def catchEncryptFilter(self, lpath, node, isDir):
if self.efilters is not None:
# get the encryption information we need
einfo = self.efilters.check(lpath, node, isDir)
# build and name some important stuff for readability
etag = einfo[0]
plugid = einfo[1]
plugopts = einfo[2]
plugtag = '%s.%s' % (plugid, plugopts)
plug = getPM().getPluginInstance(plugid, plugtag, (None, plugopts,))
else:
# this should rarely be used.. the caller will likely be providing
# the efilter object when calling this function, but it is here
# in the event that they do not..
etag = b''
plug = getPM().getPluginInstance('crypt.null', '', (c, []))
plugopts = (c, [])
return (etag, plug, plugopts)
def catchEncryptFilter(self, lpath, node, isDir):
if self.efilters is not None:
# get the encryption information we need
einfo = self.efilters.check(lpath, node, isDir)
# build and name some important stuff for readability
etag = einfo[0]
plugid = einfo[1]
plugopts = einfo[2]
plugtag = '%s.%s' % (plugid, plugopts)
plug = getPM().getPluginInstance(plugid, plugtag,
(None, plugopts))
else:
# this should rarely be used.. the caller will likely be providing
# the efilter object when calling this function, but it is here
# in the event that they do not..
etag = b''
plug = getPM().getPluginInstance('crypt.null', '',
(None, []))
plugopts = (c, [])
return (etag, plug, plugopts)
def Pull(rhost, rport, sac, lpath, rpath = None, ssl = True, eventfunc = dummy):
if rpath is None:
rpath = b'/'
# the default metadata size
metasize = 128
sac = bytes(sac, 'utf8')
c = Client2(rhost, rport, sac, metasize = metasize)
c.Connect(essl = ssl)
rpbsz = len(rpath)
jobFileTime = []
jobFileSize = []
jobDownload = []
'''
Transform each node into it's full remote path, then
place it into the master node list for processing.
'''
def __eventDirEnum(pkg, result, vector):
rpath = pkg[0]
nodes = pkg[1]
if result is None:
return
for node in result:
# skip any other revision but current
if node[3] is not None:
continue
# grab the meta data and determine the encoding/encryption tag if any
meta = node[2]
if meta is not None:
tag = meta[1:1+32].strip(b'\x00')
else:
tag = None
# check that its not a special revision or system folder
name = node[0]
name = rpath + b'/' + name
nodes.append((name, node[1], tag))
def __eventFileTime(pkg, result, vector):
jobFileTime.append((pkg, result))
def __eventFileSize(pkg, result, vector):
jobFileSize.append((pkg, result))
def __eventFileRead(pkg, result, vector):
success = result[0]
if success != 1:
# for now.. just let someone know shit screwed up.. if they
# really need it to work they can come edit the code and skip
# this and continue onward..
raise Exception('Error On Read From Remote For [%s] At Offset [%x]' % (_lpath, _off))
return
data = result[1]
_lpath = pkg[0]
_off = pkg[1]
_fo = pkg[2]
_opcount = pkg[3]
logger.debug('write:%s:%x' % (_lpath, _off))
_fo.write(_off, data)
_opcount[0] -= 1
'''
The [0] is the current operation count and [1]
is the init flag. If the init flag is True then
the system is still creating pending requests
and therefore we should not terminate based on
an operation count of zero.
'''
if _opcount[0] < 1 and _opcount[1] is False:
# we are finished
_fo.finish()
echo = { 'echo': False }
def __eventEcho(pkg, result, vector):
pkg['echo'] = True
# first enumerate the remote directory
_nodes = c.DirList(rpath, Client.IOMode.Block)
nodes = []
__eventDirEnum((rpath, nodes), _nodes, 0)
sentEcho = False
while echo['echo'] is False:
c.handleOrSend()
quecount = len(nodes) + len(jobFileTime) + len(jobFileSize) + len(jobDownload)
if quecount == 0 and c.waitCount() == 0 and sentEcho is False:
sentEcho = True
c.Echo(Client.IOMode.Callback, (__eventEcho, echo))
# iterate through files
for x in range(0, min(100, len(nodes))):
# might be faste to pop from end of list
# but this will ensure more expected order
# of operations..
node = nodes.pop(0)
_rpath = node[0]
#_lpath = '%s/%s' % (lpath, node[0][rpbsz:].decode('utf8'))
_lpath = lpath + b'/' + node[0][rpbsz:]
# if directory issue enumerate call
if node[1] == 1:
pkg = (_rpath, nodes)
c.DirList(_rpath, Client.IOMode.Callback, (__eventDirEnum, pkg))
continue
# if file issue time check
pkg = (_rpath, _lpath, node[2])
c.FileTime(_rpath, Client.IOMode.Callback, (__eventFileTime, pkg))
# iterate time responses
for job in jobFileTime:
_rpath = job[0][0]
_lpath = job[0][1]
_etag = job[0][2] # etag from meta-data
_rmtime = job[1]
if os.path.exists(_lpath):
stat = os.stat(_lpath)
_lsize = stat.st_size
_lmtime = stat.st_mtime
else:
_lsize = -1
_lmtime = 0
# create the local file (0 sized)
# if newer then get file size so we can
# truncate the local file
if _rmtime >= _lmtime:
logger.debug('date failed for %s with local:%s remote:%s' % (_lpath, _lmtime, _rmtime))
pkg = (_rpath, _lpath, _lsize, _etag)
c.FileSize(_rpath, Client.IOMode.Callback, (__eventFileSize, pkg))
jobFileTime = []
# iterate size responses
for job in jobFileSize:
_rpath = job[0][0]
_lpath = job[0][1]
_lsize = job[0][2]
_etag = job[0][3]
_rsize = job[1]
# if size different truncate local file to match
if _rsize[0] != 1:
raise Exception('_rsize for %s failed' % _rpath)
logger.debug('[size] %s lsize:%s rsize:%s' % (_lpath, _lsize, _rsize))
_rsize = _rsize[1]
if _lsize != _rsize:
# truncate local file
TruncateFile(_lpath, _rsize)
# queue a download operation
pkg = [_rpath, _lpath, _rsize, 0, _etag]
jobDownload.append(pkg)
jobFileSize = []
# iterate download operations
tr = []
chunksize = 1024 * 1024 * 4 # default to 4MB chunk
for job in jobDownload:
_rpath = job[0]
_lpath = job[1]
_rsize = job[2]
_curoff = job[3]
_etag = job[4]
'''
We are going to download this file. We know the etag which is
used by modification plugins to alter the file for encryption
or compression for example. So we need to try to match the tag
back with the plugin and the options for it. Then create a
write object so as it is written it is unmodified back to it's
original form.
