mirror of
https://github.com/Grasscutters/mitmproxy.git
synced 2024-11-27 02:24:18 +00:00
405 lines
12 KiB
Python
405 lines
12 KiB
Python
from __future__ import (absolute_import, print_function, division)
|
|
import string, urlparse, binascii
|
|
import sys
|
|
from . import odict, utils
|
|
|
|
|
|
class HttpError(Exception):
|
|
def __init__(self, code, message):
|
|
super(HttpError, self).__init__(message)
|
|
self.code = code
|
|
|
|
|
|
class HttpErrorConnClosed(HttpError):
|
|
pass
|
|
|
|
|
|
def _is_valid_port(port):
|
|
if not 0 <= port <= 65535:
|
|
return False
|
|
return True
|
|
|
|
|
|
def _is_valid_host(host):
|
|
try:
|
|
host.decode("idna")
|
|
except ValueError:
|
|
return False
|
|
if "\0" in host:
|
|
return None
|
|
return True
|
|
|
|
|
|
def parse_url(url):
|
|
"""
|
|
Returns a (scheme, host, port, path) tuple, or None on error.
|
|
|
|
Checks that:
|
|
port is an integer 0-65535
|
|
host is a valid IDNA-encoded hostname with no null-bytes
|
|
path is valid ASCII
|
|
"""
|
|
try:
|
|
scheme, netloc, path, params, query, fragment = urlparse.urlparse(url)
|
|
except ValueError:
|
|
return None
|
|
if not scheme:
|
|
return None
|
|
if '@' in netloc:
|
|
# FIXME: Consider what to do with the discarded credentials here Most
|
|
# probably we should extend the signature to return these as a separate
|
|
# value.
|
|
_, netloc = string.rsplit(netloc, '@', maxsplit=1)
|
|
if ':' in netloc:
|
|
host, port = string.rsplit(netloc, ':', maxsplit=1)
|
|
try:
|
|
port = int(port)
|
|
except ValueError:
|
|
return None
|
|
else:
|
|
host = netloc
|
|
if scheme == "https":
|
|
port = 443
|
|
else:
|
|
port = 80
|
|
path = urlparse.urlunparse(('', '', path, params, query, fragment))
|
|
if not path.startswith("/"):
|
|
path = "/" + path
|
|
if not _is_valid_host(host):
|
|
return None
|
|
if not utils.isascii(path):
|
|
return None
|
|
if not _is_valid_port(port):
|
|
return None
|
|
return scheme, host, port, path
|
|
|
|
|
|
def read_headers(fp):
|
|
"""
|
|
Read a set of headers from a file pointer. Stop once a blank line is
|
|
reached. Return a ODictCaseless object, or None if headers are invalid.
|
|
"""
|
|
ret = []
|
|
name = ''
|
|
while 1:
|
|
line = fp.readline()
|
|
if not line or line == '\r\n' or line == '\n':
|
|
break
|
|
if line[0] in ' \t':
|
|
if not ret:
|
|
return None
|
|
# continued header
|
|
ret[-1][1] = ret[-1][1] + '\r\n ' + line.strip()
|
|
else:
|
|
i = line.find(':')
|
|
# We're being liberal in what we accept, here.
|
|
if i > 0:
|
|
name = line[:i]
|
|
value = line[i + 1:].strip()
|
|
ret.append([name, value])
|
|
else:
|
|
return None
|
|
return odict.ODictCaseless(ret)
|
|
|
|
|
|
def read_chunked(fp, limit, is_request):
|
|
"""
|
|
Read a chunked HTTP body.
|
|
|
|
May raise HttpError.
|
|
"""
|
|
# FIXME: Should check if chunked is the final encoding in the headers
|
|
# http://tools.ietf.org/html/draft-ietf-httpbis-p1-messaging-16#section-3.3 3.3 2.
|
|
total = 0
|
|
code = 400 if is_request else 502
|
|
while True:
|
|
line = fp.readline(128)
|
|
if line == "":
|
|
raise HttpErrorConnClosed(code, "Connection closed prematurely")
|
|
if line != '\r\n' and line != '\n':
|
|
try:
|
|
length = int(line, 16)
|
|
except ValueError:
|
|
raise HttpError(code, "Invalid chunked encoding length: %s" % line)
|
|
total += length
|
|
if limit is not None and total > limit:
|
|
msg = "HTTP Body too large." \
|
|
" Limit is %s, chunked content length was at least %s" % (limit, total)
|
|
raise HttpError(code, msg)
|
|
chunk = fp.read(length)
|
|
suffix = fp.readline(5)
|
|
if suffix != '\r\n':
|
|
raise HttpError(code, "Malformed chunked body")
|
|
yield line, chunk, '\r\n'
|
|
if length == 0:
|
|
return
|
|
|
|
|
|
def get_header_tokens(headers, key):
|
|
"""
|
|
Retrieve all tokens for a header key. A number of different headers
|
|
follow a pattern where each header line can containe comma-separated
|
|
tokens, and headers can be set multiple times.
