apt_dht/MirrorManager.py

   1
   2 from bz2 import BZ2Decompressor
   3 from zlib import decompressobj, MAX_WBITS
   4 from gzip import FCOMMENT, FEXTRA, FHCRC, FNAME, FTEXT
   5 from urlparse import urlparse
   6 import os
   7
   8 from twisted.python import log, filepath
   9 from twisted.internet import defer
  10 from twisted.trial import unittest
  11 from twisted.web2 import stream
  12 from twisted.web2.http import splitHostPort
  13
  14 from AptPackages import AptPackages
  15
  16 aptpkg_dir='.apt-dht'
  17
  18 DECOMPRESS_EXTS = ['.gz', '.bz2']
  19 DECOMPRESS_FILES = ['release', 'sources', 'packages']
  20
  21 class MirrorError(Exception):
  22     """Exception raised when there's a problem with the mirror."""
  23
  24 class ProxyFileStream(stream.SimpleStream):
  25     """Saves a stream to a file while providing a new stream."""
  26
  27     def __init__(self, stream, outFile, decompress = None, decFile = None):
  28         """Initializes the proxy.
  29
  30         @type stream: C{twisted.web2.stream.IByteStream}
  31         @param stream: the input stream to read from
  32         @type outFile: C{twisted.python.filepath.FilePath}
  33         @param outFile: the file to write to
  34         @type decompress: C{string}
  35         @param decompress: also decompress the file as this type
  36             (currently only '.gz' and '.bz2' are supported)
  37         @type decFile: C{twisted.python.filepath.FilePath}
  38         @param decFile: the file to write the decompressed data to
  39         """
  40         self.stream = stream
  41         self.outFile = outFile.open('w')
  42         self.gzfile = None
  43         self.bz2file = None
  44         if decompress == ".gz":
  45             self.gzheader = True
  46             self.gzfile = decFile.open('w')
  47             self.gzdec = decompressobj(-MAX_WBITS)
  48         elif decompress == ".bz2":
  49             self.bz2file = decFile.open('w')
  50             self.bz2dec = BZ2Decompressor()
  51         self.length = self.stream.length
  52         self.start = 0
  53         self.doneDefer = defer.Deferred()
  54
  55     def _done(self):
  56         """Close the output file."""
  57         if not self.outFile.closed:
  58             self.outFile.close()
  59             if self.gzfile:
  60                 data_dec = self.gzdec.flush()
  61                 self.gzfile.write(data_dec)
  62                 self.gzfile.close()
  63                 self.gzfile = None
  64             if self.bz2file:
  65                 self.bz2file.close()
  66                 self.bz2file = None
  67
  68             self.doneDefer.callback(1)
  69
  70     def read(self):
  71         """Read some data from the stream."""
  72         if self.outFile.closed:
  73             return None
  74
  75         data = self.stream.read()
  76         if isinstance(data, defer.Deferred):
  77             data.addCallbacks(self._write, self._done)
  78             return data
  79
  80         self._write(data)
  81         return data
  82
  83     def _write(self, data):
  84         """Write the stream data to the file and return it for others to use."""
  85         if data is None:
  86             self._done()
  87             return data
  88
  89         self.outFile.write(data)
  90         if self.gzfile:
  91             if self.gzheader:
  92                 self.gzheader = False
  93                 new_data = self._remove_gzip_header(data)
  94                 dec_data = self.gzdec.decompress(new_data)
  95             else:
  96                 dec_data = self.gzdec.decompress(data)
  97             self.gzfile.write(dec_data)
  98         if self.bz2file:
  99             dec_data = self.bz2dec.decompress(data)
 100             self.bz2file.write(dec_data)
 101         return data
 102
 103     def _remove_gzip_header(self, data):
 104         if data[:2] != '\037\213':
 105             raise IOError, 'Not a gzipped file'
 106         if ord(data[2]) != 8:
 107             raise IOError, 'Unknown compression method'
 108         flag = ord(data[3])
 109         # modtime = self.fileobj.read(4)
 110         # extraflag = self.fileobj.read(1)
 111         # os = self.fileobj.read(1)
 112
 113         skip = 10
 114         if flag & FEXTRA:
 115             # Read & discard the extra field, if present
 116             xlen = ord(data[10])
 117             xlen = xlen + 256*ord(data[11])
 118             skip = skip + 2 + xlen
 119         if flag & FNAME:
 120             # Read and discard a null-terminated string containing the filename
 121             while True:
 122                 if not data[skip] or data[skip] == '\000':
 123                     break
 124                 skip += 1
 125             skip += 1
 126         if flag & FCOMMENT:
 127             # Read and discard a null-terminated string containing a comment
 128             while True:
 129                 if not data[skip] or data[skip] == '\000':
 130                     break
 131                 skip += 1
 132             skip += 1
 133         if flag & FHCRC:
 134             skip += 2     # Read & discard the 16-bit header CRC
 135         return data[skip:]
 136
 137     def close(self):
 138         """Clean everything up and return None to future reads."""
