Statistics
| Branch: | Tag: | Revision:

root / lib / http / __init__.py @ dcd511c8

History | View | Annotate | Download (26.8 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
HTTP_WWW_AUTHENTICATE = "WWW-Authenticate"
65
HTTP_AUTHORIZATION = "Authorization"
66
HTTP_AUTHENTICATION_INFO = "Authentication-Info"
67
HTTP_ALLOW = "Allow"
68

    
69
_SSL_UNEXPECTED_EOF = "Unexpected EOF"
70

    
71
# Socket operations
72
(SOCKOP_SEND,
73
 SOCKOP_RECV,
74
 SOCKOP_SHUTDOWN,
75
 SOCKOP_HANDSHAKE) = range(4)
76

    
77
# send/receive quantum
78
SOCK_BUF_SIZE = 32768
79

    
80

    
81
class HttpError(Exception):
82
  """Internal exception for HTTP errors.
83

84
  This should only be used for internal error reporting.
85

86
  """
87

    
88

    
89
class HttpConnectionClosed(Exception):
90
  """Internal exception for a closed connection.
91

92
  This should only be used for internal error reporting. Only use
93
  it if there's no other way to report this condition.
94

95
  """
96

    
97

    
98
class HttpSessionHandshakeUnexpectedEOF(HttpError):
99
  """Internal exception for errors during SSL handshake.
100

101
  This should only be used for internal error reporting.
102

103
  """
104

    
105

    
106
class HttpSocketTimeout(Exception):
107
  """Internal exception for socket timeouts.
108

109
  This should only be used for internal error reporting.
110

111
  """
112

    
113

    
114
class HttpException(Exception):
115
  code = None
116
  message = None
117

    
118
  def __init__(self, message=None, headers=None):
119
    Exception.__init__(self)
120
    self.message = message
121
    self.headers = headers
122

    
123

    
124
class HttpBadRequest(HttpException):
125
  """400 Bad Request
126

127
  RFC2616, 10.4.1: The request could not be understood by the server
128
  due to malformed syntax. The client SHOULD NOT repeat the request
129
  without modifications.
130

131
  """
132
  code = 400
133

    
134

    
135
class HttpUnauthorized(HttpException):
136
  """401 Unauthorized
137

138
  RFC2616, section 10.4.2: The request requires user
139
  authentication. The response MUST include a WWW-Authenticate header
140
  field (section 14.47) containing a challenge applicable to the
141
  requested resource.
142

143
  """
144
  code = 401
145

    
146

    
147
class HttpForbidden(HttpException):
148
  """403 Forbidden
149

150
  RFC2616, 10.4.4: The server understood the request, but is refusing
151
  to fulfill it.  Authorization will not help and the request SHOULD
152
  NOT be repeated.
153

154
  """
155
  code = 403
156

    
157

    
158
class HttpNotFound(HttpException):
159
  """404 Not Found
160

161
  RFC2616, 10.4.5: The server has not found anything matching the
162
  Request-URI.  No indication is given of whether the condition is
163
  temporary or permanent.
164

165
  """
166
  code = 404
167

    
168

    
169
class HttpMethodNotAllowed(HttpException):
170
  """405 Method Not Allowed
171

172
  RFC2616, 10.4.6: The method specified in the Request-Line is not
173
  allowed for the resource identified by the Request-URI. The response
174
  MUST include an Allow header containing a list of valid methods for
175
  the requested resource.
176

177
  """
178
  code = 405
179

    
180

    
181
class HttpRequestTimeout(HttpException):
182
  """408 Request Timeout
183

184
  RFC2616, 10.4.9: The client did not produce a request within the
185
  time that the server was prepared to wait. The client MAY repeat the
186
  request without modifications at any later time.
187

188
  """
189
  code = 408
190

    
191

    
192
class HttpConflict(HttpException):
193
  """409 Conflict
194

195
  RFC2616, 10.4.10: The request could not be completed due to a
196
  conflict with the current state of the resource. This code is only
197
  allowed in situations where it is expected that the user might be
198
  able to resolve the conflict and resubmit the request.
199

