updateFallbackDirs.py 89 KB

1234567891011121314151617181920212223242526272829303132333435363738394041424344454647484950515253545556575859606162636465666768697071727374757677787980818283848586878889909192939495969798991001011021031041051061071081091101111121131141151161171181191201211221231241251261271281291301311321331341351361371381391401411421431441451461471481491501511521531541551561571581591601611621631641651661671681691701711721731741751761771781791801811821831841851861871881891901911921931941951961971981992002012022032042052062072082092102112122132142152162172182192202212222232242252262272282292302312322332342352362372382392402412422432442452462472482492502512522532542552562572582592602612622632642652662672682692702712722732742752762772782792802812822832842852862872882892902912922932942952962972982993003013023033043053063073083093103113123133143153163173183193203213223233243253263273283293303313323333343353363373383393403413423433443453463473483493503513523533543553563573583593603613623633643653663673683693703713723733743753763773783793803813823833843853863873883893903913923933943953963973983994004014024034044054064074084094104114124134144154164174184194204214224234244254264274284294304314324334344354364374384394404414424434444454464474484494504514524534544554564574584594604614624634644654664674684694704714724734744754764774784794804814824834844854864874884894904914924934944954964974984995005015025035045055065075085095105115125135145155165175185195205215225235245255265275285295305315325335345355365375385395405415425435445455465475485495505515525535545555565575585595605615625635645655665675685695705715725735745755765775785795805815825835845855865875885895905915925935945955965975985996006016026036046056066076086096106116126136146156166176186196206216226236246256266276286296306316326336346356366376386396406416426436446456466476486496506516526536546556566576586596606616626636646656666676686696706716726736746756766776786796806816826836846856866876886896906916926936946956966976986997007017027037047057067077087097107117127137147157167177187197207217227237247257267277287297307317327337347357367377387397407417427437447457467477487497507517527537547557567577587597607617627637647657667677687697707717727737747757767777787797807817827837847857867877887897907917927937947957967977987998008018028038048058068078088098108118128138148158168178188198208218228238248258268278288298308318328338348358368378388398408418428438448458468478488498508518528538548558568578588598608618628638648658668678688698708718728738748758768778788798808818828838848858868878888898908918928938948958968978988999009019029039049059069079089099109119129139149159169179189199209219229239249259269279289299309319329339349359369379389399409419429439449459469479489499509519529539549559569579589599609619629639649659669679689699709719729739749759769779789799809819829839849859869879889899909919929939949959969979989991000100110021003100410051006100710081009101010111012101310141015101610171018101910201021102210231024102510261027102810291030103110321033103410351036103710381039104010411042104310441045104610471048104910501051105210531054105510561057105810591060106110621063106410651066106710681069107010711072107310741075107610771078107910801081108210831084108510861087108810891090109110921093109410951096109710981099110011011102110311041105110611071108110911101111111211131114111511161117111811191120112111221123112411251126112711281129113011311132113311341135113611371138113911401141114211431144114511461147114811491150115111521153115411551156115711581159116011611162116311641165116611671168116911701171117211731174117511761177117811791180118111821183118411851186118711881189119011911192119311941195119611971198119912001201120212031204120512061207120812091210121112121213121412151216121712181219122012211222122312241225122612271228122912301231123212331234123512361237123812391240124112421243124412451246124712481249125012511252125312541255125612571258125912601261126212631264126512661267126812691270127112721273127412751276127712781279128012811282128312841285128612871288128912901291129212931294129512961297129812991300130113021303130413051306130713081309131013111312131313141315131613171318131913201321132213231324132513261327132813291330133113321333133413351336133713381339134013411342134313441345134613471348134913501351135213531354135513561357135813591360136113621363136413651366136713681369137013711372137313741375137613771378137913801381138213831384138513861387138813891390139113921393139413951396139713981399140014011402140314041405140614071408140914101411141214131414141514161417141814191420142114221423142414251426142714281429143014311432143314341435143614371438143914401441144214431444144514461447144814491450145114521453145414551456145714581459146014611462146314641465146614671468146914701471147214731474147514761477147814791480148114821483148414851486148714881489149014911492149314941495149614971498149915001501150215031504150515061507150815091510151115121513151415151516151715181519152015211522152315241525152615271528152915301531153215331534153515361537153815391540154115421543154415451546154715481549155015511552155315541555155615571558155915601561156215631564156515661567156815691570157115721573157415751576157715781579158015811582158315841585158615871588158915901591159215931594159515961597159815991600160116021603160416051606160716081609161016111612161316141615161616171618161916201621162216231624162516261627162816291630163116321633163416351636163716381639164016411642164316441645164616471648164916501651165216531654165516561657