src/allmydata/download.py

   1
   2 import os, random, weakref
   3 from zope.interface import implements
   4 from twisted.internet import defer
   5 from twisted.internet.interfaces import IPushProducer, IConsumer
   6 from twisted.application import service
   7 from foolscap.eventual import eventually
   8
   9 from allmydata.util import base32, mathutil, hashutil, log
  10 from allmydata.util.assertutil import _assert
  11 from allmydata import codec, hashtree, storage, uri
  12 from allmydata.interfaces import IDownloadTarget, IDownloader, IFileURI, \
  13      IDownloadStatus
  14 from allmydata.encode import NotEnoughPeersError
  15 from pycryptopp.cipher.aes import AES
  16
  17 class HaveAllPeersError(Exception):
  18     # we use this to jump out of the loop
  19     pass
  20
  21 class BadURIExtensionHashValue(Exception):
  22     pass
  23 class BadPlaintextHashValue(Exception):
  24     pass
  25 class BadCrypttextHashValue(Exception):
  26     pass
  27
  28 class DownloadStopped(Exception):
  29     pass
  30
  31 class Output:
  32     def __init__(self, downloadable, key, total_length, log_parent,
  33                  download_status):
  34         self.downloadable = downloadable
  35         self._decryptor = AES(key)
  36         self._crypttext_hasher = hashutil.crypttext_hasher()
  37         self._plaintext_hasher = hashutil.plaintext_hasher()
  38         self.length = 0
  39         self.total_length = total_length
  40         self._segment_number = 0
  41         self._plaintext_hash_tree = None
  42         self._crypttext_hash_tree = None
  43         self._opened = False
  44         self._log_parent = log_parent
  45         self._status = download_status
  46         self._status.set_progress(0.0)
  47
  48     def log(self, *args, **kwargs):
  49         if "parent" not in kwargs:
  50             kwargs["parent"] = self._log_parent
  51         if "facility" not in kwargs:
  52             kwargs["facility"] = "download.output"
  53         return log.msg(*args, **kwargs)
  54
  55     def setup_hashtrees(self, plaintext_hashtree, crypttext_hashtree):
  56         self._plaintext_hash_tree = plaintext_hashtree
  57         self._crypttext_hash_tree = crypttext_hashtree
  58
  59     def write_segment(self, crypttext):
  60         self.length += len(crypttext)
  61         self._status.set_progress( float(self.length) / self.total_length )
  62
  63         # memory footprint: 'crypttext' is the only segment_size usage
  64         # outstanding. While we decrypt it into 'plaintext', we hit
  65         # 2*segment_size.
  66         self._crypttext_hasher.update(crypttext)
  67         if self._crypttext_hash_tree:
  68             ch = hashutil.crypttext_segment_hasher()
  69             ch.update(crypttext)
  70             crypttext_leaves = {self._segment_number: ch.digest()}
  71             self.log(format="crypttext leaf hash (%(bytes)sB) [%(segnum)d] is %(hash)s",
  72                      bytes=len(crypttext),
  73                      segnum=self._segment_number, hash=base32.b2a(ch.digest()),
  74                      level=log.NOISY)
  75             self._crypttext_hash_tree.set_hashes(leaves=crypttext_leaves)
  76
  77         plaintext = self._decryptor.process(crypttext)
  78         del crypttext
  79
  80         # now we're back down to 1*segment_size.
  81
  82         self._plaintext_hasher.update(plaintext)
  83         if self._plaintext_hash_tree:
  84             ph = hashutil.plaintext_segment_hasher()
  85             ph.update(plaintext)
  86             plaintext_leaves = {self._segment_number: ph.digest()}
  87             self.log(format="plaintext leaf hash (%(bytes)sB) [%(segnum)d] is %(hash)s",
  88                      bytes=len(plaintext),
  89                      segnum=self._segment_number, hash=base32.b2a(ph.digest()),
  90                      level=log.NOISY)
  91             self._plaintext_hash_tree.set_hashes(leaves=plaintext_leaves)
  92
  93         self._segment_number += 1
  94         # We're still at 1*segment_size. The Downloadable is responsible for
  95         # any memory usage beyond this.
  96         if not self._opened:
  97             self._opened = True
  98             self.downloadable.open(self.total_length)
  99         self.downloadable.write(plaintext)
 100
 101     def fail(self, why):
 102         # this is really unusual, and deserves maximum forensics
 103         self.log("download failed!", failure=why, level=log.SCARY)
 104         self.downloadable.fail(why)
 105
 106     def close(self):
 107         self.crypttext_hash = self._crypttext_hasher.digest()
 108         self.plaintext_hash = self._plaintext_hasher.digest()
 109         self.log("download finished, closing IDownloadable", level=log.NOISY)
 110         self.downloadable.close()
 111
 112     def finish(self):
 113         return self.downloadable.finish()
 114
 115 class ValidatedBucket:
 116     """I am a front-end for a remote storage bucket, responsible for
 117     retrieving and validating data from that bucket.
