def _got_results_one_share(self, shnum, peerid, data):
self.check_prefix(peerid, shnum, data)
# the [seqnum:signature] pieces are validated by _compare_prefix,
# which checks their signature against the pubkey known to be
# associated with this file.
(seqnum, root_hash, IV, k, N, segsize, datalen, pubkey, signature,
share_hash_chain, block_hash_tree, share_data,
enc_privkey) = unpack_share(data)
# validate [share_hash_chain,block_hash_tree,share_data]
leaves = [hashutil.block_hash(share_data)]
t = hashtree.HashTree(leaves)
if list(t) != block_hash_tree:
raise CorruptShareError(peerid, shnum, "block hash tree failure")
share_hash_leaf = t[0]
t2 = hashtree.IncompleteHashTree(N)
# root_hash was checked by the signature
t2.set_hashes({0: root_hash})
try:
t2.set_hashes(hashes=share_hash_chain,
leaves={shnum: share_hash_leaf})
except (hashtree.BadHashError, hashtree.NotEnoughHashesError,
IndexError), e:
msg = "corrupt hashes: %s" % (e,)
raise CorruptShareError(peerid, shnum, msg)
def _got_results_one_share(self, shnum, peerid,
got_prefix, got_hash_and_data):
self.log("_got_results: got shnum #%d from peerid %s"
% (shnum, idlib.shortnodeid_b2a(peerid)))
(seqnum, root_hash, IV, segsize, datalength, k, N, prefix,
offsets_tuple) = self.verinfo
assert len(got_prefix) == len(prefix), (len(got_prefix), len(prefix))
if got_prefix != prefix:
msg = "someone wrote to the data since we read the servermap: prefix changed"
raise UncoordinatedWriteError(msg)
(share_hash_chain, block_hash_tree,
share_data) = unpack_share_data(self.verinfo, got_hash_and_data)
assert isinstance(share_data, str)
# build the block hash tree. SDMF has only one leaf.
leaves = [hashutil.block_hash(share_data)]
t = hashtree.HashTree(leaves)
if list(t) != block_hash_tree:
raise CorruptShareError(peerid, shnum, "block hash tree failure")
share_hash_leaf = t[0]
t2 = hashtree.IncompleteHashTree(N)
# root_hash was checked by the signature
t2.set_hashes({0: root_hash})
try:
t2.set_hashes(hashes=share_hash_chain,
leaves={shnum: share_hash_leaf})
except (hashtree.BadHashError, hashtree.NotEnoughHashesError,
IndexError), e:
msg = "corrupt hashes: %s" % (e,)
raise CorruptShareError(peerid, shnum, msg)
def _send_segment(self, shares_and_shareids, segnum):
# To generate the URI, we must generate the roothash, so we must
# generate all shares, even if we aren't actually giving them to
# anybody. This means that the set of shares we create will be equal
# to or larger than the set of landlords. If we have any landlord who
# *doesn't* have a share, that's an error.
(shares, shareids) = shares_and_shareids
_assert(set(self.landlords.keys()).issubset(set(shareids)),
shareids=shareids, landlords=self.landlords)
start = time.time()
dl = []
self.set_status("Sending segment %d of %d" % (segnum+1,
self.num_segments))
self.set_encode_and_push_progress(segnum)
lognum = self.log("send_segment(%d)" % segnum, level=log.NOISY)
for i in range(len(shares)):
block = shares[i]
shareid = shareids[i]
d = self.send_block(shareid, segnum, block, lognum)
dl.append(d)
block_hash = hashutil.block_hash(block)
#from allmydata.util import base32
#log.msg("creating block (shareid=%d, blocknum=%d) "
# "len=%d %r .. %r: %s" %
# (shareid, segnum, len(block),
# block[:50], block[-50:], base32.b2a(block_hash)))
self.block_hashes[shareid].append(block_hash)
dl = self._gather_responses(dl)
def do_progress(ign):
done = self.segment_size * (segnum + 1)
if self._progress:
self._progress.set_progress(done)
return ign
dl.addCallback(do_progress)
def _logit(res):
self.log("%s uploaded %s / %s bytes (%d%%) of your file." %
(self,
self.segment_size*(segnum+1),
self.segment_size*self.num_segments,
100 * (segnum+1) / self.num_segments,
),
level=log.OPERATIONAL)
elapsed = time.time() - start
self._times["cumulative_sending"] += elapsed
return res
dl.addCallback(_logit)
return dl
def _got_data(self, results, blocknum):
precondition(blocknum < self.num_blocks,
self, blocknum, self.num_blocks)
sharehashes, blockhashes, blockdata = results
try:
sharehashes = dict(sharehashes)
except ValueError as le:
le.args = tuple(le.args + (sharehashes,))
raise
blockhashes = dict(enumerate(blockhashes))
candidate_share_hash = None # in case we log it in the except block below
blockhash = None # in case we log it in the except block below
try:
if self.share_hash_tree.needed_hashes(self.sharenum):
# This will raise exception if the values being passed do not
# match the root node of self.share_hash_tree.
try:
self.share_hash_tree.set_hashes(sharehashes)
except IndexError as le:
# Weird -- sharehashes contained index numbers outside of
# the range that fit into this hash tree.
raise BadOrMissingHash(le)
# To validate a block we need the root of the block hash tree,
# which is also one of the leafs of the share hash tree, and is
# called "the share hash".
if not self.block_hash_tree[0]: # empty -- no root node yet
# Get the share hash from the share hash tree.
share_hash = self.share_hash_tree.get_leaf(self.sharenum)
if not share_hash:
# No root node in block_hash_tree and also the share hash
# wasn't sent by the server.
raise hashtree.NotEnoughHashesError
self.block_hash_tree.set_hashes({0: share_hash})
if self.block_hash_tree.needed_hashes(blocknum):
self.block_hash_tree.set_hashes(blockhashes)
blockhash = block_hash(blockdata)
self.block_hash_tree.set_hashes(leaves={blocknum: blockhash})
#self.log("checking block_hash(shareid=%d, blocknum=%d) len=%d "
# "%r .. %r: %s" %
# (self.sharenum, blocknum, len(blockdata),
# blockdata[:50], blockdata[-50:], base32.b2a(blockhash)))
except (hashtree.BadHashError, hashtree.NotEnoughHashesError) as le:
# log.WEIRD: indicates undetected disk/network error, or more
# likely a programming error
self.log("hash failure in block=%d, shnum=%d on %s" %
(blocknum, self.sharenum, self.bucket))
if self.block_hash_tree.needed_hashes(blocknum):
self.log(""" failure occurred when checking the block_hash_tree.