165816591660166116621663166416651666166716681669167016711672167316741675167616771678167916801681168216831684168516861687168816891690169116921693169416951696169716981699170017011702170317041705170617071708170917101711171217131714171517161717171817191720172117221723172417251726172717281729173017311732173317341735173617371738173917401741174217431744174517461747174817491750175117521753175417551756175717581759176017611762176317641765176617671768176917701771177217731774177517761777177817791780178117821783178417851786178717881789179017911792179317941795179617971798179918001801180218031804180518061807180818091810181118121813181418151816181718181819182018211822182318241825182618271828182918301831183218331834183518361837183818391840184118421843184418451846184718481849185018511852185318541855185618571858185918601861186218631864186518661867186818691870187118721873187418751876187718781879188018811882188318841885188618871888188918901891189218931894189518961897189818991900190119021903190419051906190719081909191019111912191319141915191619171918191919201921192219231924192519261927192819291930193119321933193419351936193719381939194019411942194319441945194619471948194919501951195219531954195519561957195819591960196119621963196419651966196719681969197019711972197319741975197619771978197919801981198219831984198519861987198819891990199119921993199419951996199719981999200020012002200320042005200620072008200920102011201220132014201520162017201820192020202120222023202420252026202720282029203020312032203320342035203620372038203920402041204220432044204520462047204820492050205120522053205420552056205720582059206020612062206320642065206620672068206920702071207220732074207520762077207820792080208120822083208420852086208720882089209020912092209320942095209620972098209921002101210221032104210521062107210821092110211121122113211421152116211721182119212021212122212321242125212621272128212921302131213221332134213521362137213821392140214121422143214421452146214721482149215021512152215321542155215621572158215921602161216221632164216521662167216821692170217121722173217421752176217721782179218021812182218321842185218621872188218921902191219221932194219521962197219821992200220122022203220422052206220722082209221022112212221322142215221622172218221922202221222222232224222522262227
  1. #!/usr/bin/python
  2. # Usage:
  3. #
  4. # Regenerate the list:
  5. # scripts/maint/updateFallbackDirs.py > src/or/fallback_dirs.inc 2> fallback_dirs.log
  6. #
  7. # Check the existing list:
  8. # scripts/maint/updateFallbackDirs.py check_existing > fallback_dirs.inc.ok 2> fallback_dirs.log
  9. # mv fallback_dirs.inc.ok src/or/fallback_dirs.inc
  10. #
  11. # This script should be run from a stable, reliable network connection,
  12. # with no other network activity (and not over tor).
  13. # If this is not possible, please disable:
  14. # PERFORM_IPV4_DIRPORT_CHECKS and PERFORM_IPV6_DIRPORT_CHECKS
  15. #
  16. # Needs dateutil (and potentially other python packages)
  17. # Needs stem available in your PYTHONPATH, or just ln -s ../stem/stem .
  18. # Optionally uses ipaddress (python 3 builtin) or py2-ipaddress (package)
  19. # for netblock analysis, in PYTHONPATH, or just
  20. # ln -s ../py2-ipaddress-3.4.1/ipaddress.py .
  21. #
  22. # Then read the logs to make sure the fallbacks aren't dominated by a single
  23. # netblock or port
  24. # Script by weasel, April 2015
  25. # Portions by gsathya & karsten, 2013
  26. # https://trac.torproject.org/projects/tor/attachment/ticket/8374/dir_list.2.py
  27. # Modifications by teor, 2015
  28. import StringIO
  29. import string
  30. import re
  31. import datetime
  32. import gzip
  33. import os.path
  34. import json
  35. import math
  36. import sys
  37. import urllib
  38. import urllib2
  39. import hashlib
  40. import dateutil.parser
  41. # bson_lazy provides bson
  42. #from bson import json_util
  43. import copy
  44. import re
  45. from stem.descriptor import DocumentHandler
  46. from stem.descriptor.remote import get_consensus
  47. import logging
  48. logging.root.name = ''
  49. HAVE_IPADDRESS = False
  50. try:
  51. # python 3 builtin, or install package py2-ipaddress
  52. # there are several ipaddress implementations for python 2
  53. # with slightly different semantics with str typed text
  54. # fortunately, all our IP addresses are in unicode
  55. import ipaddress
  56. HAVE_IPADDRESS = True
  57. except ImportError:
  58. # if this happens, we avoid doing netblock analysis
  59. logging.warning('Unable to import ipaddress, please install py2-ipaddress.' +
  60. ' A fallback list will be created, but optional netblock' +
  61. ' analysis will not be performed.')
  62. ## Top-Level Configuration
  63. # Output all candidate fallbacks, or only output selected fallbacks?
  64. OUTPUT_CANDIDATES = False
  65. # Perform DirPort checks over IPv4?
  66. # Change this to False if IPv4 doesn't work for you, or if you don't want to
  67. # download a consensus for each fallback
  68. # Don't check ~1000 candidates when OUTPUT_CANDIDATES is True
  69. PERFORM_IPV4_DIRPORT_CHECKS = False if OUTPUT_CANDIDATES else True
  70. # Perform DirPort checks over IPv6?
  71. # If you know IPv6 works for you, set this to True
  72. # This will exclude IPv6 relays without an IPv6 DirPort configured
  73. # So it's best left at False until #18394 is implemented
  74. # Don't check ~1000 candidates when OUTPUT_CANDIDATES is True
  75. PERFORM_IPV6_DIRPORT_CHECKS = False if OUTPUT_CANDIDATES else False
  76. # Must relays be running now?