'''
if len(job) < 6:
_etag = _etag.decode('utf8', 'ignore')
_, _plugid, _plugopts = eventfunc('DecryptByTag', _etag)
if _ is None and _etag is not None and len(_etag) > 0:
# well, we apparently have no entry for this file so we
# need to alert the user or the calling code that there
# is a problem that needs to be addressed
raise Exception('Tag specified as "%s" but no plugin found.' % _etag)
if _ is None:
# just use the null plugin
_ = None
_plugid = 'crypt.null'
_plugopts = (None, [])
plug = getPM().getPluginInstance(_plugid, _etag, (None, _plugopts,))
_fo = plug.beginwrite(_lpath)
job.append(_fo)
_opcount = [0, True]
job.append(_opcount)
else:
_fo = job[5]
_opcount = job[6]
# increment operation count
_opcount[0] += 1
# determine amount we can read and choose maximum
_rem = _rsize - _curoff
rsz = min(_rem, chunksize)
pkg = (_lpath, _curoff, _fo, _opcount)
# this should *not* read messages and execute callbacks, because
# if it does then technically it could call the callback before
# we have set the init flag to False meaning the file never gets
# closed
c.FileRead(_rpath, _curoff, rsz, Client.IOMode.Callback, (__eventFileRead, pkg))
if _curoff + rsz >= _rsize:
tr.append(job)
logger.debug('finish:%s' % (_lpath))
# set the init flag to False so the callback
# code knows when the count reaches zero that
# the file is done
_opcount[1] = False
continue
job[3] = _curoff + rsz
# remove completed jobs
for t in tr:
jobDownload.remove(t)
def Push(rhost, rport, sac, lpath, rpath = None, ssl = True, eventfunc = dummy, chunksize = 1024 * 32):
if rpath is None:
rpath = b'/'
metasize = 128
sac = bytes(sac, 'utf8')
c = Client2(rhost, rport, sac, metasize = metasize, eventfunc = eventfunc)
c.Connect(essl = ssl)
# produce remote and local paths
lpbsz = len(lpath)
#logger.debug('test')
#print('testing')
#c.FileTrun(b'/grape', 5, mode = Client.IOMode.Block)
#c.FileWrite(b'/grape', 0, b'hello', mode = Client.IOMode.Block)
#print('read', c.FileRead(b'/grape', 0, 5, mode = Client.IOMode.Block))
#print('hash', c.FileHash(b'/grape', 0, 5, mode = Client.IOMode.Block))
#_lhash = b'hello'
#_lhash = c.HashKmc(_lhash, 128)
#print('local-hash', _lhash)
#sys.stdout.flush()
#exit()
# encryption plugin instances of options
eplugs = {}
jobDirEnum = [] # to be enumerated
jobPendingFiles = [] # files pending processing
jobGetRemoteSize = [] #
jobGetModifiedDate = [] #
jobPatch = [] # patch jobs
jobPatchQueryDelayed = [] # delayed patch hash requests
jobUpload = [] # upload jobs
jobPatchOperations = []
def getJobCount():
return len(jobDirEnum) + len(jobPendingFiles) + len(jobGetRemoteSize) + \
len(jobGetModifiedDate) + len(jobPatch) + len(jobPatchQueryDelayed) + \
len(jobUpload)
jobDirEnum.append(lpath)
maxque = 2
echo = { 'echo': False }
sentEcho = False
'''
These turn the async polling model into a async callback model, at
least to some extent. We still poll but we do not poll individual
objects which reduces polling time (CPU burn)..
'''
def __eventFileSize(pkg, result, vector):
jobGetRemoteSize.append((pkg, result))
def __eventFileTime(pkg, result, vector):
jobGetModifiedDate.append((pkg, result))
def __eventEcho(pkg, result, vector):
logger.debug('ECHO')
pkg['echo'] = True
def __eventFileWrite(pkg, result, vector):
if result == 0:
raise Exception('WRITE FAILED')
def __eventHashReply(pkg, result, vector):
# hash local file now and compare
_success = result[0]
_rhash = result[1]
_rpath = pkg[0]
_lpath = pkg[1]
_rsize = pkg[2]
_lsize = pkg[3]
_offset = pkg[4]
_size = pkg[5]
_shrstate = pkg[6]
_fo = pkg[7]
if _success == 0:
raise Exception('Hash Error')
#gc.collect()
_data = _fo.read(_offset, _size)
#_lhash = _fo.read(_offset, _size)
#fd = open(_lpath, 'rb')
#fd.seek(_offset)
#_data = fd.read(_size)
#fd.seek(_offset)
#_lhash = fd.read(_size)
#fd.close()
# this returns a new bytes instance representing
# the hash
_lhash = c.HashKmc(_data, 128)
# if we have used more bytes than the file's actual size then we are
# basically just wasting bandwidth and should immediantly force all
# remaining operations to perform a write to turn this into an upload
if _shrstate.bytesProtoUsed > (_shrstate.firstSize * 0.5) - _shrstate.bytesPatched:
_shrstate.bytesPatched += len(_data)
c.FileWrite(_rpath, _offset, _data, Client.IOMode.Discard)
_shrstate.dec()
if _shrstate.opCount < 1 and _shrstate.init is False:
ct = time.time()
c.FileSetTime(_rpath, int(ct), int(ct), Client.IOMode.Discard)
eventfunc('PatchFinish', _rpath, _lpath)
_fo.finish()
return
# if the hashes are the same do nothing
if _lhash != _rhash:
eventfunc('HashBad', _rpath, _lpath, _offset, _size, _rhash, _lhash)
# let us decide if it is worth breaking down futher
# or we should just cut our loses and force a patch
# this section of the remote file is different so,
# now we must decide to either patch this section
# or continue breaking it down further in hopes that
# we find where it differs -- we could end up spending
# more bandwidth on
#
logger.debug('hash-bad:%s:%x:%x\n' % (_lpath, _offset, _size))
# limit the minimum size of a hash
if _size > 1024 * 128:
# we have not decremented for this operation and we are
# going to create two new (but we are going to die) so
# only increment it by one
logger.debug('hash; [%s] plus one' % id(_shrstate))
_nsz = int(_size / 2)
if _size % 2 == 0:
_aoff = int(_offset + _nsz)
_alen = _nsz
_boff = _offset
_blen = _nsz
else:
_aoff = int(_offset + _nsz)
_alen = _nsz
_boff = _offset
# adjust by one since we had odd number of bytes
_blen = _nsz + 1
# start the queries
logger.debug('patching-split:%s:%x:%x (a:%x:%x) (b:%x:%x)' % (_lpath, _offset, _size, _aoff, _alen, _boff, _blen))
# either execute it or delay it
subjob = [_rpath, _lpath, _rsize, _lsize, _aoff, _alen, _shrstate, _fo]
jobPatchQueryDelayed.append((_rpath, _aoff, _alen, subjob))
# represent this sub-job/job as sleeping
_shrstate.incsleeping()
subjob = [_rpath, _lpath, _rsize, _lsize, _boff, _blen, _shrstate, _fo]
jobPatchQueryDelayed.append((_rpath, _boff, _blen, subjob))
# represent this sub-job/job as sleeping
_shrstate.incsleeping()
# this one turned into two (so only increment once)
_shrstate.inc()
# just an estimate of how much we are about to use
_shrstate.bytesProtoUsed += 128 + (8 * 2 + 32) * 2
return
# just upload this section..
eventfunc('Write', _rpath, _lpath, _offset, _size)
_shrstate.bytesPatched += _size
logger.debug('patching-section:%s:%x:%x' % (_lpath, _offset, _size))
c.FileWrite(_rpath, _offset, _data, Client.IOMode.Callback, (__eventFileWrite, None))
_shrstate.dec()
# exit here
return
else:
# track how many bytes we saved!
_shrstate.bytesSaved += _size
eventfunc('HashGood', _rpath, _lpath, _offset, _size)
logger.debug('patching-match:%s:%x:%x' % (_lpath, _offset, _size))
# decrement since we just died and spawned no sub-hash jobs
logger.debug('hash; [%s] minus one' % id(_shrstate))
_shrstate.dec()
if _shrstate.opCount < 1 and _shrstate.init is False:
# set the time to the current time to keep it from being check every
# time by way of hashing
ct = time.time()
c.FileSetTime(_rpath, int(ct), int(ct), Client.IOMode.Discard)
eventfunc('PatchFinish', _rpath, _lpath, _size)
_fo.finish()
# statistics
databytesout = 0
stat_uptodate = 0
stat_uploaded = 0
stat_patched = 0
stat_checked = 0
patchrunning = [0]
dd = time.time()
# the soft limit for application level buffer size
buflimit = 1024 * 1024 * 4
lt = time.time()
# keep going until we get the echo
while echo['echo'] is False:
# read any messages
c.handleOrSend()
# update our throughput
eventfunc('Cycle', c.getThroughput())
########################################################
##################### LIMITERS #########################
########################################################
#
# This section sets and adjusts the limits which directly
# effect memory consumption mainly, and also help to finish
# things before starting new things. I placed everything
# into a central location to make it easier to adjust stuff
# because you can look at everything as a whole.
#
# JOB-DIR-ENUM
if len(jobPendingFiles) < 1000:
# only produce more pending files when low
jobDirEnumLimit = min(500, len(jobDirEnum))
else:
jobDirEnumLimit = 0
# JOB-PENDING-FILES
# do not process any new files unless we have no patch operations
# and no more than 19 upload operations.. otherwise we start spreading
# things out and nothing really gets completely done
if c.waitCount() < 500 and len(jobUpload) < 4:
# only throw files into pipeline when wait count is low
jobPendingFilesLimit = min(500, len(jobPendingFiles))
else:
jobPendingFilesLimit = 0
# JOB-PATCH-QUERY-DELAYED
# always process delayed patch queries
jobPatchQueryDelayedLimit = min(2, len(jobPatchQueryDelayed))
# JOB-PATCH
# always process patch jobs
jobPatchLimit = min(2, len(jobPatch))
#########################################################
# if our output buffer is empty, we have no jobs, and we are not
# waiting on any requests
if c.getBytesToSend() < 1 and getJobCount() < 1 and c.waitCount() < 1 and sentEcho is False:
sentEcho = True
c.Echo(Client.IOMode.Callback, (__eventEcho, echo))
# if we let this continue to grow it could eventually consume all
# physical memory so we limit it and once it reaches that soft
# limit it will empty our buffers completely
boutbuf = c.getBytesToSend()
if boutbuf > buflimit:
logger.debug('emptying outbound buffer..')
# i decided not to force the buffer to completely drain
# in hopes that if we didnt it would be easier to keep
# the server working - the more time the server is idle
# is lost work time
while c.getBytesToSend() > 1024 * 1024:
# STALL/NET-DEADLOCK PREVENTION
# this will still cause callbacks to be fired and async results
# to be stored at the same time it will send any data from our
# application level buffers, which is desired - if we only sent
# pending data and never read the server's outgoing buffer would
# fill up because our incoming would fill up these creating a
# data lock because we could never send from our buffer because
# the remote side could not read because it can not write
c.handleOrSend()
eventfunc('DumpCycle', c.getThroughput(), c.getBytesToSend())
time.sleep(0.01)
else:
# just send what we can right now
c.send()
#
# JOB-DIR-ENUM
#
# DESCRIPTION:
# this fills the pending files list; we limit it because if
# we do not it could overwhelm physical memory on the machine
# so we only produce more when there is room to do so
x = -1
for x in range(0, jobDirEnumLimit):
dej = jobDirEnum[x]
try:
nodes = os.listdir(dej)
except:
continue
for node in nodes:
_lpath = b'/'.join((dej, node))
if os.path.isdir(_lpath):
# delay this..
res = eventfunc('Filter', _lpath, node, True)
if res or res is None:
jobDirEnum.append(_lpath)
continue
res = eventfunc('Filter', _lpath, node, False)
if res or res is None:
jobPendingFiles.append(_lpath)
# drop what we completed
jobDirEnum = jobDirEnum[x + 1:]
#
# JOB-PATCH-QUERY-DELAYED
#
# DESCRIPTION:
# this happens first because while we wish to work on multiple
# files at once we also wish to finish up in progress operations
# so in order to do that we first service pending hash requests
# which come from hash operations that get delayed because there
# was too many outstanding requests; our code path for hashing
# can cause physical memory to be overwhelmed if not limited and
# thus that is what this is doing - holding patch operations until
# our queue count is under the limit.
x = -1
keep = []
#print('len(jobPatchQueryDelayed):%s patchrunning:%s' % (len(jobPatchQueryDelayed), patchrunning[0]))
for x in range(0, len(jobPatchQueryDelayed)):
job = jobPatchQueryDelayed[x]
_rpath = job[0]
_off = job[1]
_size = job[2]
subjob = job[3]
# track how many are active [sent to server] (versus sleeping)
if (subjob[6].opCount - subjob[6].opSleeping) > jobPatchQueryDelayedLimit:
# we have too many active requests so wait until they complete
# and are then represented as sleeping..
keep.append(job)
continue
v = c.FileHash(_rpath, _off, _size, Client.IOMode.Callback, (__eventHashReply, subjob))
# decrement sleeping to represent one that has been sent to the server
subjob[6].decsleeping()
jobPatchQueryDelayed = keep
# drop the ones we executed
#jobPatchQueryDelayed = jobPatchQueryDelayed[x + 1:]
#
# JOB-PATCH
#
# DESCRIPTION:
# this will issue new hash requests which starting the patching
# process only if the queue is below a certain number of items
# which helps keep the memory consumption limited because these
# can get out of hand fast if you have a multi-GB file or even
# a TB can create one million requests which is a lot
# LOCATION:
# this happens before any new files are placed into the queues
# so we can finish our patch jobs before moving on the to next
# file
#if patchrunning[0] > 999999:
# _jobPatchLimit = 0
#else:
# _jobPatchLimit = jobPatchLimit
if patchrunning[0] > 10:
_max = 0
else:
_max = min(jobPatchLimit, len(jobPatch))
x = 0
while x < _max:
job = jobPatch[x]
_rpath = job[0]
_lpath = job[1]
_rsize = job[2]
_lsize = job[3]
_curoff = job[4]
# make sure to create the shared state container
if len(job) < 6:
shrstate = PatchJobSharedState()
shrstate.firstOffset = _curoff
shrstate.firstSize = _lsize
shrstate.bytesProtoUsed = 0
shrstate.bytesPatched = 0
shrstate.bytesSaved = 0
shrstate.opCount = 0
shrstate.opSleeping = 0
shrstate.startTime = time.time()
shrstate.lpath = _lpath
shrstate.rpath = _rpath
shrstate.init = True
shrstate.patchrunning = patchrunning
job.append(shrstate)
jobPatchOperations.append(shrstate)
'''
This _fo object replaces the old code that used to open a file. The _fo
is short for file object. It provides the functionality of reading and
writing to a file except the implementation can do additional things.
See ./plugins/ and especially ./plugins/crypt for examples of the
implementation of this.