This suggests that either the block data was bad, or that the
block hashes we received along with it were bad.""")
else:
self.log(""" the failure probably occurred when checking the
share_hash_tree, which suggests that the share hashes we
received from the remote peer were bad.""")
self.log(" have candidate_share_hash: %s" % bool(candidate_share_hash))
self.log(" block length: %d" % len(blockdata))
self.log(" block hash: %s" % base32.b2a_or_none(blockhash))
if len(blockdata) < 100:
self.log(" block data: %r" % (blockdata,))
else:
self.log(" block data start/end: %r .. %r" %
(blockdata[:50], blockdata[-50:]))
self.log(" share hash tree:\n" + self.share_hash_tree.dump())
self.log(" block hash tree:\n" + self.block_hash_tree.dump())
lines = []
for i,h in sorted(sharehashes.items()):
lines.append("%3d: %s" % (i, base32.b2a_or_none(h)))
self.log(" sharehashes:\n" + "\n".join(lines) + "\n")
lines = []
for i,h in blockhashes.items():
lines.append("%3d: %s" % (i, base32.b2a_or_none(h)))
log.msg(" blockhashes:\n" + "\n".join(lines) + "\n")
raise BadOrMissingHash(le)
# If we made it here, the block is good. If the hash trees didn't
# like what they saw, they would have raised a BadHashError, causing
# our caller to see a Failure and thus ignore this block (as well as
# dropping this bucket).
return blockdata
# To validate a block we need the root of the block hash tree,
# which is also one of the leafs of the share hash tree, and is
# called "the share hash".
if not self.block_hash_tree[0]: # empty -- no root node yet
# Get the share hash from the share hash tree.
share_hash = self.share_hash_tree.get_leaf(self.sharenum)
if not share_hash:
# No root node in block_hash_tree and also the share hash
# wasn't sent by the server.
raise hashtree.NotEnoughHashesError
self.block_hash_tree.set_hashes({0: share_hash})
if self.block_hash_tree.needed_hashes(blocknum):
self.block_hash_tree.set_hashes(blockhashes)
blockhash = block_hash(blockdata)
self.block_hash_tree.set_hashes(leaves={blocknum: blockhash})
#self.log("checking block_hash(shareid=%d, blocknum=%d) len=%d "
# "%r .. %r: %s" %
# (self.sharenum, blocknum, len(blockdata),
# blockdata[:50], blockdata[-50:], base32.b2a(blockhash)))
except (hashtree.BadHashError, hashtree.NotEnoughHashesError), le:
# log.WEIRD: indicates undetected disk/network error, or more
# likely a programming error
self.log("hash failure in block=%d, shnum=%d on %s" %
(blocknum, self.sharenum, self.bucket))
if self.block_hash_tree.needed_hashes(blocknum):
self.log(""" failure occurred when checking the block_hash_tree.
This suggests that either the block data was bad, or that the
block hashes we received along with it were bad.""")
# To validate a block we need the root of the block hash tree,
# which is also one of the leafs of the share hash tree, and is
# called "the share hash".
if not self.block_hash_tree[0]: # empty -- no root node yet
# Get the share hash from the share hash tree.
share_hash = self.share_hash_tree.get_leaf(self.sharenum)
if not share_hash:
# No root node in block_hash_tree and also the share hash
# wasn't sent by the server.
raise hashtree.NotEnoughHashesError
self.block_hash_tree.set_hashes({0: share_hash})
if self.block_hash_tree.needed_hashes(blocknum):
self.block_hash_tree.set_hashes(blockhashes)
blockhash = block_hash(blockdata)
self.block_hash_tree.set_hashes(leaves={blocknum: blockhash})
#self.log("checking block_hash(shareid=%d, blocknum=%d) len=%d "
# "%r .. %r: %s" %
# (self.sharenum, blocknum, len(blockdata),
# blockdata[:50], blockdata[-50:], base32.b2a(blockhash)))
except (hashtree.BadHashError, hashtree.NotEnoughHashesError), le:
# log.WEIRD: indicates undetected disk/network error, or more
# likely a programming error
self.log("hash failure in block=%d, shnum=%d on %s" %
(blocknum, self.sharenum, self.bucket))
if self.block_hash_tree.needed_hashes(blocknum):
self.log(
""" failure occurred when checking the block_hash_tree.
This suggests that either the block data was bad, or that the
class Encoder(object):
implements(IEncoder)
def __init__(self, log_parent=None, upload_status=None):
object.__init__(self)
self.uri_extension_data = {}
self._codec = None
self._status = None
if upload_status:
self._status = IUploadStatus(upload_status)
precondition(log_parent is None or isinstance(log_parent, int),
log_parent)
self._log_number = log.msg("creating Encoder %s" % self,
facility="tahoe.encoder",
parent=log_parent)
self._aborted = False
def __repr__(self):
if hasattr(self, "_storage_index"):
return "<Encoder for %s>" % si_b2a(self._storage_index)[:5]
return "<Encoder for unknown storage index>"
def log(self, *args, **kwargs):
if "parent" not in kwargs:
kwargs["parent"] = self._log_number
if "facility" not in kwargs:
kwargs["facility"] = "tahoe.encoder"
return log.msg(*args, **kwargs)
def set_encrypted_uploadable(self, uploadable):
eu = self._uploadable = IEncryptedUploadable(uploadable)
d = eu.get_size()
def _got_size(size):
self.log(format="file size: %(size)d", size=size)
self.file_size = size
d.addCallback(_got_size)
d.addCallback(lambda res: eu.get_all_encoding_parameters())
d.addCallback(self._got_all_encoding_parameters)
d.addCallback(lambda res: eu.get_storage_index())
def _done(storage_index):
self._storage_index = storage_index
return self
d.addCallback(_done)
return d
def _got_all_encoding_parameters(self, params):
assert not self._codec
k, happy, n, segsize = params
self.required_shares = k
self.servers_of_happiness = happy
self.num_shares = n
self.segment_size = segsize
self.log("got encoding parameters: %d/%d/%d %d" %
(k, happy, n, segsize))
self.log("now setting up codec")
assert self.segment_size % self.required_shares == 0
self.num_segments = mathutil.div_ceil(self.file_size,
self.segment_size)
self._codec = CRSEncoder()
self._codec.set_params(self.segment_size, self.required_shares,
self.num_shares)
data = self.uri_extension_data
data['codec_name'] = self._codec.get_encoder_type()
data['codec_params'] = self._codec.get_serialized_params()
data['size'] = self.file_size
data['segment_size'] = self.segment_size
self.share_size = mathutil.div_ceil(self.file_size,
self.required_shares)
data['num_segments'] = self.num_segments
data['needed_shares'] = self.required_shares
data['total_shares'] = self.num_shares
# the "tail" is the last segment. This segment may or may not be
# shorter than all other segments. We use the "tail codec" to handle
# it. If the tail is short, we use a different codec instance. In
# addition, the tail codec must be fed data which has been padded out
# to the right size.
tail_size = self.file_size % self.segment_size
if not tail_size:
tail_size = self.segment_size
# the tail codec is responsible for encoding tail_size bytes
padded_tail_size = mathutil.next_multiple(tail_size,
self.required_shares)
self._tail_codec = CRSEncoder()
self._tail_codec.set_params(padded_tail_size, self.required_shares,
self.num_shares)
data['tail_codec_params'] = self._tail_codec.get_serialized_params()
def _get_share_size(self):
share_size = mathutil.div_ceil(self.file_size, self.required_shares)
overhead = self._compute_overhead()
return share_size + overhead
def _compute_overhead(self):
return 0
def get_param(self, name):
assert self._codec
if name == "storage_index":
return self._storage_index
elif name == "share_counts":
return (self.required_shares, self.servers_of_happiness,
self.num_shares)
elif name == "num_segments":
return self.num_segments
elif name == "segment_size":
return self.segment_size
elif name == "block_size":
return self._codec.get_block_size()
elif name == "share_size":
return self._get_share_size()
elif name == "serialized_params":
return self._codec.get_serialized_params()
else:
raise KeyError("unknown parameter name '%s'" % name)
def set_shareholders(self, landlords, servermap):
assert isinstance(landlords, dict)
for k in landlords:
assert IStorageBucketWriter.providedBy(landlords[k])
self.landlords = landlords.copy()
assert isinstance(servermap, dict)
for v in servermap.itervalues():
assert isinstance(v, set)
self.servermap = servermap.copy()
def start(self):
""" Returns a Deferred that will fire with the verify cap (an instance of
uri.CHKFileVerifierURI)."""