 118
 119     My get_block() method is used by BlockDownloaders.
 120     """
 121
 122     def __init__(self, sharenum, bucket,
 123                  share_hash_tree, roothash,
 124                  num_blocks):
 125         self.sharenum = sharenum
 126         self.bucket = bucket
 127         self._share_hash = None # None means not validated yet
 128         self.share_hash_tree = share_hash_tree
 129         self._roothash = roothash
 130         self.block_hash_tree = hashtree.IncompleteHashTree(num_blocks)
 131         self.started = False
 132
 133     def get_block(self, blocknum):
 134         if not self.started:
 135             d = self.bucket.start()
 136             def _started(res):
 137                 self.started = True
 138                 return self.get_block(blocknum)
 139             d.addCallback(_started)
 140             return d
 141
 142         # the first time we use this bucket, we need to fetch enough elements
 143         # of the share hash tree to validate it from our share hash up to the
 144         # hashroot.
 145         if not self._share_hash:
 146             d1 = self.bucket.get_share_hashes()
 147         else:
 148             d1 = defer.succeed([])
 149
 150         # we might need to grab some elements of our block hash tree, to
 151         # validate the requested block up to the share hash
 152         needed = self.block_hash_tree.needed_hashes(blocknum)
 153         if needed:
 154             # TODO: get fewer hashes, use get_block_hashes(needed)
 155             d2 = self.bucket.get_block_hashes()
 156         else:
 157             d2 = defer.succeed([])
 158
 159         d3 = self.bucket.get_block(blocknum)
 160
 161         d = defer.gatherResults([d1, d2, d3])
 162         d.addCallback(self._got_data, blocknum)
 163         return d
 164
 165     def _got_data(self, res, blocknum):
 166         sharehashes, blockhashes, blockdata = res
 167         blockhash = None # to make logging it safe
 168
 169         try:
 170             if not self._share_hash:
 171                 sh = dict(sharehashes)
 172                 sh[0] = self._roothash # always use our own root, from the URI
 173                 sht = self.share_hash_tree
 174                 if sht.get_leaf_index(self.sharenum) not in sh:
 175                     raise hashtree.NotEnoughHashesError
 176                 sht.set_hashes(sh)
 177                 self._share_hash = sht.get_leaf(self.sharenum)
 178
 179             blockhash = hashutil.block_hash(blockdata)
 180             #log.msg("checking block_hash(shareid=%d, blocknum=%d) len=%d "
 181             #        "%r .. %r: %s" %
 182             #        (self.sharenum, blocknum, len(blockdata),
 183             #         blockdata[:50], blockdata[-50:], base32.b2a(blockhash)))
 184
 185             # we always validate the blockhash
 186             bh = dict(enumerate(blockhashes))
 187             # replace blockhash root with validated value
 188             bh[0] = self._share_hash
 189             self.block_hash_tree.set_hashes(bh, {blocknum: blockhash})
 190
 191         except (hashtree.BadHashError, hashtree.NotEnoughHashesError):
 192             # log.WEIRD: indicates undetected disk/network error, or more
 193             # likely a programming error
 194             log.msg("hash failure in block=%d, shnum=%d on %s" %
 195                     (blocknum, self.sharenum, self.bucket))
 196             if self._share_hash:
 197                 log.msg(""" failure occurred when checking the block_hash_tree.
 198                 This suggests that either the block data was bad, or that the
 199                 block hashes we received along with it were bad.""")
 200             else:
 201                 log.msg(""" the failure probably occurred when checking the
 202                 share_hash_tree, which suggests that the share hashes we
 203                 received from the remote peer were bad.""")
 204             log.msg(" have self._share_hash: %s" % bool(self._share_hash))
 205             log.msg(" block length: %d" % len(blockdata))
 206             log.msg(" block hash: %s" % base32.b2a_or_none(blockhash))
 207             if len(blockdata) < 100:
 208                 log.msg(" block data: %r" % (blockdata,))
 209             else:
 210                 log.msg(" block data start/end: %r .. %r" %
 211                         (blockdata[:50], blockdata[-50:]))
 212             log.msg(" root hash: %s" % base32.b2a(self._roothash))
 213             log.msg(" share hash tree:\n" + self.share_hash_tree.dump())
 214             log.msg(" block hash tree:\n" + self.block_hash_tree.dump())
 215             lines = []
 216             for i,h in sorted(sharehashes):
 217                 lines.append("%3d: %s" % (i, base32.b2a_or_none(h)))
 218             log.msg(" sharehashes:\n" + "\n".join(lines) + "\n")
 219             lines = []
 220             for i,h in enumerate(blockhashes):
 221                 lines.append("%3d: %s" % (i, base32.b2a_or_none(h)))
 222             log.msg(" blockhashes:\n" + "\n".join(lines) + "\n")
 223             raise
 224
 225         # If we made it here, the block is good. If the hash trees didn't
 226         # like what they saw, they would have raised a BadHashError, causing
 227         # our caller to see a Failure and thus ignore this block (as well as
 228         # dropping this bucket).
