root / lib / http / __init__.py @ 39cfc25b
History | View | Annotate | Download (22.7 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) = range(3) |
71 |
|
72 |
# send/receive quantum
|
73 |
SOCK_BUF_SIZE = 32768
|
74 |
|
75 |
|
76 |
class HttpError(Exception): |
77 |
"""Internal exception for HTTP errors.
|
78 |
|
79 |
This should only be used for internal error reporting.
|
80 |
|
81 |
"""
|
82 |
|
83 |
|
84 |
class HttpConnectionClosed(Exception): |
85 |
"""Internal exception for a closed connection.
|
86 |
|
87 |
This should only be used for internal error reporting. Only use
|
88 |
it if there's no other way to report this condition.
|
89 |
|
90 |
"""
|
91 |
|
92 |
|
93 |
class HttpSocketTimeout(Exception): |
94 |
"""Internal exception for socket timeouts.
|
95 |
|
96 |
This should only be used for internal error reporting.
|
97 |
|
98 |
"""
|
99 |
|
100 |
|
101 |
class HttpException(Exception): |
102 |
code = None
|
103 |
message = None
|
104 |
|
105 |
def __init__(self, message=None): |
106 |
Exception.__init__(self) |
107 |
if message is not None: |
108 |
self.message = message
|
109 |
|
110 |
|
111 |
class HttpBadRequest(HttpException): |
112 |
code = 400
|
113 |
|
114 |
|
115 |
class HttpForbidden(HttpException): |
116 |
code = 403
|
117 |
|
118 |
|
119 |
class HttpNotFound(HttpException): |
120 |
code = 404
|
121 |
|
122 |
|
123 |
class HttpGone(HttpException): |
124 |
code = 410
|
125 |
|
126 |
|
127 |
class HttpLengthRequired(HttpException): |
128 |
code = 411
|
129 |
|
130 |
|
131 |
class HttpInternalError(HttpException): |
132 |
code = 500
|
133 |
|
134 |
|
135 |
class HttpNotImplemented(HttpException): |
136 |
code = 501
|
137 |
|
138 |
|
139 |
class HttpServiceUnavailable(HttpException): |
140 |
code = 503
|
141 |
|
142 |
|
143 |
class HttpVersionNotSupported(HttpException): |
144 |
code = 505
|
145 |
|
146 |
|
147 |
class HttpJsonConverter: |
148 |
CONTENT_TYPE = "application/json"
|
149 |
|
150 |
def Encode(self, data): |
151 |
return serializer.DumpJson(data)
|
152 |
|
153 |
def Decode(self, data): |
154 |
return serializer.LoadJson(data)
|
155 |
|
156 |
|
157 |
def WaitForSocketCondition(poller, sock, event, timeout): |
158 |
"""Waits for a condition to occur on the socket.
|
159 |
|
160 |
@type poller: select.Poller
|
161 |
@param poller: Poller object as created by select.poll()
|
162 |
@type sock: socket
|
163 |
@param sock: Wait for events on this socket
|
164 |
@type event: int
|
165 |
@param event: ORed condition (see select module)
|
166 |
@type timeout: float or None
|
167 |
@param timeout: Timeout in seconds
|
168 |
@rtype: int or None
|
169 |
@return: None for timeout, otherwise occured conditions
|
170 |
|
171 |
"""
|
172 |
check = (event | select.POLLPRI | |
173 |
select.POLLNVAL | select.POLLHUP | select.POLLERR) |
174 |
|
175 |
if timeout is not None: |
176 |
# Poller object expects milliseconds
|
177 |
timeout *= 1000
|
178 |
|
179 |
poller.register(sock, event) |
180 |
try:
|
181 |
while True: |
182 |
# TODO: If the main thread receives a signal and we have no timeout, we
|
183 |
# could wait forever. This should check a global "quit" flag or
|
184 |
# something every so often.
|
185 |
io_events = poller.poll(timeout) |
186 |
if not io_events: |
187 |
# Timeout
|
188 |
return None |
189 |
for (evfd, evcond) in io_events: |
190 |
if evcond & check:
|
191 |
return evcond
|
192 |
finally:
|
193 |
poller.unregister(sock) |
194 |
|
195 |
|
196 |
def SocketOperation(poller, sock, op, arg1, timeout): |
197 |
"""Wrapper around socket functions.