200
  """
201
  code = 409
202

    
203

    
204
class HttpGone(HttpException):
205
  """410 Gone
206

207
  RFC2616, 10.4.11: The requested resource is no longer available at
208
  the server and no forwarding address is known. This condition is
209
  expected to be considered permanent.
210

211
  """
212
  code = 410
213

    
214

    
215
class HttpLengthRequired(HttpException):
216
  """411 Length Required
217

218
  RFC2616, 10.4.12: The server refuses to accept the request without a
219
  defined Content-Length. The client MAY repeat the request if it adds
220
  a valid Content-Length header field containing the length of the
221
  message-body in the request message.
222

223
  """
224
  code = 411
225

    
226

    
227
class HttpPreconditionFailed(HttpException):
228
  """412 Precondition Failed
229

230
  RFC2616, 10.4.13: The precondition given in one or more of the
231
  request-header fields evaluated to false when it was tested on the
232
  server.
233

234
  """
235
  code = 412
236

    
237

    
238
class HttpInternalServerError(HttpException):
239
  """500 Internal Server Error
240

241
  RFC2616, 10.5.1: The server encountered an unexpected condition
242
  which prevented it from fulfilling the request.
243

244
  """
245
  code = 500
246

    
247

    
248
class HttpNotImplemented(HttpException):
249
  """501 Not Implemented
250

251
  RFC2616, 10.5.2: The server does not support the functionality
252
  required to fulfill the request.
253

254
  """
255
  code = 501
256

    
257

    
258
class HttpBadGateway(HttpException):
259
  """502 Bad Gateway
260

261
  RFC2616, 10.5.3: The server, while acting as a gateway or proxy,
262
  received an invalid response from the upstream server it accessed in
263
  attempting to fulfill the request.
264

265
  """
266
  code = 502
267

    
268

    
269
class HttpServiceUnavailable(HttpException):
270
  """503 Service Unavailable
271

272
  RFC2616, 10.5.4: The server is currently unable to handle the
273
  request due to a temporary overloading or maintenance of the server.
274

275
  """
276
  code = 503
277

    
278

    
279
class HttpGatewayTimeout(HttpException):
280
  """504 Gateway Timeout
281

282
  RFC2616, 10.5.5: The server, while acting as a gateway or proxy, did
283
  not receive a timely response from the upstream server specified by
284
  the URI (e.g.  HTTP, FTP, LDAP) or some other auxiliary server
285
  (e.g. DNS) it needed to access in attempting to complete the
286
  request.
287

288
  """
289
  code = 504
290

    
291

    
292
class HttpVersionNotSupported(HttpException):
293
  """505 HTTP Version Not Supported
294

295
  RFC2616, 10.5.6: The server does not support, or refuses to support,
296
  the HTTP protocol version that was used in the request message.
297

298
  """
299
  code = 505
300

    
301

    
302
class HttpJsonConverter: # pylint: disable-msg=W0232
303
  CONTENT_TYPE = "application/json"
304

    
305
  @staticmethod
306
  def Encode(data):
307
    return serializer.DumpJson(data)
308

    
309
  @staticmethod
310
  def Decode(data):
311
    return serializer.LoadJson(data)
312

    
313

    
314
def SocketOperation(sock, op, arg1, timeout):
315
  """Wrapper around socket functions.
316

317
  This function abstracts error handling for socket operations, especially
318
  for the complicated interaction with OpenSSL.
319

320
  @type sock: socket
321
  @param sock: Socket for the operation
322
  @type op: int
323
  @param op: Operation to execute (SOCKOP_* constants)
324
  @type arg1: any
325
  @param arg1: Parameter for function (if needed)
326
  @type timeout: None or float
327
  @param timeout: Timeout in seconds or None
328
  @return: Return value of socket function
329

