tahoe-lafs/src/allmydata/test/test_encode.py

411 lines
16 KiB
Python
Raw Normal View History

from zope.interface import implementer
from twisted.trial import unittest
from twisted.internet import defer
from twisted.python.failure import Failure
from foolscap.api import fireEventually
from allmydata import uri
from allmydata.immutable import encode, upload, checker
from allmydata.util import hashutil
from allmydata.util.assertutil import _assert
from allmydata.util.consumer import download_to_data
from allmydata.interfaces import IStorageBucketWriter, IStorageBucketReader
from allmydata.test.no_network import GridTestMixin
class LostPeerError(Exception):
pass
def flip_bit(good): # flips the last bit
return good[:-1] + chr(ord(good[-1]) ^ 0x01)
@implementer(IStorageBucketWriter, IStorageBucketReader)
class FakeBucketReaderWriterProxy(object):
# these are used for both reading and writing
def __init__(self, mode="good", peerid="peer"):
self.mode = mode
self.blocks = {}
self.plaintext_hashes = []
self.crypttext_hashes = []
self.block_hashes = None
self.share_hashes = None
self.closed = False
self.peerid = peerid
def get_peerid(self):
return self.peerid
immutable: refactor downloader to be more reusable for checker/verifier/repairer (and better) The code for validating the share hash tree and the block hash tree has been rewritten to make sure it handles all cases, to share metadata about the file (such as the share hash tree, block hash trees, and UEB) among different share downloads, and not to require hashes to be stored on the server unnecessarily, such as the roots of the block hash trees (not needed since they are also the leaves of the share hash tree), and the root of the share hash tree (not needed since it is also included in the UEB). It also passes the latest tests including handling corrupted shares well. ValidatedReadBucketProxy takes a share_hash_tree argument to its constructor, which is a reference to a share hash tree shared by all ValidatedReadBucketProxies for that immutable file download. ValidatedReadBucketProxy requires the block_size and share_size to be provided in its constructor, and it then uses those to compute the offsets and lengths of blocks when it needs them, instead of reading those values out of the share. The user of ValidatedReadBucketProxy therefore has to have first used a ValidatedExtendedURIProxy to compute those two values from the validated contents of the URI. This is pleasingly simplifies safety analysis: the client knows which span of bytes corresponds to a given block from the validated URI data, rather than from the unvalidated data stored on the storage server. It also simplifies unit testing of verifier/repairer, because now it doesn't care about the contents of the "share size" and "block size" fields in the share. It does not relieve the need for share data v2 layout, because we still need to store and retrieve the offsets of the fields which come after the share data, therefore we still need to use share data v2 with its 8-byte fields if we want to store share data larger than about 2^32. Specify which subset of the block hashes and share hashes you need while downloading a particular share. In the future this will hopefully be used to fetch only a subset, for network efficiency, but currently all of them are fetched, regardless of which subset you specify. ReadBucketProxy hides the question of whether it has "started" or not (sent a request to the server to get metadata) from its user. Download is optimized to do as few roundtrips and as few requests as possible, hopefully speeding up download a bit.
2009-01-05 16:51:45 +00:00
def _start(self):
if self.mode == "lost-early":
f = Failure(LostPeerError("I went away early"))
return fireEventually(f)
return defer.succeed(self)
immutable: refactor downloader to be more reusable for checker/verifier/repairer (and better) The code for validating the share hash tree and the block hash tree has been rewritten to make sure it handles all cases, to share metadata about the file (such as the share hash tree, block hash trees, and UEB) among different share downloads, and not to require hashes to be stored on the server unnecessarily, such as the roots of the block hash trees (not needed since they are also the leaves of the share hash tree), and the root of the share hash tree (not needed since it is also included in the UEB). It also passes the latest tests including handling corrupted shares well. ValidatedReadBucketProxy takes a share_hash_tree argument to its constructor, which is a reference to a share hash tree shared by all ValidatedReadBucketProxies for that immutable file download. ValidatedReadBucketProxy requires the block_size and share_size to be provided in its constructor, and it then uses those to compute the offsets and lengths of blocks when it needs them, instead of reading those values out of the share. The user of ValidatedReadBucketProxy therefore has to have first used a ValidatedExtendedURIProxy to compute those two values from the validated contents of the URI. This is pleasingly simplifies safety analysis: the client knows which span of bytes corresponds to a given block from the validated URI data, rather than from the unvalidated data stored on the storage server. It also simplifies unit testing of verifier/repairer, because now it doesn't care about the contents of the "share size" and "block size" fields in the share. It does not relieve the need for share data v2 layout, because we still need to store and retrieve the offsets of the fields which come after the share data, therefore we still need to use share data v2 with its 8-byte fields if we want to store share data larger than about 2^32. Specify which subset of the block hashes and share hashes you need while downloading a particular share. In the future this will hopefully be used to fetch only a subset, for network efficiency, but currently all of them are fetched, regardless of which subset you specify. ReadBucketProxy hides the question of whether it has "started" or not (sent a request to the server to get metadata) from its user. Download is optimized to do as few roundtrips and as few requests as possible, hopefully speeding up download a bit.