  77. MUST_BE_RUNNING_NOW = (PERFORM_IPV4_DIRPORT_CHECKS
  78. or PERFORM_IPV6_DIRPORT_CHECKS)
  79. # Clients have been using microdesc consensuses by default for a while now
  80. DOWNLOAD_MICRODESC_CONSENSUS = True
  81. # If a relay delivers an expired consensus, if it expired less than this many
  82. # seconds ago, we still allow the relay. This should never be less than -90,
  83. # as all directory mirrors should have downloaded a consensus 90 minutes
  84. # before it expires. It should never be more than 24 hours, because clients
  85. # reject consensuses that are older than REASONABLY_LIVE_TIME.
  86. # For the consensus expiry check to be accurate, the machine running this
  87. # script needs an accurate clock.
  88. #
  89. # Relays on 0.3.0 and later return a 404 when they are about to serve an
  90. # expired consensus. This makes them fail the download check.
  91. # We use a tolerance of 0, so that 0.2.x series relays also fail the download
  92. # check if they serve an expired consensus.
  93. CONSENSUS_EXPIRY_TOLERANCE = 0
  94. # Output fallback name, flags, bandwidth, and ContactInfo in a C comment?
  95. OUTPUT_COMMENTS = True if OUTPUT_CANDIDATES else False
  96. # Output matching ContactInfo in fallbacks list or the blacklist?
  97. # Useful if you're trying to contact operators
  98. CONTACT_COUNT = True if OUTPUT_CANDIDATES else False
  99. CONTACT_BLACKLIST_COUNT = True if OUTPUT_CANDIDATES else False
  100. # How the list should be sorted:
  101. # fingerprint: is useful for stable diffs of fallback lists
  102. # measured_bandwidth: is useful when pruning the list based on bandwidth
  103. # contact: is useful for contacting operators once the list has been pruned
  104. OUTPUT_SORT_FIELD = 'contact' if OUTPUT_CANDIDATES else 'fingerprint'
  105. ## OnionOO Settings
  106. ONIONOO = 'https://onionoo.torproject.org/'
  107. #ONIONOO = 'https://onionoo.thecthulhu.com/'
  108. # Don't bother going out to the Internet, just use the files available locally,
  109. # even if they're very old
  110. LOCAL_FILES_ONLY = False
  111. ## Whitelist / Blacklist Filter Settings
  112. # The whitelist contains entries that are included if all attributes match
  113. # (IPv4, dirport, orport, id, and optionally IPv6 and IPv6 orport)
  114. # The blacklist contains (partial) entries that are excluded if any
  115. # sufficiently specific group of attributes matches:
  116. # IPv4 & DirPort
  117. # IPv4 & ORPort
  118. # ID
  119. # IPv6 & DirPort
  120. # IPv6 & IPv6 ORPort
  121. # If neither port is included in the blacklist, the entire IP address is
  122. # blacklisted.
  123. # What happens to entries in neither list?
  124. # When True, they are included, when False, they are excluded
  125. INCLUDE_UNLISTED_ENTRIES = True if OUTPUT_CANDIDATES else False
  126. # If an entry is in both lists, what happens?
  127. # When True, it is excluded, when False, it is included
  128. BLACKLIST_EXCLUDES_WHITELIST_ENTRIES = True
  129. WHITELIST_FILE_NAME = 'scripts/maint/fallback.whitelist'
  130. BLACKLIST_FILE_NAME = 'scripts/maint/fallback.blacklist'
  131. FALLBACK_FILE_NAME = 'src/or/fallback_dirs.inc'
  132. # The number of bytes we'll read from a filter file before giving up
  133. MAX_LIST_FILE_SIZE = 1024 * 1024
  134. ## Eligibility Settings
  135. # Require fallbacks to have the same address and port for a set amount of time
  136. # We used to have this at 1 week, but that caused many fallback failures, which
  137. # meant that we had to rebuild the list more often.
  138. #
  139. # There was a bug in Tor 0.2.8.1-alpha and earlier where a relay temporarily
  140. # submits a 0 DirPort when restarted.
  141. # This causes OnionOO to (correctly) reset its stability timer.
  142. # Affected relays should upgrade to Tor 0.2.8.7 or later, which has a fix
  143. # for this issue.
  144. ADDRESS_AND_PORT_STABLE_DAYS = 30
  145. # We ignore relays that have been down for more than this period
  146. MAX_DOWNTIME_DAYS = 0 if MUST_BE_RUNNING_NOW else 7
  147. # What time-weighted-fraction of these flags must FallbackDirs
  148. # Equal or Exceed?
  149. CUTOFF_RUNNING = .90
  150. CUTOFF_V2DIR = .90
  151. # Tolerate lower guard flag averages, as guard flags are removed for some time
  152. # after a relay restarts
  153. CUTOFF_GUARD = .80
  154. # What time-weighted-fraction of these flags must FallbackDirs
  155. # Equal or Fall Under?