 139         self.length = 0
 140         self._done()
 141         self.stream.close()
 142
 143 class MirrorManager:
 144     """Manages all requests for mirror objects."""
 145
 146     def __init__(self, cache_dir):
 147         self.cache_dir = cache_dir
 148         self.cache = filepath.FilePath(self.cache_dir)
 149         self.apt_caches = {}
 150
 151     def extractPath(self, url):
 152         parsed = urlparse(url)
 153         host, port = splitHostPort(parsed[0], parsed[1])
 154         site = host + ":" + str(port)
 155         path = parsed[2]
 156
 157         i = max(path.rfind('/dists/'), path.rfind('/pool/'))
 158         if i >= 0:
 159             baseDir = path[:i]
 160             path = path[i:]
 161         else:
 162             # Uh oh, this is not good
 163             log.msg("Couldn't find a good base directory for path: %s" % (site + path))
 164             baseDir = ''
 165             if site in self.apt_caches:
 166                 longest_match = 0
 167                 for base in self.apt_caches[site]:
 168                     base_match = ''
 169                     for dirs in path.split('/'):
 170                         if base.startswith(base_match + '/' + dirs):
 171                             base_match += '/' + dirs
 172                         else:
 173                             break
 174                     if len(base_match) > longest_match:
 175                         longest_match = len(base_match)
 176                         baseDir = base_match
 177             log.msg("Settled on baseDir: %s" % baseDir)
 178
 179         return site, baseDir, path
 180
 181     def init(self, site, baseDir):
 182         if site not in self.apt_caches:
 183             self.apt_caches[site] = {}
 184
 185         if baseDir not in self.apt_caches[site]:
 186             site_cache = os.path.join(self.cache_dir, aptpkg_dir, 'mirrors', site + baseDir.replace('/', '_'))
 187             self.apt_caches[site][baseDir] = AptPackages(site_cache)
 188
 189     def findHash(self, url):
 190         site, baseDir, path = self.extractPath(url)
 191         if site in self.apt_caches and baseDir in self.apt_caches[site]:
 192             return self.apt_caches[site][baseDir].findHash(path)
 193         d = defer.Deferred()
 194         d.errback(MirrorError("Site Not Found"))
 195         return d
 196
 197     def save_file(self, response, hash, size, url):
 198         """Save a downloaded file to the cache and stream it."""
 199         log.msg('Returning file: %s' % url)
 200
 201         parsed = urlparse(url)
 202         destFile = self.cache.preauthChild(parsed[1] + parsed[2])
 203         log.msg('Saving returned %r byte file to cache: %s' % (response.stream.length, destFile.path))
 204
 205         if destFile.exists():
 206             log.msg('File already exists, removing: %s' % destFile.path)
 207             destFile.remove()
 208         else:
 209             destFile.parent().makedirs()
 210
 211         root, ext = os.path.splitext(destFile.basename())
 212         if root.lower() in DECOMPRESS_FILES and ext.lower() in DECOMPRESS_EXTS:
 213             ext = ext.lower()
 214             decFile = destFile.sibling(root)
 215             log.msg('Decompressing to: %s' % decFile.path)
 216             if decFile.exists():
 217                 log.msg('File already exists, removing: %s' % decFile.path)
 218                 decFile.remove()
 219         else:
 220             ext = None
 221             decFile = None
 222
 223         orig_stream = response.stream
 224         response.stream = ProxyFileStream(orig_stream, destFile, ext, decFile)
 225         response.stream.doneDefer.addCallback(self.save_complete, url, destFile,
 226                                               response.headers.getHeader('Last-Modified'),
 227                                               ext, decFile)
 228         response.stream.doneDefer.addErrback(self.save_error, url)
 229         return response
 230
 231     def save_complete(self, result, url, destFile, modtime = None, ext = None, decFile = None):
 232         """Update the modification time and AptPackages."""