self.log("%s starting" % (self, ))
#paddedsize = self._size + mathutil.pad_size(self._size, self.needed_shares)
assert self._codec
self._crypttext_hasher = hashutil.crypttext_hasher()
self._crypttext_hashes = []
self.segment_num = 0
self.block_hashes = [[] for x in range(self.num_shares)]
# block_hashes[i] is a list that will be accumulated and then send
# to landlord[i]. This list contains a hash of each segment_share
# that we sent to that landlord.
self.share_root_hashes = [None] * self.num_shares
self._times = {
"cumulative_encoding": 0.0,
"cumulative_sending": 0.0,
"hashes_and_close": 0.0,
"total_encode_and_push": 0.0,
}
self._start_total_timestamp = time.time()
d = fireEventually()
d.addCallback(lambda res: self.start_all_shareholders())
for i in range(self.num_segments - 1):
# note to self: this form doesn't work, because lambda only
# captures the slot, not the value
#d.addCallback(lambda res: self.do_segment(i))
# use this form instead:
d.addCallback(lambda res, i=i: self._encode_segment(i))
d.addCallback(self._send_segment, i)
d.addCallback(self._turn_barrier)
last_segnum = self.num_segments - 1
d.addCallback(lambda res: self._encode_tail_segment(last_segnum))
d.addCallback(self._send_segment, last_segnum)
d.addCallback(self._turn_barrier)
d.addCallback(lambda res: self.finish_hashing())
d.addCallback(
lambda res: self.send_crypttext_hash_tree_to_all_shareholders())
d.addCallback(lambda res: self.send_all_block_hash_trees())
d.addCallback(lambda res: self.send_all_share_hash_trees())
d.addCallback(
lambda res: self.send_uri_extension_to_all_shareholders())
d.addCallback(lambda res: self.close_all_shareholders())
d.addCallbacks(self.done, self.err)
return d
def set_status(self, status):
if self._status:
self._status.set_status(status)
def set_encode_and_push_progress(self, sent_segments=None, extra=0.0):
if self._status:
# we treat the final hash+close as an extra segment
if sent_segments is None:
sent_segments = self.num_segments
progress = float(sent_segments + extra) / (self.num_segments + 1)
self._status.set_progress(2, progress)
def abort(self):
self.log("aborting upload", level=log.UNUSUAL)
assert self._codec, "don't call abort before start"
self._aborted = True
# the next segment read (in _gather_data inside _encode_segment) will
# raise UploadAborted(), which will bypass the rest of the upload
# chain. If we've sent the final segment's shares, it's too late to
# abort. TODO: allow abort any time up to close_all_shareholders.
def _turn_barrier(self, res):
# putting this method in a Deferred chain imposes a guaranteed
# reactor turn between the pre- and post- portions of that chain.
# This can be useful to limit memory consumption: since Deferreds do
# not do tail recursion, code which uses defer.succeed(result) for
# consistency will cause objects to live for longer than you might
# normally expect.
return fireEventually(res)
def start_all_shareholders(self):
self.log("starting shareholders", level=log.NOISY)
self.set_status("Starting shareholders")
dl = []
for shareid in list(self.landlords):
d = self.landlords[shareid].put_header()
d.addErrback(self._remove_shareholder, shareid, "start")
dl.append(d)
return self._gather_responses(dl)
def _encode_segment(self, segnum):
codec = self._codec
start = time.time()
# the ICodecEncoder API wants to receive a total of self.segment_size
# bytes on each encode() call, broken up into a number of
# identically-sized pieces. Due to the way the codec algorithm works,
# these pieces need to be the same size as the share which the codec
# will generate. Therefore we must feed it with input_piece_size that
# equals the output share size.
input_piece_size = codec.get_block_size()
# as a result, the number of input pieces per encode() call will be
# equal to the number of required shares with which the codec was
# constructed. You can think of the codec as chopping up a
# 'segment_size' of data into 'required_shares' shares (not doing any
# fancy math at all, just doing a split), then creating some number
# of additional shares which can be substituted if the primary ones
# are unavailable
# we read data from the source one segment at a time, and then chop
# it into 'input_piece_size' pieces before handing it to the codec
crypttext_segment_hasher = hashutil.crypttext_segment_hasher()
# memory footprint: we only hold a tiny piece of the plaintext at any
# given time. We build up a segment's worth of cryptttext, then hand
# it to the encoder. Assuming 3-of-10 encoding (3.3x expansion) and
# 1MiB max_segment_size, we get a peak memory footprint of 4.3*1MiB =
# 4.3MiB. Lowering max_segment_size to, say, 100KiB would drop the
# footprint to 430KiB at the expense of more hash-tree overhead.
d = self._gather_data(self.required_shares, input_piece_size,
crypttext_segment_hasher)
def _done_gathering(chunks):
for c in chunks:
assert len(c) == input_piece_size
self._crypttext_hashes.append(crypttext_segment_hasher.digest())
# during this call, we hit 5*segsize memory
return codec.encode(chunks)
d.addCallback(_done_gathering)
def _done(res):
elapsed = time.time() - start
self._times["cumulative_encoding"] += elapsed
return res
d.addCallback(_done)
return d
def _encode_tail_segment(self, segnum):
start = time.time()
codec = self._tail_codec
input_piece_size = codec.get_block_size()
crypttext_segment_hasher = hashutil.crypttext_segment_hasher()
d = self._gather_data(self.required_shares,
input_piece_size,
crypttext_segment_hasher,
allow_short=True)
def _done_gathering(chunks):
for c in chunks:
# a short trailing chunk will have been padded by
# _gather_data
assert len(c) == input_piece_size
self._crypttext_hashes.append(crypttext_segment_hasher.digest())
return codec.encode(chunks)
d.addCallback(_done_gathering)
def _done(res):
elapsed = time.time() - start
self._times["cumulative_encoding"] += elapsed
return res
d.addCallback(_done)
return d
def _gather_data(self,
num_chunks,
input_chunk_size,
crypttext_segment_hasher,
allow_short=False):
"""Return a Deferred that will fire when the required number of
chunks have been read (and hashed and encrypted). The Deferred fires
with a list of chunks, each of size input_chunk_size."""
# I originally built this to allow read_encrypted() to behave badly:
# to let it return more or less data than you asked for. It would
# stash the leftovers until later, and then recurse until it got
# enough. I don't think that was actually useful.
#
# who defines read_encrypted?
# offloaded.LocalCiphertextReader: real disk file: exact
# upload.EncryptAnUploadable: Uploadable, but a wrapper that makes
# it exact. The return value is a list of 50KiB chunks, to reduce
# the memory footprint of the encryption process.
# repairer.Repairer: immutable.filenode.CiphertextFileNode: exact
#
# This has been redefined to require read_encrypted() to behave like
# a local file: return exactly the amount requested unless it hits
# EOF.
# -warner
if self._aborted:
raise UploadAborted()
read_size = num_chunks * input_chunk_size
d = self._uploadable.read_encrypted(read_size, hash_only=False)
def _got(data):
assert isinstance(data, (list, tuple))
if self._aborted:
raise UploadAborted()
data = "".join(data)
precondition(len(data) <= read_size, len(data), read_size)
if not allow_short:
precondition(len(data) == read_size, len(data), read_size)
crypttext_segment_hasher.update(data)
self._crypttext_hasher.update(data)
if allow_short and len(data) < read_size:
# padding
data += "\x00" * (read_size - len(data))
encrypted_pieces = [
data[i:i + input_chunk_size]
for i in range(0, len(data), input_chunk_size)
]
return encrypted_pieces
d.addCallback(_got)
return d
def _send_segment(self, (shares, shareids), segnum):