 229         return blockdata
 230
 231
 232
 233 class BlockDownloader:
 234     """I am responsible for downloading a single block (from a single bucket)
 235     for a single segment.
 236
 237     I am a child of the SegmentDownloader.
 238     """
 239
 240     def __init__(self, vbucket, blocknum, parent):
 241         self.vbucket = vbucket
 242         self.blocknum = blocknum
 243         self.parent = parent
 244         self._log_number = self.parent.log("starting block %d" % blocknum)
 245
 246     def log(self, msg, parent=None):
 247         if parent is None:
 248             parent = self._log_number
 249         return self.parent.log(msg, parent=parent)
 250
 251     def start(self, segnum):
 252         lognum = self.log("get_block(segnum=%d)" % segnum)
 253         d = self.vbucket.get_block(segnum)
 254         d.addCallbacks(self._hold_block, self._got_block_error,
 255                        callbackArgs=(lognum,), errbackArgs=(lognum,))
 256         return d
 257
 258     def _hold_block(self, data, lognum):
 259         self.log("got block", parent=lognum)
 260         self.parent.hold_block(self.blocknum, data)
 261
 262     def _got_block_error(self, f, lognum):
 263         self.log("BlockDownloader[%d] got error: %s" % (self.blocknum, f),
 264                  parent=lognum)
 265         self.parent.bucket_failed(self.vbucket)
 266
 267 class SegmentDownloader:
 268     """I am responsible for downloading all the blocks for a single segment
 269     of data.
 270
 271     I am a child of the FileDownloader.
 272     """
 273
 274     def __init__(self, parent, segmentnumber, needed_shares):
 275         self.parent = parent
 276         self.segmentnumber = segmentnumber
 277         self.needed_blocks = needed_shares
 278         self.blocks = {} # k: blocknum, v: data
 279         self._log_number = self.parent.log("starting segment %d" %
 280                                            segmentnumber)
 281
 282     def log(self, msg, parent=None):
 283         if parent is None:
 284             parent = self._log_number
 285         return self.parent.log(msg, parent=parent)
 286
 287     def start(self):
 288         return self._download()
 289
 290     def _download(self):
 291         d = self._try()
 292         def _done(res):
 293             if len(self.blocks) >= self.needed_blocks:
 294                 # we only need self.needed_blocks blocks
 295                 # we want to get the smallest blockids, because they are
 296                 # more likely to be fast "primary blocks"
 297                 blockids = sorted(self.blocks.keys())[:self.needed_blocks]
 298                 blocks = []
 299                 for blocknum in blockids:
 300                     blocks.append(self.blocks[blocknum])
 301                 return (blocks, blockids)
 302             else:
 303                 return self._download()
 304         d.addCallback(_done)
 305         return d
 306
 307     def _try(self):
 308         # fill our set of active buckets, maybe raising NotEnoughPeersError
 309         active_buckets = self.parent._activate_enough_buckets()
 310         # Now we have enough buckets, in self.parent.active_buckets.
 311
 312         # in test cases, bd.start might mutate active_buckets right away, so
 313         # we need to put off calling start() until we've iterated all the way
 314         # through it.
