lbry-sdk/lbrynet/blob/blob_file.py

218 lines
8 KiB
Python
Raw Normal View History

2015-08-20 17:27:15 +02:00
import os
2018-11-07 21:15:05 +01:00
import logging
from twisted.internet import defer
2017-07-24 09:04:15 +02:00
from twisted.web.client import FileBodyProducer
2015-08-20 17:27:15 +02:00
from twisted.python.failure import Failure
from lbrynet.cryptoutils import get_lbry_hash_obj
from lbrynet.p2p.Error import DownloadCanceledError, InvalidDataError, InvalidBlobHashError
2017-09-13 21:46:39 +02:00
from lbrynet.blob.writer import HashBlobWriter
from lbrynet.blob.reader import HashBlobReader
2015-08-20 17:27:15 +02:00
log = logging.getLogger(__name__)
MAX_BLOB_SIZE = 2 * 2 ** 20
2017-09-13 21:27:43 +02:00
# digest_size is in bytes, and blob hashes are hex encoded
blobhash_length = get_lbry_hash_obj().digest_size * 2
def is_valid_hashcharacter(char):
return char in "0123456789abcdef"
def is_valid_blobhash(blobhash):
"""Checks whether the blobhash is the correct length and contains only
valid characters (0-9, a-f)
@param blobhash: string, the blobhash to check
@return: True/False
"""
return len(blobhash) == blobhash_length and all(is_valid_hashcharacter(l) for l in blobhash)
2018-02-12 20:16:43 +01:00
class BlobFile:
2017-09-13 21:27:43 +02:00
"""
A chunk of data available on the network which is specified by a hashsum
2015-08-20 17:27:15 +02:00
2017-09-13 21:27:43 +02:00
This class is used to create blobs on the local filesystem
when we already know the blob hash before hand (i.e., when downloading blobs)
Also can be used for reading from blobs on the local filesystem
"""
def __str__(self):
return self.blob_hash[:16]
def __repr__(self):
return '<{}({})>'.format(self.__class__.__name__, str(self))
def __init__(self, blob_dir, blob_hash, length=None):
if not is_valid_blobhash(blob_hash):
raise InvalidBlobHashError(blob_hash)
2015-08-20 17:27:15 +02:00
self.blob_hash = blob_hash
self.length = length
self.writers = {} # {Peer: writer, finished_deferred}
self._verified = False
2015-08-20 17:27:15 +02:00
self.readers = 0
2017-09-13 21:27:43 +02:00
self.blob_dir = blob_dir
self.file_path = os.path.join(blob_dir, self.blob_hash)
self.blob_write_lock = defer.DeferredLock()
self.saved_verified_blob = False
2017-09-13 21:27:43 +02:00
if os.path.isfile(self.file_path):
self.set_length(os.path.getsize(self.file_path))
# This assumes that the hash of the blob has already been
# checked as part of the blob creation process. It might
# be worth having a function that checks the actual hash;
# its probably too expensive to have that check be part of
# this call.
self._verified = True
def open_for_writing(self, peer):
"""
open a blob file to be written by peer, supports concurrent
2018-10-18 13:40:37 +02:00
writers, as long as they are from different peers.
2017-09-13 21:27:43 +02:00
returns tuple of (writer, finished_deferred)
writer - a file like object with a write() function, close() when finished
finished_deferred - deferred that is fired when write is finished and returns
a instance of itself as HashBlob
"""
if peer not in self.writers:
2017-09-13 21:27:43 +02:00
log.debug("Opening %s to be written by %s", str(self), str(peer))
finished_deferred = defer.Deferred()
writer = HashBlobWriter(self.get_length, self.writer_finished)
self.writers[peer] = (writer, finished_deferred)
return writer, finished_deferred
2017-09-13 21:27:43 +02:00
log.warning("Tried to download the same file twice simultaneously from the same peer")
return None, None
def open_for_reading(self):
"""
open blob for reading
returns a file like object that can be read() from, and closed() when
finished
2017-09-13 21:27:43 +02:00
"""
if self._verified is True:
f = open(self.file_path, 'rb')
reader = HashBlobReader(f, self.reader_finished)
self.readers += 1
return reader
2017-09-13 21:27:43 +02:00
return None
def delete(self):
"""
delete blob file from file system, prevent deletion
if a blob is being read from or written to
returns a deferred that firesback when delete is completed
"""
if not self.writers and not self.readers:
self._verified = False
self.saved_verified_blob = False
try:
if os.path.isfile(self.file_path):
os.remove(self.file_path)
except Exception as e:
log.exception("An error occurred deleting %s:", str(self.file_path), exc_info=e)
2017-09-13 21:27:43 +02:00
else:
raise ValueError("File is currently being read or written and cannot be deleted")
2015-08-20 17:27:15 +02:00
@property
def verified(self):
"""
Protect verified from being modified by other classes.
