src/allmydata/mutable/retrieve.py

   1
   2 import time
   3 from itertools import count
   4 from zope.interface import implements
   5 from twisted.internet import defer
   6 from twisted.python import failure
   7 from twisted.internet.interfaces import IPushProducer, IConsumer
   8 from foolscap.api import eventually, fireEventually
   9 from allmydata.interfaces import IRetrieveStatus, NotEnoughSharesError, \
  10                                  MDMF_VERSION, SDMF_VERSION
  11 from allmydata.util import hashutil, log, mathutil
  12 from allmydata.util.dictutil import DictOfSets
  13 from allmydata import hashtree, codec
  14 from allmydata.storage.server import si_b2a
  15 from pycryptopp.cipher.aes import AES
  16 from pycryptopp.publickey import rsa
  17
  18 from allmydata.mutable.common import CorruptShareError, UncoordinatedWriteError
  19 from allmydata.mutable.layout import MDMFSlotReadProxy
  20
  21 class RetrieveStatus:
  22     implements(IRetrieveStatus)
  23     statusid_counter = count(0)
  24     def __init__(self):
  25         self.timings = {}
  26         self.timings["fetch_per_server"] = {}
  27         self.timings["decode"] = 0.0
  28         self.timings["decrypt"] = 0.0
  29         self.timings["cumulative_verify"] = 0.0
  30         self.problems = {}
  31         self.active = True
  32         self.storage_index = None
  33         self.helper = False
  34         self.encoding = ("?","?")
  35         self.size = None
  36         self.status = "Not started"
  37         self.progress = 0.0
  38         self.counter = self.statusid_counter.next()
  39         self.started = time.time()
  40
  41     def get_started(self):
  42         return self.started
  43     def get_storage_index(self):
  44         return self.storage_index
  45     def get_encoding(self):
  46         return self.encoding
  47     def using_helper(self):
  48         return self.helper
  49     def get_size(self):
  50         return self.size
  51     def get_status(self):
  52         return self.status
  53     def get_progress(self):
  54         return self.progress
  55     def get_active(self):
  56         return self.active
  57     def get_counter(self):
  58         return self.counter
  59
  60     def add_fetch_timing(self, peerid, elapsed):
  61         if peerid not in self.timings["fetch_per_server"]:
  62             self.timings["fetch_per_server"][peerid] = []
  63         self.timings["fetch_per_server"][peerid].append(elapsed)
  64     def accumulate_decode_time(self, elapsed):
  65         self.timings["decode"] += elapsed
  66     def accumulate_decrypt_time(self, elapsed):
  67         self.timings["decrypt"] += elapsed
  68     def set_storage_index(self, si):
  69         self.storage_index = si
  70     def set_helper(self, helper):
  71         self.helper = helper
  72     def set_encoding(self, k, n):
  73         self.encoding = (k, n)
  74     def set_size(self, size):
  75         self.size = size
  76     def set_status(self, status):
  77         self.status = status
  78     def set_progress(self, value):
  79         self.progress = value
  80     def set_active(self, value):
  81         self.active = value
  82
  83 class Marker:
  84     pass
  85
  86 class Retrieve:
  87     # this class is currently single-use. Eventually (in MDMF) we will make
  88     # it multi-use, in which case you can call download(range) multiple
  89     # times, and each will have a separate response chain. However the
  90     # Retrieve object will remain tied to a specific version of the file, and
  91     # will use a single ServerMap instance.
  92     implements(IPushProducer)
  93
  94     def __init__(self, filenode, servermap, verinfo, fetch_privkey=False,
  95                  verify=False):
  96         self._node = filenode
  97         assert self._node.get_pubkey()
  98         self._storage_index = filenode.get_storage_index()
  99         assert self._node.get_readkey()
 100         self._last_failure = None
 101         prefix = si_b2a(self._storage_index)[:5]
 102         self._log_number = log.msg("Retrieve(%s): starting" % prefix)
 103         self._outstanding_queries = {} # maps (peerid,shnum) to start_time
 104         self._running = True
 105         self._decoding = False
 106         self._bad_shares = set()
 107
 108         self.servermap = servermap
 109         assert self._node.get_pubkey()
 110         self.verinfo = verinfo
 111         # during repair, we may be called upon to grab the private key, since
 112         # it wasn't picked up during a verify=False checker run, and we'll
 113         # need it for repair to generate a new version.
 114         self._need_privkey = verify or (fetch_privkey
 115                                         and not self._node.get_privkey())
 116
 117         if self._need_privkey:
 118             # TODO: Evaluate the need for this. We'll use it if we want
 119             # to limit how many queries are on the wire for the privkey
 120             # at once.
 121             self._privkey_query_markers = [] # one Marker for each time we've
 122                                              # tried to get the privkey.
 123
 124         # verify means that we are using the downloader logic to verify all
 125         # of our shares. This tells the downloader a few things.
 126         #
 127         # 1. We need to download all of the shares.
 128         # 2. We don't need to decode or decrypt the shares, since our
 129         #    caller doesn't care about the plaintext, only the
 130         #    information about which shares are or are not valid.
