mitmproxy/netlib/http/models.py

538 lines
17 KiB
Python
Raw Normal View History

2015-09-15 17:12:15 +00:00
from __future__ import absolute_import, print_function, division
2015-09-05 16:15:47 +00:00
import copy
2015-07-14 21:02:14 +00:00
2015-09-15 17:12:15 +00:00
from ..odict import ODict
from .. import utils, encoding
from ..utils import always_bytes, always_byte_args
from . import cookies
2015-08-01 08:39:14 +00:00
2015-09-15 17:12:15 +00:00
import six
from six.moves import urllib
try:
from collections import MutableMapping
except ImportError:
from collections.abc import MutableMapping
2015-09-16 16:43:24 +00:00
# TODO: Move somewhere else?
ALPN_PROTO_HTTP1 = b'http/1.1'
ALPN_PROTO_H2 = b'h2'
2015-09-15 17:12:15 +00:00
HDR_FORM_URLENCODED = b"application/x-www-form-urlencoded"
HDR_FORM_MULTIPART = b"multipart/form-data"
2015-07-14 21:02:14 +00:00
2015-07-29 09:27:43 +00:00
CONTENT_MISSING = 0
2015-09-15 17:12:15 +00:00
class Headers(MutableMapping, object):
2015-09-05 11:26:36 +00:00
"""
Header class which allows both convenient access to individual headers as well as
direct access to the underlying raw data. Provides a full dictionary interface.
Example:
.. code-block:: python
# Create header from a list of (header_name, header_value) tuples
>>> h = Headers([
["Host","example.com"],
["Accept","text/html"],
["accept","application/xml"]
])
# Headers mostly behave like a normal dict.
>>> h["Host"]
"example.com"
# HTTP Headers are case insensitive
>>> h["host"]
"example.com"
# Multiple headers are folded into a single header as per RFC7230
>>> h["Accept"]
"text/html, application/xml"
# Setting a header removes all existing headers with the same name.
>>> h["Accept"] = "application/text"
>>> h["Accept"]
"application/text"
# str(h) returns a HTTP1 header block.
>>> print(h)
Host: example.com
Accept: application/text
2015-09-05 13:27:48 +00:00
# For full control, the raw header fields can be accessed
>>> h.fields
2015-09-05 11:26:36 +00:00
# Headers can also be crated from keyword arguments
>>> h = Headers(host="example.com", content_type="application/xml")
Caveats:
For use with the "Set-Cookie" header, see :py:meth:`get_all`.
"""
2015-09-15 17:12:15 +00:00
@always_byte_args("ascii")
2015-09-05 13:27:48 +00:00
def __init__(self, fields=None, **headers):
2015-09-05 11:26:36 +00:00
"""
2015-09-05 13:27:48 +00:00
Args:
2015-09-15 17:12:15 +00:00
fields: (optional) list of ``(name, value)`` header tuples,
e.g. ``[("Host","example.com")]``. All names and values must be bytes.
2015-09-05 13:27:48 +00:00
**headers: Additional headers to set. Will overwrite existing values from `fields`.
For convenience, underscores in header names will be transformed to dashes -
this behaviour does not extend to other methods.
If ``**headers`` contains multiple keys that have equal ``.lower()`` s,
the behavior is undefined.
2015-09-05 11:26:36 +00:00
"""
2015-09-05 13:27:48 +00:00
self.fields = fields or []
2015-09-05 11:26:36 +00:00
# content_type -> content-type
2015-09-05 13:27:48 +00:00
headers = {
2015-09-15 17:12:15 +00:00
name.encode("ascii").replace(b"_", b"-"): value
for name, value in six.iteritems(headers)
2015-09-05 16:15:47 +00:00
}
2015-09-05 11:26:36 +00:00
self.update(headers)
2015-07-29 09:27:43 +00:00
2015-09-15 17:12:15 +00:00
def __bytes__(self):
return b"\r\n".join(b": ".join(field) for field in self.fields) + b"\r\n"
if six.PY2:
__str__ = __bytes__
2015-09-05 11:26:36 +00:00
2015-09-15 17:12:15 +00:00
@always_byte_args("ascii")
2015-09-05 13:27:48 +00:00
def __getitem__(self, name):
values = self.get_all(name)
2015-09-05 11:26:36 +00:00
if not values:
2015-09-05 13:27:48 +00:00
raise KeyError(name)
2015-09-15 17:12:15 +00:00
return b", ".join(values)
2015-09-05 11:26:36 +00:00
2015-09-15 17:12:15 +00:00
@always_byte_args("ascii")
2015-09-05 13:27:48 +00:00
def __setitem__(self, name, value):
idx = self._index(name)