 315         downloaders = []
 316         for blocknum, vbucket in active_buckets.iteritems():
 317             bd = BlockDownloader(vbucket, blocknum, self)
 318             downloaders.append(bd)
 319         l = [bd.start(self.segmentnumber) for bd in downloaders]
 320         return defer.DeferredList(l, fireOnOneErrback=True)
 321
 322     def hold_block(self, blocknum, data):
 323         self.blocks[blocknum] = data
 324
 325     def bucket_failed(self, vbucket):
 326         self.parent.bucket_failed(vbucket)
 327
 328 class DownloadStatus:
 329     implements(IDownloadStatus)
 330
 331     def __init__(self):
 332         self.storage_index = None
 333         self.size = None
 334         self.helper = False
 335         self.status = "Not started"
 336         self.progress = 0.0
 337         self.paused = False
 338         self.stopped = False
 339
 340     def get_storage_index(self):
 341         return self.storage_index
 342     def get_size(self):
 343         return self.size
 344     def using_helper(self):
 345         return self.helper
 346     def get_status(self):
 347         status = self.status
 348         if self.paused:
 349             status += " (output paused)"
 350         if self.stopped:
 351             status += " (output stopped)"
 352         return status
 353     def get_progress(self):
 354         return self.progress
 355
 356     def set_storage_index(self, si):
 357         self.storage_index = si
 358     def set_size(self, size):
 359         self.size = size
 360     def set_helper(self, helper):
 361         self.helper = helper
 362     def set_status(self, status):
 363         self.status = status
 364     def set_paused(self, paused):
 365         self.paused = paused
 366     def set_stopped(self, stopped):
 367         self.stopped = stopped
 368     def set_progress(self, value):
 369         self.progress = value
 370
 371
 372 class FileDownloader:
 373     implements(IPushProducer)
 374     check_crypttext_hash = True
 375     check_plaintext_hash = True
 376     _status = None
 377
 378     def __init__(self, client, u, downloadable):
 379         self._client = client
 380
 381         u = IFileURI(u)
 382         self._storage_index = u.storage_index
 383         self._uri_extension_hash = u.uri_extension_hash
 384         self._total_shares = u.total_shares
 385         self._size = u.size
 386         self._num_needed_shares = u.needed_shares
 387
 388         self.init_logging()
 389
 390         self._status = s = DownloadStatus()
 391         s.set_status("Starting")
 392         s.set_storage_index(self._storage_index)
 393         s.set_size(self._size)
 394         s.set_helper(False)
 395
 396         if IConsumer.providedBy(downloadable):
 397             downloadable.registerProducer(self, True)
 398         self._downloadable = downloadable
 399         self._output = Output(downloadable, u.key, self._size, self._log_number,
 400                               self._status)
 401         self._paused = False
 402         self._stopped = False
 403
 404         self.active_buckets = {} # k: shnum, v: bucket
 405         self._share_buckets = [] # list of (sharenum, bucket) tuples
 406         self._share_vbuckets = {} # k: shnum, v: set of ValidatedBuckets
 407         self._uri_extension_sources = []
 408
 409         self._uri_extension_data = None
 410
 411         self._fetch_failures = {"uri_extension": 0,
 412                                 "plaintext_hashroot": 0,
 413                                 "plaintext_hashtree": 0,
 414                                 "crypttext_hashroot": 0,
 415                                 "crypttext_hashtree": 0,
 416                                 }
 417
 418     def init_logging(self):
 419         self._log_prefix = prefix = storage.si_b2a(self._storage_index)[:5]
 420         num = self._client.log(format="FileDownloader(%(si)s): starting",
 421                                si=storage.si_b2a(self._storage_index))
 422         self._log_number = num
 423
 424     def log(self, *args, **kwargs):
 425         if "parent" not in kwargs:
 426             kwargs["parent"] = self._log_number
 427         if "facility" not in kwargs:
 428             kwargs["facility"] = "tahoe.download"
 429         return log.msg(*args, **kwargs)
 430
 431     def pauseProducing(self):
 432         if self._paused:
 433             return
 434         self._paused = defer.Deferred()
 435         if self._status:
 436             self._status.set_paused(True)
 437
 438     def resumeProducing(self):
 439         if self._paused:
 440             p = self._paused
 441             self._paused = None
 442             eventually(p.callback, None)
 443             if self._status:
 444                 self._status.set_paused(False)
 445
 446     def stopProducing(self):
 447         self.log("Download.stopProducing")
 448         self._stopped = True
 449         if self._status:
 450             self._status.set_stopped(True)
 451
 452     def start(self):
 453         self.log("starting download")
 454
 455         # first step: who should we download from?
