root / lib / http / __init__.py @ a8e01e9f
History | View | Annotate | Download (24.2 kB)
1 |
#
|
---|---|
2 |
#
|
3 |
|
4 |
# Copyright (C) 2007, 2008 Google Inc.
|
5 |
#
|
6 |
# This program is free software; you can redistribute it and/or modify
|
7 |
# it under the terms of the GNU General Public License as published by
|
8 |
# the Free Software Foundation; either version 2 of the License, or
|
9 |
# (at your option) any later version.
|
10 |
#
|
11 |
# This program is distributed in the hope that it will be useful, but
|
12 |
# WITHOUT ANY WARRANTY; without even the implied warranty of
|
13 |
# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
|
14 |
# General Public License for more details.
|
15 |
#
|
16 |
# You should have received a copy of the GNU General Public License
|
17 |
# along with this program; if not, write to the Free Software
|
18 |
# Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA
|
19 |
# 02110-1301, USA.
|
20 |
|
21 |
"""HTTP module.
|
22 |
|
23 |
"""
|
24 |
|
25 |
import logging |
26 |
import mimetools |
27 |
import OpenSSL |
28 |
import select |
29 |
import socket |
30 |
import errno |
31 |
|
32 |
from cStringIO import StringIO |
33 |
|
34 |
from ganeti import constants |
35 |
from ganeti import serializer |
36 |
from ganeti import utils |
37 |
|
38 |
|
39 |
HTTP_GANETI_VERSION = "Ganeti %s" % constants.RELEASE_VERSION
|
40 |
|
41 |
HTTP_OK = 200
|
42 |
HTTP_NO_CONTENT = 204
|
43 |
HTTP_NOT_MODIFIED = 304
|
44 |
|
45 |
HTTP_0_9 = "HTTP/0.9"
|
46 |
HTTP_1_0 = "HTTP/1.0"
|
47 |
HTTP_1_1 = "HTTP/1.1"
|
48 |
|
49 |
HTTP_GET = "GET"
|
50 |
HTTP_HEAD = "HEAD"
|
51 |
HTTP_POST = "POST"
|
52 |
HTTP_PUT = "PUT"
|
53 |
HTTP_DELETE = "DELETE"
|
54 |
|
55 |
HTTP_ETAG = "ETag"
|
56 |
HTTP_HOST = "Host"
|
57 |
HTTP_SERVER = "Server"
|
58 |
HTTP_DATE = "Date"
|
59 |
HTTP_USER_AGENT = "User-Agent"
|
60 |
HTTP_CONTENT_TYPE = "Content-Type"
|
61 |
HTTP_CONTENT_LENGTH = "Content-Length"
|
62 |
HTTP_CONNECTION = "Connection"
|
63 |
HTTP_KEEP_ALIVE = "Keep-Alive"
|
64 |
|
65 |
_SSL_UNEXPECTED_EOF = "Unexpected EOF"
|
66 |
|
67 |
# Socket operations
|
68 |
(SOCKOP_SEND, |
69 |
SOCKOP_RECV, |
70 |
SOCKOP_SHUTDOWN, |
71 |
SOCKOP_HANDSHAKE) = range(4) |
72 |
|
73 |
# send/receive quantum
|
74 |
SOCK_BUF_SIZE = 32768
|
75 |
|
76 |
|
77 |
class HttpError(Exception): |
78 |
"""Internal exception for HTTP errors.
|
79 |
|
80 |
This should only be used for internal error reporting.
|
81 |
|
82 |
"""
|
83 |
|
84 |
|
85 |
class HttpConnectionClosed(Exception): |
86 |
"""Internal exception for a closed connection.
|
87 |
|
88 |
This should only be used for internal error reporting. Only use
|
89 |
it if there's no other way to report this condition.
|
90 |
|
91 |
"""
|
92 |
|
93 |
|
94 |
class HttpSessionHandshakeUnexpectedEOF(HttpError): |
95 |
"""Internal exception for errors during SSL handshake.
|
96 |
|
97 |
This should only be used for internal error reporting.
|
98 |
|
99 |
"""
|
100 |
|
101 |
|
102 |
class HttpSocketTimeout(Exception): |
103 |
"""Internal exception for socket timeouts.
|
104 |
|
105 |
This should only be used for internal error reporting.