  156. # .00 means no bad exits
  157. PERMITTED_BADEXIT = .00
  158. # older entries' weights are adjusted with ALPHA^(age in days)
  159. AGE_ALPHA = 0.99
  160. # this factor is used to scale OnionOO entries to [0,1]
  161. ONIONOO_SCALE_ONE = 999.
  162. ## Fallback Count Limits
  163. # The target for these parameters is 20% of the guards in the network
  164. # This is around 200 as of October 2015
  165. _FB_POG = 0.2
  166. FALLBACK_PROPORTION_OF_GUARDS = None if OUTPUT_CANDIDATES else _FB_POG
  167. # Limit the number of fallbacks (eliminating lowest by advertised bandwidth)
  168. MAX_FALLBACK_COUNT = None if OUTPUT_CANDIDATES else 200
  169. # Emit a C #error if the number of fallbacks is less than expected
  170. MIN_FALLBACK_COUNT = 0 if OUTPUT_CANDIDATES else MAX_FALLBACK_COUNT*0.5
  171. # The maximum number of fallbacks on the same address, contact, or family
  172. # With 200 fallbacks, this means each operator can see 1% of client bootstraps
  173. # (The directory authorities used to see ~12% of client bootstraps each.)
  174. MAX_FALLBACKS_PER_IP = 1
  175. MAX_FALLBACKS_PER_IPV4 = MAX_FALLBACKS_PER_IP
  176. MAX_FALLBACKS_PER_IPV6 = MAX_FALLBACKS_PER_IP
  177. MAX_FALLBACKS_PER_CONTACT = 3
  178. MAX_FALLBACKS_PER_FAMILY = 3
  179. ## Fallback Bandwidth Requirements
  180. # Any fallback with the Exit flag has its bandwidth multipled by this fraction
  181. # to make sure we aren't further overloading exits
  182. # (Set to 1.0, because we asked that only lightly loaded exits opt-in,
  183. # and the extra load really isn't that much for large relays.)
  184. EXIT_BANDWIDTH_FRACTION = 1.0
  185. # If a single fallback's bandwidth is too low, it's pointless adding it
  186. # We expect fallbacks to handle an extra 10 kilobytes per second of traffic
  187. # Make sure they can support a hundred times the expected extra load
  188. # (Use 102.4 to make it come out nicely in MByte/s)
  189. # We convert this to a consensus weight before applying the filter,
  190. # because all the bandwidth amounts are specified by the relay
  191. MIN_BANDWIDTH = 102.4 * 10.0 * 1024.0
  192. # Clients will time out after 30 seconds trying to download a consensus
  193. # So allow fallback directories half that to deliver a consensus
  194. # The exact download times might change based on the network connection
  195. # running this script, but only by a few seconds
  196. # There is also about a second of python overhead
  197. CONSENSUS_DOWNLOAD_SPEED_MAX = 15.0
  198. # If the relay fails a consensus check, retry the download
  199. # This avoids delisting a relay due to transient network conditions
  200. CONSENSUS_DOWNLOAD_RETRY = True
  201. ## Fallback Weights for Client Selection
  202. # All fallback weights are equal, and set to the value below
  203. # Authorities are weighted 1.0 by default
  204. # Clients use these weights to select fallbacks and authorities at random
  205. # If there are 100 fallbacks and 9 authorities:
  206. # - each fallback is chosen with probability 10.0/(10.0*100 + 1.0*9) ~= 0.99%
  207. # - each authority is chosen with probability 1.0/(10.0*100 + 1.0*9) ~= 0.09%
  208. # A client choosing a bootstrap directory server will choose a fallback for
  209. # 10.0/(10.0*100 + 1.0*9) * 100 = 99.1% of attempts, and an authority for
  210. # 1.0/(10.0*100 + 1.0*9) * 9 = 0.9% of attempts.
  211. # (This disregards the bootstrap schedules, where clients start by choosing
  212. # from fallbacks & authoritites, then later choose from only authorities.)