 131         # 3. When we are validating readers, we need to validate the
 132         #    signature on the prefix. Do we? We already do this in the
 133         #    servermap update?
 134         self._verify = verify
 135
 136         self._status = RetrieveStatus()
 137         self._status.set_storage_index(self._storage_index)
 138         self._status.set_helper(False)
 139         self._status.set_progress(0.0)
 140         self._status.set_active(True)
 141         (seqnum, root_hash, IV, segsize, datalength, k, N, prefix,
 142          offsets_tuple) = self.verinfo
 143         self._status.set_size(datalength)
 144         self._status.set_encoding(k, N)
 145         self.readers = {}
 146         self._pause_deferred = None
 147         self._offset = None
 148         self._read_length = None
 149         self.log("got seqnum %d" % self.verinfo[0])
 150
 151
 152     def get_status(self):
 153         return self._status
 154
 155     def log(self, *args, **kwargs):
 156         if "parent" not in kwargs:
 157             kwargs["parent"] = self._log_number
 158         if "facility" not in kwargs:
 159             kwargs["facility"] = "tahoe.mutable.retrieve"
 160         return log.msg(*args, **kwargs)
 161
 162     def _set_current_status(self, state):
 163         seg = "%d/%d" % (self._current_segment, self._last_segment)
 164         self._status.set_status("segment %s (%s)" % (seg, state))
 165
 166     ###################
 167     # IPushProducer
 168
 169     def pauseProducing(self):
 170         """
 171         I am called by my download target if we have produced too much
 172         data for it to handle. I make the downloader stop producing new
 173         data until my resumeProducing method is called.
 174         """
 175         if self._pause_deferred is not None:
 176             return
 177
 178         # fired when the download is unpaused.
 179         self._old_status = self._status.get_status()
 180         self._set_current_status("paused")
 181
 182         self._pause_deferred = defer.Deferred()
 183
 184
 185     def resumeProducing(self):
 186         """
 187         I am called by my download target once it is ready to begin
 188         receiving data again.
 189         """
 190         if self._pause_deferred is None:
 191             return
 192
 193         p = self._pause_deferred
 194         self._pause_deferred = None
 195         self._status.set_status(self._old_status)
 196
 197         eventually(p.callback, None)
 198
 199
 200     def _check_for_paused(self, res):
 201         """
 202         I am called just before a write to the consumer. I return a
 203         Deferred that eventually fires with the data that is to be
 204         written to the consumer. If the download has not been paused,
 205         the Deferred fires immediately. Otherwise, the Deferred fires
 206         when the downloader is unpaused.
 207         """
 208         if self._pause_deferred is not None:
 209             d = defer.Deferred()
 210             self._pause_deferred.addCallback(lambda ignored: d.callback(res))
 211             return d
 212         return defer.succeed(res)
 213
 214
 215     def download(self, consumer=None, offset=0, size=None):
 216         assert IConsumer.providedBy(consumer) or self._verify
 217
 218         if consumer:
 219             self._consumer = consumer
 220             # we provide IPushProducer, so streaming=True, per
 221             # IConsumer.
 222             self._consumer.registerProducer(self, streaming=True)
 223
 224         self._done_deferred = defer.Deferred()
 225         self._offset = offset
 226         self._read_length = size
 227         self._setup_download()
 228         self._setup_encoding_parameters()
 229         self.log("starting download")
 230         self._started_fetching = time.time()
 231         # The download process beyond this is a state machine.
 232         # _add_active_peers will select the peers that we want to use
 233         # for the download, and then attempt to start downloading. After
 234         # each segment, it will check for doneness, reacting to broken
 235         # peers and corrupt shares as necessary. If it runs out of good
 236         # peers before downloading all of the segments, _done_deferred
 237         # will errback.  Otherwise, it will eventually callback with the
 238         # contents of the mutable file.
 239         self.loop()
 240         return self._done_deferred
 241
 242     def loop(self):
 243         d = fireEventually(None) # avoid #237 recursion limit problem
 244         d.addCallback(lambda ign: self._activate_enough_peers())
 245         d.addCallback(lambda ign: self._download_current_segment())
 246         # when we're done, _download_current_segment will call _done. If we
 247         # aren't, it will call loop() again.
 248         d.addErrback(self._error)
 249
 250     def _setup_download(self):
 251         self._started = time.time()
 252         self._status.set_status("Retrieving Shares")
 253
 254         # how many shares do we need?
 255         (seqnum,
 256          root_hash,
 257          IV,
 258          segsize,
 259          datalength,
 260          k,
 261          N,
 262          prefix,
 263          offsets_tuple) = self.verinfo
 264
 265         # first, which servers can we use?
 266         versionmap = self.servermap.make_versionmap()
 267         shares = versionmap[self.verinfo]
 268         # this sharemap is consumed as we decide to send requests
 269         self.remaining_sharemap = DictOfSets()
 270         for (shnum, peerid, timestamp) in shares:
 271             self.remaining_sharemap.add(shnum, peerid)
 272             # If the servermap update fetched anything, it fetched at least 1
 273             # KiB, so we ask for that much.
 274             # TODO: Change the cache methods to allow us to fetch all of the
 275             # data that they have, then change this method to do that.