|
106 |
|
107 |
"""
|
108 |
|
109 |
|
110 |
class HttpException(Exception): |
111 |
code = None
|
112 |
message = None
|
113 |
|
114 |
def __init__(self, message=None, headers=None): |
115 |
Exception.__init__(self) |
116 |
self.message = message
|
117 |
self.headers = headers
|
118 |
|
119 |
|
120 |
class HttpBadRequest(HttpException): |
121 |
code = 400
|
122 |
|
123 |
|
124 |
class HttpForbidden(HttpException): |
125 |
code = 403
|
126 |
|
127 |
|
128 |
class HttpNotFound(HttpException): |
129 |
code = 404
|
130 |
|
131 |
|
132 |
class HttpGone(HttpException): |
133 |
code = 410
|
134 |
|
135 |
|
136 |
class HttpLengthRequired(HttpException): |
137 |
code = 411
|
138 |
|
139 |
|
140 |
class HttpInternalError(HttpException): |
141 |
code = 500
|
142 |
|
143 |
|
144 |
class HttpNotImplemented(HttpException): |
145 |
code = 501
|
146 |
|
147 |
|
148 |
class HttpServiceUnavailable(HttpException): |
149 |
code = 503
|
150 |
|
151 |
|
152 |
class HttpVersionNotSupported(HttpException): |
153 |
code = 505
|
154 |
|
155 |
|
156 |
class HttpJsonConverter: |
157 |
CONTENT_TYPE = "application/json"
|
158 |
|
159 |
def Encode(self, data): |
160 |
return serializer.DumpJson(data)
|
161 |
|
162 |
def Decode(self, data): |
163 |
return serializer.LoadJson(data)
|
164 |
|
165 |
|
166 |
def WaitForSocketCondition(poller, sock, event, timeout): |
167 |
"""Waits for a condition to occur on the socket.
|
168 |
|
169 |
@type poller: select.Poller
|
170 |
@param poller: Poller object as created by select.poll()
|
171 |
@type sock: socket
|
172 |
@param sock: Wait for events on this socket
|
173 |
@type event: int
|
174 |
@param event: ORed condition (see select module)
|
175 |
@type timeout: float or None
|
176 |
@param timeout: Timeout in seconds
|
177 |
@rtype: int or None
|
178 |
@return: None for timeout, otherwise occured conditions
|
179 |
|
180 |
"""
|
181 |
check = (event | select.POLLPRI | |
182 |
select.POLLNVAL | select.POLLHUP | select.POLLERR) |
183 |
|
184 |
if timeout is not None: |
185 |
# Poller object expects milliseconds
|
186 |
timeout *= 1000
|
187 |
|
188 |
poller.register(sock, event) |
189 |
try:
|
190 |
while True: |
191 |
# TODO: If the main thread receives a signal and we have no timeout, we
|
192 |
# could wait forever. This should check a global "quit" flag or
|
193 |
# something every so often.
|
194 |
io_events = poller.poll(timeout) |
195 |
if not io_events: |
196 |
# Timeout
|
197 |
return None |
198 |
for (evfd, evcond) in io_events: |
199 |
if evcond & check:
|
200 |
return evcond
|
201 |
finally:
|
202 |
poller.unregister(sock) |
203 |
|
204 |
|
205 |
def SocketOperation(poller, sock, op, arg1, timeout): |
206 |
"""Wrapper around socket functions.
|
207 |
|
208 |
This function abstracts error handling for socket operations, especially
|
209 |
for the complicated interaction with OpenSSL.
|
210 |
|
211 |
@type poller: select.Poller
|
212 |
@param poller: Poller object as created by select.poll()
|
213 |
@type sock: socket
|
214 |
@param sock: Socket for the operation
|
215 |
@type op: int
|
216 |
@param op: Operation to execute (SOCKOP_* constants)
|
217 |
@type arg1: any
|
218 |
@param arg1: Parameter for function (if needed)
|
219 |
@type timeout: None or float
|
220 |
@param timeout: Timeout in seconds or None
|
221 |
@return: Return value of socket function
|
222 |
|
223 |
"""
|
224 |
# TODO: event_poll/event_check/override
|
225 |
if op in (SOCKOP_SEND, SOCKOP_HANDSHAKE): |
226 |
event_poll = select.POLLOUT |
227 |
event_check = select.POLLOUT |
228 |
|
229 |
elif op == SOCKOP_RECV:
|
230 |
event_poll = select.POLLIN |
231 |
event_check = select.POLLIN | select.POLLPRI |
232 |
|
233 |
elif op == SOCKOP_SHUTDOWN:
|
234 |
event_poll = None
|
235 |
event_check = None
|
236 |
|
237 |
# The timeout is only used when OpenSSL requests polling for a condition.
|
238 |
# It is not advisable to have no timeout for shutdown.
|
239 |
assert timeout
|
240 |
|
241 |
else:
|
242 |
raise AssertionError("Invalid socket operation") |
243 |
|
244 |
# Handshake is only supported by SSL sockets
|
245 |
if (op == SOCKOP_HANDSHAKE and |
246 |
not isinstance(sock, OpenSSL.SSL.ConnectionType)): |
247 |
return
|
248 |
|
249 |
# No override by default
|
250 |
event_override = 0
|
251 |
|
252 |
while True: |
253 |
# Poll only for certain operations and when asked for by an override
|
254 |
if event_override or op in (SOCKOP_SEND, SOCKOP_RECV, SOCKOP_HANDSHAKE): |
255 |
if event_override:
|
256 |
wait_for_event = event_override |
257 |
else:
|
258 |
wait_for_event = event_poll |
259 |
|
260 |
event = WaitForSocketCondition(poller, sock, wait_for_event, timeout) |
261 |
if event is None: |
262 |
raise HttpSocketTimeout()
|
263 |
|
264 |
if (op == SOCKOP_RECV and |
265 |
event & (select.POLLNVAL | select.POLLHUP | select.POLLERR)): |
266 |
return "" |
267 |
|
268 |
if not event & wait_for_event: |
269 |
continue
|
270 |
|
271 |
# Reset override
|
272 |
event_override = 0
|
273 |
|
274 |
try:
|
275 |
try:
|
276 |
if op == SOCKOP_SEND:
|
277 |
return sock.send(arg1)
|
278 |
|
279 |
elif op == SOCKOP_RECV:
|
280 |
return sock.recv(arg1)
|
281 |
|
282 |
elif op == SOCKOP_SHUTDOWN:
|
283 |
if isinstance(sock, OpenSSL.SSL.ConnectionType): |
284 |
# PyOpenSSL's shutdown() doesn't take arguments
|
285 |
return sock.shutdown()
|
286 |
else:
|
287 |
return sock.shutdown(arg1)
|
288 |
|
289 |
elif op == SOCKOP_HANDSHAKE:
|
290 |
return sock.do_handshake()
|
291 |
|
292 |
except OpenSSL.SSL.WantWriteError:
|
293 |
# OpenSSL wants to write, poll for POLLOUT
|
294 |
event_override = select.POLLOUT |
295 |
continue
|
296 |
|
297 |
except OpenSSL.SSL.WantReadError:
|
298 |
# OpenSSL wants to read, poll for POLLIN
|
299 |
event_override = select.POLLIN | select.POLLPRI |
300 |
continue
|
301 |
|
302 |
except OpenSSL.SSL.WantX509LookupError:
|
303 |
continue
|
304 |
|
305 |
except OpenSSL.SSL.ZeroReturnError, err:
|
306 |
# SSL Connection has been closed. In SSL 3.0 and TLS 1.0, this only
|
307 |
# occurs if a closure alert has occurred in the protocol, i.e. the
|
308 |
# connection has been closed cleanly. Note that this does not
|
309 |
# necessarily mean that the transport layer (e.g. a socket) has been
|
310 |
# closed.
|
311 |
if op == SOCKOP_SEND:
|
312 |
# Can happen during a renegotiation
|
313 |
raise HttpConnectionClosed(err.args)
|
314 |
elif op == SOCKOP_RECV:
|
315 |
return "" |
316 |
|
317 |
# SSL_shutdown shouldn't return SSL_ERROR_ZERO_RETURN
|
318 |
raise socket.error(err.args)
|
319 |
|
320 |
except OpenSSL.SSL.SysCallError, err:
|
321 |
if op == SOCKOP_SEND:
|
322 |
# arg1 is the data when writing
|
323 |
if err.args and err.args[0] == -1 and arg1 == "": |
324 |
# errors when writing empty strings are expected
|
325 |
# and can be ignored
|
326 |
return 0 |
327 |
|
328 |
if err.args == (-1, _SSL_UNEXPECTED_EOF): |
329 |
if op == SOCKOP_RECV:
|
330 |
return "" |
331 |
elif op == SOCKOP_HANDSHAKE:
|
332 |
# Can happen if peer disconnects directly after the connection is
|
333 |
# opened.
|
334 |
raise HttpSessionHandshakeUnexpectedEOF(err.args)
|
335 |
|
336 |
raise socket.error(err.args)
|
337 |
|
338 |
except OpenSSL.SSL.Error, err:
|
339 |
raise socket.error(err.args)
|
340 |
|
341 |
except socket.error, err:
|
342 |
if err.args and err.args[0] == errno.EAGAIN: |
343 |
# Ignore EAGAIN
|
344 |
continue
|
345 |
|
346 |
raise
|
347 |
|
348 |
|
349 |
def ShutdownConnection(poller, sock, close_timeout, write_timeout, msgreader, |
350 |
force): |
351 |
"""Closes the connection.