330
  """
331
  # TODO: event_poll/event_check/override
332
  if op in (SOCKOP_SEND, SOCKOP_HANDSHAKE):
333
    event_poll = select.POLLOUT
334

    
335
  elif op == SOCKOP_RECV:
336
    event_poll = select.POLLIN
337

    
338
  elif op == SOCKOP_SHUTDOWN:
339
    event_poll = None
340

    
341
    # The timeout is only used when OpenSSL requests polling for a condition.
342
    # It is not advisable to have no timeout for shutdown.
343
    assert timeout
344

    
345
  else:
346
    raise AssertionError("Invalid socket operation")
347

    
348
  # Handshake is only supported by SSL sockets
349
  if (op == SOCKOP_HANDSHAKE and
350
      not isinstance(sock, OpenSSL.SSL.ConnectionType)):
351
    return
352

    
353
  # No override by default
354
  event_override = 0
355

    
356
  while True:
357
    # Poll only for certain operations and when asked for by an override
358
    if event_override or op in (SOCKOP_SEND, SOCKOP_RECV, SOCKOP_HANDSHAKE):
359
      if event_override:
360
        wait_for_event = event_override
361
      else:
362
        wait_for_event = event_poll
363

    
364
      event = utils.WaitForSocketCondition(sock, wait_for_event, timeout)
365
      if event is None:
366
        raise HttpSocketTimeout()
367

    
368
      if event & (select.POLLNVAL | select.POLLHUP | select.POLLERR):
369
        # Let the socket functions handle these
370
        break
371

    
372
      if not event & wait_for_event:
373
        continue
374

    
375
    # Reset override
376
    event_override = 0
377

    
378
    try:
379
      try:
380
        if op == SOCKOP_SEND:
381
          return sock.send(arg1)
382

    
383
        elif op == SOCKOP_RECV:
384
          return sock.recv(arg1)
385

    
386
        elif op == SOCKOP_SHUTDOWN:
387
          if isinstance(sock, OpenSSL.SSL.ConnectionType):
388
            # PyOpenSSL's shutdown() doesn't take arguments
389
            return sock.shutdown()
390
          else:
391
            return sock.shutdown(arg1)
392

    
393
        elif op == SOCKOP_HANDSHAKE:
394
          return sock.do_handshake()
395

    
396
      except OpenSSL.SSL.WantWriteError:
397
        # OpenSSL wants to write, poll for POLLOUT
398
        event_override = select.POLLOUT
399
        continue
400

    
401
      except OpenSSL.SSL.WantReadError:
402
        # OpenSSL wants to read, poll for POLLIN
403
        event_override = select.POLLIN | select.POLLPRI
404
        continue
405

    
406
      except OpenSSL.SSL.WantX509LookupError:
407
        continue
408

    
409
      except OpenSSL.SSL.ZeroReturnError, err:
410
        # SSL Connection has been closed. In SSL 3.0 and TLS 1.0, this only
411
        # occurs if a closure alert has occurred in the protocol, i.e. the
412
        # connection has been closed cleanly. Note that this does not
413
        # necessarily mean that the transport layer (e.g. a socket) has been
414
        # closed.
415
        if op == SOCKOP_SEND:
416
          # Can happen during a renegotiation
417
          raise HttpConnectionClosed(err.args)
418
        elif op == SOCKOP_RECV:
419
          return ""
420

    
421
        # SSL_shutdown shouldn't return SSL_ERROR_ZERO_RETURN
422
        raise socket.error(err.args)
423

    
424
      except OpenSSL.SSL.SysCallError, err:
425
        if op == SOCKOP_SEND:
426
          # arg1 is the data when writing
427
          if err.args and err.args[0] == -1 and arg1 == "":
428
            # errors when writing empty strings are expected
429
            # and can be ignored
430
            return 0
431