2015-09-05 11:26:36 +00:00
# To please the human eye, we insert at the same position the first existing header occured.
if idx is not None:
2015-09-05 13:27:48 +00:00
del self[name]
self.fields.insert(idx, [name, value])
2015-09-05 11:26:36 +00:00
else:
2015-09-05 13:27:48 +00:00
self.fields.append([name, value])
2015-09-15 17:12:15 +00:00
@always_byte_args("ascii")
2015-09-05 13:27:48 +00:00
def __delitem__(self, name):
if name not in self:
raise KeyError(name)
name = name.lower()
self.fields = [
field for field in self.fields
if name != field[0].lower()
2015-09-05 16:15:47 +00:00
]
2015-09-05 13:27:48 +00:00
2015-09-15 17:12:15 +00:00
def __iter__(self):
seen = set()
for name, _ in self.fields:
name_lower = name.lower()
if name_lower not in seen:
seen.add(name_lower)
yield name
def __len__(self):
return len(set(name.lower() for name, _ in self.fields))
#__hash__ = object.__hash__
2015-09-05 13:27:48 +00:00
def _index(self, name):
name = name.lower()
for i, field in enumerate(self.fields):
if field[0].lower() == name:
2015-09-05 11:26:36 +00:00
return i
return None
def __eq__(self, other):
2015-09-05 13:27:48 +00:00
if isinstance(other, Headers):
return self.fields == other.fields
return False
2015-09-05 11:26:36 +00:00
def __ne__(self, other):
return not self.__eq__(other)
2015-09-15 17:12:15 +00:00
@always_byte_args("ascii")
def get_all(self, name):
2015-09-05 11:26:36 +00:00
"""
Like :py:meth:`get`, but does not fold multiple headers into a single one.
This is useful for Set-Cookie headers, which do not support folding.
See also: https://tools.ietf.org/html/rfc7230#section-3.2.2
"""
2015-09-15 17:12:15 +00:00
name_lower = name.lower()
values = [value for n, value in self.fields if n.lower() == name_lower]
return values
2015-09-05 11:26:36 +00:00
2015-09-05 13:27:48 +00:00
def set_all(self, name, values):
2015-09-05 11:26:36 +00:00
"""
Explicitly set multiple headers for the given key.
See: :py:meth:`get_all`
"""
2015-09-15 17:12:15 +00:00
name = always_bytes(name, "ascii")
values = (always_bytes(value, "ascii") for value in values)
2015-09-05 13:27:48 +00:00
if name in self:
del self[name]
self.fields.extend(
[name, value] for value in values
2015-09-05 11:26:36 +00:00
)
2015-09-05 16:15:47 +00:00
def copy(self):
return Headers(copy.copy(self.fields))
2015-09-05 13:27:48 +00:00
# Implement the StateObject protocol from mitmproxy
def get_state(self, short=False):
return tuple(tuple(field) for field in self.fields)
def load_state(self, state):
self.fields = [list(field) for field in state]
@classmethod
def from_state(cls, state):
return cls([list(field) for field in state])
2015-09-05 11:26:36 +00:00
2015-09-17 13:16:12 +00:00
class Message(object):
def __init__(self, http_version, headers, body, timestamp_start, timestamp_end):
self.http_version = http_version
if not headers:
headers = Headers()
assert isinstance(headers, Headers)
self.headers = headers
self._body = body
self.timestamp_start = timestamp_start
self.timestamp_end = timestamp_end
@property
def body(self):
return self._body
@body.setter
def body(self, body):
self._body = body
if isinstance(body, bytes):
self.headers[b"Content-Length"] = str(len(body)).encode()
content = body
def __eq__(self, other):
if isinstance(other, Message):
return self.__dict__ == other.__dict__
return False
class Request(Message):