2009-01-05 16:51:45 +00:00
def put_header(self):
return self._start()
def put_block(self, segmentnum, data):
if self.mode == "lost-early":
f = Failure(LostPeerError("I went away early"))
return fireEventually(f)
def _try():
assert not self.closed
assert segmentnum not in self.blocks
if self.mode == "lost" and segmentnum >= 1:
raise LostPeerError("I'm going away now")
self.blocks[segmentnum] = data
return defer.maybeDeferred(_try)
def put_crypttext_hashes(self, hashes):
def _try():
assert not self.closed
assert not self.crypttext_hashes
self.crypttext_hashes = hashes
return defer.maybeDeferred(_try)
def put_block_hashes(self, blockhashes):
def _try():
assert not self.closed
assert self.block_hashes is None
self.block_hashes = blockhashes
return defer.maybeDeferred(_try)
def put_share_hashes(self, sharehashes):
def _try():
assert not self.closed
assert self.share_hashes is None
self.share_hashes = sharehashes
return defer.maybeDeferred(_try)
2007-06-08 22:59:16 +00:00
def put_uri_extension(self, uri_extension):
def _try():
assert not self.closed
self.uri_extension = uri_extension
return defer.maybeDeferred(_try)
def close(self):
def _try():
assert not self.closed
self.closed = True
return defer.maybeDeferred(_try)
def abort(self):
return defer.succeed(None)
immutable: refactor downloader to be more reusable for checker/verifier/repairer (and better) The code for validating the share hash tree and the block hash tree has been rewritten to make sure it handles all cases, to share metadata about the file (such as the share hash tree, block hash trees, and UEB) among different share downloads, and not to require hashes to be stored on the server unnecessarily, such as the roots of the block hash trees (not needed since they are also the leaves of the share hash tree), and the root of the share hash tree (not needed since it is also included in the UEB). It also passes the latest tests including handling corrupted shares well. ValidatedReadBucketProxy takes a share_hash_tree argument to its constructor, which is a reference to a share hash tree shared by all ValidatedReadBucketProxies for that immutable file download. ValidatedReadBucketProxy requires the block_size and share_size to be provided in its constructor, and it then uses those to compute the offsets and lengths of blocks when it needs them, instead of reading those values out of the share. The user of ValidatedReadBucketProxy therefore has to have first used a ValidatedExtendedURIProxy to compute those two values from the validated contents of the URI. This is pleasingly simplifies safety analysis: the client knows which span of bytes corresponds to a given block from the validated URI data, rather than from the unvalidated data stored on the storage server. It also simplifies unit testing of verifier/repairer, because now it doesn't care about the contents of the "share size" and "block size" fields in the share. It does not relieve the need for share data v2 layout, because we still need to store and retrieve the offsets of the fields which come after the share data, therefore we still need to use share data v2 with its 8-byte fields if we want to store share data larger than about 2^32. Specify which subset of the block hashes and share hashes you need while downloading a particular share. In the future this will hopefully be used to fetch only a subset, for network efficiency, but currently all of them are fetched, regardless of which subset you specify. ReadBucketProxy hides the question of whether it has "started" or not (sent a request to the server to get metadata) from its user. Download is optimized to do as few roundtrips and as few requests as possible, hopefully speeding up download a bit.
2009-01-05 16:51:45 +00:00
def get_block_data(self, blocknum, blocksize, size):
d = self._start()
def _try(unused=None):
assert isinstance(blocknum, (int, long))
if self.mode == "bad block":
return flip_bit(self.blocks[blocknum])
return self.blocks[blocknum]
immutable: refactor downloader to be more reusable for checker/verifier/repairer (and better) The code for validating the share hash tree and the block hash tree has been rewritten to make sure it handles all cases, to share metadata about the file (such as the share hash tree, block hash trees, and UEB) among different share downloads, and not to require hashes to be stored on the server unnecessarily, such as the roots of the block hash trees (not needed since they are also the leaves of the share hash tree), and the root of the share hash tree (not needed since it is also included in the UEB). It also passes the latest tests including handling corrupted shares well. ValidatedReadBucketProxy takes a share_hash_tree argument to its constructor, which is a reference to a share hash tree shared by all ValidatedReadBucketProxies for that immutable file download. ValidatedReadBucketProxy requires the block_size and share_size to be provided in its constructor, and it then uses those to compute the offsets and lengths of blocks when it needs them, instead of reading those values out of the share. The user of ValidatedReadBucketProxy therefore has to have first used a ValidatedExtendedURIProxy to compute those two values from the validated contents of the URI. This is pleasingly simplifies safety analysis: the client knows which span of bytes corresponds to a given block from the validated URI data, rather than from the unvalidated data stored on the storage server. It also simplifies unit testing of verifier/repairer, because now it doesn't care about the contents of the "share size" and "block size" fields in the share. It does not relieve the need for share data v2 layout, because we still need to store and retrieve the offsets of the fields which come after the share data, therefore we still need to use share data v2 with its 8-byte fields if we want to store share data larger than about 2^32. Specify which subset of the block hashes and share hashes you need while downloading a particular share. In the future this will hopefully be used to fetch only a subset, for network efficiency, but currently all of them are fetched, regardless of which subset you specify. ReadBucketProxy hides the question of whether it has "started" or not (sent a request to the server to get metadata) from its user. Download is optimized to do as few roundtrips and as few requests as possible, hopefully speeding up download a bit.
2009-01-05 16:51:45 +00:00
d.addCallback(_try)
return d
def get_plaintext_hashes(self):
immutable: refactor downloader to be more reusable for checker/verifier/repairer (and better) The code for validating the share hash tree and the block hash tree has been rewritten to make sure it handles all cases, to share metadata about the file (such as the share hash tree, block hash trees, and UEB) among different share downloads, and not to require hashes to be stored on the server unnecessarily, such as the roots of the block hash trees (not needed since they are also the leaves of the share hash tree), and the root of the share hash tree (not needed since it is also included in the UEB). It also passes the latest tests including handling corrupted shares well. ValidatedReadBucketProxy takes a share_hash_tree argument to its constructor, which is a reference to a share hash tree shared by all ValidatedReadBucketProxies for that immutable file download. ValidatedReadBucketProxy requires the block_size and share_size to be provided in its constructor, and it then uses those to compute the offsets and lengths of blocks when it needs them, instead of reading those values out of the share. The user of ValidatedReadBucketProxy therefore has to have first used a ValidatedExtendedURIProxy to compute those two values from the validated contents of the URI. This is pleasingly simplifies safety analysis: the client knows which span of bytes corresponds to a given block from the validated URI data, rather than from the unvalidated data stored on the storage server. It also simplifies unit testing of verifier/repairer, because now it doesn't care about the contents of the "share size" and "block size" fields in the share. It does not relieve the need for share data v2 layout, because we still need to store and retrieve the offsets of the fields which come after the share data, therefore we still need to use share data v2 with its 8-byte fields if we want to store share data larger than about 2^32. Specify which subset of the block hashes and share hashes you need while downloading a particular share. In the future this will hopefully be used to fetch only a subset, for network efficiency, but currently all of them are fetched, regardless of which subset you specify. ReadBucketProxy hides the question of whether it has "started" or not (sent a request to the server to get metadata) from its user. Download is optimized to do as few roundtrips and as few requests as possible, hopefully speeding up download a bit.