|
198 |
|
199 |
This function abstracts error handling for socket operations, especially
|
200 |
for the complicated interaction with OpenSSL.
|
201 |
|
202 |
@type poller: select.Poller
|
203 |
@param poller: Poller object as created by select.poll()
|
204 |
@type sock: socket
|
205 |
@param sock: Socket for the operation
|
206 |
@type op: int
|
207 |
@param op: Operation to execute (SOCKOP_* constants)
|
208 |
@type arg1: any
|
209 |
@param arg1: Parameter for function (if needed)
|
210 |
@type timeout: None or float
|
211 |
@param timeout: Timeout in seconds or None
|
212 |
@return: Return value of socket function
|
213 |
|
214 |
"""
|
215 |
# TODO: event_poll/event_check/override
|
216 |
if op == SOCKOP_SEND:
|
217 |
event_poll = select.POLLOUT |
218 |
event_check = select.POLLOUT |
219 |
|
220 |
elif op == SOCKOP_RECV:
|
221 |
event_poll = select.POLLIN |
222 |
event_check = select.POLLIN | select.POLLPRI |
223 |
|
224 |
elif op == SOCKOP_SHUTDOWN:
|
225 |
event_poll = None
|
226 |
event_check = None
|
227 |
|
228 |
# The timeout is only used when OpenSSL requests polling for a condition.
|
229 |
# It is not advisable to have no timeout for shutdown.
|
230 |
assert timeout
|
231 |
|
232 |
else:
|
233 |
raise AssertionError("Invalid socket operation") |
234 |
|
235 |
# No override by default
|
236 |
event_override = 0
|
237 |
|
238 |
while True: |
239 |
# Poll only for certain operations and when asked for by an override
|
240 |
if event_override or op in (SOCKOP_SEND, SOCKOP_RECV): |
241 |
if event_override:
|
242 |
wait_for_event = event_override |
243 |
else:
|
244 |
wait_for_event = event_poll |
245 |
|
246 |
event = WaitForSocketCondition(poller, sock, wait_for_event, timeout) |
247 |
if event is None: |
248 |
raise HttpSocketTimeout()
|
249 |
|
250 |
if (op == SOCKOP_RECV and |
251 |
event & (select.POLLNVAL | select.POLLHUP | select.POLLERR)): |
252 |
return "" |
253 |
|
254 |
if not event & wait_for_event: |
255 |
continue
|
256 |
|
257 |
# Reset override
|
258 |
event_override = 0
|
259 |
|
260 |
try:
|
261 |
try:
|
262 |
if op == SOCKOP_SEND:
|
263 |
return sock.send(arg1)
|
264 |
|
265 |
elif op == SOCKOP_RECV:
|
266 |
return sock.recv(arg1)
|
267 |
|
268 |
elif op == SOCKOP_SHUTDOWN:
|
269 |
if isinstance(sock, OpenSSL.SSL.ConnectionType): |
270 |
# PyOpenSSL's shutdown() doesn't take arguments
|
271 |
return sock.shutdown()
|
272 |
else:
|
273 |
return sock.shutdown(arg1)
|
274 |
|
275 |
except OpenSSL.SSL.WantWriteError:
|
276 |
# OpenSSL wants to write, poll for POLLOUT
|
277 |
event_override = select.POLLOUT |
278 |
continue
|
279 |
|
280 |
except OpenSSL.SSL.WantReadError:
|
281 |
# OpenSSL wants to read, poll for POLLIN
|
282 |
event_override = select.POLLIN | select.POLLPRI |
283 |
continue
|
284 |
|
285 |
except OpenSSL.SSL.WantX509LookupError:
|
286 |
continue
|
287 |
|
288 |
except OpenSSL.SSL.ZeroReturnError, err:
|
289 |
# SSL Connection has been closed. In SSL 3.0 and TLS 1.0, this only
|
290 |
# occurs if a closure alert has occurred in the protocol, i.e. the
|
291 |
# connection has been closed cleanly. Note that this does not
|
292 |
# necessarily mean that the transport layer (e.g. a socket) has been
|
293 |
# closed.
|
294 |
if op == SOCKOP_SEND:
|
295 |
# Can happen during a renegotiation
|
296 |
raise HttpConnectionClosed(err.args)
|
297 |
elif op == SOCKOP_RECV:
|
298 |
return "" |
299 |
|
300 |
# SSL_shutdown shouldn't return SSL_ERROR_ZERO_RETURN
|
301 |
raise socket.error(err.args)
|
302 |
|
303 |
except OpenSSL.SSL.SysCallError, err:
|
304 |
if op == SOCKOP_SEND:
|
305 |
# arg1 is the data when writing
|
306 |
if err.args and err.args[0] == -1 and arg1 == "": |
307 |
# errors when writing empty strings are expected
|
308 |
# and can be ignored
|
309 |
return 0 |
310 |
|
311 |
elif op == SOCKOP_RECV:
|
312 |
if err.args == (-1, _SSL_UNEXPECTED_EOF): |
313 |
return "" |
314 |
|
315 |
raise socket.error(err.args)
|
316 |
|
317 |
except OpenSSL.SSL.Error, err:
|
318 |
raise socket.error(err.args)
|
319 |
|
320 |
except socket.error, err:
|
321 |
if err.args and err.args[0] == errno.EAGAIN: |
322 |
# Ignore EAGAIN
|
323 |
continue
|
324 |
|
325 |
raise
|
326 |
|
327 |
|
328 |
def ShutdownConnection(poller, sock, close_timeout, write_timeout, msgreader, |
329 |
force): |
330 |
"""Closes the connection.
|
331 |
|
332 |
@type poller: select.Poller
|
333 |
@param poller: Poller object as created by select.poll()
|
334 |
@type sock: socket
|
335 |
@param sock: Socket to be shut down
|
336 |
@type close_timeout: float
|
337 |
@param close_timeout: How long to wait for the peer to close the connection
|
338 |
@type write_timeout: float
|
339 |
@param write_timeout: Write timeout for shutdown
|
340 |
@type msgreader: http.HttpMessageReader
|
341 |
@param msgreader: Request message reader, used to determine whether peer
|
342 |
should close connection
|
343 |
@type force: bool
|
344 |
@param force: Whether to forcibly close the connection without waiting
|
345 |
for peer
|
346 |
|
347 |
"""
|
348 |
poller = select.poll() |
349 |
|
350 |
#print msgreader.peer_will_close, force
|
351 |
if msgreader and msgreader.peer_will_close and not force: |
352 |
# Wait for peer to close
|
353 |
try:
|
354 |
# Check whether it's actually closed
|
355 |
if not SocketOperation(poller, sock, SOCKOP_RECV, 1, close_timeout): |
356 |
return
|
357 |
except (socket.error, HttpError, HttpSocketTimeout):
|
358 |
# Ignore errors at this stage
|
359 |
pass
|
360 |
|
361 |
# Close the connection from our side
|
362 |
try:
|
363 |
# We don't care about the return value, see NOTES in SSL_shutdown(3).
|
364 |
SocketOperation(poller, sock, SOCKOP_SHUTDOWN, socket.SHUT_RDWR, |
365 |
write_timeout) |
366 |
except HttpSocketTimeout:
|
367 |
raise HttpError("Timeout while shutting down connection") |
368 |
except socket.error, err:
|
369 |
raise HttpError("Error while shutting down connection: %s" % err) |
370 |
|
371 |
|
372 |
class HttpSslParams(object): |
373 |
"""Data class for SSL key and certificate.