|
|
"""
|
|
toks = []
|
|
for i in headers[key]:
|
|
for j in i.split(","):
|
|
toks.append(j.strip())
|
|
return toks
|
|
|
|
|
|
def has_chunked_encoding(headers):
|
|
return "chunked" in [i.lower() for i in get_header_tokens(headers, "transfer-encoding")]
|
|
|
|
|
|
def parse_http_protocol(s):
|
|
"""
|
|
Parse an HTTP protocol declaration. Returns a (major, minor) tuple, or
|
|
None.
|
|
"""
|
|
if not s.startswith("HTTP/"):
|
|
return None
|
|
_, version = s.split('/', 1)
|
|
if "." not in version:
|
|
return None
|
|
major, minor = version.split('.', 1)
|
|
try:
|
|
major = int(major)
|
|
minor = int(minor)
|
|
except ValueError:
|
|
return None
|
|
return major, minor
|
|
|
|
|
|
def parse_http_basic_auth(s):
|
|
words = s.split()
|
|
if len(words) != 2:
|
|
return None
|
|
scheme = words[0]
|
|
try:
|
|
user = binascii.a2b_base64(words[1])
|
|
except binascii.Error:
|
|
return None
|
|
parts = user.split(':')
|
|
if len(parts) != 2:
|
|
return None
|
|
return scheme, parts[0], parts[1]
|
|
|
|
|
|
def assemble_http_basic_auth(scheme, username, password):
|
|
v = binascii.b2a_base64(username + ":" + password)
|
|
return scheme + " " + v
|
|
|
|
|
|
def parse_init(line):
|
|
try:
|
|
method, url, protocol = string.split(line)
|
|
except ValueError:
|
|
return None
|
|
httpversion = parse_http_protocol(protocol)
|
|
if not httpversion:
|
|
return None
|
|
if not utils.isascii(method):
|
|
return None
|
|
return method, url, httpversion
|
|
|
|
|
|
def parse_init_connect(line):
|
|
"""
|
|
Returns (host, port, httpversion) if line is a valid CONNECT line.
|
|
http://tools.ietf.org/html/draft-luotonen-web-proxy-tunneling-01 section 3.1
|
|
"""
|
|
v = parse_init(line)
|
|
if not v:
|
|
return None
|
|
method, url, httpversion = v
|
|
|
|
if method.upper() != 'CONNECT':
|
|
return None
|
|
try:
|
|
host, port = url.split(":")
|
|
except ValueError:
|
|
return None
|
|
try:
|
|
port = int(port)
|
|
except ValueError:
|
|
return None
|
|
if not _is_valid_port(port):
|
|
return None
|
|
if not _is_valid_host(host):
|
|
return None
|
|
return host, port, httpversion
|
|
|
|
|
|
def parse_init_proxy(line):
|
|
v = parse_init(line)
|
|
if not v:
|
|
return None
|
|
method, url, httpversion = v
|
|
|
|
parts = parse_url(url)
|
|
if not parts:
|
|
return None
|
|
scheme, host, port, path = parts
|
|
return method, scheme, host, port, path, httpversion
|
|
|
|
|
|
def parse_init_http(line):
|
|
"""
|
|
Returns (method, url, httpversion)
|
|
"""
|
|
v = parse_init(line)
|
|
if not v:
|
|
return None
|
|
method, url, httpversion = v
|
|
if not utils.isascii(url):
|
|
return None
|
|
if not (url.startswith("/") or url == "*"):
|
|
return None
|
|
return method, url, httpversion
|
|
|
|
|
|
def connection_close(httpversion, headers):
|
|
"""
|
|
Checks the message to see if the client connection should be closed according to RFC 2616 Section 8.1
|
|
Note that a connection should be closed as well if the response has been read until end of the stream.
|
|
"""
|
|
# At first, check if we have an explicit Connection header.
|
|
if "connection" in headers:
|
|
toks = get_header_tokens(headers, "connection")
|
|
if "close" in toks:
|
|
return True
|
|
elif "keep-alive" in toks:
|
|
return False
|
|
# If we don't have a Connection header, HTTP 1.1 connections are assumed to be persistent
|
|
if httpversion == (1, 1):
|
|
return False
|
|
return True
|
|
|
|
|
|
def parse_response_line(line):
|
|
parts = line.strip().split(" ", 2)
|
|
if len(parts) == 2: # handle missing message gracefully
|
|
parts.append("")
|
|
if len(parts) != 3:
|
|
return None
|
|
proto, code, msg = parts
|
|
try:
|
|
code = int(code)
|
|
except ValueError:
|
|
return None
|
|
return (proto, code, msg)
|
|
|
|
|
|
def read_response(rfile, request_method, body_size_limit, include_body=True):
|
|
"""
|
|
Return an (httpversion, code, msg, headers, content) tuple.