2009-01-05 16:51:45 +00:00
d = self._start()
def _try(unused=None):
hashes = self.plaintext_hashes[:]
return hashes
immutable: refactor downloader to be more reusable for checker/verifier/repairer (and better) The code for validating the share hash tree and the block hash tree has been rewritten to make sure it handles all cases, to share metadata about the file (such as the share hash tree, block hash trees, and UEB) among different share downloads, and not to require hashes to be stored on the server unnecessarily, such as the roots of the block hash trees (not needed since they are also the leaves of the share hash tree), and the root of the share hash tree (not needed since it is also included in the UEB). It also passes the latest tests including handling corrupted shares well. ValidatedReadBucketProxy takes a share_hash_tree argument to its constructor, which is a reference to a share hash tree shared by all ValidatedReadBucketProxies for that immutable file download. ValidatedReadBucketProxy requires the block_size and share_size to be provided in its constructor, and it then uses those to compute the offsets and lengths of blocks when it needs them, instead of reading those values out of the share. The user of ValidatedReadBucketProxy therefore has to have first used a ValidatedExtendedURIProxy to compute those two values from the validated contents of the URI. This is pleasingly simplifies safety analysis: the client knows which span of bytes corresponds to a given block from the validated URI data, rather than from the unvalidated data stored on the storage server. It also simplifies unit testing of verifier/repairer, because now it doesn't care about the contents of the "share size" and "block size" fields in the share. It does not relieve the need for share data v2 layout, because we still need to store and retrieve the offsets of the fields which come after the share data, therefore we still need to use share data v2 with its 8-byte fields if we want to store share data larger than about 2^32. Specify which subset of the block hashes and share hashes you need while downloading a particular share. In the future this will hopefully be used to fetch only a subset, for network efficiency, but currently all of them are fetched, regardless of which subset you specify. ReadBucketProxy hides the question of whether it has "started" or not (sent a request to the server to get metadata) from its user. Download is optimized to do as few roundtrips and as few requests as possible, hopefully speeding up download a bit.
2009-01-05 16:51:45 +00:00
d.addCallback(_try)
return d
def get_crypttext_hashes(self):
immutable: refactor downloader to be more reusable for checker/verifier/repairer (and better) The code for validating the share hash tree and the block hash tree has been rewritten to make sure it handles all cases, to share metadata about the file (such as the share hash tree, block hash trees, and UEB) among different share downloads, and not to require hashes to be stored on the server unnecessarily, such as the roots of the block hash trees (not needed since they are also the leaves of the share hash tree), and the root of the share hash tree (not needed since it is also included in the UEB). It also passes the latest tests including handling corrupted shares well. ValidatedReadBucketProxy takes a share_hash_tree argument to its constructor, which is a reference to a share hash tree shared by all ValidatedReadBucketProxies for that immutable file download. ValidatedReadBucketProxy requires the block_size and share_size to be provided in its constructor, and it then uses those to compute the offsets and lengths of blocks when it needs them, instead of reading those values out of the share. The user of ValidatedReadBucketProxy therefore has to have first used a ValidatedExtendedURIProxy to compute those two values from the validated contents of the URI. This is pleasingly simplifies safety analysis: the client knows which span of bytes corresponds to a given block from the validated URI data, rather than from the unvalidated data stored on the storage server. It also simplifies unit testing of verifier/repairer, because now it doesn't care about the contents of the "share size" and "block size" fields in the share. It does not relieve the need for share data v2 layout, because we still need to store and retrieve the offsets of the fields which come after the share data, therefore we still need to use share data v2 with its 8-byte fields if we want to store share data larger than about 2^32. Specify which subset of the block hashes and share hashes you need while downloading a particular share. In the future this will hopefully be used to fetch only a subset, for network efficiency, but currently all of them are fetched, regardless of which subset you specify. ReadBucketProxy hides the question of whether it has "started" or not (sent a request to the server to get metadata) from its user. Download is optimized to do as few roundtrips and as few requests as possible, hopefully speeding up download a bit.
2009-01-05 16:51:45 +00:00
d = self._start()
def _try(unused=None):
hashes = self.crypttext_hashes[:]
if self.mode == "bad crypttext hashroot":
hashes[0] = flip_bit(hashes[0])
if self.mode == "bad crypttext hash":
hashes[1] = flip_bit(hashes[1])
return hashes
immutable: refactor downloader to be more reusable for checker/verifier/repairer (and better) The code for validating the share hash tree and the block hash tree has been rewritten to make sure it handles all cases, to share metadata about the file (such as the share hash tree, block hash trees, and UEB) among different share downloads, and not to require hashes to be stored on the server unnecessarily, such as the roots of the block hash trees (not needed since they are also the leaves of the share hash tree), and the root of the share hash tree (not needed since it is also included in the UEB). It also passes the latest tests including handling corrupted shares well. ValidatedReadBucketProxy takes a share_hash_tree argument to its constructor, which is a reference to a share hash tree shared by all ValidatedReadBucketProxies for that immutable file download. ValidatedReadBucketProxy requires the block_size and share_size to be provided in its constructor, and it then uses those to compute the offsets and lengths of blocks when it needs them, instead of reading those values out of the share. The user of ValidatedReadBucketProxy therefore has to have first used a ValidatedExtendedURIProxy to compute those two values from the validated contents of the URI. This is pleasingly simplifies safety analysis: the client knows which span of bytes corresponds to a given block from the validated URI data, rather than from the unvalidated data stored on the storage server. It also simplifies unit testing of verifier/repairer, because now it doesn't care about the contents of the "share size" and "block size" fields in the share. It does not relieve the need for share data v2 layout, because we still need to store and retrieve the offsets of the fields which come after the share data, therefore we still need to use share data v2 with its 8-byte fields if we want to store share data larger than about 2^32. Specify which subset of the block hashes and share hashes you need while downloading a particular share. In the future this will hopefully be used to fetch only a subset, for network efficiency, but currently all of them are fetched, regardless of which subset you specify. ReadBucketProxy hides the question of whether it has "started" or not (sent a request to the server to get metadata) from its user. Download is optimized to do as few roundtrips and as few requests as possible, hopefully speeding up download a bit.
2009-01-05 16:51:45 +00:00
d.addCallback(_try)
return d
immutable: refactor downloader to be more reusable for checker/verifier/repairer (and better) The code for validating the share hash tree and the block hash tree has been rewritten to make sure it handles all cases, to share metadata about the file (such as the share hash tree, block hash trees, and UEB) among different share downloads, and not to require hashes to be stored on the server unnecessarily, such as the roots of the block hash trees (not needed since they are also the leaves of the share hash tree), and the root of the share hash tree (not needed since it is also included in the UEB). It also passes the latest tests including handling corrupted shares well. ValidatedReadBucketProxy takes a share_hash_tree argument to its constructor, which is a reference to a share hash tree shared by all ValidatedReadBucketProxies for that immutable file download. ValidatedReadBucketProxy requires the block_size and share_size to be provided in its constructor, and it then uses those to compute the offsets and lengths of blocks when it needs them, instead of reading those values out of the share. The user of ValidatedReadBucketProxy therefore has to have first used a ValidatedExtendedURIProxy to compute those two values from the validated contents of the URI. This is pleasingly simplifies safety analysis: the client knows which span of bytes corresponds to a given block from the validated URI data, rather than from the unvalidated data stored on the storage server. It also simplifies unit testing of verifier/repairer, because now it doesn't care about the contents of the "share size" and "block size" fields in the share. It does not relieve the need for share data v2 layout, because we still need to store and retrieve the offsets of the fields which come after the share data, therefore we still need to use share data v2 with its 8-byte fields if we want to store share data larger than about 2^32. Specify which subset of the block hashes and share hashes you need while downloading a particular share. In the future this will hopefully be used to fetch only a subset, for network efficiency, but currently all of them are fetched, regardless of which subset you specify. ReadBucketProxy hides the question of whether it has "started" or not (sent a request to the server to get metadata) from its user. Download is optimized to do as few roundtrips and as few requests as possible, hopefully speeding up download a bit.