|
374 |
|
375 |
"""
|
376 |
def __init__(self, ssl_key_path, ssl_cert_path): |
377 |
"""Initializes this class.
|
378 |
|
379 |
@type ssl_key_path: string
|
380 |
@param ssl_key_path: Path to file containing SSL key in PEM format
|
381 |
@type ssl_cert_path: string
|
382 |
@param ssl_cert_path: Path to file containing SSL certificate in PEM format
|
383 |
|
384 |
"""
|
385 |
self.ssl_key_pem = utils.ReadFile(ssl_key_path)
|
386 |
self.ssl_cert_pem = utils.ReadFile(ssl_cert_path)
|
387 |
|
388 |
def GetKey(self): |
389 |
return OpenSSL.crypto.load_privatekey(OpenSSL.crypto.FILETYPE_PEM,
|
390 |
self.ssl_key_pem)
|
391 |
|
392 |
def GetCertificate(self): |
393 |
return OpenSSL.crypto.load_certificate(OpenSSL.crypto.FILETYPE_PEM,
|
394 |
self.ssl_cert_pem)
|
395 |
|
396 |
|
397 |
class HttpBase(object): |
398 |
"""Base class for HTTP server and client.
|
399 |
|
400 |
"""
|
401 |
def __init__(self): |
402 |
self.using_ssl = None |
403 |
self._ssl_params = None |
404 |
self._ssl_key = None |
405 |
self._ssl_cert = None |
406 |
|
407 |
def _CreateSocket(self, ssl_params, ssl_verify_peer): |
408 |
"""Creates a TCP socket and initializes SSL if needed.
|
409 |
|
410 |
@type ssl_params: HttpSslParams
|
411 |
@param ssl_params: SSL key and certificate
|
412 |
@type ssl_verify_peer: bool
|
413 |
@param ssl_verify_peer: Whether to require client certificate and compare
|
414 |
it with our certificate
|
415 |
|
416 |
"""
|
417 |
self._ssl_params = ssl_params
|
418 |
|
419 |
sock = socket.socket(socket.AF_INET, socket.SOCK_STREAM) |
420 |
|
421 |
# Should we enable SSL?
|
422 |
self.using_ssl = ssl_params is not None |
423 |
|
424 |
if not self.using_ssl: |
425 |
return sock
|
426 |
|
427 |
self._ssl_key = ssl_params.GetKey()
|
428 |
self._ssl_cert = ssl_params.GetCertificate()
|
429 |
|
430 |
ctx = OpenSSL.SSL.Context(OpenSSL.SSL.SSLv23_METHOD) |
431 |
ctx.set_options(OpenSSL.SSL.OP_NO_SSLv2) |
432 |
|
433 |
ctx.use_privatekey(self._ssl_key)
|
434 |
ctx.use_certificate(self._ssl_cert)
|
435 |
ctx.check_privatekey() |
436 |
|
437 |
if ssl_verify_peer:
|
438 |
ctx.set_verify(OpenSSL.SSL.VERIFY_PEER | |
439 |
OpenSSL.SSL.VERIFY_FAIL_IF_NO_PEER_CERT, |
440 |
self._SSLVerifyCallback)
|
441 |
|
442 |
return OpenSSL.SSL.Connection(ctx, sock)
|
443 |
|
444 |
def _SSLVerifyCallback(self, conn, cert, errnum, errdepth, ok): |
445 |
"""Verify the certificate provided by the peer
|
446 |
|
447 |
We only compare fingerprints. The client must use the same certificate as
|
448 |
we do on our side.