'''
tag, plugid, plugopts = eventfunc('EncryptFilter', _lpath, _lpath[_lpath.rfind(b'/') + 1:], False)
_fo = plug.beginread(_lpath)
job.append(_fo)
# make sure the correct metadata type/version (VERSION 1) byte is written
c.FileWriteMeta(_rpath, 0, b'\xAA', Client.IOMode.Discard)
# write in the encryption tag (for automated reversal using encryption filter file)
btag = bytes(tag, 'utf8')
if len(btag) > 32:
raise Exception('The encryption tag "%s" is longer than 32 bytes!' % tag)
c.FileWriteMeta(_rpath, 1, btag.ljust(32, b'\x00'), Client.IOMode.Discard)
else:
# get existing shared state container
shrstate = job[5]
_fo = job[6]
# hash 32MB chunks (server is threaded so it should be okay)
csz = 1024 * 1024 * 4
# get actual size due to what we have remaining of the file
_tsz = min(csz, _lsize - _curoff)
# increment number of operations ongoing
shrstate.inc()
# create job and request with callback
subjob = [_rpath, _lpath, _rsize, _lsize, _curoff, _tsz, shrstate, _fo]
c.FileHash(_rpath, _curoff, _tsz, Client.IOMode.Callback, (__eventHashReply, subjob))
# determine and track if we are finished
job[4] = job[4] + _tsz
if job[4] >= _lsize:
# do not increment x after removing it; this
# feels kind of hacky but it does work right
# now at least but i think its on the edge
# of being bad code
shrstate.init = False
del jobPatch[x]
_max = _max - 1
continue
# increment x since we did not remove anything
x = x = 1
#
# JOB-PENDING-FILES
#
# DESCRIPTION:
# This takes pending files created by enumeration of local directories,
# and starts processin by first issuing a file size request to determine
# if the file exist and if it is the correct size.
x = -1
for x in range(0, jobPendingFilesLimit):
_lpath = jobPendingFiles[x]
##### DEBUG (NO FILES GREATER THAN 200MB) #####
try:
stat = os.stat(_lpath)
except:
continue
_lsize = stat.st_size
if _lsize > 1024 * 1024 * 200:
continue
###############################################
'''
We have to get the modified size of the local file
since we will be comparing this to the remote. For
the NULL modification plugin it will be exactly the
same, but for others it will be different. So we do
that here. The plugin instance is cached by the plugin
manager (PM).
'''
tag, plug, plugopts = eventfunc('EncryptFilter', _lpath, _lpath[_lpath.rfind(b'/') + 1:], False)
if tag is None:
# if none specified then default to null
plug = getPM().getPluginInstance('crypt.null', '', (c, []))
if plug is None:
raise Exception('Apparently, we are missing a plugin referenced by "%s".' % plugid)
_esize = plug.getencryptedsize(_lpath)
_lsize = _esize
_rpath = rpath + _lpath[lpbsz:]
stat_checked = stat_checked + 1
pkg = (_rpath, _lpath, _lsize, None, int(stat.st_mtime))
eventfunc('Start', _rpath, _lpath)
c.FileSize(_rpath, Client.IOMode.Callback, (__eventFileSize, pkg))
# drop what we completed
jobPendingFiles = jobPendingFiles[x + 1:]
########################################################
############## NO LIMIT SECTIONS BELOW #################
########################################################
#
# JOB-GET-REMOTE-SIZE
#
for rsj in jobGetRemoteSize:
pkg = rsj[0]
_result = rsj[1]
_rpath = pkg[0]
_lpath = pkg[1]
_lsize = pkg[2]
_vector = pkg[3]
_lmtime = pkg[4]
eventfunc('SizeReply', _rpath, _lpath)
# result[0] = success code is non-zero and result[1] = size (0 on failure code)
_rsize = _result[1]
if _lsize == _rsize and _result[0] == 1:
# need to check modified date
pkg = (_rpath, _lpath, _rsize, _lsize, _vector, _lmtime)
c.FileTime(_rpath, Client.IOMode.Callback, (__eventFileTime, pkg))
else:
# first make the remote size match the local size
c.FileTrun(_rpath, _lsize, Client.IOMode.Discard)
if max(_rsize, _lsize) < 1:
eventfunc('Finished', _rpath, _lpath)
continue
# need to decide if we want to upload or patch
if min(_rsize, _lsize) / max(_rsize, _lsize) < 0.5:
# make upload job
jobUpload.append([_rpath, _lpath, _rsize, _lsize, 0])
else:
# make patch job
jobPatch.append([_rpath, _lpath, _rsize, _lsize, 0])
jobGetRemoteSize = []
#
# JOB-GET-MODIFIED-DATE
#
for rtj in jobGetModifiedDate:
pkg = rtj[0]
_rmtime = rtj[1]
_rpath = pkg[0]
_lpath = pkg[1]
_rsize = pkg[2]
_lsize = pkg[3]
_vector = pkg[4]
_lmtime = pkg[5]
eventfunc('DateReply', _rpath, _lpath)
if _rmtime < _lmtime:
# need to decide if we want to upload or patch
if min(_rsize, _lsize) / (max(_rsize, _lsize) + 1) < 0.5:
# make upload job
jobUpload.append([_rpath, _lpath, _rsize, _lsize, 0])
else:
# make patch job
jobPatch.append([_rpath, _lpath, _rsize, _lsize, 0])
else:
# just drop it since its either up to date or newer
eventfunc('Finished', _rpath, _lpath)
stat_uptodate = stat_uptodate + 1
continue
jobGetModifiedDate = []
#
# JOB-UPLOAD
#
# Limit, ourselves to only 4 at a time.
tr = []
cjc = 0
#for uj in jobUpload:
for x in range(0, min(4, len(jobUpload))):
uj = jobUpload[x]
_rpath = uj[0]
_lpath = uj[1]
_rsize = uj[2]
_lsize = uj[3]
_curoff = uj[4]
'''
Here we have to determine what modification plugin to use
and then get an instance of it, and use that instance to
do read operations instead of directly on the file.
'''
if len(uj) < 6:
tag, plug, plugopts = eventfunc('EncryptFilter', _lpath, _lpath[_lpath.rfind(b'/') + 1:], False)
if plug is None:
# if none specified then default to null
plug = getPM().getPluginInstance('crypt.null', '', (c, []))
tag = ''
try:
_fo = plug.beginread(_lpath)
except PermissionError:
eventfunc('OpenFailed', _rpath, _lpath)
tr.append(uj)
continue
uj.append(_fo)
# make sure the correct metadata type/version (VERSION 1) byte is written
c.FileWriteMeta(_rpath, 0, b'\xAA', Client.IOMode.Discard)
# write in the encryption tag (for automated reversal using encryption filter file)
btag = bytes(tag, 'utf8')
if len(btag) > 32:
raise Exception('The encryption tag "%s" is longer than 32 bytes!' % tag)
c.FileWriteMeta(_rpath, 1, btag.ljust(32, b'\x00'), Client.IOMode.Discard)
else:
_fo = uj[5]
'''
At this point we have the modification plugin instance
read object and we are treating it like a normal file.