# To generate the URI, we must generate the roothash, so we must
# generate all shares, even if we aren't actually giving them to
# anybody. This means that the set of shares we create will be equal
# to or larger than the set of landlords. If we have any landlord who
# *doesn't* have a share, that's an error.
_assert(set(self.landlords.keys()).issubset(set(shareids)),
shareids=shareids,
landlords=self.landlords)
start = time.time()
dl = []
self.set_status("Sending segment %d of %d" %
(segnum + 1, self.num_segments))
self.set_encode_and_push_progress(segnum)
lognum = self.log("send_segment(%d)" % segnum, level=log.NOISY)
for i in range(len(shares)):
block = shares[i]
shareid = shareids[i]
d = self.send_block(shareid, segnum, block, lognum)
dl.append(d)
block_hash = hashutil.block_hash(block)
#from allmydata.util import base32
#log.msg("creating block (shareid=%d, blocknum=%d) "
# "len=%d %r .. %r: %s" %
# (shareid, segnum, len(block),
# block[:50], block[-50:], base32.b2a(block_hash)))
self.block_hashes[shareid].append(block_hash)
dl = self._gather_responses(dl)
def _logit(res):
self.log("%s uploaded %s / %s bytes (%d%%) of your file." % (
self,
self.segment_size * (segnum + 1),
self.segment_size * self.num_segments,
100 * (segnum + 1) / self.num_segments,
),
level=log.OPERATIONAL)
elapsed = time.time() - start
self._times["cumulative_sending"] += elapsed
return res
dl.addCallback(_logit)
return dl
def _validate_block(self, results, segnum, reader, server, started):
"""
I validate a block from one share on a remote server.
"""
# Grab the part of the block hash tree that is necessary to
# validate this block, then generate the block hash root.
self.log("validating share %d for segment %d" % (reader.shnum, segnum))
elapsed = time.time() - started
self._status.add_fetch_timing(server, elapsed)
self._set_current_status("validating blocks")
block_and_salt, blockhashes, sharehashes = results
block, salt = block_and_salt
_assert(isinstance(block, bytes), (block, salt))
blockhashes = dict(enumerate(blockhashes))
self.log("the reader gave me the following blockhashes: %s" % \
list(blockhashes.keys()))
self.log("the reader gave me the following sharehashes: %s" % \
list(sharehashes.keys()))
bht = self._block_hash_trees[reader.shnum]
if bht.needed_hashes(segnum, include_leaf=True):
try:
bht.set_hashes(blockhashes)
except (hashtree.BadHashError, hashtree.NotEnoughHashesError, \
IndexError) as e:
raise CorruptShareError(server, reader.shnum,
"block hash tree failure: %s" % e)
if self._version == MDMF_VERSION:
blockhash = hashutil.block_hash(salt + block)
else:
blockhash = hashutil.block_hash(block)
# If this works without an error, then validation is
# successful.
try:
bht.set_hashes(leaves={segnum: blockhash})
except (hashtree.BadHashError, hashtree.NotEnoughHashesError, \
IndexError) as e:
raise CorruptShareError(server, reader.shnum,
"block hash tree failure: %s" % e)
# Reaching this point means that we know that this segment
# is correct. Now we need to check to see whether the share
# hash chain is also correct.
# SDMF wrote share hash chains that didn't contain the
# leaves, which would be produced from the block hash tree.
# So we need to validate the block hash tree first. If
# successful, then bht[0] will contain the root for the
# shnum, which will be a leaf in the share hash tree, which
# will allow us to validate the rest of the tree.
try:
self.share_hash_tree.set_hashes(hashes=sharehashes,
leaves={reader.shnum: bht[0]})
except (hashtree.BadHashError, hashtree.NotEnoughHashesError, \
IndexError) as e:
raise CorruptShareError(server, reader.shnum,
"corrupt hashes: %s" % e)
self.log('share %d is valid for segment %d' % (reader.shnum, segnum))
return {reader.shnum: (block, salt)}
def _got_data(self, results, blocknum):
precondition(blocknum < self.num_blocks, self, blocknum,
self.num_blocks)
sharehashes, blockhashes, blockdata = results
try:
sharehashes = dict(sharehashes)
except ValueError as le:
le.args = tuple(le.args + (sharehashes, ))
raise
blockhashes = dict(enumerate(blockhashes))
candidate_share_hash = None # in case we log it in the except block below
blockhash = None # in case we log it in the except block below
try:
if self.share_hash_tree.needed_hashes(self.sharenum):
# This will raise exception if the values being passed do not
# match the root node of self.share_hash_tree.
try:
self.share_hash_tree.set_hashes(sharehashes)
except IndexError as le:
# Weird -- sharehashes contained index numbers outside of
# the range that fit into this hash tree.
raise BadOrMissingHash(le)
# To validate a block we need the root of the block hash tree,
# which is also one of the leafs of the share hash tree, and is
# called "the share hash".
if not self.block_hash_tree[0]: # empty -- no root node yet
# Get the share hash from the share hash tree.
share_hash = self.share_hash_tree.get_leaf(self.sharenum)
if not share_hash:
# No root node in block_hash_tree and also the share hash
# wasn't sent by the server.
raise hashtree.NotEnoughHashesError
self.block_hash_tree.set_hashes({0: share_hash})
if self.block_hash_tree.needed_hashes(blocknum):
self.block_hash_tree.set_hashes(blockhashes)
blockhash = block_hash(blockdata)
self.block_hash_tree.set_hashes(leaves={blocknum: blockhash})
#self.log("checking block_hash(shareid=%d, blocknum=%d) len=%d "
# "%r .. %r: %s" %
# (self.sharenum, blocknum, len(blockdata),
# blockdata[:50], blockdata[-50:], base32.b2a(blockhash)))
except (hashtree.BadHashError, hashtree.NotEnoughHashesError) as le:
# log.WEIRD: indicates undetected disk/network error, or more
# likely a programming error
self.log("hash failure in block=%d, shnum=%d on %s" %
(blocknum, self.sharenum, self.bucket))
if self.block_hash_tree.needed_hashes(blocknum):
self.log(
""" failure occurred when checking the block_hash_tree.
This suggests that either the block data was bad, or that the
block hashes we received along with it were bad.""")
else:
self.log(""" the failure probably occurred when checking the
share_hash_tree, which suggests that the share hashes we
received from the remote peer were bad.""")