|
449 |
|
450 |
"""
|
451 |
assert self._ssl_params, "SSL not initialized" |
452 |
|
453 |
return (self._ssl_cert.digest("sha1") == cert.digest("sha1") and |
454 |
self._ssl_cert.digest("md5") == cert.digest("md5")) |
455 |
|
456 |
|
457 |
class HttpMessage(object): |
458 |
"""Data structure for HTTP message.
|
459 |
|
460 |
"""
|
461 |
def __init__(self): |
462 |
self.start_line = None |
463 |
self.headers = None |
464 |
self.body = None |
465 |
self.decoded_body = None |
466 |
|
467 |
|
468 |
class HttpClientToServerStartLine(object): |
469 |
"""Data structure for HTTP request start line.
|
470 |
|
471 |
"""
|
472 |
def __init__(self, method, path, version): |
473 |
self.method = method
|
474 |
self.path = path
|
475 |
self.version = version
|
476 |
|
477 |
def __str__(self): |
478 |
return "%s %s %s" % (self.method, self.path, self.version) |
479 |
|
480 |
|
481 |
class HttpServerToClientStartLine(object): |
482 |
"""Data structure for HTTP response start line.
|
483 |
|
484 |
"""
|
485 |
def __init__(self, version, code, reason): |
486 |
self.version = version
|
487 |
self.code = code
|
488 |
self.reason = reason
|
489 |
|
490 |
def __str__(self): |
491 |
return "%s %s %s" % (self.version, self.code, self.reason) |
492 |
|
493 |
|
494 |
class HttpMessageWriter(object): |
495 |
"""Writes an HTTP message to a socket.
|
496 |
|
497 |
"""
|
498 |
def __init__(self, sock, msg, write_timeout): |
499 |
"""Initializes this class and writes an HTTP message to a socket.
|
500 |
|
501 |
@type sock: socket
|
502 |
@param sock: Socket to be written to
|
503 |
@type msg: http.HttpMessage
|
504 |
@param msg: HTTP message to be written
|
505 |
@type write_timeout: float
|
506 |
@param write_timeout: Write timeout for socket
|
507 |
|
508 |
"""
|
509 |
self._msg = msg
|
510 |
|
511 |
self._PrepareMessage()
|
512 |
|
513 |
buf = self._FormatMessage()
|
514 |
|
515 |
poller = select.poll() |
516 |
|
517 |
pos = 0
|
518 |
end = len(buf)
|
519 |
while pos < end:
|
520 |
# Send only SOCK_BUF_SIZE bytes at a time
|
521 |
data = buf[pos:(pos + SOCK_BUF_SIZE)] |
522 |
|
523 |
sent = SocketOperation(poller, sock, SOCKOP_SEND, data, |
524 |
write_timeout) |
525 |
|
526 |
# Remove sent bytes
|
527 |
pos += sent |
528 |
|
529 |
assert pos == end, "Message wasn't sent completely" |
530 |
|
531 |
def _PrepareMessage(self): |
532 |
"""Prepares the HTTP message by setting mandatory headers.
|
533 |
|
534 |
"""
|
535 |
# RFC2616, section 4.3: "The presence of a message-body in a request is
|
536 |
# signaled by the inclusion of a Content-Length or Transfer-Encoding header
|
537 |
# field in the request's message-headers."
|
538 |
if self._msg.body: |
539 |
self._msg.headers[HTTP_CONTENT_LENGTH] = len(self._msg.body) |
540 |
|
541 |
def _FormatMessage(self): |
542 |
"""Serializes the HTTP message into a string.
|
543 |
|
544 |
"""
|
545 |
buf = StringIO() |
546 |
|
547 |
# Add start line
|
548 |
buf.write(str(self._msg.start_line)) |
549 |
buf.write("\r\n")
|
550 |
|
551 |
# Add headers
|
552 |
if self._msg.start_line.version != HTTP_0_9: |
553 |
for name, value in self._msg.headers.iteritems(): |
554 |
buf.write("%s: %s\r\n" % (name, value))
|
555 |
|
556 |
buf.write("\r\n")
|
557 |
|
558 |
# Add message body if needed
|
559 |
if self.HasMessageBody(): |
560 |
buf.write(self._msg.body)
|
561 |
|
562 |
elif self._msg.body: |
563 |
logging.warning("Ignoring message body")
|
564 |
|
565 |
return buf.getvalue()
|
566 |
|
567 |
def HasMessageBody(self): |
568 |
"""Checks whether the HTTP message contains a body.