2009-01-05 16:51:45 +00:00
def get_block_hashes(self, at_least_these=()):
d = self._start()
def _try(unused=None):
if self.mode == "bad blockhash":
hashes = self.block_hashes[:]
hashes[1] = flip_bit(hashes[1])
return hashes
return self.block_hashes
immutable: refactor downloader to be more reusable for checker/verifier/repairer (and better) The code for validating the share hash tree and the block hash tree has been rewritten to make sure it handles all cases, to share metadata about the file (such as the share hash tree, block hash trees, and UEB) among different share downloads, and not to require hashes to be stored on the server unnecessarily, such as the roots of the block hash trees (not needed since they are also the leaves of the share hash tree), and the root of the share hash tree (not needed since it is also included in the UEB). It also passes the latest tests including handling corrupted shares well. ValidatedReadBucketProxy takes a share_hash_tree argument to its constructor, which is a reference to a share hash tree shared by all ValidatedReadBucketProxies for that immutable file download. ValidatedReadBucketProxy requires the block_size and share_size to be provided in its constructor, and it then uses those to compute the offsets and lengths of blocks when it needs them, instead of reading those values out of the share. The user of ValidatedReadBucketProxy therefore has to have first used a ValidatedExtendedURIProxy to compute those two values from the validated contents of the URI. This is pleasingly simplifies safety analysis: the client knows which span of bytes corresponds to a given block from the validated URI data, rather than from the unvalidated data stored on the storage server. It also simplifies unit testing of verifier/repairer, because now it doesn't care about the contents of the "share size" and "block size" fields in the share. It does not relieve the need for share data v2 layout, because we still need to store and retrieve the offsets of the fields which come after the share data, therefore we still need to use share data v2 with its 8-byte fields if we want to store share data larger than about 2^32. Specify which subset of the block hashes and share hashes you need while downloading a particular share. In the future this will hopefully be used to fetch only a subset, for network efficiency, but currently all of them are fetched, regardless of which subset you specify. ReadBucketProxy hides the question of whether it has "started" or not (sent a request to the server to get metadata) from its user. Download is optimized to do as few roundtrips and as few requests as possible, hopefully speeding up download a bit.
2009-01-05 16:51:45 +00:00
d.addCallback(_try)
return d
immutable: refactor downloader to be more reusable for checker/verifier/repairer (and better) The code for validating the share hash tree and the block hash tree has been rewritten to make sure it handles all cases, to share metadata about the file (such as the share hash tree, block hash trees, and UEB) among different share downloads, and not to require hashes to be stored on the server unnecessarily, such as the roots of the block hash trees (not needed since they are also the leaves of the share hash tree), and the root of the share hash tree (not needed since it is also included in the UEB). It also passes the latest tests including handling corrupted shares well. ValidatedReadBucketProxy takes a share_hash_tree argument to its constructor, which is a reference to a share hash tree shared by all ValidatedReadBucketProxies for that immutable file download. ValidatedReadBucketProxy requires the block_size and share_size to be provided in its constructor, and it then uses those to compute the offsets and lengths of blocks when it needs them, instead of reading those values out of the share. The user of ValidatedReadBucketProxy therefore has to have first used a ValidatedExtendedURIProxy to compute those two values from the validated contents of the URI. This is pleasingly simplifies safety analysis: the client knows which span of bytes corresponds to a given block from the validated URI data, rather than from the unvalidated data stored on the storage server. It also simplifies unit testing of verifier/repairer, because now it doesn't care about the contents of the "share size" and "block size" fields in the share. It does not relieve the need for share data v2 layout, because we still need to store and retrieve the offsets of the fields which come after the share data, therefore we still need to use share data v2 with its 8-byte fields if we want to store share data larger than about 2^32. Specify which subset of the block hashes and share hashes you need while downloading a particular share. In the future this will hopefully be used to fetch only a subset, for network efficiency, but currently all of them are fetched, regardless of which subset you specify. ReadBucketProxy hides the question of whether it has "started" or not (sent a request to the server to get metadata) from its user. Download is optimized to do as few roundtrips and as few requests as possible, hopefully speeding up download a bit.
2009-01-05 16:51:45 +00:00
def get_share_hashes(self, at_least_these=()):
d = self._start()
def _try(unused=None):
if self.mode == "bad sharehash":
hashes = self.share_hashes[:]
hashes[1] = (hashes[1][0], flip_bit(hashes[1][1]))
return hashes
if self.mode == "missing sharehash":
# one sneaky attack would be to pretend we don't know our own
# sharehash, which could manage to frame someone else.
# download.py is supposed to guard against this case.
return []
return self.share_hashes
immutable: refactor downloader to be more reusable for checker/verifier/repairer (and better) The code for validating the share hash tree and the block hash tree has been rewritten to make sure it handles all cases, to share metadata about the file (such as the share hash tree, block hash trees, and UEB) among different share downloads, and not to require hashes to be stored on the server unnecessarily, such as the roots of the block hash trees (not needed since they are also the leaves of the share hash tree), and the root of the share hash tree (not needed since it is also included in the UEB). It also passes the latest tests including handling corrupted shares well. ValidatedReadBucketProxy takes a share_hash_tree argument to its constructor, which is a reference to a share hash tree shared by all ValidatedReadBucketProxies for that immutable file download. ValidatedReadBucketProxy requires the block_size and share_size to be provided in its constructor, and it then uses those to compute the offsets and lengths of blocks when it needs them, instead of reading those values out of the share. The user of ValidatedReadBucketProxy therefore has to have first used a ValidatedExtendedURIProxy to compute those two values from the validated contents of the URI. This is pleasingly simplifies safety analysis: the client knows which span of bytes corresponds to a given block from the validated URI data, rather than from the unvalidated data stored on the storage server. It also simplifies unit testing of verifier/repairer, because now it doesn't care about the contents of the "share size" and "block size" fields in the share. It does not relieve the need for share data v2 layout, because we still need to store and retrieve the offsets of the fields which come after the share data, therefore we still need to use share data v2 with its 8-byte fields if we want to store share data larger than about 2^32. Specify which subset of the block hashes and share hashes you need while downloading a particular share. In the future this will hopefully be used to fetch only a subset, for network efficiency, but currently all of them are fetched, regardless of which subset you specify. ReadBucketProxy hides the question of whether it has "started" or not (sent a request to the server to get metadata) from its user. Download is optimized to do as few roundtrips and as few requests as possible, hopefully speeding up download a bit.