  213. FALLBACK_OUTPUT_WEIGHT = 10.0
  214. ## Parsing Functions
  215. def parse_ts(t):
  216. return datetime.datetime.strptime(t, "%Y-%m-%d %H:%M:%S")
  217. def remove_bad_chars(raw_string, bad_char_list):
  218. # Remove each character in the bad_char_list
  219. cleansed_string = raw_string
  220. for c in bad_char_list:
  221. cleansed_string = cleansed_string.replace(c, '')
  222. return cleansed_string
  223. def cleanse_unprintable(raw_string):
  224. # Remove all unprintable characters
  225. cleansed_string = ''
  226. for c in raw_string:
  227. if c in string.printable:
  228. cleansed_string += c
  229. return cleansed_string
  230. def cleanse_whitespace(raw_string):
  231. # Replace all whitespace characters with a space
  232. cleansed_string = raw_string
  233. for c in string.whitespace:
  234. cleansed_string = cleansed_string.replace(c, ' ')
  235. return cleansed_string
  236. def cleanse_c_multiline_comment(raw_string):
  237. cleansed_string = raw_string
  238. # Embedded newlines should be removed by tor/onionoo, but let's be paranoid
  239. cleansed_string = cleanse_whitespace(cleansed_string)
  240. # ContactInfo and Version can be arbitrary binary data
  241. cleansed_string = cleanse_unprintable(cleansed_string)
  242. # Prevent a malicious / unanticipated string from breaking out
  243. # of a C-style multiline comment
  244. # This removes '/*' and '*/' and '//'
  245. bad_char_list = '*/'
  246. # Prevent a malicious string from using C nulls
  247. bad_char_list += '\0'
  248. # Be safer by removing bad characters entirely
  249. cleansed_string = remove_bad_chars(cleansed_string, bad_char_list)
  250. # Some compilers may further process the content of comments
  251. # There isn't much we can do to cover every possible case
  252. # But comment-based directives are typically only advisory
  253. return cleansed_string
  254. def cleanse_c_string(raw_string):
  255. cleansed_string = raw_string
  256. # Embedded newlines should be removed by tor/onionoo, but let's be paranoid
  257. cleansed_string = cleanse_whitespace(cleansed_string)
  258. # ContactInfo and Version can be arbitrary binary data
  259. cleansed_string = cleanse_unprintable(cleansed_string)
  260. # Prevent a malicious address/fingerprint string from breaking out
  261. # of a C-style string
  262. bad_char_list = '"'
  263. # Prevent a malicious string from using escapes
  264. bad_char_list += '\\'
  265. # Prevent a malicious string from using C nulls
  266. bad_char_list += '\0'
  267. # Be safer by removing bad characters entirely
  268. cleansed_string = remove_bad_chars(cleansed_string, bad_char_list)
  269. # Some compilers may further process the content of strings
  270. # There isn't much we can do to cover every possible case
  271. # But this typically only results in changes to the string data
  272. return cleansed_string
  273. ## OnionOO Source Functions
  274. # a dictionary of source metadata for each onionoo query we've made
  275. fetch_source = {}
  276. # register source metadata for 'what'
  277. # assumes we only retrieve one document for each 'what'
  278. def register_fetch_source(what, url, relays_published, version):
  279. fetch_source[what] = {}
  280. fetch_source[what]['url'] = url
  281. fetch_source[what]['relays_published'] = relays_published
  282. fetch_source[what]['version'] = version
  283. # list each registered source's 'what'
  284. def fetch_source_list():
  285. return sorted(fetch_source.keys())
  286. # given 'what', provide a multiline C comment describing the source
  287. def describe_fetch_source(what):
  288. desc = '/*'
  289. desc += '\n'
  290. desc += 'Onionoo Source: '
  291. desc += cleanse_c_multiline_comment(what)
  292. desc += ' Date: '
  293. desc += cleanse_c_multiline_comment(fetch_source[what]['relays_published'])
  294. desc += ' Version: '
  295. desc += cleanse_c_multiline_comment(fetch_source[what]['version'])
  296. desc += '\n'
  297. desc += 'URL: '
  298. desc += cleanse_c_multiline_comment(fetch_source[what]['url'])
  299. desc += '\n'
  300. desc += '*/'
  301. return desc
  302. ## File Processing Functions
  303. def write_to_file(str, file_name, max_len):
  304. try:
  305. with open(file_name, 'w') as f:
  306. f.write(str[0:max_len])
  307. except EnvironmentError, error:
  308. logging.error('Writing file %s failed: %d: %s'%
  309. (file_name,
  310. error.errno,
  311. error.strerror)
  312. )
  313. def read_from_file(file_name, max_len):
  314. try:
  315. if os.path.isfile(file_name):
  316. with open(file_name, 'r') as f:
  317. return f.read(max_len)
  318. except EnvironmentError, error:
  319. logging.info('Loading file %s failed: %d: %s'%
  320. (file_name,
  321. error.errno,
  322. error.strerror)
  323. )
  324. return None
  325. def parse_fallback_file(file_name):
  326. file_data = read_from_file(file_name, MAX_LIST_FILE_SIZE)
  327. file_data = cleanse_unprintable(file_data)
  328. file_data = remove_bad_chars(file_data, '\n"\0')
  329. file_data = re.sub('/\*.*?\*/', '', file_data)
  330. file_data = file_data.replace(',', '\n')
  331. file_data = file_data.replace(' weight=10', '')
  332. return file_data
  333. def load_possibly_compressed_response_json(response):
  334. if response.info().get('Content-Encoding') == 'gzip':
  335. buf = StringIO.StringIO( response.read() )
  336. f = gzip.GzipFile(fileobj=buf)
  337. return json.load(f)
  338. else:
  339. return json.load(response)
  340. def load_json_from_file(json_file_name):
  341. # An exception here may be resolved by deleting the .last_modified
  342. # and .json files, and re-running the script
  343. try:
  344. with open(json_file_name, 'r') as f:
  345. return json.load(f)
  346. except EnvironmentError, error:
  347. raise Exception('Reading not-modified json file %s failed: %d: %s'%
  348. (json_file_name,