 456         d = defer.maybeDeferred(self._get_all_shareholders)
 457         d.addCallback(self._got_all_shareholders)
 458         # now get the uri_extension block from somebody and validate it
 459         d.addCallback(self._obtain_uri_extension)
 460         d.addCallback(self._got_uri_extension)
 461         d.addCallback(self._get_hashtrees)
 462         d.addCallback(self._create_validated_buckets)
 463         # once we know that, we can download blocks from everybody
 464         d.addCallback(self._download_all_segments)
 465         def _finished(res):
 466             if self._status:
 467                 self._status.set_status("Finished")
 468             if IConsumer.providedBy(self._downloadable):
 469                 self._downloadable.unregisterProducer()
 470             return res
 471         d.addBoth(_finished)
 472         def _failed(why):
 473             if self._status:
 474                 self._status.set_status("Failed")
 475             self._output.fail(why)
 476             return why
 477         d.addErrback(_failed)
 478         d.addCallback(self._done)
 479         return d
 480
 481     def _get_all_shareholders(self):
 482         dl = []
 483         for (peerid,ss) in self._client.get_permuted_peers("storage",
 484                                                            self._storage_index):
 485             d = ss.callRemote("get_buckets", self._storage_index)
 486             d.addCallbacks(self._got_response, self._got_error)
 487             dl.append(d)
 488         self._responses_received = 0
 489         self._queries_sent = len(dl)
 490         if self._status:
 491             self._status.set_status("Locating Shares (%d/%d)" %
 492                                     (self._responses_received,
 493                                      self._queries_sent))
 494         return defer.DeferredList(dl)
 495
 496     def _got_response(self, buckets):
 497         self._responses_received += 1
 498         if self._status:
 499             self._status.set_status("Locating Shares (%d/%d)" %
 500                                     (self._responses_received,
 501                                      self._queries_sent))
 502         for sharenum, bucket in buckets.iteritems():
 503             b = storage.ReadBucketProxy(bucket)
 504             self.add_share_bucket(sharenum, b)
 505             self._uri_extension_sources.append(b)
 506
 507     def add_share_bucket(self, sharenum, bucket):
 508         # this is split out for the benefit of test_encode.py
 509         self._share_buckets.append( (sharenum, bucket) )
 510
 511     def _got_error(self, f):
 512         self._client.log("Somebody failed. -- %s" % (f,))
 513
 514     def bucket_failed(self, vbucket):
 515         shnum = vbucket.sharenum
 516         del self.active_buckets[shnum]
 517         s = self._share_vbuckets[shnum]
 518         # s is a set of ValidatedBucket instances
 519         s.remove(vbucket)
 520         # ... which might now be empty
 521         if not s:
 522             # there are no more buckets which can provide this share, so
 523             # remove the key. This may prompt us to use a different share.
 524             del self._share_vbuckets[shnum]
 525
 526     def _got_all_shareholders(self, res):
 527         if len(self._share_buckets) < self._num_needed_shares:
 528             raise NotEnoughPeersError
 529
 530         #for s in self._share_vbuckets.values():
 531         #    for vb in s:
 532         #        assert isinstance(vb, ValidatedBucket), \
 533         #               "vb is %s but should be a ValidatedBucket" % (vb,)
 534
 535     def _unpack_uri_extension_data(self, data):
 536         return uri.unpack_extension(data)
 537
 538     def _obtain_uri_extension(self, ignored):
 539         # all shareholders are supposed to have a copy of uri_extension, and
 540         # all are supposed to be identical. We compute the hash of the data
 541         # that comes back, and compare it against the version in our URI. If
 542         # they don't match, ignore their data and try someone else.
 543         if self._status:
 544             self._status.set_status("Obtaining URI Extension")
 545
 546         def _validate(proposal, bucket):
 547             h = hashutil.uri_extension_hash(proposal)
 548             if h != self._uri_extension_hash:
 549                 self._fetch_failures["uri_extension"] += 1
 550                 msg = ("The copy of uri_extension we received from "
 551                        "%s was bad" % bucket)
 552                 raise BadURIExtensionHashValue(msg)
 553             return self._unpack_uri_extension_data(proposal)
 554         return self._obtain_validated_thing(None,
 555                                             self._uri_extension_sources,
 556                                             "uri_extension",
 557                                             "get_uri_extension", (), _validate)
 558
 559     def _obtain_validated_thing(self, ignored, sources, name, methname, args,
 560                                 validatorfunc):
 561         if not sources:
 562             raise NotEnoughPeersError("started with zero peers while fetching "
 563                                       "%s" % name)
 564         bucket = sources[0]
 565         sources = sources[1:]
 566         #d = bucket.callRemote(methname, *args)
 567         d = bucket.startIfNecessary()
 568         d.addCallback(lambda res: getattr(bucket, methname)(*args))
 569         d.addCallback(validatorfunc, bucket)
 570         def _bad(f):
 571             self.log("WEIRD: %s from vbucket %s failed: %s" % (name, bucket, f))
 572             if not sources:
 573                 raise NotEnoughPeersError("ran out of peers, last error was %s"
 574                                           % (f,))
 575             # try again with a different one
 576             return self._obtain_validated_thing(None, sources, name,
 577                                                 methname, args, validatorfunc)
 578         d.addErrback(_bad)