|
352 |
|
353 |
@type poller: select.Poller
|
354 |
@param poller: Poller object as created by select.poll()
|
355 |
@type sock: socket
|
356 |
@param sock: Socket to be shut down
|
357 |
@type close_timeout: float
|
358 |
@param close_timeout: How long to wait for the peer to close the connection
|
359 |
@type write_timeout: float
|
360 |
@param write_timeout: Write timeout for shutdown
|
361 |
@type msgreader: http.HttpMessageReader
|
362 |
@param msgreader: Request message reader, used to determine whether peer
|
363 |
should close connection
|
364 |
@type force: bool
|
365 |
@param force: Whether to forcibly close the connection without waiting
|
366 |
for peer
|
367 |
|
368 |
"""
|
369 |
poller = select.poll() |
370 |
|
371 |
#print msgreader.peer_will_close, force
|
372 |
if msgreader and msgreader.peer_will_close and not force: |
373 |
# Wait for peer to close
|
374 |
try:
|
375 |
# Check whether it's actually closed
|
376 |
if not SocketOperation(poller, sock, SOCKOP_RECV, 1, close_timeout): |
377 |
return
|
378 |
except (socket.error, HttpError, HttpSocketTimeout):
|
379 |
# Ignore errors at this stage
|
380 |
pass
|
381 |
|
382 |
# Close the connection from our side
|
383 |
try:
|
384 |
# We don't care about the return value, see NOTES in SSL_shutdown(3).
|
385 |
SocketOperation(poller, sock, SOCKOP_SHUTDOWN, socket.SHUT_RDWR, |
386 |
write_timeout) |
387 |
except HttpSocketTimeout:
|
388 |
raise HttpError("Timeout while shutting down connection") |
389 |
except socket.error, err:
|
390 |
raise HttpError("Error while shutting down connection: %s" % err) |
391 |
|
392 |
|
393 |
def Handshake(poller, sock, write_timeout): |
394 |
"""Shakes peer's hands.
|
395 |
|
396 |
@type poller: select.Poller
|
397 |
@param poller: Poller object as created by select.poll()
|
398 |
@type sock: socket
|
399 |
@param sock: Socket to be shut down
|
400 |
@type write_timeout: float
|
401 |
@param write_timeout: Write timeout for handshake
|
402 |
|
403 |
"""
|
404 |
try:
|
405 |
return SocketOperation(poller, sock, SOCKOP_HANDSHAKE, None, write_timeout) |
406 |
except HttpSocketTimeout:
|
407 |
raise HttpError("Timeout during SSL handshake") |
408 |
except socket.error, err:
|
409 |
raise HttpError("Error in SSL handshake: %s" % err) |
410 |
|
411 |
|
412 |
class HttpSslParams(object): |
413 |
"""Data class for SSL key and certificate.
|
414 |
|
415 |
"""
|
416 |
def __init__(self, ssl_key_path, ssl_cert_path): |
417 |
"""Initializes this class.
|
418 |
|
419 |
@type ssl_key_path: string
|
420 |
@param ssl_key_path: Path to file containing SSL key in PEM format
|
421 |
@type ssl_cert_path: string
|
422 |
@param ssl_cert_path: Path to file containing SSL certificate in PEM format
|
423 |
|
424 |
"""
|
425 |
self.ssl_key_pem = utils.ReadFile(ssl_key_path)
|
426 |
self.ssl_cert_pem = utils.ReadFile(ssl_cert_path)
|
427 |
|
428 |
def GetKey(self): |
429 |
return OpenSSL.crypto.load_privatekey(OpenSSL.crypto.FILETYPE_PEM,
|
430 |
self.ssl_key_pem)
|
431 |
|
432 |
def GetCertificate(self): |
433 |
return OpenSSL.crypto.load_certificate(OpenSSL.crypto.FILETYPE_PEM,
|
434 |
self.ssl_cert_pem)
|
435 |
|
436 |
|
437 |
class HttpBase(object): |
438 |
"""Base class for HTTP server and client.
|
439 |
|
440 |
"""
|
441 |
def __init__(self): |
442 |
self.using_ssl = None |
443 |
self._ssl_params = None |
444 |
self._ssl_key = None |
445 |
self._ssl_cert = None |
446 |
|
447 |
def _CreateSocket(self, ssl_params, ssl_verify_peer): |
448 |
"""Creates a TCP socket and initializes SSL if needed.