 276             any_cache = self._node._read_from_cache(self.verinfo, shnum,
 277                                                     0, 1000)
 278             ss = self.servermap.connections[peerid]
 279             reader = MDMFSlotReadProxy(ss,
 280                                        self._storage_index,
 281                                        shnum,
 282                                        any_cache)
 283             reader.peerid = peerid
 284             self.readers[shnum] = reader
 285         assert len(self.remaining_sharemap) >= k
 286
 287         self.shares = {} # maps shnum to validated blocks
 288         self._active_readers = [] # list of active readers for this dl.
 289         self._validated_readers = set() # set of readers that we have
 290                                         # validated the prefix of
 291         self._block_hash_trees = {} # shnum => hashtree
 292
 293         # We need one share hash tree for the entire file; its leaves
 294         # are the roots of the block hash trees for the shares that
 295         # comprise it, and its root is in the verinfo.
 296         self.share_hash_tree = hashtree.IncompleteHashTree(N)
 297         self.share_hash_tree.set_hashes({0: root_hash})
 298
 299     def decode(self, blocks_and_salts, segnum):
 300         """
 301         I am a helper method that the mutable file update process uses
 302         as a shortcut to decode and decrypt the segments that it needs
 303         to fetch in order to perform a file update. I take in a
 304         collection of blocks and salts, and pick some of those to make a
 305         segment with. I return the plaintext associated with that
 306         segment.
 307         """
 308         # shnum => block hash tree. Unused, but setup_encoding_parameters will
 309         # want to set this.
 310         self._block_hash_trees = None
 311         self._setup_encoding_parameters()
 312
 313         # This is the form expected by decode.
 314         blocks_and_salts = blocks_and_salts.items()
 315         blocks_and_salts = [(True, [d]) for d in blocks_and_salts]
 316
 317         d = self._decode_blocks(blocks_and_salts, segnum)
 318         d.addCallback(self._decrypt_segment)
 319         return d
 320
 321
 322     def _setup_encoding_parameters(self):
 323         """
 324         I set up the encoding parameters, including k, n, the number
 325         of segments associated with this file, and the segment decoders.
 326         """
 327         (seqnum,
 328          root_hash,
 329          IV,
 330          segsize,
 331          datalength,
 332          k,
 333          n,
 334          known_prefix,
 335          offsets_tuple) = self.verinfo
 336         self._required_shares = k
 337         self._total_shares = n
 338         self._segment_size = segsize
 339         self._data_length = datalength
 340
 341         if not IV:
 342             self._version = MDMF_VERSION
 343         else:
 344             self._version = SDMF_VERSION
 345
 346         if datalength and segsize:
 347             self._num_segments = mathutil.div_ceil(datalength, segsize)
 348             self._tail_data_size = datalength % segsize
 349         else:
 350             self._num_segments = 0
 351             self._tail_data_size = 0
 352
 353         self._segment_decoder = codec.CRSDecoder()
 354         self._segment_decoder.set_params(segsize, k, n)
 355
 356         if  not self._tail_data_size:
 357             self._tail_data_size = segsize
 358
 359         self._tail_segment_size = mathutil.next_multiple(self._tail_data_size,
 360                                                          self._required_shares)
 361         if self._tail_segment_size == self._segment_size:
 362             self._tail_decoder = self._segment_decoder
 363         else:
 364             self._tail_decoder = codec.CRSDecoder()
 365             self._tail_decoder.set_params(self._tail_segment_size,
 366                                           self._required_shares,
 367                                           self._total_shares)
 368
 369         self.log("got encoding parameters: "
 370                  "k: %d "
 371                  "n: %d "
 372                  "%d segments of %d bytes each (%d byte tail segment)" % \
 373                  (k, n, self._num_segments, self._segment_size,
 374                   self._tail_segment_size))
 375
 376         if self._block_hash_trees is not None:
 377             for i in xrange(self._total_shares):
 378                 # So we don't have to do this later.
 379                 self._block_hash_trees[i] = hashtree.IncompleteHashTree(self._num_segments)
 380
 381         # Our last task is to tell the downloader where to start and
 382         # where to stop. We use three parameters for that:
 383         #   - self._start_segment: the segment that we need to start
 384         #     downloading from.
 385         #   - self._current_segment: the next segment that we need to
 386         #     download.
 387         #   - self._last_segment: The last segment that we were asked to
 388         #     download.
 389         #
 390         #  We say that the download is complete when
 391         #  self._current_segment > self._last_segment. We use
 392         #  self._start_segment and self._last_segment to know when to
 393         #  strip things off of segments, and how much to strip.
 394         if self._offset:
 395             self.log("got offset: %d" % self._offset)
 396             # our start segment is the first segment containing the
 397             # offset we were given.
 398             start = self._offset // self._segment_size
 399
 400             assert start < self._num_segments
 401             self._start_segment = start
 402             self.log("got start segment: %d" % self._start_segment)
 403         else:
 404             self._start_segment = 0
 405
 406
 407         # If self._read_length is None, then we want to read the whole
 408         # file. Otherwise, we want to read only part of the file, and
 409         # need to figure out where to stop reading.