self.log(" have candidate_share_hash: %s" %
bool(candidate_share_hash))
self.log(" block length: %d" % len(blockdata))
self.log(" block hash: %s" % base32.b2a_or_none(blockhash))
if len(blockdata) < 100:
self.log(" block data: %r" % (blockdata, ))
else:
self.log(" block data start/end: %r .. %r" %
(blockdata[:50], blockdata[-50:]))
self.log(" share hash tree:\n" + self.share_hash_tree.dump())
self.log(" block hash tree:\n" + self.block_hash_tree.dump())
lines = []
for i, h in sorted(sharehashes.items()):
lines.append("%3d: %s" % (i, base32.b2a_or_none(h)))
self.log(" sharehashes:\n" + "\n".join(lines) + "\n")
lines = []
for i, h in blockhashes.items():
lines.append("%3d: %s" % (i, base32.b2a_or_none(h)))
log.msg(" blockhashes:\n" + "\n".join(lines) + "\n")
raise BadOrMissingHash(le)
# If we made it here, the block is good. If the hash trees didn't
# like what they saw, they would have raised a BadHashError, causing
# our caller to see a Failure and thus ignore this block (as well as
# dropping this bucket).
return blockdata
class Retrieve:
# this class is currently single-use. Eventually (in MDMF) we will make
# it multi-use, in which case you can call download(range) multiple
# times, and each will have a separate response chain. However the
# Retrieve object will remain tied to a specific version of the file, and
# will use a single ServerMap instance.
implements(IPushProducer)
def __init__(self, filenode, storage_broker, servermap, verinfo,
fetch_privkey=False, verify=False):
self._node = filenode
_assert(self._node.get_pubkey())
self._storage_broker = storage_broker
self._storage_index = filenode.get_storage_index()
_assert(self._node.get_readkey())
self._last_failure = None
prefix = si_b2a(self._storage_index)[:5]
self._log_number = log.msg("Retrieve(%s): starting" % prefix)
self._running = True
self._decoding = False
self._bad_shares = set()
self.servermap = servermap
self.verinfo = verinfo
# TODO: make it possible to use self.verinfo.datalength instead
(seqnum, root_hash, IV, segsize, datalength, k, N, prefix,
offsets_tuple) = self.verinfo
self._data_length = datalength
# during repair, we may be called upon to grab the private key, since
# it wasn't picked up during a verify=False checker run, and we'll
# need it for repair to generate a new version.
self._need_privkey = verify or (fetch_privkey
and not self._node.get_privkey())
if self._need_privkey:
# TODO: Evaluate the need for this. We'll use it if we want
# to limit how many queries are on the wire for the privkey
# at once.
self._privkey_query_markers = [] # one Marker for each time we've
# tried to get the privkey.
# verify means that we are using the downloader logic to verify all
# of our shares. This tells the downloader a few things.
#
# 1. We need to download all of the shares.
# 2. We don't need to decode or decrypt the shares, since our
# caller doesn't care about the plaintext, only the
# information about which shares are or are not valid.
# 3. When we are validating readers, we need to validate the
# signature on the prefix. Do we? We already do this in the
# servermap update?
self._verify = verify
self._status = RetrieveStatus()
self._status.set_storage_index(self._storage_index)
self._status.set_helper(False)
self._status.set_progress(0.0)
self._status.set_active(True)
self._status.set_size(datalength)
self._status.set_encoding(k, N)
self.readers = {}
self._stopped = False
self._pause_deferred = None
self._offset = None
self._read_length = None
self.log("got seqnum %d" % self.verinfo[0])
def get_status(self):
return self._status
def log(self, *args, **kwargs):
if "parent" not in kwargs:
kwargs["parent"] = self._log_number
if "facility" not in kwargs:
kwargs["facility"] = "tahoe.mutable.retrieve"
return log.msg(*args, **kwargs)
def _set_current_status(self, state):
seg = "%d/%d" % (self._current_segment, self._last_segment)
self._status.set_status("segment %s (%s)" % (seg, state))
###################
# IPushProducer
def pauseProducing(self):
"""
I am called by my download target if we have produced too much
data for it to handle. I make the downloader stop producing new
data until my resumeProducing method is called.
"""
if self._pause_deferred is not None:
return
# fired when the download is unpaused.
self._old_status = self._status.get_status()
self._set_current_status("paused")
self._pause_deferred = defer.Deferred()
def resumeProducing(self):
"""
I am called by my download target once it is ready to begin
receiving data again.
"""
if self._pause_deferred is None:
return
p = self._pause_deferred
self._pause_deferred = None
self._status.set_status(self._old_status)
eventually(p.callback, None)
def stopProducing(self):
self._stopped = True
self.resumeProducing()
def _check_for_paused(self, res):
"""
I am called just before a write to the consumer. I return a
Deferred that eventually fires with the data that is to be
written to the consumer. If the download has not been paused,
the Deferred fires immediately. Otherwise, the Deferred fires
when the downloader is unpaused.
"""
if self._pause_deferred is not None:
d = defer.Deferred()
self._pause_deferred.addCallback(lambda ignored: d.callback(res))
return d
return res
def _check_for_stopped(self, res):
if self._stopped:
raise DownloadStopped("our Consumer called stopProducing()")
return res
def download(self, consumer=None, offset=0, size=None):
precondition(self._verify or IConsumer.providedBy(consumer))
if size is None:
size = self._data_length - offset
if self._verify:
_assert(size == self._data_length, (size, self._data_length))
self.log("starting download")
self._done_deferred = defer.Deferred()
if consumer:
self._consumer = consumer
# we provide IPushProducer, so streaming=True, per IConsumer.
self._consumer.registerProducer(self, streaming=True)
self._started = time.time()
self._started_fetching = time.time()
if size == 0:
# short-circuit the rest of the process
self._done()
else:
self._start_download(consumer, offset, size)
return self._done_deferred
def _start_download(self, consumer, offset, size):
precondition((0 <= offset < self._data_length)
and (size > 0)
and (offset+size <= self._data_length),
(offset, size, self._data_length))
self._offset = offset
self._read_length = size
self._setup_encoding_parameters()
self._setup_download()
# The download process beyond this is a state machine.
# _add_active_servers will select the servers that we want to use
# for the download, and then attempt to start downloading. After
# each segment, it will check for doneness, reacting to broken
# servers and corrupt shares as necessary. If it runs out of good
# servers before downloading all of the segments, _done_deferred
# will errback. Otherwise, it will eventually callback with the
# contents of the mutable file.
self.loop()
def loop(self):
d = fireEventually(None) # avoid #237 recursion limit problem
d.addCallback(lambda ign: self._activate_enough_servers())
d.addCallback(lambda ign: self._download_current_segment())
# when we're done, _download_current_segment will call _done. If we
# aren't, it will call loop() again.
d.addErrback(self._error)
def _setup_download(self):
self._status.set_status("Retrieving Shares")
# how many shares do we need?
(seqnum,
root_hash,
IV,
segsize,
datalength,
k,
N,
prefix,
offsets_tuple) = self.verinfo
# first, which servers can we use?
versionmap = self.servermap.make_versionmap()
shares = versionmap[self.verinfo]
# this sharemap is consumed as we decide to send requests
self.remaining_sharemap = DictOfSets()
for (shnum, server, timestamp) in shares:
self.remaining_sharemap.add(shnum, server)
# Reuse the SlotReader from the servermap.
key = (self.verinfo, server.get_serverid(),
self._storage_index, shnum)
if key in self.servermap.proxies:
reader = self.servermap.proxies[key]
else:
reader = MDMFSlotReadProxy(server.get_rref(),
self._storage_index, shnum, None)
reader.server = server
self.readers[shnum] = reader
if len(self.remaining_sharemap) < k:
self._raise_notenoughshareserror()
self.shares = {} # maps shnum to validated blocks
self._active_readers = [] # list of active readers for this dl.
self._block_hash_trees = {} # shnum => hashtree
for i in xrange(self._total_shares):
# So we don't have to do this later.
self._block_hash_trees[i] = hashtree.IncompleteHashTree(self._num_segments)
# We need one share hash tree for the entire file; its leaves
# are the roots of the block hash trees for the shares that
# comprise it, and its root is in the verinfo.
self.share_hash_tree = hashtree.IncompleteHashTree(N)
self.share_hash_tree.set_hashes({0: root_hash})
def decode(self, blocks_and_salts, segnum):
"""
I am a helper method that the mutable file update process uses
as a shortcut to decode and decrypt the segments that it needs
to fetch in order to perform a file update. I take in a
collection of blocks and salts, and pick some of those to make a
segment with. I return the plaintext associated with that
segment.