|
449 |
|
450 |
@type ssl_params: HttpSslParams
|
451 |
@param ssl_params: SSL key and certificate
|
452 |
@type ssl_verify_peer: bool
|
453 |
@param ssl_verify_peer: Whether to require client certificate and compare
|
454 |
it with our certificate
|
455 |
|
456 |
"""
|
457 |
self._ssl_params = ssl_params
|
458 |
|
459 |
sock = socket.socket(socket.AF_INET, socket.SOCK_STREAM) |
460 |
|
461 |
# Should we enable SSL?
|
462 |
self.using_ssl = ssl_params is not None |
463 |
|
464 |
if not self.using_ssl: |
465 |
return sock
|
466 |
|
467 |
self._ssl_key = ssl_params.GetKey()
|
468 |
self._ssl_cert = ssl_params.GetCertificate()
|
469 |
|
470 |
ctx = OpenSSL.SSL.Context(OpenSSL.SSL.SSLv23_METHOD) |
471 |
ctx.set_options(OpenSSL.SSL.OP_NO_SSLv2) |
472 |
|
473 |
ctx.use_privatekey(self._ssl_key)
|
474 |
ctx.use_certificate(self._ssl_cert)
|
475 |
ctx.check_privatekey() |
476 |
|
477 |
if ssl_verify_peer:
|
478 |
ctx.set_verify(OpenSSL.SSL.VERIFY_PEER | |
479 |
OpenSSL.SSL.VERIFY_FAIL_IF_NO_PEER_CERT, |
480 |
self._SSLVerifyCallback)
|
481 |
|
482 |
return OpenSSL.SSL.Connection(ctx, sock)
|
483 |
|
484 |
def _SSLVerifyCallback(self, conn, cert, errnum, errdepth, ok): |
485 |
"""Verify the certificate provided by the peer
|
486 |
|
487 |
We only compare fingerprints. The client must use the same certificate as
|
488 |
we do on our side.
|
489 |
|
490 |
"""
|
491 |
assert self._ssl_params, "SSL not initialized" |
492 |
|
493 |
return (self._ssl_cert.digest("sha1") == cert.digest("sha1") and |
494 |
self._ssl_cert.digest("md5") == cert.digest("md5")) |
495 |
|
496 |
|
497 |
class HttpMessage(object): |
498 |
"""Data structure for HTTP message.
|
499 |
|
500 |
"""
|
501 |
def __init__(self): |
502 |
self.start_line = None |
503 |
self.headers = None |
504 |
self.body = None |
505 |
self.decoded_body = None |
506 |
|
507 |
|
508 |
class HttpClientToServerStartLine(object): |
509 |
"""Data structure for HTTP request start line.
|
510 |
|
511 |
"""
|
512 |
def __init__(self, method, path, version): |
513 |
self.method = method
|
514 |
self.path = path
|
515 |
self.version = version
|
516 |
|
517 |
def __str__(self): |
518 |
return "%s %s %s" % (self.method, self.path, self.version) |
519 |
|
520 |
|
521 |
class HttpServerToClientStartLine(object): |
522 |
"""Data structure for HTTP response start line.
|
523 |
|
524 |
"""
|
525 |
def __init__(self, version, code, reason): |
526 |
self.version = version
|
527 |
self.code = code
|
528 |
self.reason = reason
|
529 |
|
530 |
def __str__(self): |
531 |
return "%s %s %s" % (self.version, self.code, self.reason) |
532 |
|
533 |
|
534 |
class HttpMessageWriter(object): |
535 |
"""Writes an HTTP message to a socket.
|
536 |
|
537 |
"""
|
538 |
def __init__(self, sock, msg, write_timeout): |
539 |
"""Initializes this class and writes an HTTP message to a socket.
|
540 |
|
541 |
@type sock: socket
|
542 |
@param sock: Socket to be written to
|
543 |
@type msg: http.HttpMessage
|
544 |
@param msg: HTTP message to be written
|
545 |
@type write_timeout: float
|
546 |
@param write_timeout: Write timeout for socket
|
547 |
|
548 |
"""
|
549 |
self._msg = msg
|
550 |
|
551 |
self._PrepareMessage()
|
552 |
|
553 |
buf = self._FormatMessage()
|
554 |
|
555 |
poller = select.poll() |
556 |
|
557 |
pos = 0
|
558 |
end = len(buf)
|
559 |
while pos < end:
|
560 |
# Send only SOCK_BUF_SIZE bytes at a time
|
561 |
data = buf[pos:(pos + SOCK_BUF_SIZE)] |
562 |
|
563 |
sent = SocketOperation(poller, sock, SOCKOP_SEND, data, |
564 |
write_timeout) |
565 |
|
566 |
# Remove sent bytes
|
567 |
pos += sent |
568 |
|
569 |
assert pos == end, "Message wasn't sent completely" |
570 |
|
571 |
def _PrepareMessage(self): |
572 |
"""Prepares the HTTP message by setting mandatory headers.