2009-01-05 16:51:45 +00:00
d.addCallback(_try)
return d
2007-06-08 22:59:16 +00:00
def get_uri_extension(self):
immutable: refactor downloader to be more reusable for checker/verifier/repairer (and better) The code for validating the share hash tree and the block hash tree has been rewritten to make sure it handles all cases, to share metadata about the file (such as the share hash tree, block hash trees, and UEB) among different share downloads, and not to require hashes to be stored on the server unnecessarily, such as the roots of the block hash trees (not needed since they are also the leaves of the share hash tree), and the root of the share hash tree (not needed since it is also included in the UEB). It also passes the latest tests including handling corrupted shares well. ValidatedReadBucketProxy takes a share_hash_tree argument to its constructor, which is a reference to a share hash tree shared by all ValidatedReadBucketProxies for that immutable file download. ValidatedReadBucketProxy requires the block_size and share_size to be provided in its constructor, and it then uses those to compute the offsets and lengths of blocks when it needs them, instead of reading those values out of the share. The user of ValidatedReadBucketProxy therefore has to have first used a ValidatedExtendedURIProxy to compute those two values from the validated contents of the URI. This is pleasingly simplifies safety analysis: the client knows which span of bytes corresponds to a given block from the validated URI data, rather than from the unvalidated data stored on the storage server. It also simplifies unit testing of verifier/repairer, because now it doesn't care about the contents of the "share size" and "block size" fields in the share. It does not relieve the need for share data v2 layout, because we still need to store and retrieve the offsets of the fields which come after the share data, therefore we still need to use share data v2 with its 8-byte fields if we want to store share data larger than about 2^32. Specify which subset of the block hashes and share hashes you need while downloading a particular share. In the future this will hopefully be used to fetch only a subset, for network efficiency, but currently all of them are fetched, regardless of which subset you specify. ReadBucketProxy hides the question of whether it has "started" or not (sent a request to the server to get metadata) from its user. Download is optimized to do as few roundtrips and as few requests as possible, hopefully speeding up download a bit.
2009-01-05 16:51:45 +00:00
d = self._start()
def _try(unused=None):
if self.mode == "bad uri_extension":
return flip_bit(self.uri_extension)
return self.uri_extension
immutable: refactor downloader to be more reusable for checker/verifier/repairer (and better) The code for validating the share hash tree and the block hash tree has been rewritten to make sure it handles all cases, to share metadata about the file (such as the share hash tree, block hash trees, and UEB) among different share downloads, and not to require hashes to be stored on the server unnecessarily, such as the roots of the block hash trees (not needed since they are also the leaves of the share hash tree), and the root of the share hash tree (not needed since it is also included in the UEB). It also passes the latest tests including handling corrupted shares well. ValidatedReadBucketProxy takes a share_hash_tree argument to its constructor, which is a reference to a share hash tree shared by all ValidatedReadBucketProxies for that immutable file download. ValidatedReadBucketProxy requires the block_size and share_size to be provided in its constructor, and it then uses those to compute the offsets and lengths of blocks when it needs them, instead of reading those values out of the share. The user of ValidatedReadBucketProxy therefore has to have first used a ValidatedExtendedURIProxy to compute those two values from the validated contents of the URI. This is pleasingly simplifies safety analysis: the client knows which span of bytes corresponds to a given block from the validated URI data, rather than from the unvalidated data stored on the storage server. It also simplifies unit testing of verifier/repairer, because now it doesn't care about the contents of the "share size" and "block size" fields in the share. It does not relieve the need for share data v2 layout, because we still need to store and retrieve the offsets of the fields which come after the share data, therefore we still need to use share data v2 with its 8-byte fields if we want to store share data larger than about 2^32. Specify which subset of the block hashes and share hashes you need while downloading a particular share. In the future this will hopefully be used to fetch only a subset, for network efficiency, but currently all of them are fetched, regardless of which subset you specify. ReadBucketProxy hides the question of whether it has "started" or not (sent a request to the server to get metadata) from its user. Download is optimized to do as few roundtrips and as few requests as possible, hopefully speeding up download a bit.
2009-01-05 16:51:45 +00:00
d.addCallback(_try)
return d
def make_data(length):
data = "happy happy joy joy" * 100
assert length <= len(data)
return data[:length]
download: refactor handling of URI Extension Block and crypttext hash tree, simplify things Refactor into a class the logic of asking each server in turn until one of them gives an answer that validates. It is called ValidatedThingObtainer. Refactor the downloading and verification of the URI Extension Block into a class named ValidatedExtendedURIProxy. The new logic of validating UEBs is minimalist: it doesn't require the UEB to contain any unncessary information, but of course it still accepts such information for backwards compatibility (so that this new download code is able to download files uploaded with old, and for that matter with current, upload code). The new logic of validating UEBs follows the practice of doing all validation up front. This practice advises one to isolate the validation of incoming data into one place, so that all of the rest of the code can assume only valid data. If any redundant information is present in the UEB+URI, the new code cross-checks and asserts that it is all fully consistent. This closes some issues where the uploader could have uploaded inconsistent redundant data, which would probably have caused the old downloader to simply reject that download after getting a Python exception, but perhaps could have caused greater harm to the old downloader. I removed the notion of selecting an erasure codec from codec.py based on the string that was passed in the UEB. Currently "crs" is the only such string that works, so "_assert(codec_name == 'crs')" is simpler and more explicit. This is also in keeping with the "validate up front" strategy -- now if someone sets a different string than "crs" in their UEB, the downloader will reject the download in the "validate this UEB" function instead of in a separate "select the codec instance" function. I removed the code to check plaintext hashes and plaintext Merkle Trees. Uploaders do not produce this information any more (since it potentially exposes confidential information about the file), and the unit tests for it were disabled. The downloader before this patch would check that plaintext hash or plaintext merkle tree if they were present, but not complain if they were absent. The new downloader in this patch complains if they are present and doesn't check them. (We might in the future re-introduce such hashes over the plaintext, but encrypt the hashes which are stored in the UEB to preserve confidentiality. This would be a double- check on the correctness of our own source code -- the current Merkle Tree over the ciphertext is already sufficient to guarantee the integrity of the download unless there is a bug in our Merkle Tree or AES implementation.) This patch increases the lines-of-code count by 8 (from 17,770 to 17,778), and reduces the uncovered-by-tests lines-of-code count by 24 (from 1408 to 1384). Those numbers would be more meaningful if we omitted src/allmydata/util/ from the test-coverage statistics.