    
432
        if err.args == (-1, _SSL_UNEXPECTED_EOF):
433
          if op == SOCKOP_RECV:
434
            return ""
435
          elif op == SOCKOP_HANDSHAKE:
436
            # Can happen if peer disconnects directly after the connection is
437
            # opened.
438
            raise HttpSessionHandshakeUnexpectedEOF(err.args)
439

    
440
        raise socket.error(err.args)
441

    
442
      except OpenSSL.SSL.Error, err:
443
        raise socket.error(err.args)
444

    
445
    except socket.error, err:
446
      if err.args and err.args[0] == errno.EAGAIN:
447
        # Ignore EAGAIN
448
        continue
449

    
450
      raise
451

    
452

    
453
def ShutdownConnection(sock, close_timeout, write_timeout, msgreader, force):
454
  """Closes the connection.
455

456
  @type sock: socket
457
  @param sock: Socket to be shut down
458
  @type close_timeout: float
459
  @param close_timeout: How long to wait for the peer to close
460
      the connection
461
  @type write_timeout: float
462
  @param write_timeout: Write timeout for shutdown
463
  @type msgreader: http.HttpMessageReader
464
  @param msgreader: Request message reader, used to determine whether
465
      peer should close connection
466
  @type force: bool
467
  @param force: Whether to forcibly close the connection without
468
      waiting for peer
469

470
  """
471
  #print msgreader.peer_will_close, force
472
  if msgreader and msgreader.peer_will_close and not force:
473
    # Wait for peer to close
474
    try:
475
      # Check whether it's actually closed
476
      if not SocketOperation(sock, SOCKOP_RECV, 1, close_timeout):
477
        return
478
    except (socket.error, HttpError, HttpSocketTimeout):
479
      # Ignore errors at this stage
480
      pass
481

    
482
  # Close the connection from our side
483
  try:
484
    # We don't care about the return value, see NOTES in SSL_shutdown(3).
485
    SocketOperation(sock, SOCKOP_SHUTDOWN, socket.SHUT_RDWR,
486
                    write_timeout)
487
  except HttpSocketTimeout:
488
    raise HttpError("Timeout while shutting down connection")
489
  except socket.error, err:
490
    # Ignore ENOTCONN
491
    if not (err.args and err.args[0] == errno.ENOTCONN):
492
      raise HttpError("Error while shutting down connection: %s" % err)
493

    
494

    
495
def Handshake(sock, write_timeout):
496
  """Shakes peer's hands.
497

498
  @type sock: socket
499
  @param sock: Socket to be shut down
500
  @type write_timeout: float
501
  @param write_timeout: Write timeout for handshake
502

503
  """
504
  try:
505
    return SocketOperation(sock, SOCKOP_HANDSHAKE, None, write_timeout)
506
  except HttpSocketTimeout:
507
    raise HttpError("Timeout during SSL handshake")
508
  except socket.error, err:
509
    raise HttpError("Error in SSL handshake: %s" % err)
510

    
511

    
512
def InitSsl():
513
  """Initializes the SSL infrastructure.
514

515
  This function is idempotent.
516

517
  """
518
  if not OpenSSL.rand.status():
519
    raise EnvironmentError("OpenSSL could not collect enough entropy"
520
                           " for the PRNG")
521

    
522
  # TODO: Maybe add some additional seeding for OpenSSL's PRNG
523

    
524

    
525
class HttpSslParams(object):
526
  """Data class for SSL key and certificate.
527

528
  """
529
  def __init__(self, ssl_key_path, ssl_cert_path):
530
    """Initializes this class.
531

532
    @type ssl_key_path: string
533
    @param ssl_key_path: Path to file containing SSL key in PEM format
534
    @type ssl_cert_path: string
535
    @param ssl_cert_path: Path to file containing SSL certificate
536
        in PEM format
537

538
    """
539
    self.ssl_key_pem = utils.ReadFile(ssl_key_path)
540
    self.ssl_cert_pem = utils.ReadFile(ssl_cert_path)
541

    
542
  def GetKey(self):
543
    return OpenSSL.crypto.load_privatekey(OpenSSL.crypto.FILETYPE_PEM,
544
                                          self.ssl_key_pem)
545