|
573 |
|
574 |
"""
|
575 |
# RFC2616, section 4.3: "The presence of a message-body in a request is
|
576 |
# signaled by the inclusion of a Content-Length or Transfer-Encoding header
|
577 |
# field in the request's message-headers."
|
578 |
if self._msg.body: |
579 |
self._msg.headers[HTTP_CONTENT_LENGTH] = len(self._msg.body) |
580 |
|
581 |
def _FormatMessage(self): |
582 |
"""Serializes the HTTP message into a string.
|
583 |
|
584 |
"""
|
585 |
buf = StringIO() |
586 |
|
587 |
# Add start line
|
588 |
buf.write(str(self._msg.start_line)) |
589 |
buf.write("\r\n")
|
590 |
|
591 |
# Add headers
|
592 |
if self._msg.start_line.version != HTTP_0_9: |
593 |
for name, value in self._msg.headers.iteritems(): |
594 |
buf.write("%s: %s\r\n" % (name, value))
|
595 |
|
596 |
buf.write("\r\n")
|
597 |
|
598 |
# Add message body if needed
|
599 |
if self.HasMessageBody(): |
600 |
buf.write(self._msg.body)
|
601 |
|
602 |
elif self._msg.body: |
603 |
logging.warning("Ignoring message body")
|
604 |
|
605 |
return buf.getvalue()
|
606 |
|
607 |
def HasMessageBody(self): |
608 |
"""Checks whether the HTTP message contains a body.
|
609 |
|
610 |
Can be overriden by subclasses.
|
611 |
|
612 |
"""
|
613 |
return bool(self._msg.body) |
614 |
|
615 |
|
616 |
class HttpMessageReader(object): |
617 |
"""Reads HTTP message from socket.
|
618 |
|
619 |
"""
|
620 |
# Length limits
|
621 |
START_LINE_LENGTH_MAX = None
|
622 |
HEADER_LENGTH_MAX = None
|
623 |
|
624 |
# Parser state machine
|
625 |
PS_START_LINE = "start-line"
|
626 |
PS_HEADERS = "headers"
|
627 |
PS_BODY = "entity-body"
|
628 |
PS_COMPLETE = "complete"
|
629 |
|
630 |
def __init__(self, sock, msg, read_timeout): |
631 |
"""Reads an HTTP message from a socket.
|
632 |
|
633 |
@type sock: socket
|
634 |
@param sock: Socket to be read from
|
635 |
@type msg: http.HttpMessage
|
636 |
@param msg: Object for the read message
|
637 |
@type read_timeout: float
|
638 |
@param read_timeout: Read timeout for socket
|
639 |
|
640 |
"""
|
641 |
self.sock = sock
|
642 |
self.msg = msg
|
643 |
|
644 |
self.poller = select.poll()
|
645 |
self.start_line_buffer = None |
646 |
self.header_buffer = StringIO()
|
647 |
self.body_buffer = StringIO()
|
648 |
self.parser_status = self.PS_START_LINE |
649 |
self.content_length = None |
650 |
self.peer_will_close = None |
651 |
|
652 |
buf = ""
|
653 |
eof = False
|
654 |
while self.parser_status != self.PS_COMPLETE: |
655 |
# TODO: Don't read more than necessary (Content-Length), otherwise
|
656 |
# data might be lost and/or an error could occur
|
657 |
data = SocketOperation(self.poller, sock, SOCKOP_RECV, SOCK_BUF_SIZE,
|
658 |
read_timeout) |
659 |
|
660 |
if data:
|
661 |
buf += data |
662 |
else:
|
663 |
eof = True
|
664 |
|
665 |
# Do some parsing and error checking while more data arrives
|
666 |
buf = self._ContinueParsing(buf, eof)
|
667 |
|
668 |
# Must be done only after the buffer has been evaluated
|
669 |
# TODO: Connection-length < len(data read) and connection closed
|
670 |
if (eof and |
671 |
self.parser_status in (self.PS_START_LINE, |
672 |
self.PS_HEADERS)):
|
673 |
raise HttpError("Connection closed prematurely") |
674 |
|
675 |
# Parse rest
|
676 |
buf = self._ContinueParsing(buf, True) |
677 |
|
678 |
assert self.parser_status == self.PS_COMPLETE |
679 |
assert not buf, "Parser didn't read full response" |
680 |
|
681 |
msg.body = self.body_buffer.getvalue()
|
682 |
|
683 |
# TODO: Content-type, error handling
|
684 |
if msg.body:
|
685 |
msg.decoded_body = HttpJsonConverter().Decode(msg.body) |
686 |
else:
|
687 |
msg.decoded_body = None
|
688 |
|
689 |
if msg.decoded_body:
|
690 |
logging.debug("Message body: %s", msg.decoded_body)
|
691 |
|
692 |
def _ContinueParsing(self, buf, eof): |
693 |
"""Main function for HTTP message state machine.