2008-12-05 15:17:54 +00:00
class ValidatedExtendedURIProxy(unittest.TestCase):
timeout = 240 # It takes longer than 120 seconds on Francois's arm box.
download: refactor handling of URI Extension Block and crypttext hash tree, simplify things Refactor into a class the logic of asking each server in turn until one of them gives an answer that validates. It is called ValidatedThingObtainer. Refactor the downloading and verification of the URI Extension Block into a class named ValidatedExtendedURIProxy. The new logic of validating UEBs is minimalist: it doesn't require the UEB to contain any unncessary information, but of course it still accepts such information for backwards compatibility (so that this new download code is able to download files uploaded with old, and for that matter with current, upload code). The new logic of validating UEBs follows the practice of doing all validation up front. This practice advises one to isolate the validation of incoming data into one place, so that all of the rest of the code can assume only valid data. If any redundant information is present in the UEB+URI, the new code cross-checks and asserts that it is all fully consistent. This closes some issues where the uploader could have uploaded inconsistent redundant data, which would probably have caused the old downloader to simply reject that download after getting a Python exception, but perhaps could have caused greater harm to the old downloader. I removed the notion of selecting an erasure codec from codec.py based on the string that was passed in the UEB. Currently "crs" is the only such string that works, so "_assert(codec_name == 'crs')" is simpler and more explicit. This is also in keeping with the "validate up front" strategy -- now if someone sets a different string than "crs" in their UEB, the downloader will reject the download in the "validate this UEB" function instead of in a separate "select the codec instance" function. I removed the code to check plaintext hashes and plaintext Merkle Trees. Uploaders do not produce this information any more (since it potentially exposes confidential information about the file), and the unit tests for it were disabled. The downloader before this patch would check that plaintext hash or plaintext merkle tree if they were present, but not complain if they were absent. The new downloader in this patch complains if they are present and doesn't check them. (We might in the future re-introduce such hashes over the plaintext, but encrypt the hashes which are stored in the UEB to preserve confidentiality. This would be a double- check on the correctness of our own source code -- the current Merkle Tree over the ciphertext is already sufficient to guarantee the integrity of the download unless there is a bug in our Merkle Tree or AES implementation.) This patch increases the lines-of-code count by 8 (from 17,770 to 17,778), and reduces the uncovered-by-tests lines-of-code count by 24 (from 1408 to 1384). Those numbers would be more meaningful if we omitted src/allmydata/util/ from the test-coverage statistics.
2008-12-05 15:17:54 +00:00
K = 4
M = 10
SIZE = 200
SEGSIZE = 72
_TMP = SIZE%SEGSIZE
if _TMP == 0:
_TMP = SEGSIZE
if _TMP % K != 0:
_TMP += (K - (_TMP % K))
TAIL_SEGSIZE = _TMP
_TMP = SIZE / SEGSIZE
if SIZE % SEGSIZE != 0:
_TMP += 1
NUM_SEGMENTS = _TMP
mindict = { 'segment_size': SEGSIZE,
'crypttext_root_hash': '0'*hashutil.CRYPTO_VAL_SIZE,
'share_root_hash': '1'*hashutil.CRYPTO_VAL_SIZE }
optional_consistent = { 'crypttext_hash': '2'*hashutil.CRYPTO_VAL_SIZE,
'codec_name': "crs",
'codec_params': "%d-%d-%d" % (SEGSIZE, K, M),
'tail_codec_params': "%d-%d-%d" % (TAIL_SEGSIZE, K, M),
'num_segments': NUM_SEGMENTS,
'size': SIZE,
'needed_shares': K,
'total_shares': M,
'plaintext_hash': "anything",
'plaintext_root_hash': "anything", }
# optional_inconsistent = { 'crypttext_hash': ('2'*(hashutil.CRYPTO_VAL_SIZE-1), "", 77),
optional_inconsistent = { 'crypttext_hash': (77,),
'codec_name': ("digital fountain", ""),
'codec_params': ("%d-%d-%d" % (SEGSIZE, K-1, M),
"%d-%d-%d" % (SEGSIZE-1, K, M),
"%d-%d-%d" % (SEGSIZE, K, M-1)),
'tail_codec_params': ("%d-%d-%d" % (TAIL_SEGSIZE, K-1, M),
"%d-%d-%d" % (TAIL_SEGSIZE-1, K, M),
"%d-%d-%d" % (TAIL_SEGSIZE, K, M-1)),
'num_segments': (NUM_SEGMENTS-1,),
'size': (SIZE-1,),
'needed_shares': (K-1,),
'total_shares': (M-1,), }
def _test(self, uebdict):
uebstring = uri.pack_extension(uebdict)
uebhash = hashutil.uri_extension_hash(uebstring)
fb = FakeBucketReaderWriterProxy()
fb.put_uri_extension(uebstring)
verifycap = uri.CHKFileVerifierURI(storage_index='x'*16, uri_extension_hash=uebhash, needed_shares=self.K, total_shares=self.M, size=self.SIZE)
vup = checker.ValidatedExtendedURIProxy(fb, verifycap)
download: refactor handling of URI Extension Block and crypttext hash tree, simplify things Refactor into a class the logic of asking each server in turn until one of them gives an answer that validates. It is called ValidatedThingObtainer. Refactor the downloading and verification of the URI Extension Block into a class named ValidatedExtendedURIProxy. The new logic of validating UEBs is minimalist: it doesn't require the UEB to contain any unncessary information, but of course it still accepts such information for backwards compatibility (so that this new download code is able to download files uploaded with old, and for that matter with current, upload code). The new logic of validating UEBs follows the practice of doing all validation up front. This practice advises one to isolate the validation of incoming data into one place, so that all of the rest of the code can assume only valid data. If any redundant information is present in the UEB+URI, the new code cross-checks and asserts that it is all fully consistent. This closes some issues where the uploader could have uploaded inconsistent redundant data, which would probably have caused the old downloader to simply reject that download after getting a Python exception, but perhaps could have caused greater harm to the old downloader. I removed the notion of selecting an erasure codec from codec.py based on the string that was passed in the UEB. Currently "crs" is the only such string that works, so "_assert(codec_name == 'crs')" is simpler and more explicit. This is also in keeping with the "validate up front" strategy -- now if someone sets a different string than "crs" in their UEB, the downloader will reject the download in the "validate this UEB" function instead of in a separate "select the codec instance" function. I removed the code to check plaintext hashes and plaintext Merkle Trees. Uploaders do not produce this information any more (since it potentially exposes confidential information about the file), and the unit tests for it were disabled. The downloader before this patch would check that plaintext hash or plaintext merkle tree if they were present, but not complain if they were absent. The new downloader in this patch complains if they are present and doesn't check them. (We might in the future re-introduce such hashes over the plaintext, but encrypt the hashes which are stored in the UEB to preserve confidentiality. This would be a double- check on the correctness of our own source code -- the current Merkle Tree over the ciphertext is already sufficient to guarantee the integrity of the download unless there is a bug in our Merkle Tree or AES implementation.) This patch increases the lines-of-code count by 8 (from 17,770 to 17,778), and reduces the uncovered-by-tests lines-of-code count by 24 (from 1408 to 1384). Those numbers would be more meaningful if we omitted src/allmydata/util/ from the test-coverage statistics.
