tahoe-lafs/src/allmydata/immutable/checker.py

279 lines
10 KiB
Python
Raw Normal View History

"""
Given a StorageIndex, count how many shares we can find.
This does no verification of the shares whatsoever. If the peer claims to
have the share, we believe them.
"""
from twisted.internet import defer
from twisted.python import log
from allmydata import storage
from allmydata.checker_results import CheckerResults
from allmydata.immutable import download
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
from allmydata.uri import CHKFileURI
2008-09-10 02:50:17 +00:00
from allmydata.util import hashutil
from allmydata.util.assertutil import precondition
class SimpleCHKFileChecker:
"""Return a list of (needed, total, found, sharemap), where sharemap maps
share number to a list of (binary) nodeids of the shareholders."""
def __init__(self, client, uri, storage_index, needed_shares, total_shares):
self.peer_getter = client.get_permuted_peers
self.needed_shares = needed_shares
self.total_shares = total_shares
self.found_shares = set()
self.uri = uri
self.storage_index = storage_index
self.sharemap = {}
self.responded = set()
'''
def check_synchronously(self, si):
# this is how we would write this class if we were using synchronous
# messages (or if we used promises).
found = set()
for (pmpeerid, peerid, connection) in self.peer_getter(storage_index):
buckets = connection.get_buckets(si)
found.update(buckets.keys())
return len(found)
'''
def start(self):
d = self._get_all_shareholders(self.storage_index)
d.addCallback(self._done)
return d
def _get_all_shareholders(self, storage_index):
dl = []
for (peerid, ss) in self.peer_getter("storage", storage_index):
d = ss.callRemote("get_buckets", storage_index)
d.addCallbacks(self._got_response, self._got_error,
callbackArgs=(peerid,))
dl.append(d)
return defer.DeferredList(dl)
def _got_response(self, buckets, peerid):
# buckets is a dict: maps shum to an rref of the server who holds it
self.found_shares.update(buckets.keys())
for k in buckets:
if k not in self.sharemap:
self.sharemap[k] = []
self.sharemap[k].append(peerid)
self.responded.add(peerid)
def _got_error(self, f):
if f.check(KeyError):
pass
log.err(f)
pass
def _done(self, res):
r = CheckerResults(self.uri, self.storage_index)
report = []
healthy = bool(len(self.found_shares) >= self.total_shares)
r.set_healthy(healthy)
recoverable = bool(len(self.found_shares) >= self.needed_shares)
r.set_recoverable(recoverable)
data = {"count-shares-good": len(self.found_shares),
"count-shares-needed": self.needed_shares,
"count-shares-expected": self.total_shares,
"count-wrong-shares": 0,
}
if recoverable:
data["count-recoverable-versions"] = 1
data["count-unrecoverable-versions"] = 0
else:
data["count-recoverable-versions"] = 0
data["count-unrecoverable-versions"] = 1
data["count-corrupt-shares"] = 0 # non-verifier doesn't see corruption
data["list-corrupt-shares"] = []
hosts = set()
sharemap = {}
for (shnum,nodeids) in self.sharemap.items():
hosts.update(nodeids)
sharemap[shnum] = nodeids
data["count-good-share-hosts"] = len(hosts)
data["servers-responding"] = list(self.responded)
data["sharemap"] = sharemap
r.set_data(data)
r.set_needs_rebalancing(bool( len(self.found_shares) > len(hosts) ))
#r.stuff = (self.needed_shares, self.total_shares,
# len(self.found_shares), self.sharemap)
if len(self.found_shares) < self.total_shares:
wanted = set(range(self.total_shares))
missing = wanted - self.found_shares
report.append("Missing shares: %s" %
",".join(["sh%d" % shnum
for shnum in sorted(missing)]))
r.set_report(report)
if healthy:
r.set_summary("Healthy")
else:
r.set_summary("Not Healthy")
# TODO: more detail
return r
2007-10-16 19:25:09 +00:00
class VerifyingOutput:
def __init__(self, total_length, results):
2007-10-16 19:25:09 +00:00
self._crypttext_hasher = hashutil.crypttext_hasher()
self.length = 0
self.total_length = total_length
self._segment_number = 0
self._crypttext_hash_tree = None
self._opened = False
self._results = results
results.set_healthy(False)
results.set_recoverable(False)
results.set_summary("Not Healthy")
2007-10-16 19:25:09 +00:00
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
def got_crypttext_hash_tree(self, crypttext_hashtree):
2007-10-16 19:25:09 +00:00
self._crypttext_hash_tree = crypttext_hashtree
def write_segment(self, crypttext):
self.length += len(crypttext)
self._crypttext_hasher.update(crypttext)
if self._crypttext_hash_tree:
ch = hashutil.crypttext_segment_hasher()
ch.update(crypttext)
crypttext_leaves = {self._segment_number: ch.digest()}
self._crypttext_hash_tree.set_hashes(leaves=crypttext_leaves)
self._segment_number += 1
def close(self):
self.crypttext_hash = self._crypttext_hasher.digest()
def finish(self):
self._results.set_healthy(True)
self._results.set_recoverable(True)
self._results.set_summary("Healthy")