|
569 |
|
570 |
Can be overriden by subclasses.
|
571 |
|
572 |
"""
|
573 |
return bool(self._msg.body) |
574 |
|
575 |
|
576 |
class HttpMessageReader(object): |
577 |
"""Reads HTTP message from socket.
|
578 |
|
579 |
"""
|
580 |
# Length limits
|
581 |
START_LINE_LENGTH_MAX = None
|
582 |
HEADER_LENGTH_MAX = None
|
583 |
|
584 |
# Parser state machine
|
585 |
PS_START_LINE = "start-line"
|
586 |
PS_HEADERS = "headers"
|
587 |
PS_BODY = "entity-body"
|
588 |
PS_COMPLETE = "complete"
|
589 |
|
590 |
def __init__(self, sock, msg, read_timeout): |
591 |
"""Reads an HTTP message from a socket.
|
592 |
|
593 |
@type sock: socket
|
594 |
@param sock: Socket to be read from
|
595 |
@type msg: http.HttpMessage
|
596 |
@param msg: Object for the read message
|
597 |
@type read_timeout: float
|
598 |
@param read_timeout: Read timeout for socket
|
599 |
|
600 |
"""
|
601 |
self.sock = sock
|
602 |
self.msg = msg
|
603 |
|
604 |
self.poller = select.poll()
|
605 |
self.start_line_buffer = None |
606 |
self.header_buffer = StringIO()
|
607 |
self.body_buffer = StringIO()
|
608 |
self.parser_status = self.PS_START_LINE |
609 |
self.content_length = None |
610 |
self.peer_will_close = None |
611 |
|
612 |
buf = ""
|
613 |
eof = False
|
614 |
while self.parser_status != self.PS_COMPLETE: |
615 |
data = SocketOperation(self.poller, sock, SOCKOP_RECV, SOCK_BUF_SIZE,
|
616 |
read_timeout) |
617 |
|
618 |
if data:
|
619 |
buf += data |
620 |
else:
|
621 |
eof = True
|
622 |
|
623 |
# Do some parsing and error checking while more data arrives
|
624 |
buf = self._ContinueParsing(buf, eof)
|
625 |
|
626 |
# Must be done only after the buffer has been evaluated
|
627 |
# TODO: Connection-length < len(data read) and connection closed
|
628 |
if (eof and |
629 |
self.parser_status in (self.PS_START_LINE, |
630 |
self.PS_HEADERS)):
|
631 |
raise HttpError("Connection closed prematurely") |
632 |
|
633 |
# Parse rest
|
634 |
buf = self._ContinueParsing(buf, True) |
635 |
|
636 |
assert self.parser_status == self.PS_COMPLETE |
637 |
assert not buf, "Parser didn't read full response" |
638 |
|
639 |
msg.body = self.body_buffer.getvalue()
|
640 |
|
641 |
# TODO: Content-type, error handling
|
642 |
if msg.body:
|
643 |
msg.decoded_body = HttpJsonConverter().Decode(msg.body) |
644 |
else:
|
645 |
msg.decoded_body = None
|
646 |
|
647 |
if msg.decoded_body:
|
648 |
logging.debug("Message body: %s", msg.decoded_body)
|
649 |
|
650 |
def _ContinueParsing(self, buf, eof): |
651 |
"""Main function for HTTP message state machine.