  349. error.errno,
  350. error.strerror)
  351. )
  352. ## OnionOO Functions
  353. def datestr_to_datetime(datestr):
  354. # Parse datetimes like: Fri, 02 Oct 2015 13:34:14 GMT
  355. if datestr is not None:
  356. dt = dateutil.parser.parse(datestr)
  357. else:
  358. # Never modified - use start of epoch
  359. dt = datetime.datetime.utcfromtimestamp(0)
  360. # strip any timezone out (in case they're supported in future)
  361. dt = dt.replace(tzinfo=None)
  362. return dt
  363. def onionoo_fetch(what, **kwargs):
  364. params = kwargs
  365. params['type'] = 'relay'
  366. #params['limit'] = 10
  367. params['first_seen_days'] = '%d-'%(ADDRESS_AND_PORT_STABLE_DAYS)
  368. params['last_seen_days'] = '-%d'%(MAX_DOWNTIME_DAYS)
  369. params['flag'] = 'V2Dir'
  370. url = ONIONOO + what + '?' + urllib.urlencode(params)
  371. # Unfortunately, the URL is too long for some OS filenames,
  372. # but we still don't want to get files from different URLs mixed up
  373. base_file_name = what + '-' + hashlib.sha1(url).hexdigest()
  374. full_url_file_name = base_file_name + '.full_url'
  375. MAX_FULL_URL_LENGTH = 1024
  376. last_modified_file_name = base_file_name + '.last_modified'
  377. MAX_LAST_MODIFIED_LENGTH = 64
  378. json_file_name = base_file_name + '.json'
  379. if LOCAL_FILES_ONLY:
  380. # Read from the local file, don't write to anything
  381. response_json = load_json_from_file(json_file_name)
  382. else:
  383. # store the full URL to a file for debugging
  384. # no need to compare as long as you trust SHA-1
  385. write_to_file(url, full_url_file_name, MAX_FULL_URL_LENGTH)
  386. request = urllib2.Request(url)
  387. request.add_header('Accept-encoding', 'gzip')
  388. # load the last modified date from the file, if it exists
  389. last_mod_date = read_from_file(last_modified_file_name,
  390. MAX_LAST_MODIFIED_LENGTH)
  391. if last_mod_date is not None:
  392. request.add_header('If-modified-since', last_mod_date)
  393. # Parse last modified date
  394. last_mod = datestr_to_datetime(last_mod_date)
  395. # Not Modified and still recent enough to be useful
  396. # Onionoo / Globe used to use 6 hours, but we can afford a day
  397. required_freshness = datetime.datetime.utcnow()
  398. # strip any timezone out (to match dateutil.parser)
  399. required_freshness = required_freshness.replace(tzinfo=None)
  400. required_freshness -= datetime.timedelta(hours=24)
  401. # Make the OnionOO request
  402. response_code = 0
  403. try:
  404. response = urllib2.urlopen(request)
  405. response_code = response.getcode()
  406. except urllib2.HTTPError, error:
  407. response_code = error.code
  408. if response_code == 304: # not modified
  409. pass
  410. else:
  411. raise Exception("Could not get " + url + ": "
  412. + str(error.code) + ": " + error.reason)
  413. if response_code == 200: # OK
  414. last_mod = datestr_to_datetime(response.info().get('Last-Modified'))
  415. # Check for freshness
  416. if last_mod < required_freshness:
  417. if last_mod_date is not None:
  418. # This check sometimes fails transiently, retry the script if it does
  419. date_message = "Outdated data: last updated " + last_mod_date
  420. else:
  421. date_message = "No data: never downloaded "
  422. raise Exception(date_message + " from " + url)
  423. # Process the data
  424. if response_code == 200: # OK
  425. response_json = load_possibly_compressed_response_json(response)
  426. with open(json_file_name, 'w') as f:
  427. # use the most compact json representation to save space
  428. json.dump(response_json, f, separators=(',',':'))
  429. # store the last modified date in its own file
  430. if response.info().get('Last-modified') is not None:
  431. write_to_file(response.info().get('Last-Modified'),
  432. last_modified_file_name,
  433. MAX_LAST_MODIFIED_LENGTH)
  434. elif response_code == 304: # Not Modified
  435. response_json = load_json_from_file(json_file_name)
  436. else: # Unexpected HTTP response code not covered in the HTTPError above
  437. raise Exception("Unexpected HTTP response code to " + url + ": "
  438. + str(response_code))
  439. register_fetch_source(what,
  440. url,
  441. response_json['relays_published'],
  442. response_json['version'])
  443. return response_json
  444. def fetch(what, **kwargs):
  445. #x = onionoo_fetch(what, **kwargs)
  446. # don't use sort_keys, as the order of or_addresses is significant
  447. #print json.dumps(x, indent=4, separators=(',', ': '))
  448. #sys.exit(0)
  449. return onionoo_fetch(what, **kwargs)
  450. ## Fallback Candidate Class
  451. class Candidate(object):
  452. CUTOFF_ADDRESS_AND_PORT_STABLE = (datetime.datetime.utcnow()
  453. - datetime.timedelta(ADDRESS_AND_PORT_STABLE_DAYS))
  454. def __init__(self, details):
  455. for f in ['fingerprint', 'nickname', 'last_changed_address_or_port',
  456. 'consensus_weight', 'or_addresses', 'dir_address']:
  457. if not f in details: raise Exception("Document has no %s field."%(f,))
  458. if not 'contact' in details:
  459. details['contact'] = None
  460. if not 'flags' in details or details['flags'] is None:
  461. details['flags'] = []
  462. if (not 'advertised_bandwidth' in details
  463. or details['advertised_bandwidth'] is None):
  464. # relays without advertised bandwdith have it calculated from their
  465. # consensus weight
  466. details['advertised_bandwidth'] = 0
  467. if (not 'effective_family' in details
  468. or details['effective_family'] is None):
  469. details['effective_family'] = []
  470. if not 'platform' in details:
  471. details['platform'] = None
  472. details['last_changed_address_or_port'] = parse_ts(
  473. details['last_changed_address_or_port'])
  474. self._data = details
  475. self._stable_sort_or_addresses()
  476. self._fpr = self._data['fingerprint']
  477. self._running = self._guard = self._v2dir = 0.