2008-12-05 15:17:54 +00:00
return vup.start()
def _test_accept(self, uebdict):
return self._test(uebdict)
def _should_fail(self, res, expected_failures):
if isinstance(res, Failure):
res.trap(*expected_failures)
else:
self.fail("was supposed to raise %s, not get '%s'" % (expected_failures, res))
def _test_reject(self, uebdict):
d = self._test(uebdict)
d.addBoth(self._should_fail, (KeyError, checker.BadURIExtension))
download: refactor handling of URI Extension Block and crypttext hash tree, simplify things Refactor into a class the logic of asking each server in turn until one of them gives an answer that validates. It is called ValidatedThingObtainer. Refactor the downloading and verification of the URI Extension Block into a class named ValidatedExtendedURIProxy. The new logic of validating UEBs is minimalist: it doesn't require the UEB to contain any unncessary information, but of course it still accepts such information for backwards compatibility (so that this new download code is able to download files uploaded with old, and for that matter with current, upload code). The new logic of validating UEBs follows the practice of doing all validation up front. This practice advises one to isolate the validation of incoming data into one place, so that all of the rest of the code can assume only valid data. If any redundant information is present in the UEB+URI, the new code cross-checks and asserts that it is all fully consistent. This closes some issues where the uploader could have uploaded inconsistent redundant data, which would probably have caused the old downloader to simply reject that download after getting a Python exception, but perhaps could have caused greater harm to the old downloader. I removed the notion of selecting an erasure codec from codec.py based on the string that was passed in the UEB. Currently "crs" is the only such string that works, so "_assert(codec_name == 'crs')" is simpler and more explicit. This is also in keeping with the "validate up front" strategy -- now if someone sets a different string than "crs" in their UEB, the downloader will reject the download in the "validate this UEB" function instead of in a separate "select the codec instance" function. I removed the code to check plaintext hashes and plaintext Merkle Trees. Uploaders do not produce this information any more (since it potentially exposes confidential information about the file), and the unit tests for it were disabled. The downloader before this patch would check that plaintext hash or plaintext merkle tree if they were present, but not complain if they were absent. The new downloader in this patch complains if they are present and doesn't check them. (We might in the future re-introduce such hashes over the plaintext, but encrypt the hashes which are stored in the UEB to preserve confidentiality. This would be a double- check on the correctness of our own source code -- the current Merkle Tree over the ciphertext is already sufficient to guarantee the integrity of the download unless there is a bug in our Merkle Tree or AES implementation.) This patch increases the lines-of-code count by 8 (from 17,770 to 17,778), and reduces the uncovered-by-tests lines-of-code count by 24 (from 1408 to 1384). Those numbers would be more meaningful if we omitted src/allmydata/util/ from the test-coverage statistics.
2008-12-05 15:17:54 +00:00
return d
def test_accept_minimal(self):
return self._test_accept(self.mindict)
def test_reject_insufficient(self):
dl = []
for k in self.mindict.iterkeys():
insuffdict = self.mindict.copy()
del insuffdict[k]
d = self._test_reject(insuffdict)
dl.append(d)
return defer.DeferredList(dl)
def test_accept_optional(self):
dl = []
for k in self.optional_consistent.iterkeys():
mydict = self.mindict.copy()
mydict[k] = self.optional_consistent[k]
d = self._test_accept(mydict)
dl.append(d)
return defer.DeferredList(dl)
def test_reject_optional(self):
dl = []
for k in self.optional_inconsistent.iterkeys():
for v in self.optional_inconsistent[k]:
mydict = self.mindict.copy()
mydict[k] = v
d = self._test_reject(mydict)
dl.append(d)
return defer.DeferredList(dl)
class Encode(unittest.TestCase):
timeout = 2400 # It takes longer than 240 seconds on Zandr's ARM box.
def do_encode(self, max_segment_size, datalen, NUM_SHARES, NUM_SEGMENTS,
expected_block_hashes, expected_share_hashes):
data = make_data(datalen)
# force use of multiple segments
e = encode.Encoder()
u = upload.Data(data, convergence="some convergence string")
u.set_default_encoding_parameters({'max_segment_size': max_segment_size,
'k': 25, 'happy': 75, 'n': 100})
eu = upload.EncryptAnUploadable(u)
d = e.set_encrypted_uploadable(eu)
all_shareholders = []
def _ready(res):
k,happy,n = e.get_param("share_counts")
_assert(n == NUM_SHARES) # else we'll be completely confused
numsegs = e.get_param("num_segments")
_assert(numsegs == NUM_SEGMENTS, numsegs, NUM_SEGMENTS)
segsize = e.get_param("segment_size")
_assert( (NUM_SEGMENTS-1)*segsize < len(data) <= NUM_SEGMENTS*segsize,
NUM_SEGMENTS, segsize,
(NUM_SEGMENTS-1)*segsize, len(data), NUM_SEGMENTS*segsize)
shareholders = {}
servermap = {}
for shnum in range(NUM_SHARES):
download: refactor handling of URI Extension Block and crypttext hash tree, simplify things Refactor into a class the logic of asking each server in turn until one of them gives an answer that validates. It is called ValidatedThingObtainer. Refactor the downloading and verification of the URI Extension Block into a class named ValidatedExtendedURIProxy. The new logic of validating UEBs is minimalist: it doesn't require the UEB to contain any unncessary information, but of course it still accepts such information for backwards compatibility (so that this new download code is able to download files uploaded with old, and for that matter with current, upload code). The new logic of validating UEBs follows the practice of doing all validation up front. This practice advises one to isolate the validation of incoming data into one place, so that all of the rest of the code can assume only valid data. If any redundant information is present in the UEB+URI, the new code cross-checks and asserts that it is all fully consistent. This closes some issues where the uploader could have uploaded inconsistent redundant data, which would probably have caused the old downloader to simply reject that download after getting a Python exception, but perhaps could have caused greater harm to the old downloader. I removed the notion of selecting an erasure codec from codec.py based on the string that was passed in the UEB. Currently "crs" is the only such string that works, so "_assert(codec_name == 'crs')" is simpler and more explicit. This is also in keeping with the "validate up front" strategy -- now if someone sets a different string than "crs" in their UEB, the downloader will reject the download in the "validate this UEB" function instead of in a separate "select the codec instance" function. I removed the code to check plaintext hashes and plaintext Merkle Trees. Uploaders do not produce this information any more (since it potentially exposes confidential information about the file), and the unit tests for it were disabled. The downloader before this patch would check that plaintext hash or plaintext merkle tree if they were present, but not complain if they were absent. The new downloader in this patch complains if they are present and doesn't check them. (We might in the future re-introduce such hashes over the plaintext, but encrypt the hashes which are stored in the UEB to preserve confidentiality. This would be a double- check on the correctness of our own source code -- the current Merkle Tree over the ciphertext is already sufficient to guarantee the integrity of the download unless there is a bug in our Merkle Tree or AES implementation.) This patch increases the lines-of-code count by 8 (from 17,770 to 17,778), and reduces the uncovered-by-tests lines-of-code count by 24 (from 1408 to 1384). Those numbers would be more meaningful if we omitted src/allmydata/util/ from the test-coverage statistics.