|
652 |
|
653 |
@type buf: string
|
654 |
@param buf: Receive buffer
|
655 |
@type eof: bool
|
656 |
@param eof: Whether we've reached EOF on the socket
|
657 |
@rtype: string
|
658 |
@return: Updated receive buffer
|
659 |
|
660 |
"""
|
661 |
if self.parser_status == self.PS_START_LINE: |
662 |
# Expect start line
|
663 |
while True: |
664 |
idx = buf.find("\r\n")
|
665 |
|
666 |
# RFC2616, section 4.1: "In the interest of robustness, servers SHOULD
|
667 |
# ignore any empty line(s) received where a Request-Line is expected.
|
668 |
# In other words, if the server is reading the protocol stream at the
|
669 |
# beginning of a message and receives a CRLF first, it should ignore
|
670 |
# the CRLF."
|
671 |
if idx == 0: |
672 |
# TODO: Limit number of CRLFs/empty lines for safety?
|
673 |
buf = buf[:2]
|
674 |
continue
|
675 |
|
676 |
if idx > 0: |
677 |
self.start_line_buffer = buf[:idx]
|
678 |
|
679 |
self._CheckStartLineLength(len(self.start_line_buffer)) |
680 |
|
681 |
# Remove status line, including CRLF
|
682 |
buf = buf[idx + 2:]
|
683 |
|
684 |
self.msg.start_line = self.ParseStartLine(self.start_line_buffer) |
685 |
|
686 |
self.parser_status = self.PS_HEADERS |
687 |
else:
|
688 |
# Check whether incoming data is getting too large, otherwise we just
|
689 |
# fill our read buffer.
|
690 |
self._CheckStartLineLength(len(buf)) |
691 |
|
692 |
break
|
693 |
|
694 |
# TODO: Handle messages without headers
|
695 |
if self.parser_status == self.PS_HEADERS: |
696 |
# Wait for header end
|
697 |
idx = buf.find("\r\n\r\n")
|
698 |
if idx >= 0: |
699 |
self.header_buffer.write(buf[:idx + 2]) |
700 |
|
701 |
self._CheckHeaderLength(self.header_buffer.tell()) |
702 |
|
703 |
# Remove headers, including CRLF
|
704 |
buf = buf[idx + 4:]
|
705 |
|
706 |
self._ParseHeaders()
|
707 |
|
708 |
self.parser_status = self.PS_BODY |
709 |
else:
|
710 |
# Check whether incoming data is getting too large, otherwise we just
|
711 |
# fill our read buffer.
|
712 |
self._CheckHeaderLength(len(buf)) |
713 |
|
714 |
if self.parser_status == self.PS_BODY: |
715 |
# TODO: Implement max size for body_buffer
|
716 |
self.body_buffer.write(buf)
|
717 |
buf = ""
|
718 |
|
719 |
# Check whether we've read everything
|
720 |
#
|
721 |
# RFC2616, section 4.4: "When a message-body is included with a message,
|
722 |
# the transfer-length of that body is determined by one of the following
|
723 |
# [...] 5. By the server closing the connection. (Closing the connection
|
724 |
# cannot be used to indicate the end of a request body, since that would
|
725 |
# leave no possibility for the server to send back a response.)"
|
726 |
if (eof or |
727 |
self.content_length is None or |
728 |
(self.content_length is not None and |
729 |
self.body_buffer.tell() >= self.content_length)): |
730 |
self.parser_status = self.PS_COMPLETE |
731 |
|
732 |
return buf
|
733 |
|
734 |
def _CheckStartLineLength(self, length): |
735 |
"""Limits the start line buffer size.
|
736 |
|
737 |
@type length: int
|
738 |
@param length: Buffer size
|
739 |
|
740 |
"""
|
741 |
if (self.START_LINE_LENGTH_MAX is not None and |
742 |
length > self.START_LINE_LENGTH_MAX):
|
743 |
raise HttpError("Start line longer than %d chars" % |
744 |
self.START_LINE_LENGTH_MAX)
|
745 |
|
746 |
def _CheckHeaderLength(self, length): |
747 |
"""Limits the header buffer size.