'''
_chunksize = chunksize
# see what we can send
_rem = min(_lsize - _curoff, _chunksize)
# open local file and read chunk
#_fd = open(_lpath, 'rb')
#_fd.seek(_curoff)
#_data = _fd.read(_rem)
#_fd.close()
_data = _fo.read(_curoff, _rem)
# if no more data then stop and terminate this job
if not _data:
tr.append(uj)
continue
eventfunc('Write', _rpath, _lpath, _curoff, _chunksize)
c.FileWrite(_rpath, _curoff, _data, Client.IOMode.Discard)
# advance our current offset
uj[4] = _curoff + len(_data)
# (UNUSED) -- old code not we check if _data is empty
# if we reached the EOF then drop it
if uj[4] >= _lsize:
tr.append(uj)
# keep accounting for the bytes sent for statistics
databytesout = databytesout + len(_data)
# dont overfill our buffers just exit
if c.getBytesToSend() > buflimit:
break
# keep track of number of jobs processed (i think old code)
cjc = cjc + 1
# remove finished jobs
ct = time.time()
for uj in tr:
eventfunc('Finish', uj[0], uj[1])
# set the modified time on the file to the current
# time to represent it is up to date
_rpath = uj[0]
c.FileSetTime(_rpath, int(ct), int(ct), Client.IOMode.Discard)
jobUpload.remove(uj)
stat_uploaded = stat_uploaded + 1
continue
c.close()
# we are done!
return
def Pull(rhost, rport, sac, lpath, rpath=None, ssl=True, eventfunc=dummy):
if rpath is None:
rpath = b'/'
# the default metadata size
metasize = 128
sac = bytes(sac, 'utf8')
c = Client2(rhost, rport, sac, metasize=metasize)
c.Connect(essl=ssl)
rpbsz = len(rpath)
jobFileTime = []
jobFileSize = []
jobDownload = []
'''
Transform each node into it's full remote path, then
place it into the master node list for processing.
'''
def __eventDirEnum(pkg, result, vector):
rpath = pkg[0]
nodes = pkg[1]
if result is None:
return
for node in result:
# skip any other revision but current
if node[3] is not None:
continue
# grab the meta data and determine the encoding/encryption tag if any
meta = node[2]
if meta is not None:
tag = meta[1:1 + 32].strip(b'\x00')
else:
tag = None
# check that its not a special revision or system folder
name = node[0]
name = rpath + b'/' + name
nodes.append((name, node[1], tag))
def __eventFileTime(pkg, result, vector):
jobFileTime.append((pkg, result))
def __eventFileSize(pkg, result, vector):
jobFileSize.append((pkg, result))
def __eventFileRead(pkg, result, vector):
success = result[0]
if success != 1:
# for now.. just let someone know shit screwed up.. if they
# really need it to work they can come edit the code and skip
# this and continue onward..
raise Exception(
'Error On Read From Remote For [%s] At Offset [%x]' %
(_lpath, _off))
return
data = result[1]
_lpath = pkg[0]
_off = pkg[1]
_fo = pkg[2]
_opcount = pkg[3]
logger.debug('write:%s:%x' % (_lpath, _off))
_fo.write(_off, data)
_opcount[0] -= 1
'''
The [0] is the current operation count and [1]
is the init flag. If the init flag is True then
the system is still creating pending requests
and therefore we should not terminate based on
an operation count of zero.
'''
if _opcount[0] < 1 and _opcount[1] is False:
# we are finished
_fo.finish()
echo = {'echo': False}
def __eventEcho(pkg, result, vector):
pkg['echo'] = True
# first enumerate the remote directory
_nodes = c.DirList(rpath, Client.IOMode.Block)
nodes = []
__eventDirEnum((rpath, nodes), _nodes, 0)
sentEcho = False
while echo['echo'] is False:
c.handleOrSend()
quecount = len(nodes) + len(jobFileTime) + len(jobFileSize) + len(
jobDownload)
if quecount == 0 and c.waitCount() == 0 and sentEcho is False:
sentEcho = True
c.Echo(Client.IOMode.Callback, (__eventEcho, echo))
# iterate through files
for x in range(0, min(100, len(nodes))):
# might be faste to pop from end of list
# but this will ensure more expected order
# of operations..
node = nodes.pop(0)
_rpath = node[0]
#_lpath = '%s/%s' % (lpath, node[0][rpbsz:].decode('utf8'))
_lpath = lpath + b'/' + node[0][rpbsz:]
# if directory issue enumerate call
if node[1] == 1:
pkg = (_rpath, nodes)
c.DirList(_rpath, Client.IOMode.Callback,
(__eventDirEnum, pkg))
continue
# if file issue time check
pkg = (_rpath, _lpath, node[2])
c.FileTime(_rpath, Client.IOMode.Callback, (__eventFileTime, pkg))
# iterate time responses
for job in jobFileTime:
_rpath = job[0][0]
_lpath = job[0][1]
_etag = job[0][2] # etag from meta-data
_rmtime = job[1]
if os.path.exists(_lpath):
stat = os.stat(_lpath)
_lsize = stat.st_size
_lmtime = stat.st_mtime
else:
_lsize = -1
_lmtime = 0
# create the local file (0 sized)
# if newer then get file size so we can
# truncate the local file
if _rmtime >= _lmtime:
logger.debug('date failed for %s with local:%s remote:%s' %
(_lpath, _lmtime, _rmtime))
pkg = (_rpath, _lpath, _lsize, _etag)
c.FileSize(_rpath, Client.IOMode.Callback,
(__eventFileSize, pkg))
jobFileTime = []
# iterate size responses
for job in jobFileSize:
_rpath = job[0][0]
_lpath = job[0][1]
_lsize = job[0][2]
_etag = job[0][3]
_rsize = job[1]
# if size different truncate local file to match
if _rsize[0] != 1:
raise Exception('_rsize for %s failed' % _rpath)
logger.debug('[size] %s lsize:%s rsize:%s' %
(_lpath, _lsize, _rsize))
_rsize = _rsize[1]
if _lsize != _rsize:
# truncate local file
TruncateFile(_lpath, _rsize)
# queue a download operation
pkg = [_rpath, _lpath, _rsize, 0, _etag]
jobDownload.append(pkg)
jobFileSize = []
# iterate download operations
tr = []
chunksize = 1024 * 1024 * 4 # default to 4MB chunk
for job in jobDownload:
_rpath = job[0]
_lpath = job[1]
_rsize = job[2]
_curoff = job[3]
_etag = job[4]
'''
We are going to download this file. We know the etag which is
used by modification plugins to alter the file for encryption
or compression for example. So we need to try to match the tag
back with the plugin and the options for it. Then create a
write object so as it is written it is unmodified back to it's
original form.