 410         if self._read_length is not None:
 411             # our end segment is the last segment containing part of the
 412             # segment that we were asked to read.
 413             self.log("got read length %d" % self._read_length)
 414             if self._read_length != 0:
 415                 end_data = self._offset + self._read_length
 416
 417                 # We don't actually need to read the byte at end_data,
 418                 # but the one before it.
 419                 end = (end_data - 1) // self._segment_size
 420
 421                 assert end < self._num_segments
 422                 self._last_segment = end
 423             else:
 424                 self._last_segment = self._start_segment
 425             self.log("got end segment: %d" % self._last_segment)
 426         else:
 427             self._last_segment = self._num_segments - 1
 428
 429         self._current_segment = self._start_segment
 430
 431     def _activate_enough_peers(self):
 432         """
 433         I populate self._active_readers with enough active readers to
 434         retrieve the contents of this mutable file. I am called before
 435         downloading starts, and (eventually) after each validation
 436         error, connection error, or other problem in the download.
 437         """
 438         # TODO: It would be cool to investigate other heuristics for
 439         # reader selection. For instance, the cost (in time the user
 440         # spends waiting for their file) of selecting a really slow peer
 441         # that happens to have a primary share is probably more than
 442         # selecting a really fast peer that doesn't have a primary
 443         # share. Maybe the servermap could be extended to provide this
 444         # information; it could keep track of latency information while
 445         # it gathers more important data, and then this routine could
 446         # use that to select active readers.
 447         #
 448         # (these and other questions would be easier to answer with a
 449         #  robust, configurable tahoe-lafs simulator, which modeled node
 450         #  failures, differences in node speed, and other characteristics
 451         #  that we expect storage servers to have.  You could have
 452         #  presets for really stable grids (like allmydata.com),
 453         #  friendnets, make it easy to configure your own settings, and
 454         #  then simulate the effect of big changes on these use cases
 455         #  instead of just reasoning about what the effect might be. Out
 456         #  of scope for MDMF, though.)
 457
 458         # We need at least self._required_shares readers to download a
 459         # segment. If we're verifying, we need all shares.
 460         if self._verify:
 461             needed = self._total_shares
 462         else:
 463             needed = self._required_shares
 464         # XXX: Why don't format= log messages work here?
 465         self.log("adding %d peers to the active peers list" % needed)
 466
 467         if len(self._active_readers) >= needed:
 468             # enough shares are active
 469             return
 470
 471         more = needed - len(self._active_readers)
 472         known_shnums = set(self.remaining_sharemap.keys())
 473         used_shnums = set([r.shnum for r in self._active_readers])
 474         unused_shnums = known_shnums - used_shnums
 475         # We favor lower numbered shares, since FEC is faster with
 476         # primary shares than with other shares, and lower-numbered
 477         # shares are more likely to be primary than higher numbered
 478         # shares.
 479         new_shnums = sorted(unused_shnums)[:more]
 480         if len(new_shnums) < more and not self._verify:
 481             # We don't have enough readers to retrieve the file; fail.
 482             self._raise_notenoughshareserror()
 483
 484         for shnum in new_shnums:
 485             reader = self.readers[shnum]
 486             self._active_readers.append(reader)
 487             self._validated_readers.add(reader)
 488             self.log("added reader for share %d" % shnum)
 489             # Each time we validate a reader, we check to see if we need the
 490             # private key. If we do, we politely ask for it and then continue
 491             # computing. If we find that we haven't gotten it at the end of
 492             # segment decoding, then we'll take more drastic measures.
 493             if self._need_privkey and not self._node.is_readonly():
 494                 d = reader.get_encprivkey()
 495                 d.addCallback(self._try_to_validate_privkey, reader)
 496                 # XXX: don't just drop the Deferred. We need error-reporting
 497                 # but not flow-control here.
 498         assert len(self._active_readers) >= self._required_shares
 499
 500     def _try_to_validate_prefix(self, prefix, reader):
 501         """
 502         I check that the prefix returned by a candidate server for
 503         retrieval matches the prefix that the servermap knows about
 504         (and, hence, the prefix that was validated earlier). If it does,
 505         I return True, which means that I approve of the use of the
 506         candidate server for segment retrieval. If it doesn't, I return
 507         False, which means that another server must be chosen.
 508         """
 509         (seqnum,
 510          root_hash,
 511          IV,
 512          segsize,
 513          datalength,
 514          k,
 515          N,
 516          known_prefix,
 517          offsets_tuple) = self.verinfo
 518         if known_prefix != prefix:
 519             self.log("prefix from share %d doesn't match" % reader.shnum)
 520             raise UncoordinatedWriteError("Mismatched prefix -- this could "
 521                                           "indicate an uncoordinated write")
 522         # Otherwise, we're okay -- no issues.
 523
 524
 525     def _remove_reader(self, reader):
 526         """
 527         At various points, we will wish to remove a peer from
 528         consideration and/or use. These include, but are not necessarily
 529         limited to:
 530
 531             - A connection error.
 532             - A mismatched prefix (that is, a prefix that does not match
 533               our conception of the version information string).