2018-10-18 13:40:37 +02:00
verified is True if a write to a blob has completed successfully,
or a blob has been read to have the same length as specified
in init
"""
return self._verified
2015-08-20 17:27:15 +02:00
def set_length(self, length):
if self.length is not None and length == self.length:
return True
if self.length is None and 0 <= length <= MAX_BLOB_SIZE:
2015-08-20 17:27:15 +02:00
self.length = length
return True
2016-11-30 21:20:45 +01:00
log.warning("Got an invalid length. Previous length: %s, Invalid length: %s",
self.length, length)
2015-08-20 17:27:15 +02:00
return False
def get_length(self):
return self.length
def get_is_verified(self):
return self.verified
2015-08-20 17:27:15 +02:00
def is_downloading(self):
if self.writers:
return True
return False
def reader_finished(self, reader):
self.readers -= 1
return defer.succeed(True)
2015-08-20 17:27:15 +02:00
def writer_finished(self, writer, err=None):
def fire_finished_deferred():
self._verified = True
for p, (w, finished_deferred) in list(self.writers.items()):
2015-08-20 17:27:15 +02:00
if w == writer:
del self.writers[p]
finished_deferred.callback(self)
2015-08-20 17:27:15 +02:00
return True
2016-11-30 21:20:45 +01:00
log.warning(
"Somehow, the writer that was accepted as being valid was already removed: %s",
writer)
2015-08-20 17:27:15 +02:00
return False
def errback_finished_deferred(err):
for p, (w, finished_deferred) in list(self.writers.items()):
2015-08-20 17:27:15 +02:00
if w == writer:
del self.writers[p]
finished_deferred.errback(err)
2015-08-20 17:27:15 +02:00
def cancel_other_downloads():
for p, (w, finished_deferred) in self.writers.items():
w.close()
2015-08-20 17:27:15 +02:00
if err is None:
if writer.len_so_far == self.length and writer.blob_hash == self.blob_hash:
if self._verified is False:
d = self.save_verified_blob(writer)
2015-08-20 17:27:15 +02:00
d.addCallbacks(lambda _: fire_finished_deferred(), errback_finished_deferred)
d.addCallback(lambda _: cancel_other_downloads())
else:
d = defer.succeed(None)
fire_finished_deferred()
2015-08-20 17:27:15 +02:00
else:
if writer.len_so_far != self.length:
err_string = "blob length is %i vs expected %i" % (writer.len_so_far, self.length)
else:
err_string = f"blob hash is {writer.blob_hash} vs expected {self.blob_hash}"
2015-08-20 17:27:15 +02:00
errback_finished_deferred(Failure(InvalidDataError(err_string)))
d = defer.succeed(None)
2015-08-20 17:27:15 +02:00
else:
errback_finished_deferred(err)
d = defer.succeed(None)
d.addBoth(lambda _: writer.close_handle())
2015-08-20 17:27:15 +02:00
return d
def save_verified_blob(self, writer):
# we cannot have multiple _save_verified_blob interrupting
# each other, can happen since startProducing is a deferred
return self.blob_write_lock.run(self._save_verified_blob, writer)
@defer.inlineCallbacks
2015-08-20 17:27:15 +02:00
def _save_verified_blob(self, writer):
if self.saved_verified_blob is False:
writer.write_handle.seek(0)
out_path = os.path.join(self.blob_dir, self.blob_hash)
producer = FileBodyProducer(writer.write_handle)
yield producer.startProducing(open(out_path, 'wb'))
self.saved_verified_blob = True
defer.returnValue(True)
else:
raise DownloadCanceledError()