    
546
  def GetCertificate(self):
547
    return OpenSSL.crypto.load_certificate(OpenSSL.crypto.FILETYPE_PEM,
548
                                           self.ssl_cert_pem)
549

    
550

    
551
class HttpBase(object):
552
  """Base class for HTTP server and client.
553

554
  """
555
  def __init__(self):
556
    self.using_ssl = None
557
    self._ssl_params = None
558
    self._ssl_key = None
559
    self._ssl_cert = None
560

    
561
  def _CreateSocket(self, ssl_params, ssl_verify_peer):
562
    """Creates a TCP socket and initializes SSL if needed.
563

564
    @type ssl_params: HttpSslParams
565
    @param ssl_params: SSL key and certificate
566
    @type ssl_verify_peer: bool
567
    @param ssl_verify_peer: Whether to require client certificate
568
        and compare it with our certificate
569

570
    """
571
    self._ssl_params = ssl_params
572

    
573
    sock = socket.socket(socket.AF_INET, socket.SOCK_STREAM)
574

    
575
    # Should we enable SSL?
576
    self.using_ssl = ssl_params is not None
577

    
578
    if not self.using_ssl:
579
      return sock
580

    
581
    self._ssl_key = ssl_params.GetKey()
582
    self._ssl_cert = ssl_params.GetCertificate()
583

    
584
    ctx = OpenSSL.SSL.Context(OpenSSL.SSL.SSLv23_METHOD)
585
    ctx.set_options(OpenSSL.SSL.OP_NO_SSLv2)
586

    
587
    ctx.use_privatekey(self._ssl_key)
588
    ctx.use_certificate(self._ssl_cert)
589
    ctx.check_privatekey()
590

    
591
    if ssl_verify_peer:
592
      ctx.set_verify(OpenSSL.SSL.VERIFY_PEER |
593
                     OpenSSL.SSL.VERIFY_FAIL_IF_NO_PEER_CERT,
594
                     self._SSLVerifyCallback)
595

    
596
    return OpenSSL.SSL.Connection(ctx, sock)
597

    
598
  def _SSLVerifyCallback(self, conn, cert, errnum, errdepth, ok):
599
    """Verify the certificate provided by the peer
600

601
    We only compare fingerprints. The client must use the same certificate as
602
    we do on our side.
603

604
    """
605
    # some parameters are unused, but this is the API
606
    # pylint: disable-msg=W0613
607
    assert self._ssl_params, "SSL not initialized"
608

    
609
    return (self._ssl_cert.digest("sha1") == cert.digest("sha1") and
610
            self._ssl_cert.digest("md5") == cert.digest("md5"))
611

    
612

    
613
class HttpMessage(object):
614
  """Data structure for HTTP message.
615

616
  """
617
  def __init__(self):
618
    self.start_line = None
619
    self.headers = None
620
    self.body = None
621
    self.decoded_body = None
622

    
623

    
624
class HttpClientToServerStartLine(object):
625
  """Data structure for HTTP request start line.
626

627
  """
628
  def __init__(self, method, path, version):
629
    self.method = method
630
    self.path = path
631
    self.version = version
632

    
633
  def __str__(self):
634
    return "%s %s %s" % (self.method, self.path, self.version)
635

    
636

    
637
class HttpServerToClientStartLine(object):
638
  """Data structure for HTTP response start line.
639

640
  """
641
  def __init__(self, version, code, reason):
642
    self.version = version
643
    self.code = code
644
    self.reason = reason
645

    
646
  def __str__(self):
647
    return "%s %s %s" % (self.version, self.code, self.reason)
648

    
649

    
650
class HttpMessageWriter(object):
651
  """Writes an HTTP message to a socket.
652

653
  """
654
  def __init__(self, sock, msg, write_timeout):
655
    """Initializes this class and writes an HTTP message to a socket.
656