2015-08-05 19:32:53 +00:00
# This list is adopted legacy code.
# We probably don't need to strip off keep-alive.
_headers_to_strip_off = [
2015-09-17 13:16:12 +00:00
b'Proxy-Connection',
b'Keep-Alive',
b'Connection',
b'Transfer-Encoding',
b'Upgrade',
2015-08-05 19:32:53 +00:00
]
def __init__(
2015-09-05 16:15:47 +00:00
self,
form_in,
method,
scheme,
host,
port,
path,
2015-09-17 13:16:12 +00:00
http_version,
2015-09-05 16:15:47 +00:00
headers=None,
body=None,
timestamp_start=None,
timestamp_end=None,
form_out=None
):
2015-09-17 13:16:12 +00:00
super(Request, self).__init__(http_version, headers, body, timestamp_start, timestamp_end)
2015-07-27 07:36:50 +00:00
self.form_in = form_in
self.method = method
self.scheme = scheme
self.host = host
self.port = port
self.path = path
2015-08-15 18:30:22 +00:00
self.form_out = form_out or form_in
def __repr__(self):
2015-09-15 17:12:15 +00:00
if self.host and self.port:
hostport = "{}:{}".format(self.host, self.port)
2015-08-01 08:39:14 +00:00
else:
2015-09-15 17:12:15 +00:00
hostport = ""
path = self.path or ""
return "HTTPRequest({} {}{})".format(
self.method, hostport, path
)
2015-08-01 08:39:14 +00:00
def anticache(self):
"""
Modifies this request to remove headers that might produce a cached
response. That is, we remove ETags and If-Modified-Since headers.
"""
delheaders = [
2015-09-17 13:16:12 +00:00
b"if-modified-since",
b"if-none-match",
2015-08-01 08:39:14 +00:00
]
for i in delheaders:
2015-09-05 16:15:47 +00:00
self.headers.pop(i, None)
2015-08-01 08:39:14 +00:00
def anticomp(self):
"""
Modifies this request to remove headers that will compress the
resource's data.
"""
2015-09-17 13:16:12 +00:00
self.headers[b"accept-encoding"] = b"identity"
2015-08-01 08:39:14 +00:00
def constrain_encoding(self):
"""
Limits the permissible Accept-Encoding values, based on what we can
decode appropriately.
"""
2015-09-17 13:16:12 +00:00
accept_encoding = self.headers.get(b"accept-encoding")
2015-09-05 16:15:47 +00:00
if accept_encoding:
2015-09-17 13:16:12 +00:00
self.headers[b"accept-encoding"] = (
2015-08-01 08:39:14 +00:00
', '.join(
2015-09-05 16:15:47 +00:00
e
for e in encoding.ENCODINGS
if e in accept_encoding
)
)
2015-08-01 08:39:14 +00:00
def update_host_header(self):
"""
Update the host header to reflect the current target.
"""
2015-09-05 16:15:47 +00:00
self.headers["Host"] = self.host
2015-08-01 08:39:14 +00:00
def get_form(self):
"""
Retrieves the URL-encoded or multipart form data, returning an ODict object.
Returns an empty ODict if there is no data or the content-type
indicates non-form data.
"""
if self.body:
2015-09-05 16:15:47 +00:00
if HDR_FORM_URLENCODED in self.headers.get("content-type","").lower():
2015-08-01 08:39:14 +00:00
return self.get_form_urlencoded()
2015-09-05 16:15:47 +00:00
elif HDR_FORM_MULTIPART in self.headers.get("content-type","").lower():
2015-08-01 08:39:14 +00:00
return self.get_form_multipart()
2015-09-15 17:12:15 +00:00
return ODict([])
2015-08-01 08:39:14 +00:00
def get_form_urlencoded(self):
"""
Retrieves the URL-encoded form data, returning an ODict object.
Returns an empty ODict if there is no data or the content-type
indicates non-form data.