|
694 |
|
695 |
@type buf: string
|
696 |
@param buf: Receive buffer
|
697 |
@type eof: bool
|
698 |
@param eof: Whether we've reached EOF on the socket
|
699 |
@rtype: string
|
700 |
@return: Updated receive buffer
|
701 |
|
702 |
"""
|
703 |
# TODO: Use offset instead of slicing when possible
|
704 |
if self.parser_status == self.PS_START_LINE: |
705 |
# Expect start line
|
706 |
while True: |
707 |
idx = buf.find("\r\n")
|
708 |
|
709 |
# RFC2616, section 4.1: "In the interest of robustness, servers SHOULD
|
710 |
# ignore any empty line(s) received where a Request-Line is expected.
|
711 |
# In other words, if the server is reading the protocol stream at the
|
712 |
# beginning of a message and receives a CRLF first, it should ignore
|
713 |
# the CRLF."
|
714 |
if idx == 0: |
715 |
# TODO: Limit number of CRLFs/empty lines for safety?
|
716 |
buf = buf[:2]
|
717 |
continue
|
718 |
|
719 |
if idx > 0: |
720 |
self.start_line_buffer = buf[:idx]
|
721 |
|
722 |
self._CheckStartLineLength(len(self.start_line_buffer)) |
723 |
|
724 |
# Remove status line, including CRLF
|
725 |
buf = buf[idx + 2:]
|
726 |
|
727 |
self.msg.start_line = self.ParseStartLine(self.start_line_buffer) |
728 |
|
729 |
self.parser_status = self.PS_HEADERS |
730 |
else:
|
731 |
# Check whether incoming data is getting too large, otherwise we just
|
732 |
# fill our read buffer.
|
733 |
self._CheckStartLineLength(len(buf)) |
734 |
|
735 |
break
|
736 |
|
737 |
# TODO: Handle messages without headers
|
738 |
if self.parser_status == self.PS_HEADERS: |
739 |
# Wait for header end
|
740 |
idx = buf.find("\r\n\r\n")
|
741 |
if idx >= 0: |
742 |
self.header_buffer.write(buf[:idx + 2]) |
743 |
|
744 |
self._CheckHeaderLength(self.header_buffer.tell()) |
745 |
|
746 |
# Remove headers, including CRLF
|
747 |
buf = buf[idx + 4:]
|
748 |
|
749 |
self._ParseHeaders()
|
750 |
|
751 |
self.parser_status = self.PS_BODY |
752 |
else:
|
753 |
# Check whether incoming data is getting too large, otherwise we just
|
754 |
# fill our read buffer.
|
755 |
self._CheckHeaderLength(len(buf)) |
756 |
|
757 |
if self.parser_status == self.PS_BODY: |
758 |
# TODO: Implement max size for body_buffer
|
759 |
self.body_buffer.write(buf)
|
760 |
buf = ""
|
761 |
|
762 |
# Check whether we've read everything
|
763 |
#
|
764 |
# RFC2616, section 4.4: "When a message-body is included with a message,
|
765 |
# the transfer-length of that body is determined by one of the following
|
766 |
# [...] 5. By the server closing the connection. (Closing the connection
|
767 |
# cannot be used to indicate the end of a request body, since that would
|
768 |
# leave no possibility for the server to send back a response.)"
|
769 |
#
|
770 |
# TODO: Error when buffer length > Content-Length header
|
771 |
if (eof or |
772 |
self.content_length is None or |
773 |
(self.content_length is not None and |
774 |
self.body_buffer.tell() >= self.content_length)): |
775 |
self.parser_status = self.PS_COMPLETE |
776 |
|
777 |
return buf
|
778 |
|
779 |
def _CheckStartLineLength(self, length): |
780 |
"""Limits the start line buffer size.