 579         return d
 580
 581     def _got_uri_extension(self, uri_extension_data):
 582         d = self._uri_extension_data = uri_extension_data
 583
 584         self._codec = codec.get_decoder_by_name(d['codec_name'])
 585         self._codec.set_serialized_params(d['codec_params'])
 586         self._tail_codec = codec.get_decoder_by_name(d['codec_name'])
 587         self._tail_codec.set_serialized_params(d['tail_codec_params'])
 588
 589         crypttext_hash = d['crypttext_hash']
 590         assert isinstance(crypttext_hash, str)
 591         assert len(crypttext_hash) == 32
 592         self._crypttext_hash = crypttext_hash
 593         self._plaintext_hash = d['plaintext_hash']
 594         self._roothash = d['share_root_hash']
 595
 596         self._segment_size = segment_size = d['segment_size']
 597         self._total_segments = mathutil.div_ceil(self._size, segment_size)
 598         self._current_segnum = 0
 599
 600         self._share_hashtree = hashtree.IncompleteHashTree(d['total_shares'])
 601         self._share_hashtree.set_hashes({0: self._roothash})
 602
 603     def _get_hashtrees(self, res):
 604         if self._status:
 605             self._status.set_status("Retrieving Hash Trees")
 606         d = self._get_plaintext_hashtrees()
 607         d.addCallback(self._get_crypttext_hashtrees)
 608         d.addCallback(self._setup_hashtrees)
 609         return d
 610
 611     def _get_plaintext_hashtrees(self):
 612         def _validate_plaintext_hashtree(proposal, bucket):
 613             if proposal[0] != self._uri_extension_data['plaintext_root_hash']:
 614                 self._fetch_failures["plaintext_hashroot"] += 1
 615                 msg = ("The copy of the plaintext_root_hash we received from"
 616                        " %s was bad" % bucket)
 617                 raise BadPlaintextHashValue(msg)
 618             pt_hashtree = hashtree.IncompleteHashTree(self._total_segments)
 619             pt_hashes = dict(list(enumerate(proposal)))
 620             try:
 621                 pt_hashtree.set_hashes(pt_hashes)
 622             except hashtree.BadHashError:
 623                 # the hashes they gave us were not self-consistent, even
 624                 # though the root matched what we saw in the uri_extension
 625                 # block
 626                 self._fetch_failures["plaintext_hashtree"] += 1
 627                 raise
 628             self._plaintext_hashtree = pt_hashtree
 629         d = self._obtain_validated_thing(None,
 630                                          self._uri_extension_sources,
 631                                          "plaintext_hashes",
 632                                          "get_plaintext_hashes", (),
 633                                          _validate_plaintext_hashtree)
 634         return d
 635
 636     def _get_crypttext_hashtrees(self, res):
 637         def _validate_crypttext_hashtree(proposal, bucket):
 638             if proposal[0] != self._uri_extension_data['crypttext_root_hash']:
 639                 self._fetch_failures["crypttext_hashroot"] += 1
 640                 msg = ("The copy of the crypttext_root_hash we received from"
 641                        " %s was bad" % bucket)
 642                 raise BadCrypttextHashValue(msg)
 643             ct_hashtree = hashtree.IncompleteHashTree(self._total_segments)
 644             ct_hashes = dict(list(enumerate(proposal)))
 645             try:
 646                 ct_hashtree.set_hashes(ct_hashes)
 647             except hashtree.BadHashError:
 648                 self._fetch_failures["crypttext_hashtree"] += 1
 649                 raise
 650             ct_hashtree.set_hashes(ct_hashes)
 651             self._crypttext_hashtree = ct_hashtree
 652         d = self._obtain_validated_thing(None,
 653                                          self._uri_extension_sources,
 654                                          "crypttext_hashes",
 655                                          "get_crypttext_hashes", (),
 656                                          _validate_crypttext_hashtree)
 657         return d
 658
 659     def _setup_hashtrees(self, res):
 660         self._output.setup_hashtrees(self._plaintext_hashtree,
 661                                      self._crypttext_hashtree)
 662
 663
 664     def _create_validated_buckets(self, ignored=None):
 665         self._share_vbuckets = {}
 666         for sharenum, bucket in self._share_buckets:
 667             vbucket = ValidatedBucket(sharenum, bucket,
 668                                       self._share_hashtree,
 669                                       self._roothash,
 670                                       self._total_segments)
 671             s = self._share_vbuckets.setdefault(sharenum, set())
 672             s.add(vbucket)
 673
 674     def _activate_enough_buckets(self):
 675         """either return a mapping from shnum to a ValidatedBucket that can
 676         provide data for that share, or raise NotEnoughPeersError"""
 677
 678         while len(self.active_buckets) < self._num_needed_shares:
 679             # need some more
 680             handled_shnums = set(self.active_buckets.keys())
 681             available_shnums = set(self._share_vbuckets.keys())
 682             potential_shnums = list(available_shnums - handled_shnums)
 683             if not potential_shnums:
 684                 raise NotEnoughPeersError
 685             # choose a random share
 686             shnum = random.choice(potential_shnums)
 687             # and a random bucket that will provide it
 688             validated_bucket = random.choice(list(self._share_vbuckets[shnum]))
 689             self.active_buckets[shnum] = validated_bucket
 690         return self.active_buckets
 691
 692
 693     def _download_all_segments(self, res):
 694         # the promise: upon entry to this function, self._share_vbuckets
 695         # contains enough buckets to complete the download, and some extra
 696         # ones to tolerate some buckets dropping out or having errors.