  478. self._split_dirport()
  479. self._compute_orport()
  480. if self.orport is None:
  481. raise Exception("Failed to get an orport for %s."%(self._fpr,))
  482. self._compute_ipv6addr()
  483. if not self.has_ipv6():
  484. logging.debug("Failed to get an ipv6 address for %s."%(self._fpr,))
  485. self._compute_version()
  486. def _stable_sort_or_addresses(self):
  487. # replace self._data['or_addresses'] with a stable ordering,
  488. # sorting the secondary addresses in string order
  489. # leave the received order in self._data['or_addresses_raw']
  490. self._data['or_addresses_raw'] = self._data['or_addresses']
  491. or_address_primary = self._data['or_addresses'][:1]
  492. # subsequent entries in the or_addresses array are in an arbitrary order
  493. # so we stabilise the addresses by sorting them in string order
  494. or_addresses_secondaries_stable = sorted(self._data['or_addresses'][1:])
  495. or_addresses_stable = or_address_primary + or_addresses_secondaries_stable
  496. self._data['or_addresses'] = or_addresses_stable
  497. def get_fingerprint(self):
  498. return self._fpr
  499. # is_valid_ipv[46]_address by gsathya, karsten, 2013
  500. @staticmethod
  501. def is_valid_ipv4_address(address):
  502. if not isinstance(address, (str, unicode)):
  503. return False
  504. # check if there are four period separated values
  505. if address.count(".") != 3:
  506. return False
  507. # checks that each value in the octet are decimal values between 0-255
  508. for entry in address.split("."):
  509. if not entry.isdigit() or int(entry) < 0 or int(entry) > 255:
  510. return False
  511. elif entry[0] == "0" and len(entry) > 1:
  512. return False # leading zeros, for instance in "1.2.3.001"
  513. return True
  514. @staticmethod
  515. def is_valid_ipv6_address(address):
  516. if not isinstance(address, (str, unicode)):
  517. return False
  518. # remove brackets
  519. address = address[1:-1]
  520. # addresses are made up of eight colon separated groups of four hex digits
  521. # with leading zeros being optional
  522. # https://en.wikipedia.org/wiki/IPv6#Address_format
  523. colon_count = address.count(":")
  524. if colon_count > 7:
  525. return False # too many groups
  526. elif colon_count != 7 and not "::" in address:
  527. return False # not enough groups and none are collapsed
  528. elif address.count("::") > 1 or ":::" in address:
  529. return False # multiple groupings of zeros can't be collapsed
  530. found_ipv4_on_previous_entry = False
  531. for entry in address.split(":"):
  532. # If an IPv6 address has an embedded IPv4 address,
  533. # it must be the last entry
  534. if found_ipv4_on_previous_entry:
  535. return False
  536. if not re.match("^[0-9a-fA-f]{0,4}$", entry):
  537. if not Candidate.is_valid_ipv4_address(entry):
  538. return False
  539. else:
  540. found_ipv4_on_previous_entry = True
  541. return True
  542. def _split_dirport(self):
  543. # Split the dir_address into dirip and dirport
  544. (self.dirip, _dirport) = self._data['dir_address'].split(':', 2)
  545. self.dirport = int(_dirport)
  546. def _compute_orport(self):
  547. # Choose the first ORPort that's on the same IPv4 address as the DirPort.
  548. # In rare circumstances, this might not be the primary ORPort address.
  549. # However, _stable_sort_or_addresses() ensures we choose the same one
  550. # every time, even if onionoo changes the order of the secondaries.
  551. self._split_dirport()
  552. self.orport = None
  553. for i in self._data['or_addresses']:
  554. if i != self._data['or_addresses'][0]:
  555. logging.debug('Secondary IPv4 Address Used for %s: %s'%(self._fpr, i))
  556. (ipaddr, port) = i.rsplit(':', 1)
  557. if (ipaddr == self.dirip) and Candidate.is_valid_ipv4_address(ipaddr):
  558. self.orport = int(port)
  559. return
  560. def _compute_ipv6addr(self):
  561. # Choose the first IPv6 address that uses the same port as the ORPort
  562. # Or, choose the first IPv6 address in the list
  563. # _stable_sort_or_addresses() ensures we choose the same IPv6 address
  564. # every time, even if onionoo changes the order of the secondaries.