657
    @type sock: socket
658
    @param sock: Socket to be written to
659
    @type msg: http.HttpMessage
660
    @param msg: HTTP message to be written
661
    @type write_timeout: float
662
    @param write_timeout: Write timeout for socket
663

664
    """
665
    self._msg = msg
666

    
667
    self._PrepareMessage()
668

    
669
    buf = self._FormatMessage()
670

    
671
    pos = 0
672
    end = len(buf)
673
    while pos < end:
674
      # Send only SOCK_BUF_SIZE bytes at a time
675
      data = buf[pos:(pos + SOCK_BUF_SIZE)]
676

    
677
      sent = SocketOperation(sock, SOCKOP_SEND, data, write_timeout)
678

    
679
      # Remove sent bytes
680
      pos += sent
681

    
682
    assert pos == end, "Message wasn't sent completely"
683

    
684
  def _PrepareMessage(self):
685
    """Prepares the HTTP message by setting mandatory headers.
686

687
    """
688
    # RFC2616, section 4.3: "The presence of a message-body in a request is
689
    # signaled by the inclusion of a Content-Length or Transfer-Encoding header
690
    # field in the request's message-headers."
691
    if self._msg.body:
692
      self._msg.headers[HTTP_CONTENT_LENGTH] = len(self._msg.body)
693

    
694
  def _FormatMessage(self):
695
    """Serializes the HTTP message into a string.
696

697
    """
698
    buf = StringIO()
699

    
700
    # Add start line
701
    buf.write(str(self._msg.start_line))
702
    buf.write("\r\n")
703

    
704
    # Add headers
705
    if self._msg.start_line.version != HTTP_0_9:
706
      for name, value in self._msg.headers.iteritems():
707
        buf.write("%s: %s\r\n" % (name, value))
708

    
709
    buf.write("\r\n")
710

    
711
    # Add message body if needed
712
    if self.HasMessageBody():
713
      buf.write(self._msg.body)
714

    
715
    elif self._msg.body:
716
      logging.warning("Ignoring message body")
717

    
718
    return buf.getvalue()
719

    
720
  def HasMessageBody(self):
721
    """Checks whether the HTTP message contains a body.
722

723
    Can be overridden by subclasses.
724

725
    """
726
    return bool(self._msg.body)
727

    
728

    
729
class HttpMessageReader(object):
730
  """Reads HTTP message from socket.
731

732
  """
733
  # Length limits
734
  START_LINE_LENGTH_MAX = None
735
  HEADER_LENGTH_MAX = None
736

    
737
  # Parser state machine
738
  PS_START_LINE = "start-line"
739
  PS_HEADERS = "headers"
740
  PS_BODY = "entity-body"
741
  PS_COMPLETE = "complete"
742

    
743
  def __init__(self, sock, msg, read_timeout):
744
    """Reads an HTTP message from a socket.
745

746
    @type sock: socket
747
    @param sock: Socket to be read from
748
    @type msg: http.HttpMessage
749
    @param msg: Object for the read message
750
    @type read_timeout: float
751
    @param read_timeout: Read timeout for socket
752

753
    """
754
    self.sock = sock
755
    self.msg = msg
756

    
757
    self.start_line_buffer = None
758
    self.header_buffer = StringIO()
759
    self.body_buffer = StringIO()
760
    self.parser_status = self.PS_START_LINE
761
    self.content_length = None
762
    self.peer_will_close = None
763

    
764
    buf = ""
765
    eof = False
766
    while self.parser_status != self.PS_COMPLETE:
767
      # TODO: Don't read more than necessary (Content-Length), otherwise
768
      # data might be lost and/or an error could occur
769
      data = SocketOperation(sock, SOCKOP_RECV, SOCK_BUF_SIZE, read_timeout)
770

    
771
      if data:
772
        buf += data
773
      else:
774
        eof = True
775

    
776
      # Do some parsing and error checking while more data arrives
777
      buf = self._ContinueParsing(buf, eof)
778