'''
if len(job) < 6:
_etag = _etag.decode('utf8', 'ignore')
_, _plugid, _plugopts = eventfunc('DecryptByTag', _etag)
if _ is None and _etag is not None and len(_etag) > 0:
# well, we apparently have no entry for this file so we
# need to alert the user or the calling code that there
# is a problem that needs to be addressed
raise Exception(
'Tag specified as "%s" but no plugin found.' % _etag)
if _ is None:
# just use the null plugin
_ = None
_plugid = 'crypt.null'
_plugopts = (None, [])
plug = getPM().getPluginInstance(_plugid, _etag, (
None,
_plugopts,
))
_fo = plug.beginwrite(_lpath)
job.append(_fo)
_opcount = [0, True]
job.append(_opcount)
else:
_fo = job[5]
_opcount = job[6]
# increment operation count
_opcount[0] += 1
# determine amount we can read and choose maximum
_rem = _rsize - _curoff
rsz = min(_rem, chunksize)
pkg = (_lpath, _curoff, _fo, _opcount)
# this should *not* read messages and execute callbacks, because
# if it does then technically it could call the callback before
# we have set the init flag to False meaning the file never gets
# closed
c.FileRead(_rpath, _curoff, rsz, Client.IOMode.Callback,
(__eventFileRead, pkg))
if _curoff + rsz >= _rsize:
tr.append(job)
logger.debug('finish:%s' % (_lpath))
# set the init flag to False so the callback
# code knows when the count reaches zero that
# the file is done
_opcount[1] = False
continue
job[3] = _curoff + rsz
# remove completed jobs
for t in tr:
jobDownload.remove(t)
def Push(rhost,
rport,
sac,
lpath,
rpath=None,
ssl=True,
eventfunc=dummy,
chunksize=1024 * 32):
if rpath is None:
rpath = b'/'
metasize = 128
sac = bytes(sac, 'utf8')
c = Client2(rhost, rport, sac, metasize=metasize, eventfunc=eventfunc)
c.Connect(essl=ssl)
# produce remote and local paths
lpbsz = len(lpath)
#logger.debug('test')
#print('testing')
#c.FileTrun(b'/grape', 5, mode = Client.IOMode.Block)
#c.FileWrite(b'/grape', 0, b'hello', mode = Client.IOMode.Block)
#print('read', c.FileRead(b'/grape', 0, 5, mode = Client.IOMode.Block))
#print('hash', c.FileHash(b'/grape', 0, 5, mode = Client.IOMode.Block))
#_lhash = b'hello'
#_lhash = c.HashKmc(_lhash, 128)
#print('local-hash', _lhash)
#sys.stdout.flush()
#exit()
# encryption plugin instances of options
eplugs = {}
jobDirEnum = [] # to be enumerated
jobPendingFiles = [] # files pending processing
jobGetRemoteSize = [] #
jobGetModifiedDate = [] #
jobPatch = [] # patch jobs
jobPatchQueryDelayed = [] # delayed patch hash requests
jobUpload = [] # upload jobs
jobPatchOperations = []
def getJobCount():
return len(jobDirEnum) + len(jobPendingFiles) + len(jobGetRemoteSize) + \
len(jobGetModifiedDate) + len(jobPatch) + len(jobPatchQueryDelayed) + \
len(jobUpload)
jobDirEnum.append(lpath)
maxque = 2
echo = {'echo': False}
sentEcho = False
'''
These turn the async polling model into a async callback model, at
least to some extent. We still poll but we do not poll individual
objects which reduces polling time (CPU burn)..
'''
def __eventFileSize(pkg, result, vector):
jobGetRemoteSize.append((pkg, result))
def __eventFileTime(pkg, result, vector):
jobGetModifiedDate.append((pkg, result))
def __eventEcho(pkg, result, vector):
logger.debug('ECHO')
pkg['echo'] = True
def __eventFileWrite(pkg, result, vector):
if result == 0:
raise Exception('WRITE FAILED')
def __eventHashReply(pkg, result, vector):
# hash local file now and compare
_success = result[0]
_rhash = result[1]
_rpath = pkg[0]
_lpath = pkg[1]
_rsize = pkg[2]
_lsize = pkg[3]
_offset = pkg[4]
_size = pkg[5]
_shrstate = pkg[6]
_fo = pkg[7]
if _success == 0:
raise Exception('Hash Error')
#gc.collect()
_data = _fo.read(_offset, _size)
#_lhash = _fo.read(_offset, _size)
#fd = open(_lpath, 'rb')
#fd.seek(_offset)
#_data = fd.read(_size)
#fd.seek(_offset)
#_lhash = fd.read(_size)
#fd.close()
# this returns a new bytes instance representing
# the hash
_lhash = c.HashKmc(_data, 128)
# if we have used more bytes than the file's actual size then we are
# basically just wasting bandwidth and should immediantly force all
# remaining operations to perform a write to turn this into an upload
if _shrstate.bytesProtoUsed > (_shrstate.firstSize *
0.5) - _shrstate.bytesPatched:
_shrstate.bytesPatched += len(_data)
c.FileWrite(_rpath, _offset, _data, Client.IOMode.Discard)
_shrstate.dec()
if _shrstate.opCount < 1 and _shrstate.init is False:
ct = time.time()
c.FileSetTime(_rpath, int(ct), int(ct), Client.IOMode.Discard)
eventfunc('PatchFinish', _rpath, _lpath)
_fo.finish()
return
# if the hashes are the same do nothing
if _lhash != _rhash:
eventfunc('HashBad', _rpath, _lpath, _offset, _size, _rhash,
_lhash)
# let us decide if it is worth breaking down futher
# or we should just cut our loses and force a patch
# this section of the remote file is different so,
# now we must decide to either patch this section
# or continue breaking it down further in hopes that
# we find where it differs -- we could end up spending
# more bandwidth on
#
logger.debug('hash-bad:%s:%x:%x\n' % (_lpath, _offset, _size))
# limit the minimum size of a hash
if _size > 1024 * 128:
# we have not decremented for this operation and we are
# going to create two new (but we are going to die) so
# only increment it by one
logger.debug('hash; [%s] plus one' % id(_shrstate))
_nsz = int(_size / 2)
if _size % 2 == 0:
_aoff = int(_offset + _nsz)
_alen = _nsz
_boff = _offset
_blen = _nsz
else:
_aoff = int(_offset + _nsz)
_alen = _nsz
_boff = _offset
# adjust by one since we had odd number of bytes
_blen = _nsz + 1
# start the queries
logger.debug(
'patching-split:%s:%x:%x (a:%x:%x) (b:%x:%x)' %
(_lpath, _offset, _size, _aoff, _alen, _boff, _blen))
# either execute it or delay it
subjob = [
_rpath, _lpath, _rsize, _lsize, _aoff, _alen, _shrstate,
_fo
]
jobPatchQueryDelayed.append((_rpath, _aoff, _alen, subjob))
# represent this sub-job/job as sleeping
_shrstate.incsleeping()
subjob = [
_rpath, _lpath, _rsize, _lsize, _boff, _blen, _shrstate,
_fo
]
jobPatchQueryDelayed.append((_rpath, _boff, _blen, subjob))
# represent this sub-job/job as sleeping
_shrstate.incsleeping()
# this one turned into two (so only increment once)
_shrstate.inc()
# just an estimate of how much we are about to use
_shrstate.bytesProtoUsed += 128 + (8 * 2 + 32) * 2
return
# just upload this section..
eventfunc('Write', _rpath, _lpath, _offset, _size)
_shrstate.bytesPatched += _size
logger.debug('patching-section:%s:%x:%x' %
(_lpath, _offset, _size))
c.FileWrite(_rpath, _offset, _data, Client.IOMode.Callback,
(__eventFileWrite, None))
_shrstate.dec()
# exit here
return
else:
# track how many bytes we saved!
_shrstate.bytesSaved += _size
eventfunc('HashGood', _rpath, _lpath, _offset, _size)
logger.debug('patching-match:%s:%x:%x' % (_lpath, _offset, _size))
# decrement since we just died and spawned no sub-hash jobs
logger.debug('hash; [%s] minus one' % id(_shrstate))
_shrstate.dec()
if _shrstate.opCount < 1 and _shrstate.init is False:
# set the time to the current time to keep it from being check every
# time by way of hashing
ct = time.time()
c.FileSetTime(_rpath, int(ct), int(ct), Client.IOMode.Discard)
eventfunc('PatchFinish', _rpath, _lpath, _size)
_fo.finish()
# statistics
databytesout = 0
stat_uptodate = 0
stat_uploaded = 0
stat_patched = 0
stat_checked = 0
patchrunning = [0]
dd = time.time()
# the soft limit for application level buffer size
buflimit = 1024 * 1024 * 4
lt = time.time()
# keep going until we get the echo
while echo['echo'] is False:
# read any messages
c.handleOrSend()
# update our throughput
eventfunc('Cycle', c.getThroughput())