 697         # self._share_vbuckets is a dictionary that maps from shnum to a set
 698         # of ValidatedBuckets, which themselves are wrappers around
 699         # RIBucketReader references.
 700         self.active_buckets = {} # k: shnum, v: ValidatedBucket instance
 701
 702         d = defer.succeed(None)
 703         for segnum in range(self._total_segments-1):
 704             d.addCallback(self._download_segment, segnum)
 705             # this pause, at the end of write, prevents pre-fetch from
 706             # happening until the consumer is ready for more data.
 707             d.addCallback(self._check_for_pause)
 708         d.addCallback(self._download_tail_segment, self._total_segments-1)
 709         return d
 710
 711     def _check_for_pause(self, res):
 712         if self._paused:
 713             d = defer.Deferred()
 714             self._paused.addCallback(lambda ignored: d.callback(res))
 715             return d
 716         if self._stopped:
 717             raise DownloadStopped("our Consumer called stopProducing()")
 718         return res
 719
 720     def _download_segment(self, res, segnum):
 721         if self._status:
 722             self._status.set_status("Downloading segment %d of %d" %
 723                                     (segnum+1, self._total_segments))
 724         self.log("downloading seg#%d of %d (%d%%)"
 725                  % (segnum, self._total_segments,
 726                     100.0 * segnum / self._total_segments))
 727         # memory footprint: when the SegmentDownloader finishes pulling down
 728         # all shares, we have 1*segment_size of usage.
 729         segmentdler = SegmentDownloader(self, segnum, self._num_needed_shares)
 730         d = segmentdler.start()
 731         # pause before using more memory
 732         d.addCallback(self._check_for_pause)
 733         # while the codec does its job, we hit 2*segment_size
 734         d.addCallback(lambda (shares, shareids):
 735                       self._codec.decode(shares, shareids))
 736         # once the codec is done, we drop back to 1*segment_size, because
 737         # 'shares' goes out of scope. The memory usage is all in the
 738         # plaintext now, spread out into a bunch of tiny buffers.
 739
 740         # pause/check-for-stop just before writing, to honor stopProducing
 741         d.addCallback(self._check_for_pause)
 742         def _done(buffers):
 743             # we start by joining all these buffers together into a single
 744             # string. This makes Output.write easier, since it wants to hash
 745             # data one segment at a time anyways, and doesn't impact our
 746             # memory footprint since we're already peaking at 2*segment_size
 747             # inside the codec a moment ago.
 748             segment = "".join(buffers)
 749             del buffers
 750             # we're down to 1*segment_size right now, but write_segment()
 751             # will decrypt a copy of the segment internally, which will push
 752             # us up to 2*segment_size while it runs.
 753             self._output.write_segment(segment)
 754         d.addCallback(_done)
 755         return d
 756
 757     def _download_tail_segment(self, res, segnum):
 758         self.log("downloading seg#%d of %d (%d%%)"
 759                  % (segnum, self._total_segments,
 760                     100.0 * segnum / self._total_segments))
 761         segmentdler = SegmentDownloader(self, segnum, self._num_needed_shares)
 762         d = segmentdler.start()
 763         # pause before using more memory
 764         d.addCallback(self._check_for_pause)
 765         d.addCallback(lambda (shares, shareids):
 766                       self._tail_codec.decode(shares, shareids))
 767         # pause/check-for-stop just before writing, to honor stopProducing
 768         d.addCallback(self._check_for_pause)
 769         def _done(buffers):
 770             # trim off any padding added by the upload side
 771             segment = "".join(buffers)
 772             del buffers
 773             # we never send empty segments. If the data was an exact multiple
 774             # of the segment size, the last segment will be full.