"""
# We don't need the block hash trees in this case.
self._block_hash_trees = None
self._offset = 0
self._read_length = self._data_length
self._setup_encoding_parameters()
# _decode_blocks() expects the output of a gatherResults that
# contains the outputs of _validate_block() (each of which is a dict
# mapping shnum to (block,salt) bytestrings).
d = self._decode_blocks([blocks_and_salts], segnum)
d.addCallback(self._decrypt_segment)
return d
def _setup_encoding_parameters(self):
"""
I set up the encoding parameters, including k, n, the number
of segments associated with this file, and the segment decoders.
"""
(seqnum,
root_hash,
IV,
segsize,
datalength,
k,
n,
known_prefix,
offsets_tuple) = self.verinfo
self._required_shares = k
self._total_shares = n
self._segment_size = segsize
#self._data_length = datalength # set during __init__()
if not IV:
self._version = MDMF_VERSION
else:
self._version = SDMF_VERSION
if datalength and segsize:
self._num_segments = mathutil.div_ceil(datalength, segsize)
self._tail_data_size = datalength % segsize
else:
self._num_segments = 0
self._tail_data_size = 0
self._segment_decoder = codec.CRSDecoder()
self._segment_decoder.set_params(segsize, k, n)
if not self._tail_data_size:
self._tail_data_size = segsize
self._tail_segment_size = mathutil.next_multiple(self._tail_data_size,
self._required_shares)
if self._tail_segment_size == self._segment_size:
self._tail_decoder = self._segment_decoder
else:
self._tail_decoder = codec.CRSDecoder()
self._tail_decoder.set_params(self._tail_segment_size,
self._required_shares,
self._total_shares)
self.log("got encoding parameters: "
"k: %d "
"n: %d "
"%d segments of %d bytes each (%d byte tail segment)" % \
(k, n, self._num_segments, self._segment_size,
self._tail_segment_size))
# Our last task is to tell the downloader where to start and
# where to stop. We use three parameters for that:
# - self._start_segment: the segment that we need to start
# downloading from.
# - self._current_segment: the next segment that we need to
# download.
# - self._last_segment: The last segment that we were asked to
# download.
#
# We say that the download is complete when
# self._current_segment > self._last_segment. We use
# self._start_segment and self._last_segment to know when to
# strip things off of segments, and how much to strip.
if self._offset:
self.log("got offset: %d" % self._offset)
# our start segment is the first segment containing the
# offset we were given.
start = self._offset // self._segment_size
_assert(start <= self._num_segments,
start=start, num_segments=self._num_segments,
offset=self._offset, segment_size=self._segment_size)
self._start_segment = start
self.log("got start segment: %d" % self._start_segment)
else:
self._start_segment = 0
# We might want to read only part of the file, and need to figure out
# where to stop reading. Our end segment is the last segment
# containing part of the segment that we were asked to read.
_assert(self._read_length > 0, self._read_length)
end_data = self._offset + self._read_length
# We don't actually need to read the byte at end_data, but the one
# before it.
end = (end_data - 1) // self._segment_size
_assert(0 <= end < self._num_segments,
end=end, num_segments=self._num_segments,
end_data=end_data, offset=self._offset,
read_length=self._read_length, segment_size=self._segment_size)
self._last_segment = end
self.log("got end segment: %d" % self._last_segment)
self._current_segment = self._start_segment
def _activate_enough_servers(self):
"""
I populate self._active_readers with enough active readers to
retrieve the contents of this mutable file. I am called before
downloading starts, and (eventually) after each validation
error, connection error, or other problem in the download.
"""
# TODO: It would be cool to investigate other heuristics for
# reader selection. For instance, the cost (in time the user
# spends waiting for their file) of selecting a really slow server
# that happens to have a primary share is probably more than
# selecting a really fast server that doesn't have a primary
# share. Maybe the servermap could be extended to provide this
# information; it could keep track of latency information while
# it gathers more important data, and then this routine could
# use that to select active readers.
#
# (these and other questions would be easier to answer with a
# robust, configurable tahoe-lafs simulator, which modeled node
# failures, differences in node speed, and other characteristics
# that we expect storage servers to have. You could have
# presets for really stable grids (like allmydata.com),
# friendnets, make it easy to configure your own settings, and
# then simulate the effect of big changes on these use cases
# instead of just reasoning about what the effect might be. Out
# of scope for MDMF, though.)
# XXX: Why don't format= log messages work here?
known_shnums = set(self.remaining_sharemap.keys())
used_shnums = set([r.shnum for r in self._active_readers])
unused_shnums = known_shnums - used_shnums
if self._verify:
new_shnums = unused_shnums # use them all
elif len(self._active_readers) < self._required_shares:
# need more shares
more = self._required_shares - len(self._active_readers)
# We favor lower numbered shares, since FEC is faster with
# primary shares than with other shares, and lower-numbered
# shares are more likely to be primary than higher numbered
# shares.
new_shnums = sorted(unused_shnums)[:more]
if len(new_shnums) < more:
# We don't have enough readers to retrieve the file; fail.
self._raise_notenoughshareserror()
else:
new_shnums = []
self.log("adding %d new servers to the active list" % len(new_shnums))
for shnum in new_shnums:
reader = self.readers[shnum]
self._active_readers.append(reader)
self.log("added reader for share %d" % shnum)
# Each time we add a reader, we check to see if we need the
# private key. If we do, we politely ask for it and then continue
# computing. If we find that we haven't gotten it at the end of
# segment decoding, then we'll take more drastic measures.
if self._need_privkey and not self._node.is_readonly():
d = reader.get_encprivkey()
d.addCallback(self._try_to_validate_privkey, reader, reader.server)
# XXX: don't just drop the Deferred. We need error-reporting
# but not flow-control here.
def _try_to_validate_prefix(self, prefix, reader):
"""
I check that the prefix returned by a candidate server for
retrieval matches the prefix that the servermap knows about
(and, hence, the prefix that was validated earlier). If it does,
I return True, which means that I approve of the use of the
candidate server for segment retrieval. If it doesn't, I return
False, which means that another server must be chosen.
"""
(seqnum,
root_hash,
IV,
segsize,
datalength,
k,
N,
known_prefix,
offsets_tuple) = self.verinfo
if known_prefix != prefix:
self.log("prefix from share %d doesn't match" % reader.shnum)
raise UncoordinatedWriteError("Mismatched prefix -- this could "
"indicate an uncoordinated write")
# Otherwise, we're okay -- no issues.
def _mark_bad_share(self, server, shnum, reader, f):
"""
I mark the given (server, shnum) as a bad share, which means that it
will not be used anywhere else.
There are several reasons to want to mark something as a bad
share. These include:
- A connection error to the server.
- A mismatched prefix (that is, a prefix that does not match
our local conception of the version information string).
- A failing block hash, salt hash, share hash, or other
integrity check.
This method will ensure that readers that we wish to mark bad
(for these reasons or other reasons) are not used for the rest
of the download. Additionally, it will attempt to tell the
remote server (with no guarantee of success) that its share is
corrupt.