########################################################
##################### LIMITERS #########################
########################################################
#
# This section sets and adjusts the limits which directly
# effect memory consumption mainly, and also help to finish
# things before starting new things. I placed everything
# into a central location to make it easier to adjust stuff
# because you can look at everything as a whole.
#
# JOB-DIR-ENUM
if len(jobPendingFiles) < 1000:
# only produce more pending files when low
jobDirEnumLimit = min(500, len(jobDirEnum))
else:
jobDirEnumLimit = 0
# JOB-PENDING-FILES
# do not process any new files unless we have no patch operations
# and no more than 19 upload operations.. otherwise we start spreading
# things out and nothing really gets completely done
if c.waitCount() < 500 and len(jobUpload) < 4:
# only throw files into pipeline when wait count is low
jobPendingFilesLimit = min(500, len(jobPendingFiles))
else:
jobPendingFilesLimit = 0
# JOB-PATCH-QUERY-DELAYED
# always process delayed patch queries
jobPatchQueryDelayedLimit = min(2, len(jobPatchQueryDelayed))
# JOB-PATCH
# always process patch jobs
jobPatchLimit = min(2, len(jobPatch))
#########################################################
# if our output buffer is empty, we have no jobs, and we are not
# waiting on any requests
if c.getBytesToSend() < 1 and getJobCount() < 1 and c.waitCount(
) < 1 and sentEcho is False:
sentEcho = True
c.Echo(Client.IOMode.Callback, (__eventEcho, echo))
# if we let this continue to grow it could eventually consume all
# physical memory so we limit it and once it reaches that soft
# limit it will empty our buffers completely
boutbuf = c.getBytesToSend()
if boutbuf > buflimit:
logger.debug('emptying outbound buffer..')
# i decided not to force the buffer to completely drain
# in hopes that if we didnt it would be easier to keep
# the server working - the more time the server is idle
# is lost work time
while c.getBytesToSend() > 1024 * 1024:
# STALL/NET-DEADLOCK PREVENTION
# this will still cause callbacks to be fired and async results
# to be stored at the same time it will send any data from our
# application level buffers, which is desired - if we only sent
# pending data and never read the server's outgoing buffer would
# fill up because our incoming would fill up these creating a
# data lock because we could never send from our buffer because
# the remote side could not read because it can not write
c.handleOrSend()
eventfunc('DumpCycle', c.getThroughput(), c.getBytesToSend())
time.sleep(0.01)
else:
# just send what we can right now
c.send()
#
# JOB-DIR-ENUM
#
# DESCRIPTION:
# this fills the pending files list; we limit it because if
# we do not it could overwhelm physical memory on the machine
# so we only produce more when there is room to do so
x = -1
for x in range(0, jobDirEnumLimit):
dej = jobDirEnum[x]
try:
nodes = os.listdir(dej)
except:
continue
for node in nodes:
_lpath = b'/'.join((dej, node))
if os.path.isdir(_lpath):
# delay this..
res = eventfunc('Filter', _lpath, node, True)
if res or res is None:
jobDirEnum.append(_lpath)
continue
res = eventfunc('Filter', _lpath, node, False)
if res or res is None:
jobPendingFiles.append(_lpath)
# drop what we completed
jobDirEnum = jobDirEnum[x + 1:]
#
# JOB-PATCH-QUERY-DELAYED
#
# DESCRIPTION:
# this happens first because while we wish to work on multiple
# files at once we also wish to finish up in progress operations
# so in order to do that we first service pending hash requests
# which come from hash operations that get delayed because there
# was too many outstanding requests; our code path for hashing
# can cause physical memory to be overwhelmed if not limited and
# thus that is what this is doing - holding patch operations until
# our queue count is under the limit.
x = -1
keep = []
#print('len(jobPatchQueryDelayed):%s patchrunning:%s' % (len(jobPatchQueryDelayed), patchrunning[0]))
for x in range(0, len(jobPatchQueryDelayed)):
job = jobPatchQueryDelayed[x]
_rpath = job[0]
_off = job[1]
_size = job[2]
subjob = job[3]
# track how many are active [sent to server] (versus sleeping)
if (subjob[6].opCount -
subjob[6].opSleeping) > jobPatchQueryDelayedLimit:
# we have too many active requests so wait until they complete
# and are then represented as sleeping..
keep.append(job)
continue
v = c.FileHash(_rpath, _off, _size, Client.IOMode.Callback,
(__eventHashReply, subjob))
# decrement sleeping to represent one that has been sent to the server
subjob[6].decsleeping()
jobPatchQueryDelayed = keep
# drop the ones we executed
#jobPatchQueryDelayed = jobPatchQueryDelayed[x + 1:]
#
# JOB-PATCH
#
# DESCRIPTION:
# this will issue new hash requests which starting the patching
# process only if the queue is below a certain number of items
# which helps keep the memory consumption limited because these
# can get out of hand fast if you have a multi-GB file or even
# a TB can create one million requests which is a lot
# LOCATION:
# this happens before any new files are placed into the queues
# so we can finish our patch jobs before moving on the to next
# file
#if patchrunning[0] > 999999:
# _jobPatchLimit = 0
#else:
# _jobPatchLimit = jobPatchLimit
if patchrunning[0] > 10:
_max = 0
else:
_max = min(jobPatchLimit, len(jobPatch))
x = 0
while x < _max:
job = jobPatch[x]
_rpath = job[0]
_lpath = job[1]
_rsize = job[2]
_lsize = job[3]
_curoff = job[4]
# make sure to create the shared state container
if len(job) < 6:
shrstate = PatchJobSharedState()
shrstate.firstOffset = _curoff
shrstate.firstSize = _lsize
shrstate.bytesProtoUsed = 0
shrstate.bytesPatched = 0
shrstate.bytesSaved = 0
shrstate.opCount = 0
shrstate.opSleeping = 0
shrstate.startTime = time.time()
shrstate.lpath = _lpath
shrstate.rpath = _rpath
shrstate.init = True
shrstate.patchrunning = patchrunning
job.append(shrstate)
jobPatchOperations.append(shrstate)
'''
This _fo object replaces the old code that used to open a file. The _fo
is short for file object. It provides the functionality of reading and
writing to a file except the implementation can do additional things.
See ./plugins/ and especially ./plugins/crypt for examples of the
implementation of this.
'''
tag, plugid, plugopts = eventfunc(
'EncryptFilter', _lpath, _lpath[_lpath.rfind(b'/') + 1:],
False)
_fo = plug.beginread(_lpath)
job.append(_fo)
# make sure the correct metadata type/version (VERSION 1) byte is written
c.FileWriteMeta(_rpath, 0, b'\xAA', Client.IOMode.Discard)
# write in the encryption tag (for automated reversal using encryption filter file)
btag = bytes(tag, 'utf8')
if len(btag) > 32:
raise Exception(
'The encryption tag "%s" is longer than 32 bytes!' %
tag)
c.FileWriteMeta(_rpath, 1, btag.ljust(32, b'\x00'),
Client.IOMode.Discard)
else:
# get existing shared state container
shrstate = job[5]
_fo = job[6]
# hash 32MB chunks (server is threaded so it should be okay)
csz = 1024 * 1024 * 4
# get actual size due to what we have remaining of the file
_tsz = min(csz, _lsize - _curoff)
# increment number of operations ongoing
shrstate.inc()
# create job and request with callback
subjob = [
_rpath, _lpath, _rsize, _lsize, _curoff, _tsz, shrstate, _fo
]
c.FileHash(_rpath, _curoff, _tsz, Client.IOMode.Callback,
(__eventHashReply, subjob))
# determine and track if we are finished
job[4] = job[4] + _tsz
if job[4] >= _lsize:
# do not increment x after removing it; this
# feels kind of hacky but it does work right
# now at least but i think its on the edge
# of being bad code
shrstate.init = False
del jobPatch[x]