|
781 |
|
782 |
@type length: int
|
783 |
@param length: Buffer size
|
784 |
|
785 |
"""
|
786 |
if (self.START_LINE_LENGTH_MAX is not None and |
787 |
length > self.START_LINE_LENGTH_MAX):
|
788 |
raise HttpError("Start line longer than %d chars" % |
789 |
self.START_LINE_LENGTH_MAX)
|
790 |
|
791 |
def _CheckHeaderLength(self, length): |
792 |
"""Limits the header buffer size.
|
793 |
|
794 |
@type length: int
|
795 |
@param length: Buffer size
|
796 |
|
797 |
"""
|
798 |
if (self.HEADER_LENGTH_MAX is not None and |
799 |
length > self.HEADER_LENGTH_MAX):
|
800 |
raise HttpError("Headers longer than %d chars" % self.HEADER_LENGTH_MAX) |
801 |
|
802 |
def ParseStartLine(self, start_line): |
803 |
"""Parses the start line of a message.
|
804 |
|
805 |
Must be overriden by subclass.
|
806 |
|
807 |
@type start_line: string
|
808 |
@param start_line: Start line string
|
809 |
|
810 |
"""
|
811 |
raise NotImplementedError() |
812 |
|
813 |
def _WillPeerCloseConnection(self): |
814 |
"""Evaluate whether peer will close the connection.
|
815 |
|
816 |
@rtype: bool
|
817 |
@return: Whether peer will close the connection
|
818 |
|
819 |
"""
|
820 |
# RFC2616, section 14.10: "HTTP/1.1 defines the "close" connection option
|
821 |
# for the sender to signal that the connection will be closed after
|
822 |
# completion of the response. For example,
|
823 |
#
|
824 |
# Connection: close
|
825 |
#
|
826 |
# in either the request or the response header fields indicates that the
|
827 |
# connection SHOULD NOT be considered `persistent' (section 8.1) after the
|
828 |
# current request/response is complete."
|
829 |
|
830 |
hdr_connection = self.msg.headers.get(HTTP_CONNECTION, None) |
831 |
if hdr_connection:
|
832 |
hdr_connection = hdr_connection.lower() |
833 |
|
834 |
# An HTTP/1.1 server is assumed to stay open unless explicitly closed.
|
835 |
if self.msg.start_line.version == HTTP_1_1: |
836 |
return (hdr_connection and "close" in hdr_connection) |
837 |
|
838 |
# Some HTTP/1.0 implementations have support for persistent connections,
|
839 |
# using rules different than HTTP/1.1.
|
840 |
|
841 |
# For older HTTP, Keep-Alive indicates persistent connection.
|
842 |
if self.msg.headers.get(HTTP_KEEP_ALIVE): |
843 |
return False |
844 |
|
845 |
# At least Akamai returns a "Connection: Keep-Alive" header, which was
|
846 |
# supposed to be sent by the client.
|
847 |
if hdr_connection and "keep-alive" in hdr_connection: |
848 |
return False |
849 |
|
850 |
return True |
851 |
|
852 |
def _ParseHeaders(self): |
853 |
"""Parses the headers.
|
854 |
|
855 |
This function also adjusts internal variables based on header values.
|
856 |
|
857 |
RFC2616, section 4.3: "The presence of a message-body in a request is
|
858 |
signaled by the inclusion of a Content-Length or Transfer-Encoding header
|
859 |
field in the request's message-headers."
|
860 |
|
861 |
"""
|
862 |
# Parse headers
|
863 |
self.header_buffer.seek(0, 0) |
864 |
self.msg.headers = mimetools.Message(self.header_buffer, 0) |
865 |
|
866 |
self.peer_will_close = self._WillPeerCloseConnection() |
867 |
|
868 |
# Do we have a Content-Length header?
|
869 |
hdr_content_length = self.msg.headers.get(HTTP_CONTENT_LENGTH, None) |
870 |
if hdr_content_length:
|
871 |
try:
|
872 |
self.content_length = int(hdr_content_length) |
873 |
except ValueError: |
874 |
self.content_length = None |
875 |
if self.content_length is not None and self.content_length < 0: |
876 |
self.content_length = None |
877 |
|
878 |
# if the connection remains open and a content-length was not provided,
|
879 |
# then assume that the connection WILL close.
|
880 |
if self.content_length is None: |
881 |
self.peer_will_close = True |