 775             pad_size = mathutil.pad_size(self._size, self._segment_size)
 776             tail_size = self._segment_size - pad_size
 777             segment = segment[:tail_size]
 778             self._output.write_segment(segment)
 779         d.addCallback(_done)
 780         return d
 781
 782     def _done(self, res):
 783         self.log("download done")
 784         self._output.close()
 785         if self.check_crypttext_hash:
 786             _assert(self._crypttext_hash == self._output.crypttext_hash,
 787                     "bad crypttext_hash: computed=%s, expected=%s" %
 788                     (base32.b2a(self._output.crypttext_hash),
 789                      base32.b2a(self._crypttext_hash)))
 790         if self.check_plaintext_hash:
 791             _assert(self._plaintext_hash == self._output.plaintext_hash,
 792                     "bad plaintext_hash: computed=%s, expected=%s" %
 793                     (base32.b2a(self._output.plaintext_hash),
 794                      base32.b2a(self._plaintext_hash)))
 795         _assert(self._output.length == self._size,
 796                 got=self._output.length, expected=self._size)
 797         return self._output.finish()
 798
 799     def get_download_status(self):
 800         return self._status
 801
 802
 803 class LiteralDownloader:
 804     def __init__(self, client, u, downloadable):
 805         self._uri = IFileURI(u)
 806         assert isinstance(self._uri, uri.LiteralFileURI)
 807         self._downloadable = downloadable
 808         self._status = s = DownloadStatus()
 809         s.set_storage_index(None)
 810         s.set_helper(False)
 811         s.set_status("Done")
 812         s.set_progress(1.0)
 813
 814     def start(self):
 815         data = self._uri.data
 816         self._status.set_size(len(data))
 817         self._downloadable.open(len(data))
 818         self._downloadable.write(data)
 819         self._downloadable.close()
 820         return defer.maybeDeferred(self._downloadable.finish)
 821
 822     def get_download_status(self):
 823         return self._status
 824
 825 class FileName:
 826     implements(IDownloadTarget)
 827     def __init__(self, filename):
 828         self._filename = filename
 829         self.f = None
 830     def open(self, size):
 831         self.f = open(self._filename, "wb")
 832         return self.f
 833     def write(self, data):
 834         self.f.write(data)
 835     def close(self):
 836         if self.f:
 837             self.f.close()
 838     def fail(self, why):
 839         if self.f:
 840             self.f.close()
 841             os.unlink(self._filename)
 842     def register_canceller(self, cb):
 843         pass # we won't use it
 844     def finish(self):
 845         pass
 846
 847 class Data:
 848     implements(IDownloadTarget)
 849     def __init__(self):
 850         self._data = []
 851     def open(self, size):
 852         pass
 853     def write(self, data):
 854         self._data.append(data)
 855     def close(self):
 856         self.data = "".join(self._data)
 857         del self._data
 858     def fail(self, why):
 859         del self._data
 860     def register_canceller(self, cb):
 861         pass # we won't use it
 862     def finish(self):
 863         return self.data
 864
 865 class FileHandle:
 866     """Use me to download data to a pre-defined filehandle-like object. I
 867     will use the target's write() method. I will *not* close the filehandle:
 868     I leave that up to the originator of the filehandle. The download process
 869     will return the filehandle when it completes.
 870     """
 871     implements(IDownloadTarget)
 872     def __init__(self, filehandle):
 873         self._filehandle = filehandle
 874     def open(self, size):
 875         pass
 876     def write(self, data):
 877         self._filehandle.write(data)
 878     def close(self):
 879         # the originator of the filehandle reserves the right to close it
 880         pass
 881     def fail(self, why):
 882         pass
 883     def register_canceller(self, cb):
 884         pass
 885     def finish(self):
 886         return self._filehandle
 887
 888 class Downloader(service.MultiService):
 889     """I am a service that allows file downloading.
 890     """
 891     implements(IDownloader)
 892     name = "downloader"
 893
 894     def __init__(self):
 895         service.MultiService.__init__(self)
 896         self._all_downloads = weakref.WeakKeyDictionary()
 897
 898     def download(self, u, t):
 899         assert self.parent
 900         assert self.running
 901         u = IFileURI(u)
 902         t = IDownloadTarget(t)
 903         assert t.write
 904         assert t.close
 905         if isinstance(u, uri.LiteralFileURI):
 906             dl = LiteralDownloader(self.parent, u, t)
 907         elif isinstance(u, uri.CHKFileURI):
 908             dl = FileDownloader(self.parent, u, t)
 909         else:
 910             raise RuntimeError("I don't know how to download a %s" % u)
 911         self._all_downloads[dl.get_download_status()] = None
 912         d = dl.start()
 913         return d
 914
 915     # utility functions
 916     def download_to_data(self, uri):
 917         return self.download(uri, Data())
 918     def download_to_filename(self, uri, filename):
 919         return self.download(uri, FileName(filename))
 920     def download_to_filehandle(self, uri, filehandle):
 921         return self.download(uri, FileHandle(filehandle))
 922
 923
 924     def list_downloads(self):
 925         return self._all_downloads.keys()