"""
2015-09-05 16:15:47 +00:00
if self.body and HDR_FORM_URLENCODED in self.headers.get("content-type","").lower():
2015-09-15 17:12:15 +00:00
return ODict(utils.urldecode(self.body))
return ODict([])
2015-08-01 08:39:14 +00:00
def get_form_multipart(self):
2015-09-05 16:15:47 +00:00
if self.body and HDR_FORM_MULTIPART in self.headers.get("content-type","").lower():
2015-09-15 17:12:15 +00:00
return ODict(
2015-08-01 08:39:14 +00:00
utils.multipartdecode(
self.headers,
self.body))
2015-09-15 17:12:15 +00:00
return ODict([])
2015-08-01 08:39:14 +00:00
def set_form_urlencoded(self, odict):
"""
Sets the body to the URL-encoded form data, and adds the
appropriate content-type header. Note that this will destory the
existing body if there is one.
"""
# FIXME: If there's an existing content-type header indicating a
# url-encoded form, leave it alone.
2015-09-17 13:16:12 +00:00
self.headers[b"Content-Type"] = HDR_FORM_URLENCODED
2015-08-01 08:39:14 +00:00
self.body = utils.urlencode(odict.lst)
def get_path_components(self):
"""
Returns the path components of the URL as a list of strings.
Components are unquoted.
"""
2015-09-15 17:12:15 +00:00
_, _, path, _, _, _ = urllib.parse.urlparse(self.url)
return [urllib.parse.unquote(i) for i in path.split(b"/") if i]
2015-08-01 08:39:14 +00:00
def set_path_components(self, lst):
"""
Takes a list of strings, and sets the path component of the URL.
Components are quoted.
"""
2015-09-15 17:12:15 +00:00
lst = [urllib.parse.quote(i, safe="") for i in lst]
path = b"/" + b"/".join(lst)
scheme, netloc, _, params, query, fragment = urllib.parse.urlparse(self.url)
self.url = urllib.parse.urlunparse(
2015-08-01 08:39:14 +00:00
[scheme, netloc, path, params, query, fragment]
)
def get_query(self):
"""
Gets the request query string. Returns an ODict object.
"""
2015-09-15 17:12:15 +00:00
_, _, _, _, query, _ = urllib.parse.urlparse(self.url)
2015-08-01 08:39:14 +00:00
if query:
2015-09-15 17:12:15 +00:00
return ODict(utils.urldecode(query))
return ODict([])
2015-08-01 08:39:14 +00:00
def set_query(self, odict):
"""
Takes an ODict object, and sets the request query string.
"""
2015-09-15 17:12:15 +00:00
scheme, netloc, path, params, _, fragment = urllib.parse.urlparse(self.url)
2015-08-01 08:39:14 +00:00
query = utils.urlencode(odict.lst)
2015-09-15 17:12:15 +00:00
self.url = urllib.parse.urlunparse(
2015-08-01 08:39:14 +00:00
[scheme, netloc, path, params, query, fragment]
)
def pretty_host(self, hostheader):
"""
Heuristic to get the host of the request.
Note that pretty_host() does not always return the TCP destination
of the request, e.g. if an upstream proxy is in place
If hostheader is set to True, the Host: header will be used as
additional (and preferred) data source. This is handy in
transparent mode, where only the IO of the destination is known,
but not the resolved name. This is disabled by default, as an
attacker may spoof the host header to confuse an analyst.
"""
2015-09-15 17:12:15 +00:00
if hostheader and b"Host" in self.headers:
2015-08-01 08:39:14 +00:00
try:
2015-09-15 17:12:15 +00:00
return self.headers[b"Host"].decode("idna")
2015-08-01 08:39:14 +00:00
except ValueError:
2015-09-15 17:12:15 +00:00
pass
if self.host:
return self.host.decode("idna")
2015-08-01 08:39:14 +00:00
def pretty_url(self, hostheader):
if self.form_out == "authority": # upstream proxy mode
return "%s:%s" % (self.pretty_host(hostheader), self.port)
return utils.unparse_url(self.scheme,
self.pretty_host(hostheader),
self.port,
self.path).encode('ascii')
def get_cookies(self):
"""
Returns a possibly empty netlib.odict.ODict object.