 534             - A failing block hash, salt hash, or share hash, which can
 535               indicate disk failure/bit flips, or network trouble.
 536
 537         This method will do that. I will make sure that the
 538         (shnum,reader) combination represented by my reader argument is
 539         not used for anything else during this download. I will not
 540         advise the reader of any corruption, something that my callers
 541         may wish to do on their own.
 542         """
 543         # TODO: When you're done writing this, see if this is ever
 544         # actually used for something that _mark_bad_share isn't. I have
 545         # a feeling that they will be used for very similar things, and
 546         # that having them both here is just going to be an epic amount
 547         # of code duplication.
 548         #
 549         # (well, okay, not epic, but meaningful)
 550         self.log("removing reader %s" % reader)
 551         # Remove the reader from _active_readers
 552         self._active_readers.remove(reader)
 553         # TODO: self.readers.remove(reader)?
 554         for shnum in list(self.remaining_sharemap.keys()):
 555             self.remaining_sharemap.discard(shnum, reader.peerid)
 556
 557
 558     def _mark_bad_share(self, reader, f):
 559         """
 560         I mark the (peerid, shnum) encapsulated by my reader argument as
 561         a bad share, which means that it will not be used anywhere else.
 562
 563         There are several reasons to want to mark something as a bad
 564         share. These include:
 565
 566             - A connection error to the peer.
 567             - A mismatched prefix (that is, a prefix that does not match
 568               our local conception of the version information string).
 569             - A failing block hash, salt hash, share hash, or other
 570               integrity check.
 571
 572         This method will ensure that readers that we wish to mark bad
 573         (for these reasons or other reasons) are not used for the rest
 574         of the download. Additionally, it will attempt to tell the
 575         remote peer (with no guarantee of success) that its share is
 576         corrupt.
 577         """
 578         self.log("marking share %d on server %s as bad" % \
 579                  (reader.shnum, reader))
 580         prefix = self.verinfo[-2]
 581         self.servermap.mark_bad_share(reader.peerid,
 582                                       reader.shnum,
 583                                       prefix)
 584         self._remove_reader(reader)
 585         self._bad_shares.add((reader.peerid, reader.shnum, f))
 586         self._status.problems[reader.peerid] = f
 587         self._last_failure = f
 588         self.notify_server_corruption(reader.peerid, reader.shnum,
 589                                       str(f.value))
 590
 591
 592     def _download_current_segment(self):
 593         """
 594         I download, validate, decode, decrypt, and assemble the segment
 595         that this Retrieve is currently responsible for downloading.
 596         """
 597         assert len(self._active_readers) >= self._required_shares
 598         if self._current_segment > self._last_segment:
 599             # No more segments to download, we're done.
 600             self.log("got plaintext, done")
 601             return self._done()
 602         self.log("on segment %d of %d" %
 603                  (self._current_segment + 1, self._num_segments))
 604         d = self._process_segment(self._current_segment)
 605         d.addCallback(lambda ign: self.loop())
 606         return d
 607
 608     def _process_segment(self, segnum):
 609         """
 610         I download, validate, decode, and decrypt one segment of the
 611         file that this Retrieve is retrieving. This means coordinating
 612         the process of getting k blocks of that file, validating them,
 613         assembling them into one segment with the decoder, and then
 614         decrypting them.
 615         """
 616         self.log("processing segment %d" % segnum)
 617
 618         # TODO: The old code uses a marker. Should this code do that
 619         # too? What did the Marker do?
 620         assert len(self._active_readers) >= self._required_shares
 621
 622         # We need to ask each of our active readers for its block and
 623         # salt. We will then validate those. If validation is
 624         # successful, we will assemble the results into plaintext.
 625         ds = []
 626         for reader in self._active_readers:
 627             started = time.time()
 628             d = reader.get_block_and_salt(segnum)
 629             d2 = self._get_needed_hashes(reader, segnum)
 630             dl = defer.DeferredList([d, d2], consumeErrors=True)
 631             dl.addCallback(self._validate_block, segnum, reader, started)
 632             dl.addErrback(self._validation_or_decoding_failed, [reader])
 633             ds.append(dl)
 634         dl = defer.DeferredList(ds)
 635         if self._verify:
 636             dl.addCallback(lambda ignored: "")
 637             dl.addCallback(self._set_segment)
 638         else:
 639             dl.addCallback(self._maybe_decode_and_decrypt_segment, segnum)
 640         return dl
 641
 642
 643     def _maybe_decode_and_decrypt_segment(self, blocks_and_salts, segnum):
 644         """
 645         I take the results of fetching and validating the blocks from a
 646         callback chain in another method. If the results are such that
 647         they tell me that validation and fetching succeeded without
 648         incident, I will proceed with decoding and decryption.
 649         Otherwise, I will do nothing.
 650         """
 651         self.log("trying to decode and decrypt segment %d" % segnum)
 652         failures = False
 653         for block_and_salt in blocks_and_salts:
 654             if not block_and_salt[0] or block_and_salt[1] == None:
 655                 self.log("some validation operations failed; not proceeding")
 656                 failures = True
 657                 break
 658         if not failures:
 659             self.log("everything looks ok, building segment %d" % segnum)
 660             d = self._decode_blocks(blocks_and_salts, segnum)
 661             d.addCallback(self._decrypt_segment)
 662             d.addErrback(self._validation_or_decoding_failed,
 663                          self._active_readers)
 664             # check to see whether we've been paused before writing
 665             # anything.