"""
self.log("marking share %d on server %s as bad" % \
(shnum, server.get_name()))
prefix = self.verinfo[-2]
self.servermap.mark_bad_share(server, shnum, prefix)
self._bad_shares.add((server, shnum, f))
self._status.add_problem(server, f)
self._last_failure = f
# Remove the reader from _active_readers
self._active_readers.remove(reader)
for shnum in list(self.remaining_sharemap.keys()):
self.remaining_sharemap.discard(shnum, reader.server)
if f.check(BadShareError):
self.notify_server_corruption(server, shnum, str(f.value))
def _download_current_segment(self):
"""
I download, validate, decode, decrypt, and assemble the segment
that this Retrieve is currently responsible for downloading.
"""
if self._current_segment > self._last_segment:
# No more segments to download, we're done.
self.log("got plaintext, done")
return self._done()
elif self._verify and len(self._active_readers) == 0:
self.log("no more good shares, no need to keep verifying")
return self._done()
self.log("on segment %d of %d" %
(self._current_segment + 1, self._num_segments))
d = self._process_segment(self._current_segment)
d.addCallback(lambda ign: self.loop())
return d
def _process_segment(self, segnum):
"""
I download, validate, decode, and decrypt one segment of the
file that this Retrieve is retrieving. This means coordinating
the process of getting k blocks of that file, validating them,
assembling them into one segment with the decoder, and then
decrypting them.
"""
self.log("processing segment %d" % segnum)
# TODO: The old code uses a marker. Should this code do that
# too? What did the Marker do?
# We need to ask each of our active readers for its block and
# salt. We will then validate those. If validation is
# successful, we will assemble the results into plaintext.
ds = []
for reader in self._active_readers:
started = time.time()
d1 = reader.get_block_and_salt(segnum)
d2,d3 = self._get_needed_hashes(reader, segnum)
d = deferredutil.gatherResults([d1,d2,d3])
d.addCallback(self._validate_block, segnum, reader, reader.server, started)
# _handle_bad_share takes care of recoverable errors (by dropping
# that share and returning None). Any other errors (i.e. code
# bugs) are passed through and cause the retrieve to fail.
d.addErrback(self._handle_bad_share, [reader])
ds.append(d)
dl = deferredutil.gatherResults(ds)
if self._verify:
dl.addCallback(lambda ignored: "")
dl.addCallback(self._set_segment)
else:
dl.addCallback(self._maybe_decode_and_decrypt_segment, segnum)
return dl
def _maybe_decode_and_decrypt_segment(self, results, segnum):
"""
I take the results of fetching and validating the blocks from
_process_segment. If validation and fetching succeeded without
incident, I will proceed with decoding and decryption. Otherwise, I
will do nothing.
"""
self.log("trying to decode and decrypt segment %d" % segnum)
# 'results' is the output of a gatherResults set up in
# _process_segment(). Each component Deferred will either contain the
# non-Failure output of _validate_block() for a single block (i.e.
# {segnum:(block,salt)}), or None if _validate_block threw an
# exception and _validation_or_decoding_failed handled it (by
# dropping that server).
if None in results:
self.log("some validation operations failed; not proceeding")
return defer.succeed(None)
self.log("everything looks ok, building segment %d" % segnum)
d = self._decode_blocks(results, segnum)
d.addCallback(self._decrypt_segment)
# check to see whether we've been paused before writing
# anything.
d.addCallback(self._check_for_paused)
d.addCallback(self._check_for_stopped)
d.addCallback(self._set_segment)
return d
def _set_segment(self, segment):
"""
Given a plaintext segment, I register that segment with the
target that is handling the file download.
"""
self.log("got plaintext for segment %d" % self._current_segment)
if self._read_length == 0:
self.log("on first+last segment, size=0, using 0 bytes")
segment = b""
if self._current_segment == self._last_segment:
# trim off the tail
wanted = (self._offset + self._read_length) % self._segment_size
if wanted != 0:
self.log("on the last segment: using first %d bytes" % wanted)
segment = segment[:wanted]
else:
self.log("on the last segment: using all %d bytes" %
len(segment))
if self._current_segment == self._start_segment:
# Trim off the head, if offset != 0. This should also work if
# start==last, because we trim the tail first.
skip = self._offset % self._segment_size
self.log("on the first segment: skipping first %d bytes" % skip)
segment = segment[skip:]
if not self._verify:
self._consumer.write(segment)
else:
# we don't care about the plaintext if we are doing a verify.
segment = None
self._current_segment += 1
def _handle_bad_share(self, f, readers):
"""
I am called when a block or a salt fails to correctly validate, or when
the decryption or decoding operation fails for some reason. I react to
this failure by notifying the remote server of corruption, and then
removing the remote server from further activity.
"""
# these are the errors we can tolerate: by giving up on this share
# and finding others to replace it. Any other errors (i.e. coding
# bugs) are re-raised, causing the download to fail.
f.trap(DeadReferenceError, RemoteException, BadShareError)
# DeadReferenceError happens when we try to fetch data from a server
# that has gone away. RemoteException happens if the server had an
# internal error. BadShareError encompasses: (UnknownVersionError,
# LayoutInvalid, struct.error) which happen when we get obviously
# wrong data, and CorruptShareError which happens later, when we
# perform integrity checks on the data.
precondition(isinstance(readers, list), readers)
bad_shnums = [reader.shnum for reader in readers]
self.log("validation or decoding failed on share(s) %s, server(s) %s "
", segment %d: %s" % \
(bad_shnums, readers, self._current_segment, str(f)))
for reader in readers:
self._mark_bad_share(reader.server, reader.shnum, reader, f)
return None
def _validate_block(self, results, segnum, reader, server, started):
"""
I validate a block from one share on a remote server.
"""
# Grab the part of the block hash tree that is necessary to
# validate this block, then generate the block hash root.
self.log("validating share %d for segment %d" % (reader.shnum,
segnum))
elapsed = time.time() - started
self._status.add_fetch_timing(server, elapsed)
self._set_current_status("validating blocks")
block_and_salt, blockhashes, sharehashes = results
block, salt = block_and_salt
_assert(type(block) is str, (block, salt))
blockhashes = dict(enumerate(blockhashes))
self.log("the reader gave me the following blockhashes: %s" % \
blockhashes.keys())
self.log("the reader gave me the following sharehashes: %s" % \
sharehashes.keys())
bht = self._block_hash_trees[reader.shnum]
if bht.needed_hashes(segnum, include_leaf=True):
try:
bht.set_hashes(blockhashes)
except (hashtree.BadHashError, hashtree.NotEnoughHashesError, \
IndexError), e:
raise CorruptShareError(server,
reader.shnum,
"block hash tree failure: %s" % e)
if self._version == MDMF_VERSION:
blockhash = hashutil.block_hash(salt + block)
else:
blockhash = hashutil.block_hash(block)
# If this works without an error, then validation is
# successful.
try:
bht.set_hashes(leaves={segnum: blockhash})
except (hashtree.BadHashError, hashtree.NotEnoughHashesError, \
IndexError), e:
raise CorruptShareError(server,
reader.shnum,
"block hash tree failure: %s" % e)
def _generate_shares(self, shares_and_shareids):
# this sets self.shares and self.root_hash
self.log("_generate_shares")
self._status.set_status("Generating Shares")
started = time.time()
# we should know these by now
privkey = self._privkey
encprivkey = self._encprivkey
pubkey = self._pubkey
(shares, share_ids) = shares_and_shareids
assert len(shares) == len(share_ids)
assert len(shares) == self.total_shares
all_shares = {}
block_hash_trees = {}
share_hash_leaves = [None] * len(shares)
for i in range(len(shares)):
share_data = shares[i]
shnum = share_ids[i]