"""
2015-09-15 17:12:15 +00:00
ret = ODict()
2015-09-05 16:15:47 +00:00
for i in self.headers.get_all("cookie"):
2015-08-01 08:39:14 +00:00
ret.extend(cookies.parse_cookie_header(i))
return ret
def set_cookies(self, odict):
"""
Takes an netlib.odict.ODict object. Over-writes any existing Cookie
headers.
"""
v = cookies.format_cookie_header(odict)
2015-09-05 16:15:47 +00:00
self.headers["Cookie"] = v
2015-08-01 08:39:14 +00:00
@property
def url(self):
"""
Returns a URL string, constructed from the Request's URL components.
"""
return utils.unparse_url(
self.scheme,
self.host,
self.port,
self.path
).encode('ascii')
@url.setter
def url(self, url):
"""
Parses a URL specification, and updates the Request's information
accordingly.
2015-09-15 17:12:15 +00:00
Raises:
ValueError if the URL was invalid
2015-08-01 08:39:14 +00:00
"""
2015-09-15 17:12:15 +00:00
# TODO: Should handle incoming unicode here.
2015-08-01 08:39:14 +00:00
parts = utils.parse_url(url)
if not parts:
raise ValueError("Invalid URL: %s" % url)
self.scheme, self.host, self.port, self.path = parts
2015-07-27 07:36:50 +00:00
2015-09-17 13:16:12 +00:00
class Response(Message):
2015-08-05 19:32:53 +00:00
_headers_to_strip_off = [
2015-09-17 13:16:12 +00:00
b'Proxy-Connection',
b'Alternate-Protocol',
b'Alt-Svc',
2015-08-05 19:32:53 +00:00
]
2015-07-14 21:02:14 +00:00
def __init__(
2015-09-05 16:15:47 +00:00
self,
2015-09-17 13:16:12 +00:00
http_version,
2015-09-05 16:15:47 +00:00
status_code,
msg=None,
headers=None,
body=None,
timestamp_start=None,
timestamp_end=None,
2015-07-14 21:02:14 +00:00
):
2015-09-17 13:16:12 +00:00
super(Response, self).__init__(http_version, headers, body, timestamp_start, timestamp_end)
2015-07-14 21:02:14 +00:00
self.status_code = status_code
self.msg = msg
def __repr__(self):
2015-08-01 08:39:14 +00:00
# return "Response(%s - %s)" % (self.status_code, self.msg)
if self.body:
size = utils.pretty_size(len(self.body))
else:
size = "content missing"
# TODO: Remove "(unknown content type, content missing)" edge-case
return "<Response: {status_code} {msg} ({contenttype}, {size})>".format(
2015-08-01 08:39:14 +00:00
status_code=self.status_code,
msg=self.msg,
2015-09-05 16:15:47 +00:00
contenttype=self.headers.get("content-type", "unknown content type"),
2015-08-10 18:44:36 +00:00
size=size)
2015-08-01 08:39:14 +00:00
def get_cookies(self):
"""
Get the contents of all Set-Cookie headers.
Returns a possibly empty ODict, where keys are cookie name strings,
and values are [value, attr] lists. Value is a string, and attr is
an ODictCaseless containing cookie attributes. Within attrs, unary
attributes (e.g. HTTPOnly) are indicated by a Null value.
"""
ret = []
2015-09-17 13:16:12 +00:00
for header in self.headers.get_all(b"set-cookie"):
2015-08-01 08:39:14 +00:00
v = cookies.parse_set_cookie_header(header)
if v:
name, value, attrs = v
ret.append([name, [value, attrs]])
2015-09-15 17:12:15 +00:00
return ODict(ret)
2015-08-01 08:39:14 +00:00
def set_cookies(self, odict):
"""
Set the Set-Cookie headers on this response, over-writing existing
headers.
Accepts an ODict of the same format as that returned by get_cookies.
"""
values = []
for i in odict.lst:
values.append(
cookies.format_set_cookie_header(
i[0],
i[1][0],
i[1][1]
)
)
2015-09-17 13:16:12 +00:00
self.headers.set_all(b"Set-Cookie", values)