 233         if modtime:
 234             os.utime(destFile.path, (modtime, modtime))
 235             if ext:
 236                 os.utime(decFile.path, (modtime, modtime))
 237
 238         site, baseDir, path = self.extractPath(url)
 239         self.init(site, baseDir)
 240         self.apt_caches[site][baseDir].file_updated(path, destFile.path)
 241         if ext:
 242             self.apt_caches[site][baseDir].file_updated(path[:-len(ext)], decFile.path)
 243
 244     def save_error(self, failure, url):
 245         """An error has occurred in downloadign or saving the file."""
 246         log.msg('Error occurred downloading %s' % url)
 247         log.err(failure)
 248         return failure
 249
 250 class TestMirrorManager(unittest.TestCase):
 251     """Unit tests for the mirror manager."""
 252
 253     timeout = 20
 254     pending_calls = []
 255     client = None
 256
 257     def setUp(self):
 258         self.client = MirrorManager('/tmp')
 259
 260     def test_extractPath(self):
 261         site, baseDir, path = self.client.extractPath('http://ftp.us.debian.org/debian/dists/unstable/Release')
 262         self.failUnless(site == "ftp.us.debian.org:80", "no match: %s" % site)
 263         self.failUnless(baseDir == "/debian", "no match: %s" % baseDir)
 264         self.failUnless(path == "/dists/unstable/Release", "no match: %s" % path)
 265
 266         site, baseDir, path = self.client.extractPath('http://ftp.us.debian.org:16999/debian/pool/d/dpkg/dpkg_1.2.1-1.tar.gz')
 267         self.failUnless(site == "ftp.us.debian.org:16999", "no match: %s" % site)
 268         self.failUnless(baseDir == "/debian", "no match: %s" % baseDir)
 269         self.failUnless(path == "/pool/d/dpkg/dpkg_1.2.1-1.tar.gz", "no match: %s" % path)
 270
 271         site, baseDir, path = self.client.extractPath('http://debian.camrdale.org/dists/unstable/Release')
 272         self.failUnless(site == "debian.camrdale.org:80", "no match: %s" % site)
 273         self.failUnless(baseDir == "", "no match: %s" % baseDir)
 274         self.failUnless(path == "/dists/unstable/Release", "no match: %s" % path)
 275
 276     def verifyHash(self, found_hash, path, true_hash):
 277         self.failUnless(found_hash[0] == true_hash,
 278                     "%s hashes don't match: %s != %s" % (path, found_hash[0], true_hash))
 279
 280     def test_findHash(self):
 281         self.packagesFile = os.popen('ls -Sr /var/lib/apt/lists/ | grep -E "_main_.*Packages$" | tail -n 1').read().rstrip('\n')
 282         self.sourcesFile = os.popen('ls -Sr /var/lib/apt/lists/ | grep -E "_main_.*Sources$" | tail -n 1').read().rstrip('\n')
 283         for f in os.walk('/var/lib/apt/lists').next()[2]:
 284             if f[-7:] == "Release" and self.packagesFile.startswith(f[:-7]):
 285                 self.releaseFile = f
 286                 break
 287
 288         self.client.updatedFile('http://' + self.releaseFile.replace('_','/'),
 289                                 '/var/lib/apt/lists/' + self.releaseFile)
 290         self.client.updatedFile('http://' + self.releaseFile[:self.releaseFile.find('_dists_')+1].replace('_','/') +
 291                                 self.packagesFile[self.packagesFile.find('_dists_')+1:].replace('_','/'),
 292                                 '/var/lib/apt/lists/' + self.packagesFile)
 293         self.client.updatedFile('http://' + self.releaseFile[:self.releaseFile.find('_dists_')+1].replace('_','/') +