 666             d.addCallback(self._check_for_paused)
 667             d.addCallback(self._set_segment)
 668             return d
 669         else:
 670             return defer.succeed(None)
 671
 672
 673     def _set_segment(self, segment):
 674         """
 675         Given a plaintext segment, I register that segment with the
 676         target that is handling the file download.
 677         """
 678         self.log("got plaintext for segment %d" % self._current_segment)
 679         if self._current_segment == self._start_segment:
 680             # We're on the first segment. It's possible that we want
 681             # only some part of the end of this segment, and that we
 682             # just downloaded the whole thing to get that part. If so,
 683             # we need to account for that and give the reader just the
 684             # data that they want.
 685             n = self._offset % self._segment_size
 686             self.log("stripping %d bytes off of the first segment" % n)
 687             self.log("original segment length: %d" % len(segment))
 688             segment = segment[n:]
 689             self.log("new segment length: %d" % len(segment))
 690
 691         if self._current_segment == self._last_segment and self._read_length is not None:
 692             # We're on the last segment. It's possible that we only want
 693             # part of the beginning of this segment, and that we
 694             # downloaded the whole thing anyway. Make sure to give the
 695             # caller only the portion of the segment that they want to
 696             # receive.
 697             extra = self._read_length
 698             if self._start_segment != self._last_segment:
 699                 extra -= self._segment_size - \
 700                             (self._offset % self._segment_size)
 701             extra %= self._segment_size
 702             self.log("original segment length: %d" % len(segment))
 703             segment = segment[:extra]
 704             self.log("new segment length: %d" % len(segment))
 705             self.log("only taking %d bytes of the last segment" % extra)
 706
 707         if not self._verify:
 708             self._consumer.write(segment)
 709         else:
 710             # we don't care about the plaintext if we are doing a verify.
 711             segment = None
 712         self._current_segment += 1
 713
 714
 715     def _validation_or_decoding_failed(self, f, readers):
 716         """
 717         I am called when a block or a salt fails to correctly validate, or when
 718         the decryption or decoding operation fails for some reason.  I react to
 719         this failure by notifying the remote server of corruption, and then
 720         removing the remote peer from further activity.
 721         """
 722         assert isinstance(readers, list)
 723         bad_shnums = [reader.shnum for reader in readers]
 724
 725         self.log("validation or decoding failed on share(s) %s, peer(s) %s "
 726                  ", segment %d: %s" % \
 727                  (bad_shnums, readers, self._current_segment, str(f)))
 728         for reader in readers:
 729             self._mark_bad_share(reader, f)
 730         return
 731
 732
 733     def _validate_block(self, results, segnum, reader, started):
 734         """
 735         I validate a block from one share on a remote server.
 736         """
 737         # Grab the part of the block hash tree that is necessary to
 738         # validate this block, then generate the block hash root.
 739         self.log("validating share %d for segment %d" % (reader.shnum,
 740                                                              segnum))
 741         elapsed = time.time() - started
 742         self._status.add_fetch_timing(reader.peerid, elapsed)
 743         self._set_current_status("validating blocks")
 744         # Did we fail to fetch either of the things that we were
 745         # supposed to? Fail if so.
 746         if not results[0][0] and results[1][0]:
 747             # handled by the errback handler.
 748
 749             # These all get batched into one query, so the resulting
 750             # failure should be the same for all of them, so we can just
 751             # use the first one.
 752             assert isinstance(results[0][1], failure.Failure)
 753
 754             f = results[0][1]
 755             raise CorruptShareError(reader.peerid,
 756                                     reader.shnum,
 757                                     "Connection error: %s" % str(f))
 758
 759         block_and_salt, block_and_sharehashes = results
 760         block, salt = block_and_salt[1]
 761         blockhashes, sharehashes = block_and_sharehashes[1]
 762
 763         blockhashes = dict(enumerate(blockhashes[1]))
 764         self.log("the reader gave me the following blockhashes: %s" % \
 765                  blockhashes.keys())
 766         self.log("the reader gave me the following sharehashes: %s" % \
 767                  sharehashes[1].keys())
 768         bht = self._block_hash_trees[reader.shnum]
 769
 770         if bht.needed_hashes(segnum, include_leaf=True):
 771             try:
 772                 bht.set_hashes(blockhashes)
 773             except (hashtree.BadHashError, hashtree.NotEnoughHashesError, \
 774                     IndexError), e:
 775                 raise CorruptShareError(reader.peerid,
 776                                         reader.shnum,
 777                                         "block hash tree failure: %s" % e)
 778
 779         if self._version == MDMF_VERSION:
 780             blockhash = hashutil.block_hash(salt + block)
 781         else:
 782             blockhash = hashutil.block_hash(block)
 783         # If this works without an error, then validation is
 784         # successful.