_max = _max - 1
continue
# increment x since we did not remove anything
x = x = 1
#
# JOB-PENDING-FILES
#
# DESCRIPTION:
# This takes pending files created by enumeration of local directories,
# and starts processin by first issuing a file size request to determine
# if the file exist and if it is the correct size.
x = -1
for x in range(0, jobPendingFilesLimit):
_lpath = jobPendingFiles[x]
##### DEBUG (NO FILES GREATER THAN 200MB) #####
try:
stat = os.stat(_lpath)
except:
continue
_lsize = stat.st_size
if _lsize > 1024 * 1024 * 200:
continue
###############################################
'''
We have to get the modified size of the local file
since we will be comparing this to the remote. For
the NULL modification plugin it will be exactly the
same, but for others it will be different. So we do
that here. The plugin instance is cached by the plugin
manager (PM).
'''
tag, plug, plugopts = eventfunc('EncryptFilter', _lpath,
_lpath[_lpath.rfind(b'/') + 1:],
False)
if tag is None:
# if none specified then default to null
plug = getPM().getPluginInstance('crypt.null', '', (c, []))
if plug is None:
raise Exception(
'Apparently, we are missing a plugin referenced by "%s".' %
plugid)
_esize = plug.getencryptedsize(_lpath)
_lsize = _esize
_rpath = rpath + _lpath[lpbsz:]
stat_checked = stat_checked + 1
pkg = (_rpath, _lpath, _lsize, None, int(stat.st_mtime))
eventfunc('Start', _rpath, _lpath)
c.FileSize(_rpath, Client.IOMode.Callback, (__eventFileSize, pkg))
# drop what we completed
jobPendingFiles = jobPendingFiles[x + 1:]
########################################################
############## NO LIMIT SECTIONS BELOW #################
########################################################
#
# JOB-GET-REMOTE-SIZE
#
for rsj in jobGetRemoteSize:
pkg = rsj[0]
_result = rsj[1]
_rpath = pkg[0]
_lpath = pkg[1]
_lsize = pkg[2]
_vector = pkg[3]
_lmtime = pkg[4]
eventfunc('SizeReply', _rpath, _lpath)
# result[0] = success code is non-zero and result[1] = size (0 on failure code)
_rsize = _result[1]
if _lsize == _rsize and _result[0] == 1:
# need to check modified date
pkg = (_rpath, _lpath, _rsize, _lsize, _vector, _lmtime)
c.FileTime(_rpath, Client.IOMode.Callback,
(__eventFileTime, pkg))
else:
# first make the remote size match the local size
c.FileTrun(_rpath, _lsize, Client.IOMode.Discard)
if max(_rsize, _lsize) < 1:
eventfunc('Finished', _rpath, _lpath)
continue
# need to decide if we want to upload or patch
if min(_rsize, _lsize) / max(_rsize, _lsize) < 0.5:
# make upload job
jobUpload.append([_rpath, _lpath, _rsize, _lsize, 0])
else:
# make patch job
jobPatch.append([_rpath, _lpath, _rsize, _lsize, 0])
jobGetRemoteSize = []
#
# JOB-GET-MODIFIED-DATE
#
for rtj in jobGetModifiedDate:
pkg = rtj[0]
_rmtime = rtj[1]
_rpath = pkg[0]
_lpath = pkg[1]
_rsize = pkg[2]
_lsize = pkg[3]
_vector = pkg[4]
_lmtime = pkg[5]
eventfunc('DateReply', _rpath, _lpath)
if _rmtime < _lmtime:
# need to decide if we want to upload or patch
if min(_rsize, _lsize) / (max(_rsize, _lsize) + 1) < 0.5:
# make upload job
jobUpload.append([_rpath, _lpath, _rsize, _lsize, 0])
else:
# make patch job
jobPatch.append([_rpath, _lpath, _rsize, _lsize, 0])
else:
# just drop it since its either up to date or newer
eventfunc('Finished', _rpath, _lpath)
stat_uptodate = stat_uptodate + 1
continue
jobGetModifiedDate = []
#
# JOB-UPLOAD
#
# Limit, ourselves to only 4 at a time.
tr = []
cjc = 0
#for uj in jobUpload:
for x in range(0, min(4, len(jobUpload))):
uj = jobUpload[x]
_rpath = uj[0]
_lpath = uj[1]
_rsize = uj[2]
_lsize = uj[3]
_curoff = uj[4]
'''
Here we have to determine what modification plugin to use
and then get an instance of it, and use that instance to
do read operations instead of directly on the file.
'''
if len(uj) < 6:
tag, plug, plugopts = eventfunc(
'EncryptFilter', _lpath, _lpath[_lpath.rfind(b'/') + 1:],
False)
if plug is None:
# if none specified then default to null
plug = getPM().getPluginInstance('crypt.null', '', (c, []))
tag = ''
try:
_fo = plug.beginread(_lpath)
except PermissionError:
eventfunc('OpenFailed', _rpath, _lpath)
tr.append(uj)
continue
uj.append(_fo)
# make sure the correct metadata type/version (VERSION 1) byte is written
c.FileWriteMeta(_rpath, 0, b'\xAA', Client.IOMode.Discard)
# write in the encryption tag (for automated reversal using encryption filter file)
btag = bytes(tag, 'utf8')
if len(btag) > 32:
raise Exception(
'The encryption tag "%s" is longer than 32 bytes!' %
tag)
c.FileWriteMeta(_rpath, 1, btag.ljust(32, b'\x00'),
Client.IOMode.Discard)
else:
_fo = uj[5]
'''
At this point we have the modification plugin instance
read object and we are treating it like a normal file.
'''
_chunksize = chunksize
# see what we can send
_rem = min(_lsize - _curoff, _chunksize)
# open local file and read chunk
#_fd = open(_lpath, 'rb')
#_fd.seek(_curoff)
#_data = _fd.read(_rem)
#_fd.close()
_data = _fo.read(_curoff, _rem)
# if no more data then stop and terminate this job
if not _data:
tr.append(uj)
continue
eventfunc('Write', _rpath, _lpath, _curoff, _chunksize)
c.FileWrite(_rpath, _curoff, _data, Client.IOMode.Discard)
# advance our current offset
uj[4] = _curoff + len(_data)
# (UNUSED) -- old code not we check if _data is empty
# if we reached the EOF then drop it
if uj[4] >= _lsize:
tr.append(uj)
# keep accounting for the bytes sent for statistics
databytesout = databytesout + len(_data)
# dont overfill our buffers just exit
if c.getBytesToSend() > buflimit:
break
# keep track of number of jobs processed (i think old code)
cjc = cjc + 1
# remove finished jobs
ct = time.time()
for uj in tr:
eventfunc('Finish', uj[0], uj[1])
# set the modified time on the file to the current
# time to represent it is up to date
_rpath = uj[0]
c.FileSetTime(_rpath, int(ct), int(ct), Client.IOMode.Discard)
jobUpload.remove(uj)
stat_uploaded = stat_uploaded + 1
continue
c.close()
# we are done!
return
