
+ASGI_APPLICATION = "passbook.app_gw.websocket.routing.application"

passbook/app_gw/websocket/consumer.py

@@ -0,0 +1,83 @@
+"""websocket proxy consumer"""
+import threading
+from logging import getLogger
+from ssl import CERT_NONE
+
+import websocket
+from channels.generic.websocket import WebsocketConsumer
+
+from passbook.app_gw.models import ApplicationGatewayProvider
+
+LOGGER = getLogger(__name__)
+
+class ProxyConsumer(WebsocketConsumer):
+    """Proxy websocket connection to upstream"""
+
+    _headers_dict = {}
+    _app_gw = None
+    _client = None
+    _thread = None
+
+    def _fix_headers(self, input_dict):
+        """Fix headers from bytestrings to normal strings"""
+        return {
+            key.decode('utf-8'): value.decode('utf-8')
+            for key, value in dict(input_dict).items()
+        }
+
+    def connect(self):
+        """Extract host header, lookup in database and proxy connection"""
+        self._headers_dict = self._fix_headers(dict(self.scope.get('headers')))
+        host = self._headers_dict.pop('host')
+        query_string = self.scope.get('query_string').decode('utf-8')
+        matches = ApplicationGatewayProvider.objects.filter(
+            server_name__contains=[host],
+            enabled=True)
+        if matches.exists():
+            self._app_gw = matches.first()
+            # TODO: Get upstream that starts with wss or
+            upstream = self._app_gw.upstream[0].replace('http', 'ws') + self.scope.get('path')
+            if query_string:
+                upstream += '?' + query_string
+            sslopt = {}
+            if not self._app_gw.upstream_ssl_verification:
+                sslopt = {"cert_reqs": CERT_NONE}
+            self._client = websocket.WebSocketApp(
+                url=upstream,
+                subprotocols=self.scope.get('subprotocols'),
+                header=self._headers_dict,
+                on_message=self._client_on_message_handler(),
+                on_error=self._client_on_error_handler(),
+                on_close=self._client_on_close_handler(),
+                on_open=self._client_on_open_handler())
+            LOGGER.debug("Accepting connection for %s", host)
+            self._thread = threading.Thread(target=lambda: self._client.run_forever(sslopt=sslopt))
+            self._thread.start()
+
+    def _client_on_open_handler(self):
+        return lambda ws: self.accept(self._client.sock.handshake_response.subprotocol)
+
+    def _client_on_message_handler(self):
+        # pylint: disable=unused-argument,invalid-name
+        def message_handler(ws, message):
+            if isinstance(message, str):
+                self.send(text_data=message)
+            else:
+                self.send(bytes_data=message)
+        return message_handler
+
+    def _client_on_error_handler(self):
+        return lambda ws, error: print(error)
+
+    def _client_on_close_handler(self):
+        return lambda ws: self.disconnect(0)
+
+    def disconnect(self, code):
+        self._client.close()
+
+    def receive(self, text_data=None, bytes_data=None):
+        if text_data:
+            opcode = websocket.ABNF.OPCODE_TEXT
+        if bytes_data:
+            opcode = websocket.ABNF.OPCODE_BINARY
+        self._client.send(text_data or bytes_data, opcode)

passbook/app_gw/websocket/routing.py

@@ -0,0 +1,17 @@
+"""app_gw websocket proxy"""
+from channels.auth import AuthMiddlewareStack
+from channels.routing import ProtocolTypeRouter, URLRouter
+from django.conf.urls import url
+
+from passbook.app_gw.websocket.consumer import ProxyConsumer
+
+websocket_urlpatterns = [
+    url(r'^(.*)$', ProxyConsumer),
+]
+
+application = ProtocolTypeRouter({
+    # (http->django views is added by default)
+    'websocket': AuthMiddlewareStack(
+        URLRouter(websocket_urlpatterns)
+    ),
+})

passbook/core/asgi.py

@@ -0,0 +1,13 @@
+"""
+ASGI entrypoint. Configures Django and then runs the application
+defined in the ASGI_APPLICATION setting.
+"""
+
+import os
+
+import django
+from channels.routing import get_default_application
+
+os.environ.setdefault("DJANGO_SETTINGS_MODULE", "passbook.core.settings")
+django.setup()
+application = get_default_application()