 785         try:
 786            bht.set_hashes(leaves={segnum: blockhash})
 787         except (hashtree.BadHashError, hashtree.NotEnoughHashesError, \
 788                 IndexError), e:
 789             raise CorruptShareError(reader.peerid,
 790                                     reader.shnum,
 791                                     "block hash tree failure: %s" % e)
 792
 793         # Reaching this point means that we know that this segment
 794         # is correct. Now we need to check to see whether the share
 795         # hash chain is also correct.
 796         # SDMF wrote share hash chains that didn't contain the
 797         # leaves, which would be produced from the block hash tree.
 798         # So we need to validate the block hash tree first. If
 799         # successful, then bht[0] will contain the root for the
 800         # shnum, which will be a leaf in the share hash tree, which
 801         # will allow us to validate the rest of the tree.
 802         if self.share_hash_tree.needed_hashes(reader.shnum,
 803                                               include_leaf=True) or \
 804                                               self._verify:
 805             try:
 806                 self.share_hash_tree.set_hashes(hashes=sharehashes[1],
 807                                             leaves={reader.shnum: bht[0]})
 808             except (hashtree.BadHashError, hashtree.NotEnoughHashesError, \
 809                     IndexError), e:
 810                 raise CorruptShareError(reader.peerid,
 811                                         reader.shnum,
 812                                         "corrupt hashes: %s" % e)
 813
 814         self.log('share %d is valid for segment %d' % (reader.shnum,
 815                                                        segnum))
 816         return {reader.shnum: (block, salt)}
 817
 818
 819     def _get_needed_hashes(self, reader, segnum):
 820         """
 821         I get the hashes needed to validate segnum from the reader, then return
 822         to my caller when this is done.
 823         """
 824         bht = self._block_hash_trees[reader.shnum]
 825         needed = bht.needed_hashes(segnum, include_leaf=True)
 826         # The root of the block hash tree is also a leaf in the share
 827         # hash tree. So we don't need to fetch it from the remote
 828         # server. In the case of files with one segment, this means that
 829         # we won't fetch any block hash tree from the remote server,
 830         # since the hash of each share of the file is the entire block
 831         # hash tree, and is a leaf in the share hash tree. This is fine,
 832         # since any share corruption will be detected in the share hash
 833         # tree.
 834         #needed.discard(0)
 835         self.log("getting blockhashes for segment %d, share %d: %s" % \
 836                  (segnum, reader.shnum, str(needed)))
 837         d1 = reader.get_blockhashes(needed, force_remote=True)
 838         if self.share_hash_tree.needed_hashes(reader.shnum):
 839             need = self.share_hash_tree.needed_hashes(reader.shnum)
 840             self.log("also need sharehashes for share %d: %s" % (reader.shnum,
 841                                                                  str(need)))
 842             d2 = reader.get_sharehashes(need, force_remote=True)
 843         else:
 844             d2 = defer.succeed({}) # the logic in the next method
 845                                    # expects a dict
 846         dl = defer.DeferredList([d1, d2], consumeErrors=True)
 847         return dl
 848
 849
 850     def _decode_blocks(self, blocks_and_salts, segnum):
 851         """
 852         I take a list of k blocks and salts, and decode that into a
 853         single encrypted segment.
 854         """
 855         d = {}
 856         # We want to merge our dictionaries to the form
 857         # {shnum: blocks_and_salts}
 858         #
 859         # The dictionaries come from validate block that way, so we just
 860         # need to merge them.
 861         for block_and_salt in blocks_and_salts:
 862             d.update(block_and_salt[1])
 863
 864         # All of these blocks should have the same salt; in SDMF, it is
 865         # the file-wide IV, while in MDMF it is the per-segment salt. In
 866         # either case, we just need to get one of them and use it.
 867         #
 868         # d.items()[0] is like (shnum, (block, salt))
 869         # d.items()[0][1] is like (block, salt)
 870         # d.items()[0][1][1] is the salt.
 871         salt = d.items()[0][1][1]
 872         # Next, extract just the blocks from the dict. We'll use the
 873         # salt in the next step.
 874         share_and_shareids = [(k, v[0]) for k, v in d.items()]
 875         d2 = dict(share_and_shareids)
 876         shareids = []
 877         shares = []
 878         for shareid, share in d2.items():
 879             shareids.append(shareid)
 880             shares.append(share)
 881
 882         self._set_current_status("decoding")
 883         started = time.time()
 884         assert len(shareids) >= self._required_shares, len(shareids)
 885         # zfec really doesn't want extra shares
 886         shareids = shareids[:self._required_shares]
 887         shares = shares[:self._required_shares]
 888         self.log("decoding segment %d" % segnum)
 889         if segnum == self._num_segments - 1:
 890             d = defer.maybeDeferred(self._tail_decoder.decode, shares, shareids)
 891         else:
 892             d = defer.maybeDeferred(self._segment_decoder.decode, shares, shareids)
 893         def _process(buffers):
 894             segment = "".join(buffers)
 895             self.log(format="now decoding segment %(segnum)s of %(numsegs)s",
 896                      segnum=segnum,
 897                      numsegs=self._num_segments,
 898                      level=log.NOISY)
 899             self.log(" joined length %d, datalength %d" %
 900                      (len(segment), self._data_length))
 901             if segnum == self._num_segments - 1:
 902                 size_to_use = self._tail_data_size
 903             else:
 904                 size_to_use = self._segment_size
 905             segment = segment[:size_to_use]
 906             self.log(" segment len=%d" % len(segment))
 907             self._status.accumulate_decode_time(time.time() - started)
 908             return segment, salt
 909         d.addCallback(_process)
 910         return d
 911
 912
 913     def _decrypt_segment(self, segment_and_salt):
 914         """
 915         I take a single segment and its salt, and decrypt it. I return
 916         the plaintext of the segment that is in my argument.