# the return value of finish() is passed out of FileDownloader._done,
# but SimpleCHKFileVerifier overrides this with the CheckerResults
# instance instead.
2007-10-16 19:25:09 +00:00
class SimpleCHKFileVerifier(download.FileDownloader):
# this reconstructs the crypttext, which verifies that at least 'k' of
# the shareholders are around and have valid data. It does not check the
# remaining shareholders, and it cannot verify the plaintext.
check_plaintext_hash = False
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
def __init__(self, client, u, storage_index, k, N, size, ueb_hash):
precondition(isinstance(u, CHKFileURI), u)
download.FileDownloader.__init__(self, client, u, None);
2007-10-16 19:25:09 +00:00
self._client = client
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
self._uri = u
self._storage_index = storage_index
self._uri_extension_hash = ueb_hash
self._total_shares = N
self._size = size
self._num_needed_shares = k
2007-10-16 19:25:09 +00:00
self._si_s = storage.si_b2a(self._storage_index)
2007-11-20 02:07:10 +00:00
self.init_logging()
self._check_results = r = CheckerResults(self._uri, self._storage_index)
r.set_data({"count-shares-needed": k,
"count-shares-expected": N,
})
self._output = VerifyingOutput(self._size, r)
2007-10-16 19:25:09 +00:00
self._paused = False
self._stopped = False
self._results = None
2007-10-16 19:25:09 +00:00
self.active_buckets = {} # k: shnum, v: bucket
self._share_buckets = [] # list of (sharenum, bucket) tuples
self._share_vbuckets = {} # k: shnum, v: set of ValidatedBuckets
self._uri_extension_sources = []
self._uri_extension_data = None
self._fetch_failures = {"uri_extension": 0,
"plaintext_hashroot": 0,
"plaintext_hashtree": 0,
"crypttext_hashroot": 0,
"crypttext_hashtree": 0,
}
2007-11-20 02:07:10 +00:00
def init_logging(self):
self._log_prefix = prefix = storage.si_b2a(self._storage_index)[:5]
2007-11-20 02:07:10 +00:00
num = self._client.log("SimpleCHKFileVerifier(%s): starting" % prefix)
self._log_number = num
def log(self, *args, **kwargs):
if not "parent" in kwargs:
kwargs['parent'] = self._log_number
# add a prefix to the message, regardless of how it is expressed
prefix = "SimpleCHKFileVerifier(%s): " % self._log_prefix
if "format" in kwargs:
kwargs["format"] = prefix + kwargs["format"]
elif "message" in kwargs:
kwargs["message"] = prefix + kwargs["message"]
elif args:
m = prefix + args[0]
args = (m,) + args[1:]
return self._client.log(*args, **kwargs)
2007-11-20 02:07:10 +00:00
2007-10-16 19:25:09 +00:00
def start(self):
log.msg("starting download [%s]" % storage.si_b2a(self._storage_index)[:5])
2007-10-16 19:25:09 +00:00
# first step: who should we download from?
d = defer.maybeDeferred(self._get_all_shareholders)
d.addCallback(self._got_all_shareholders)
# now get the uri_extension block from somebody and validate it
d.addCallback(self._obtain_uri_extension)
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
d.addCallback(self._get_crypttext_hash_tree)
2007-10-16 19:25:09 +00:00
# once we know that, we can download blocks from everybody
d.addCallback(self._download_all_segments)
d.addCallback(self._done)
d.addCallbacks(self._verify_done, self._verify_failed)
2007-10-16 19:25:09 +00:00
return d
def _verify_done(self, ignored):
# TODO: The following results are just stubs, and need to be replaced
# with actual values. These exist to make things like deep-check not
# fail.
self._check_results.set_needs_rebalancing(False)
N = self._total_shares
data = {
"count-shares-good": N,
"count-good-share-hosts": N,
"count-corrupt-shares": 0,
"list-corrupt-shares": [],
"servers-responding": [],
"sharemap": {},
"count-wrong-shares": 0,
"count-recoverable-versions": 1,
"count-unrecoverable-versions": 0,
}
self._check_results.set_data(data)
return self._check_results
def _verify_failed(self, ignored):
# TODO: The following results are just stubs, and need to be replaced
# with actual values. These exist to make things like deep-check not
# fail.
self._check_results.set_needs_rebalancing(False)
N = self._total_shares
data = {
"count-shares-good": 0,
"count-good-share-hosts": 0,
"count-corrupt-shares": 0,
"list-corrupt-shares": [],
"servers-responding": [],
"sharemap": {},
"count-wrong-shares": 0,
"count-recoverable-versions": 0,
"count-unrecoverable-versions": 1,
}
self._check_results.set_data(data)
return self._check_results