2008-12-05 15:17:54 +00:00
peer = FakeBucketReaderWriterProxy()
shareholders[shnum] = peer
servermap.setdefault(shnum, set()).add(peer.get_peerid())
all_shareholders.append(peer)
e.set_shareholders(shareholders, servermap)
return e.start()
d.addCallback(_ready)
def _check(res):
verifycap = res
self.failUnless(isinstance(verifycap.uri_extension_hash, str))
self.failUnlessEqual(len(verifycap.uri_extension_hash), 32)
for i,peer in enumerate(all_shareholders):
self.failUnless(peer.closed)
self.failUnlessEqual(len(peer.blocks), NUM_SEGMENTS)
# each peer gets a full tree of block hashes. For 3 or 4
# segments, that's 7 hashes. For 5 segments it's 15 hashes.
self.failUnlessEqual(len(peer.block_hashes),
expected_block_hashes)
for h in peer.block_hashes:
self.failUnlessEqual(len(h), 32)
# each peer also gets their necessary chain of share hashes.
# For 100 shares (rounded up to 128 leaves), that's 8 hashes
self.failUnlessEqual(len(peer.share_hashes),
expected_share_hashes)
for (hashnum, h) in peer.share_hashes:
self.failUnless(isinstance(hashnum, int))
self.failUnlessEqual(len(h), 32)
d.addCallback(_check)
return d
def test_send_74(self):
# 3 segments (25, 25, 24)
return self.do_encode(25, 74, 100, 3, 7, 8)
def test_send_75(self):
# 3 segments (25, 25, 25)
return self.do_encode(25, 75, 100, 3, 7, 8)
def test_send_51(self):
# 3 segments (25, 25, 1)
return self.do_encode(25, 51, 100, 3, 7, 8)
def test_send_76(self):
# encode a 76 byte file (in 4 segments: 25,25,25,1) to 100 shares
return self.do_encode(25, 76, 100, 4, 7, 8)
def test_send_99(self):
# 4 segments: 25,25,25,24
return self.do_encode(25, 99, 100, 4, 7, 8)
def test_send_100(self):
# 4 segments: 25,25,25,25
return self.do_encode(25, 100, 100, 4, 7, 8)
def test_send_124(self):
# 5 segments: 25, 25, 25, 25, 24
return self.do_encode(25, 124, 100, 5, 15, 8)
def test_send_125(self):
# 5 segments: 25, 25, 25, 25, 25
return self.do_encode(25, 125, 100, 5, 15, 8)
def test_send_101(self):
# 5 segments: 25, 25, 25, 25, 1
return self.do_encode(25, 101, 100, 5, 15, 8)
class Roundtrip(GridTestMixin, unittest.TestCase):
# a series of 3*3 tests to check out edge conditions. One axis is how the
# plaintext is divided into segments: kn+(-1,0,1). Another way to express
# this is n%k == -1 or 0 or 1. For example, for 25-byte segments, we
# might test 74 bytes, 75 bytes, and 76 bytes.
# on the other axis is how many leaves in the block hash tree we wind up
# with, relative to a power of 2, so 2^a+(-1,0,1). Each segment turns
# into a single leaf. So we'd like to check out, e.g., 3 segments, 4
# segments, and 5 segments.
# that results in the following series of data lengths:
# 3 segs: 74, 75, 51
# 4 segs: 99, 100, 76
# 5 segs: 124, 125, 101
# all tests encode to 100 shares, which means the share hash tree will
# have 128 leaves, which means that buckets will be given an 8-long share
# hash chain
# all 3-segment files will have a 4-leaf blockhashtree, and thus expect
# to get 7 blockhashes. 4-segment files will also get 4-leaf block hash
# trees and 7 blockhashes. 5-segment files will get 8-leaf block hash
# trees, which gets 15 blockhashes.
def test_74(self): return self.do_test_size(74)
def test_75(self): return self.do_test_size(75)
def test_51(self): return self.do_test_size(51)
def test_99(self): return self.do_test_size(99)
def test_100(self): return self.do_test_size(100)
def test_76(self): return self.do_test_size(76)
def test_124(self): return self.do_test_size(124)
def test_125(self): return self.do_test_size(125)
def test_101(self): return self.do_test_size(101)
def upload(self, data):
u = upload.Data(data, None)
u.max_segment_size = 25
u.encoding_param_k = 25
u.encoding_param_happy = 1
u.encoding_param_n = 100
d = self.c0.upload(u)
d.addCallback(lambda ur: self.c0.create_node_from_uri(ur.get_uri()))
# returns a FileNode
return d
def do_test_size(self, size):
self.basedir = self.mktemp()
self.set_up_grid()
self.c0 = self.g.clients[0]
DATA = "p"*size
d = self.upload(DATA)
d.addCallback(lambda n: download_to_data(n))
def _downloaded(newdata):
self.failUnlessEqual(newdata, DATA)
d.addCallback(_downloaded)
return d