all_shares[shnum] = share_data
# build the block hash tree. SDMF has only one leaf.
leaves = [hashutil.block_hash(share_data)]
t = hashtree.HashTree(leaves)
block_hash_trees[shnum] = list(t)
share_hash_leaves[shnum] = t[0]
for leaf in share_hash_leaves:
assert leaf is not None
share_hash_tree = hashtree.HashTree(share_hash_leaves)
share_hash_chain = {}
for shnum in range(self.total_shares):
needed_hashes = share_hash_tree.needed_hashes(shnum)
share_hash_chain[shnum] = dict([(i, share_hash_tree[i])
for i in needed_hashes])
root_hash = share_hash_tree[0]
assert len(root_hash) == 32
self.log("my new root_hash is %s" % base32.b2a(root_hash))
self._new_version_info = (self._new_seqnum, root_hash, self.salt)
prefix = pack_prefix(self._new_seqnum, root_hash, self.salt,
self.required_shares, self.total_shares,
self.segment_size, len(self.newdata))
# now pack the beginning of the share. All shares are the same up
# to the signature, then they have divergent share hash chains,
# then completely different block hash trees + salt + share data,
# then they all share the same encprivkey at the end. The sizes
# of everything are the same for all shares.
sign_started = time.time()
signature = privkey.sign(prefix)
self._status.timings["sign"] = time.time() - sign_started
verification_key = pubkey.serialize()
final_shares = {}
for shnum in range(self.total_shares):
final_share = pack_share(prefix, verification_key, signature,
share_hash_chain[shnum],
block_hash_trees[shnum],
all_shares[shnum], encprivkey)
final_shares[shnum] = final_share
elapsed = time.time() - started
self._status.timings["pack"] = elapsed
self.shares = final_shares
self.root_hash = root_hash
# we also need to build up the version identifier for what we're
# pushing. Extract the offsets from one of our shares.
assert final_shares
offsets = unpack_header(final_shares.values()[0])[-1]
offsets_tuple = tuple([(key, value) for key, value in offsets.items()])
verinfo = (self._new_seqnum, root_hash, self.salt, self.segment_size,
len(self.newdata), self.required_shares, self.total_shares,
prefix, offsets_tuple)
self.versioninfo = verinfo
def check_block(self, segnum, block):
assert self._block_hash_tree_is_authoritative
h = hashutil.block_hash(block)
# this may raise BadHashError or NotEnoughHashesError
self._block_hash_tree.set_hashes(leaves={segnum: h})
def _generate_shares(self, shares_and_shareids):
# this sets self.shares and self.root_hash
self.log("_generate_shares")
self._status.set_status("Generating Shares")
started = time.time()
# we should know these by now
privkey = self._privkey
encprivkey = self._encprivkey
pubkey = self._pubkey
(shares, share_ids) = shares_and_shareids
assert len(shares) == len(share_ids)
assert len(shares) == self.total_shares
all_shares = {}
block_hash_trees = {}
share_hash_leaves = [None] * len(shares)
for i in range(len(shares)):
share_data = shares[i]
shnum = share_ids[i]
all_shares[shnum] = share_data
# build the block hash tree. SDMF has only one leaf.
leaves = [hashutil.block_hash(share_data)]
t = hashtree.HashTree(leaves)
block_hash_trees[shnum] = list(t)
share_hash_leaves[shnum] = t[0]
for leaf in share_hash_leaves:
assert leaf is not None
share_hash_tree = hashtree.HashTree(share_hash_leaves)
share_hash_chain = {}
for shnum in range(self.total_shares):
needed_hashes = share_hash_tree.needed_hashes(shnum)
share_hash_chain[shnum] = dict( [ (i, share_hash_tree[i])
for i in needed_hashes ] )
root_hash = share_hash_tree[0]
assert len(root_hash) == 32
self.log("my new root_hash is %s" % base32.b2a(root_hash))
self._new_version_info = (self._new_seqnum, root_hash, self.salt)
prefix = pack_prefix(self._new_seqnum, root_hash, self.salt,
self.required_shares, self.total_shares,
self.segment_size, len(self.newdata))
# now pack the beginning of the share. All shares are the same up
# to the signature, then they have divergent share hash chains,
# then completely different block hash trees + salt + share data,
# then they all share the same encprivkey at the end. The sizes
# of everything are the same for all shares.
sign_started = time.time()
signature = privkey.sign(prefix)
self._status.timings["sign"] = time.time() - sign_started
verification_key = pubkey.serialize()
final_shares = {}
for shnum in range(self.total_shares):
final_share = pack_share(prefix,
verification_key,
signature,
share_hash_chain[shnum],
block_hash_trees[shnum],
all_shares[shnum],
encprivkey)
final_shares[shnum] = final_share
elapsed = time.time() - started
self._status.timings["pack"] = elapsed
self.shares = final_shares
self.root_hash = root_hash
# we also need to build up the version identifier for what we're
# pushing. Extract the offsets from one of our shares.
assert final_shares
offsets = unpack_header(final_shares.values()[0])[-1]
offsets_tuple = tuple( [(key,value) for key,value in offsets.items()] )
verinfo = (self._new_seqnum, root_hash, self.salt,
self.segment_size, len(self.newdata),
self.required_shares, self.total_shares,
prefix, offsets_tuple)
self.versioninfo = verinfo
def _validate_block(self, results, segnum, reader, server, started):
"""
I validate a block from one share on a remote server.
"""
# Grab the part of the block hash tree that is necessary to
# validate this block, then generate the block hash root.
self.log("validating share %d for segment %d" % (reader.shnum,
segnum))
elapsed = time.time() - started
self._status.add_fetch_timing(server, elapsed)
self._set_current_status("validating blocks")
block_and_salt, blockhashes, sharehashes = results
block, salt = block_and_salt
_assert(type(block) is str, (block, salt))
blockhashes = dict(enumerate(blockhashes))
self.log("the reader gave me the following blockhashes: %s" % \
blockhashes.keys())
self.log("the reader gave me the following sharehashes: %s" % \
sharehashes.keys())
bht = self._block_hash_trees[reader.shnum]
if bht.needed_hashes(segnum, include_leaf=True):
try:
bht.set_hashes(blockhashes)
except (hashtree.BadHashError, hashtree.NotEnoughHashesError, \
IndexError) as e:
raise CorruptShareError(server,
reader.shnum,
"block hash tree failure: %s" % e)
if self._version == MDMF_VERSION:
blockhash = hashutil.block_hash(salt + block)
else:
blockhash = hashutil.block_hash(block)
# If this works without an error, then validation is
# successful.
try:
bht.set_hashes(leaves={segnum: blockhash})
except (hashtree.BadHashError, hashtree.NotEnoughHashesError, \
IndexError) as e:
raise CorruptShareError(server,
reader.shnum,
"block hash tree failure: %s" % e)
# Reaching this point means that we know that this segment
# is correct. Now we need to check to see whether the share
# hash chain is also correct.
# SDMF wrote share hash chains that didn't contain the
# leaves, which would be produced from the block hash tree.
# So we need to validate the block hash tree first. If
# successful, then bht[0] will contain the root for the
# shnum, which will be a leaf in the share hash tree, which
# will allow us to validate the rest of the tree.
try:
self.share_hash_tree.set_hashes(hashes=sharehashes,
leaves={reader.shnum: bht[0]})
except (hashtree.BadHashError, hashtree.NotEnoughHashesError, \
IndexError) as e:
raise CorruptShareError(server,
reader.shnum,
"corrupt hashes: %s" % e)
self.log('share %d is valid for segment %d' % (reader.shnum,
segnum))
return {reader.shnum: (block, salt)}