|
|
|
|
By default, both response header and body are read.
|
|
If include_body=False is specified, content may be one of the following:
|
|
- None, if the response is technically allowed to have a response body
|
|
- "", if the response must not have a response body (e.g. it's a response to a HEAD request)
|
|
"""
|
|
line = rfile.readline()
|
|
if line == "\r\n" or line == "\n": # Possible leftover from previous message
|
|
line = rfile.readline()
|
|
if not line:
|
|
raise HttpErrorConnClosed(502, "Server disconnect.")
|
|
parts = parse_response_line(line)
|
|
if not parts:
|
|
raise HttpError(502, "Invalid server response: %s" % repr(line))
|
|
proto, code, msg = parts
|
|
httpversion = parse_http_protocol(proto)
|
|
if httpversion is None:
|
|
raise HttpError(502, "Invalid HTTP version in line: %s" % repr(proto))
|
|
headers = read_headers(rfile)
|
|
if headers is None:
|
|
raise HttpError(502, "Invalid headers.")
|
|
|
|
if include_body:
|
|
content = read_http_body(rfile, headers, body_size_limit, request_method, code, False)
|
|
else:
|
|
content = None # if include_body==False then a None content means the body should be read separately
|
|
return httpversion, code, msg, headers, content
|
|
|
|
|
|
def read_http_body(*args, **kwargs):
|
|
return "".join(content for _, content, _ in read_http_body_chunked(*args, **kwargs))
|
|
|
|
|
|
def read_http_body_chunked(rfile, headers, limit, request_method, response_code, is_request, max_chunk_size=None):
|
|
"""
|
|
Read an HTTP message body:
|
|
|
|
rfile: A file descriptor to read from
|
|
headers: An ODictCaseless object
|
|
limit: Size limit.
|
|
is_request: True if the body to read belongs to a request, False otherwise
|
|
"""
|
|
if max_chunk_size is None:
|
|
max_chunk_size = limit or sys.maxint
|
|
|
|
expected_size = expected_http_body_size(headers, is_request, request_method, response_code)
|
|
|
|
if expected_size is None:
|
|
if has_chunked_encoding(headers):
|
|
# Python 3: yield from
|
|
for x in read_chunked(rfile, limit, is_request):
|
|
yield x
|
|
else: # pragma: nocover
|
|
raise HttpError(400 if is_request else 502, "Content-Length unknown but no chunked encoding")
|
|
elif expected_size >= 0:
|
|
if limit is not None and expected_size > limit:
|
|
raise HttpError(400 if is_request else 509,
|
|
"HTTP Body too large. Limit is %s, content-length was %s" % (limit, expected_size))
|
|
bytes_left = expected_size
|
|
while bytes_left:
|
|
chunk_size = min(bytes_left, max_chunk_size)
|
|
yield "", rfile.read(chunk_size), ""
|
|
bytes_left -= chunk_size
|
|
else:
|
|
bytes_left = limit or -1
|
|
while bytes_left:
|
|
chunk_size = min(bytes_left, max_chunk_size)
|
|
content = rfile.read(chunk_size)
|
|
if not content:
|
|
return
|
|
yield "", content, ""
|
|
bytes_left -= chunk_size
|
|
not_done = rfile.read(1)
|
|
if not_done:
|
|
raise HttpError(400 if is_request else 509, "HTTP Body too large. Limit is %s," % limit)
|
|
|
|
|
|
def expected_http_body_size(headers, is_request, request_method, response_code):
|
|
"""
|
|
Returns the expected body length:
|
|
- a positive integer, if the size is known in advance
|
|
- None, if the size in unknown in advance (chunked encoding)
|
|
- -1, if all data should be read until end of stream.
|
|
"""
|
|
|
|
# Determine response size according to http://tools.ietf.org/html/rfc7230#section-3.3
|
|
if request_method:
|
|
request_method = request_method.upper()
|
|
|
|
if (not is_request and (
|
|
request_method == "HEAD" or
|
|
(request_method == "CONNECT" and response_code == 200) or
|
|
response_code in [204, 304] or
|
|
100 <= response_code <= 199)):
|
|
return 0
|
|
if has_chunked_encoding(headers):
|
|
return None
|
|
if "content-length" in headers:
|
|
try:
|
|
size = int(headers["content-length"][0])
|
|
if size < 0:
|
|
raise ValueError()
|
|
return size
|
|
except ValueError:
|
|
raise HttpError(400 if is_request else 502, "Invalid content-length header: %s" % headers["content-length"])
|
|
if is_request:
|
|
return 0
|
|
return -1
|