 294                                 self.sourcesFile[self.sourcesFile.find('_dists_')+1:].replace('_','/'),
 295                                 '/var/lib/apt/lists/' + self.sourcesFile)
 296
 297         lastDefer = defer.Deferred()
 298
 299         idx_hash = os.popen('grep -A 3000 -E "^SHA1:" ' +
 300                             '/var/lib/apt/lists/' + self.releaseFile +
 301                             ' | grep -E " main/binary-i386/Packages.bz2$"'
 302                             ' | head -n 1 | cut -d\  -f 2').read().rstrip('\n')
 303         idx_path = 'http://' + self.releaseFile.replace('_','/')[:-7] + 'main/binary-i386/Packages.bz2'
 304
 305         d = self.client.findHash(idx_path)
 306         d.addCallback(self.verifyHash, idx_path, idx_hash)
 307
 308         pkg_hash = os.popen('grep -A 30 -E "^Package: dpkg$" ' +
 309                             '/var/lib/apt/lists/' + self.packagesFile +
 310                             ' | grep -E "^SHA1:" | head -n 1' +
 311                             ' | cut -d\  -f 2').read().rstrip('\n')
 312         pkg_path = 'http://' + self.releaseFile[:self.releaseFile.find('_dists_')+1].replace('_','/') + \
 313                    os.popen('grep -A 30 -E "^Package: dpkg$" ' +
 314                             '/var/lib/apt/lists/' + self.packagesFile +
 315                             ' | grep -E "^Filename:" | head -n 1' +
 316                             ' | cut -d\  -f 2').read().rstrip('\n')
 317
 318         d = self.client.findHash(pkg_path)
 319         d.addCallback(self.verifyHash, pkg_path, pkg_hash)
 320
 321         src_dir = os.popen('grep -A 30 -E "^Package: dpkg$" ' +
 322                             '/var/lib/apt/lists/' + self.sourcesFile +
 323                             ' | grep -E "^Directory:" | head -n 1' +
 324                             ' | cut -d\  -f 2').read().rstrip('\n')
 325         src_hashes = os.popen('grep -A 20 -E "^Package: dpkg$" ' +
 326                             '/var/lib/apt/lists/' + self.sourcesFile +
 327                             ' | grep -A 4 -E "^Files:" | grep -E "^ " ' +
 328                             ' | cut -d\  -f 2').read().split('\n')[:-1]
 329         src_paths = os.popen('grep -A 20 -E "^Package: dpkg$" ' +
 330                             '/var/lib/apt/lists/' + self.sourcesFile +
 331                             ' | grep -A 4 -E "^Files:" | grep -E "^ " ' +
 332                             ' | cut -d\  -f 4').read().split('\n')[:-1]
 333
 334         for i in range(len(src_hashes)):
 335             src_path = 'http://' + self.releaseFile[:self.releaseFile.find('_dists_')+1].replace('_','/') + src_dir + '/' + src_paths[i]
 336             d = self.client.findHash(src_path)
 337             d.addCallback(self.verifyHash, src_path, src_hashes[i])
 338
 339         idx_hash = os.popen('grep -A 3000 -E "^SHA1:" ' +
 340                             '/var/lib/apt/lists/' + self.releaseFile +
 341                             ' | grep -E " main/source/Sources.bz2$"'
 342                             ' | head -n 1 | cut -d\  -f 2').read().rstrip('\n')
 343         idx_path = 'http://' + self.releaseFile.replace('_','/')[:-7] + 'main/source/Sources.bz2'
 344
 345         d = self.client.findHash(idx_path)
 346         d.addCallback(self.verifyHash, idx_path, idx_hash)
 347
 348         d.addBoth(lastDefer.callback)
 349         return lastDefer
 350
 351     def tearDown(self):
 352         for p in self.pending_calls:
 353             if p.active():
 354                 p.cancel()
 355         self.client = None
 356