 917         """
 918         segment, salt = segment_and_salt
 919         self._set_current_status("decrypting")
 920         self.log("decrypting segment %d" % self._current_segment)
 921         started = time.time()
 922         key = hashutil.ssk_readkey_data_hash(salt, self._node.get_readkey())
 923         decryptor = AES(key)
 924         plaintext = decryptor.process(segment)
 925         self._status.accumulate_decrypt_time(time.time() - started)
 926         return plaintext
 927
 928
 929     def notify_server_corruption(self, peerid, shnum, reason):
 930         ss = self.servermap.connections[peerid]
 931         ss.callRemoteOnly("advise_corrupt_share",
 932                           "mutable", self._storage_index, shnum, reason)
 933
 934
 935     def _try_to_validate_privkey(self, enc_privkey, reader):
 936         alleged_privkey_s = self._node._decrypt_privkey(enc_privkey)
 937         alleged_writekey = hashutil.ssk_writekey_hash(alleged_privkey_s)
 938         if alleged_writekey != self._node.get_writekey():
 939             self.log("invalid privkey from %s shnum %d" %
 940                      (reader, reader.shnum),
 941                      level=log.WEIRD, umid="YIw4tA")
 942             if self._verify:
 943                 self.servermap.mark_bad_share(reader.peerid, reader.shnum,
 944                                               self.verinfo[-2])
 945                 e = CorruptShareError(reader.peerid,
 946                                       reader.shnum,
 947                                       "invalid privkey")
 948                 f = failure.Failure(e)
 949                 self._bad_shares.add((reader.peerid, reader.shnum, f))
 950             return
 951
 952         # it's good
 953         self.log("got valid privkey from shnum %d on reader %s" %
 954                  (reader.shnum, reader))
 955         privkey = rsa.create_signing_key_from_string(alleged_privkey_s)
 956         self._node._populate_encprivkey(enc_privkey)
 957         self._node._populate_privkey(privkey)
 958         self._need_privkey = False
 959
 960
 961
 962     def _done(self):
 963         """
 964         I am called by _download_current_segment when the download process
 965         has finished successfully. After making some useful logging
 966         statements, I return the decrypted contents to the owner of this
 967         Retrieve object through self._done_deferred.
 968         """
 969         self._running = False
 970         self._status.set_active(False)
 971         now = time.time()
 972         self._status.timings['total'] = now - self._started
 973         self._status.timings['fetch'] = now - self._started_fetching
 974         self._status.set_status("Finished")
 975         self._status.set_progress(1.0)
 976
 977         # remember the encoding parameters, use them again next time
 978         (seqnum, root_hash, IV, segsize, datalength, k, N, prefix,
 979          offsets_tuple) = self.verinfo
 980         self._node._populate_required_shares(k)
 981         self._node._populate_total_shares(N)
 982
 983         if self._verify:
 984             ret = list(self._bad_shares)
 985             self.log("done verifying, found %d bad shares" % len(ret))
 986         else:
 987             # TODO: upload status here?
 988             ret = self._consumer
 989             self._consumer.unregisterProducer()
 990         eventually(self._done_deferred.callback, ret)
 991
 992
 993     def _raise_notenoughshareserror(self):
 994         """
 995         I am called by _activate_enough_peers when there are not enough
 996         active peers left to complete the download. After making some
 997         useful logging statements, I throw an exception to that effect
 998         to the caller of this Retrieve object through
 999         self._done_deferred.
1000         """
1001
1002         format = ("ran out of peers: "
1003                   "have %(have)d of %(total)d segments "
1004                   "found %(bad)d bad shares "
1005                   "encoding %(k)d-of-%(n)d")
1006         args = {"have": self._current_segment,
1007                 "total": self._num_segments,
1008                 "need": self._last_segment,
1009                 "k": self._required_shares,
1010                 "n": self._total_shares,
1011                 "bad": len(self._bad_shares)}
1012         raise NotEnoughSharesError("%s, last failure: %s" %
1013                                    (format % args, str(self._last_failure)))
1014
1015     def _error(self, f):
1016         # all errors, including NotEnoughSharesError, land here
1017         self._running = False
1018         self._status.set_active(False)
1019         now = time.time()
1020         self._status.timings['total'] = now - self._started
1021         self._status.timings['fetch'] = now - self._started_fetching
1022         self._status.set_status("Failed")
1023         eventually(self._done_deferred.errback, f)