    
779
      # Must be done only after the buffer has been evaluated
780
      # TODO: Connection-length < len(data read) and connection closed
781
      if (eof and
782
          self.parser_status in (self.PS_START_LINE,
783
                                 self.PS_HEADERS)):
784
        raise HttpError("Connection closed prematurely")
785

    
786
    # Parse rest
787
    buf = self._ContinueParsing(buf, True)
788

    
789
    assert self.parser_status == self.PS_COMPLETE
790
    assert not buf, "Parser didn't read full response"
791

    
792
    msg.body = self.body_buffer.getvalue()
793

    
794
    # TODO: Content-type, error handling
795
    if msg.body:
796
      msg.decoded_body = HttpJsonConverter().Decode(msg.body)
797
    else:
798
      msg.decoded_body = None
799

    
800
    if msg.decoded_body:
801
      logging.debug("Message body: %s", msg.decoded_body)
802

    
803
  def _ContinueParsing(self, buf, eof):
804
    """Main function for HTTP message state machine.
805

806
    @type buf: string
807
    @param buf: Receive buffer
808
    @type eof: bool
809
    @param eof: Whether we've reached EOF on the socket
810
    @rtype: string
811
    @return: Updated receive buffer
812

813
    """
814
    # TODO: Use offset instead of slicing when possible
815
    if self.parser_status == self.PS_START_LINE:
816
      # Expect start line
817
      while True:
818
        idx = buf.find("\r\n")
819

    
820
        # RFC2616, section 4.1: "In the interest of robustness, servers SHOULD
821
        # ignore any empty line(s) received where a Request-Line is expected.
822
        # In other words, if the server is reading the protocol stream at the
823
        # beginning of a message and receives a CRLF first, it should ignore
824
        # the CRLF."
825
        if idx == 0:
826
          # TODO: Limit number of CRLFs/empty lines for safety?
827
          buf = buf[2:]
828
          continue
829

    
830
        if idx > 0:
831
          self.start_line_buffer = buf[:idx]
832

    
833
          self._CheckStartLineLength(len(self.start_line_buffer))
834

    
835
          # Remove status line, including CRLF
836
          buf = buf[idx + 2:]
837

    
838
          self.msg.start_line = self.ParseStartLine(self.start_line_buffer)
839

    
840
          self.parser_status = self.PS_HEADERS
841
        else:
842
          # Check whether incoming data is getting too large, otherwise we just
843
          # fill our read buffer.
844
          self._CheckStartLineLength(len(buf))
845

    
846
        break
847

    
848
    # TODO: Handle messages without headers
849
    if self.parser_status == self.PS_HEADERS:
850
      # Wait for header end
851
      idx = buf.find("\r\n\r\n")
852
      if idx >= 0:
853
        self.header_buffer.write(buf[:idx + 2])
854

    
855
        self._CheckHeaderLength(self.header_buffer.tell())
856

    
857
        # Remove headers, including CRLF
858
        buf = buf[idx + 4:]
859

    
860
        self._ParseHeaders()
861

    
862
        self.parser_status = self.PS_BODY
863
      else:
864
        # Check whether incoming data is getting too large, otherwise we just
865
        # fill our read buffer.
866
        self._CheckHeaderLength(len(buf))
867

    
868
    if self.parser_status == self.PS_BODY:
869
      # TODO: Implement max size for body_buffer
870
      self.body_buffer.write(buf)
871
      buf = ""
872

    
873
      # Check whether we've read everything
874
      #
875
      # RFC2616, section 4.4: "When a message-body is included with a message,
876
      # the transfer-length of that body is determined by one of the following
877
      # [...] 5. By the server closing the connection. (Closing the connection
878
      # cannot be used to indicate the end of a request body, since that would
879
      # leave no possibility for the server to send back a response.)"
880
      #
881
      # TODO: Error when buffer length > Content-Length header
882
      if (eof or
883
          self.content_length is None or
884
          (self.content_length is not None and
885
           self.body_buffer.tell() >= self.content_length)):
886
        self.parser_status = self.PS_COMPLETE
887