|
748 |
|
749 |
@type length: int
|
750 |
@param length: Buffer size
|
751 |
|
752 |
"""
|
753 |
if (self.HEADER_LENGTH_MAX is not None and |
754 |
length > self.HEADER_LENGTH_MAX):
|
755 |
raise HttpError("Headers longer than %d chars" % self.HEADER_LENGTH_MAX) |
756 |
|
757 |
def ParseStartLine(self, start_line): |
758 |
"""Parses the start line of a message.
|
759 |
|
760 |
Must be overriden by subclass.
|
761 |
|
762 |
@type start_line: string
|
763 |
@param start_line: Start line string
|
764 |
|
765 |
"""
|
766 |
raise NotImplementedError() |
767 |
|
768 |
def _WillPeerCloseConnection(self): |
769 |
"""Evaluate whether peer will close the connection.
|
770 |
|
771 |
@rtype: bool
|
772 |
@return: Whether peer will close the connection
|
773 |
|
774 |
"""
|
775 |
# RFC2616, section 14.10: "HTTP/1.1 defines the "close" connection option
|
776 |
# for the sender to signal that the connection will be closed after
|
777 |
# completion of the response. For example,
|
778 |
#
|
779 |
# Connection: close
|
780 |
#
|
781 |
# in either the request or the response header fields indicates that the
|
782 |
# connection SHOULD NOT be considered `persistent' (section 8.1) after the
|
783 |
# current request/response is complete."
|
784 |
|
785 |
hdr_connection = self.msg.headers.get(HTTP_CONNECTION, None) |
786 |
if hdr_connection:
|
787 |
hdr_connection = hdr_connection.lower() |
788 |
|
789 |
# An HTTP/1.1 server is assumed to stay open unless explicitly closed.
|
790 |
if self.msg.start_line.version == HTTP_1_1: |
791 |
return (hdr_connection and "close" in hdr_connection) |
792 |
|
793 |
# Some HTTP/1.0 implementations have support for persistent connections,
|
794 |
# using rules different than HTTP/1.1.
|
795 |
|
796 |
# For older HTTP, Keep-Alive indicates persistent connection.
|
797 |
if self.msg.headers.get(HTTP_KEEP_ALIVE): |
798 |
return False |
799 |
|
800 |
# At least Akamai returns a "Connection: Keep-Alive" header, which was
|
801 |
# supposed to be sent by the client.
|
802 |
if hdr_connection and "keep-alive" in hdr_connection: |
803 |
return False |
804 |
|
805 |
return True |
806 |
|
807 |
def _ParseHeaders(self): |
808 |
"""Parses the headers.
|
809 |
|
810 |
This function also adjusts internal variables based on header values.
|
811 |
|
812 |
RFC2616, section 4.3: "The presence of a message-body in a request is
|
813 |
signaled by the inclusion of a Content-Length or Transfer-Encoding header
|
814 |
field in the request's message-headers."
|
815 |
|
816 |
"""
|
817 |
# Parse headers
|
818 |
self.header_buffer.seek(0, 0) |
819 |
self.msg.headers = mimetools.Message(self.header_buffer, 0) |
820 |
|
821 |
self.peer_will_close = self._WillPeerCloseConnection() |
822 |
|
823 |
# Do we have a Content-Length header?
|
824 |
hdr_content_length = self.msg.headers.get(HTTP_CONTENT_LENGTH, None) |
825 |
if hdr_content_length:
|
826 |
try:
|
827 |
self.content_length = int(hdr_content_length) |
828 |
except ValueError: |
829 |
self.content_length = None |
830 |
if self.content_length is not None and self.content_length < 0: |
831 |
self.content_length = None |
832 |
|
833 |
# if the connection remains open and a content-length was not provided,
|
834 |
# then assume that the connection WILL close.
|
835 |
if self.content_length is None: |
836 |
self.peer_will_close = True |