    
888
    return buf
889

    
890
  def _CheckStartLineLength(self, length):
891
    """Limits the start line buffer size.
892

893
    @type length: int
894
    @param length: Buffer size
895

896
    """
897
    if (self.START_LINE_LENGTH_MAX is not None and
898
        length > self.START_LINE_LENGTH_MAX):
899
      raise HttpError("Start line longer than %d chars" %
900
                       self.START_LINE_LENGTH_MAX)
901

    
902
  def _CheckHeaderLength(self, length):
903
    """Limits the header buffer size.
904

905
    @type length: int
906
    @param length: Buffer size
907

908
    """
909
    if (self.HEADER_LENGTH_MAX is not None and
910
        length > self.HEADER_LENGTH_MAX):
911
      raise HttpError("Headers longer than %d chars" % self.HEADER_LENGTH_MAX)
912

    
913
  def ParseStartLine(self, start_line):
914
    """Parses the start line of a message.
915

916
    Must be overridden by subclass.
917

918
    @type start_line: string
919
    @param start_line: Start line string
920

921
    """
922
    raise NotImplementedError()
923

    
924
  def _WillPeerCloseConnection(self):
925
    """Evaluate whether peer will close the connection.
926

927
    @rtype: bool
928
    @return: Whether peer will close the connection
929

930
    """
931
    # RFC2616, section 14.10: "HTTP/1.1 defines the "close" connection option
932
    # for the sender to signal that the connection will be closed after
933
    # completion of the response. For example,
934
    #
935
    #        Connection: close
936
    #
937
    # in either the request or the response header fields indicates that the
938
    # connection SHOULD NOT be considered `persistent' (section 8.1) after the
939
    # current request/response is complete."
940

    
941
    hdr_connection = self.msg.headers.get(HTTP_CONNECTION, None)
942
    if hdr_connection:
943
      hdr_connection = hdr_connection.lower()
944

    
945
    # An HTTP/1.1 server is assumed to stay open unless explicitly closed.
946
    if self.msg.start_line.version == HTTP_1_1:
947
      return (hdr_connection and "close" in hdr_connection)
948

    
949
    # Some HTTP/1.0 implementations have support for persistent connections,
950
    # using rules different than HTTP/1.1.
951

    
952
    # For older HTTP, Keep-Alive indicates persistent connection.
953
    if self.msg.headers.get(HTTP_KEEP_ALIVE):
954
      return False
955

    
956
    # At least Akamai returns a "Connection: Keep-Alive" header, which was
957
    # supposed to be sent by the client.
958
    if hdr_connection and "keep-alive" in hdr_connection:
959
      return False
960

    
961
    return True
962

    
963
  def _ParseHeaders(self):
964
    """Parses the headers.
965

966
    This function also adjusts internal variables based on header values.
967

968
    RFC2616, section 4.3: The presence of a message-body in a request is
969
    signaled by the inclusion of a Content-Length or Transfer-Encoding header
970
    field in the request's message-headers.
971

972
    """
973
    # Parse headers
974
    self.header_buffer.seek(0, 0)
975
    self.msg.headers = mimetools.Message(self.header_buffer, 0)
976

    
977
    self.peer_will_close = self._WillPeerCloseConnection()
978

    
979
    # Do we have a Content-Length header?
980
    hdr_content_length = self.msg.headers.get(HTTP_CONTENT_LENGTH, None)
981
    if hdr_content_length:
982
      try:
983
        self.content_length = int(hdr_content_length)
984
      except (TypeError, ValueError):
985
        self.content_length = None
986
      if self.content_length is not None and self.content_length < 0:
987
        self.content_length = None
988

    
989
    # if the connection remains open and a content-length was not provided,
990
    # then assume that the connection WILL close.
991
    if self.content_length is None:
992
      self.peer_will_close = True