  565. self.ipv6addr = None
  566. self.ipv6orport = None
  567. # Choose the first IPv6 address that uses the same port as the ORPort
  568. for i in self._data['or_addresses']:
  569. (ipaddr, port) = i.rsplit(':', 1)
  570. if (port == self.orport) and Candidate.is_valid_ipv6_address(ipaddr):
  571. self.ipv6addr = ipaddr
  572. self.ipv6orport = int(port)
  573. return
  574. # Choose the first IPv6 address in the list
  575. for i in self._data['or_addresses']:
  576. (ipaddr, port) = i.rsplit(':', 1)
  577. if Candidate.is_valid_ipv6_address(ipaddr):
  578. self.ipv6addr = ipaddr
  579. self.ipv6orport = int(port)
  580. return
  581. def _compute_version(self):
  582. # parse the version out of the platform string
  583. # The platform looks like: "Tor 0.2.7.6 on Linux"
  584. self._data['version'] = None
  585. if self._data['platform'] is None:
  586. return
  587. # be tolerant of weird whitespacing, use a whitespace split
  588. tokens = self._data['platform'].split()
  589. for token in tokens:
  590. vnums = token.split('.')
  591. # if it's at least a.b.c.d, with potentially an -alpha-dev, -alpha, -rc
  592. if (len(vnums) >= 4 and vnums[0].isdigit() and vnums[1].isdigit() and
  593. vnums[2].isdigit()):
  594. self._data['version'] = token
  595. return
  596. # From #20509
  597. # bug #20499 affects versions from 0.2.9.1-alpha-dev to 0.2.9.4-alpha-dev
  598. # and version 0.3.0.0-alpha-dev
  599. # Exhaustive lists are hard to get wrong
  600. STALE_CONSENSUS_VERSIONS = ['0.2.9.1-alpha-dev',
  601. '0.2.9.2-alpha',
  602. '0.2.9.2-alpha-dev',
  603. '0.2.9.3-alpha',
  604. '0.2.9.3-alpha-dev',
  605. '0.2.9.4-alpha',
  606. '0.2.9.4-alpha-dev',
  607. '0.3.0.0-alpha-dev'
  608. ]
  609. def is_valid_version(self):
  610. # call _compute_version before calling this
  611. # is the version of the relay a version we want as a fallback?
  612. # checks both recommended versions and bug #20499 / #20509
  613. #
  614. # if the relay doesn't have a recommended version field, exclude the relay
  615. if not self._data.has_key('recommended_version'):
  616. log_excluded('%s not a candidate: no recommended_version field',
  617. self._fpr)
  618. return False
  619. if not self._data['recommended_version']:
  620. log_excluded('%s not a candidate: version not recommended', self._fpr)
  621. return False
  622. # if the relay doesn't have version field, exclude the relay
  623. if not self._data.has_key('version'):
  624. log_excluded('%s not a candidate: no version field', self._fpr)
  625. return False
  626. if self._data['version'] in Candidate.STALE_CONSENSUS_VERSIONS:
  627. logging.warning('%s not a candidate: version delivers stale consensuses',
  628. self._fpr)
  629. return False
  630. return True
  631. @staticmethod
  632. def _extract_generic_history(history, which='unknown'):
  633. # given a tree like this:
  634. # {
  635. # "1_month": {
  636. # "count": 187,
  637. # "factor": 0.001001001001001001,
  638. # "first": "2015-02-27 06:00:00",
  639. # "interval": 14400,
  640. # "last": "2015-03-30 06:00:00",
  641. # "values": [
  642. # 999,
  643. # 999
  644. # ]
  645. # },
  646. # "1_week": {
  647. # "count": 169,
  648. # "factor": 0.001001001001001001,
  649. # "first": "2015-03-23 07:30:00",
  650. # "interval": 3600,
  651. # "last": "2015-03-30 07:30:00",
  652. # "values": [ ...]
  653. # },
  654. # "1_year": {
  655. # "count": 177,
  656. # "factor": 0.001001001001001001,
  657. # "first": "2014-04-11 00:00:00",
  658. # "interval": 172800,
  659. # "last": "2015-03-29 00:00:00",
  660. # "values": [ ...]
  661. # },
  662. # "3_months": {
  663. # "count": 185,
  664. # "factor": 0.001001001001001001,
  665. # "first": "2014-12-28 06:00:00",
  666. # "interval": 43200,
  667. # "last": "2015-03-30 06:00:00",
  668. # "values": [ ...]
  669. # }
  670. # },
  671. # extract exactly one piece of data per time interval,
  672. # using smaller intervals where available.
  673. #
  674. # returns list of (age, length, value) dictionaries.
  675. generic_history = []
  676. periods = history.keys()
  677. periods.sort(key = lambda x: history[x]['interval'])
  678. now = datetime.datetime.utcnow()
  679. newest = now
  680. for p in periods:
  681. h = history[p]
  682. interval = datetime.timedelta(seconds = h['interval'])
  683. this_ts = parse_ts(h['last'])
  684. if (len(h['values']) != h['count']):
  685. logging.warning('Inconsistent value count in %s document for %s'
  686. %(p, which))
  687. for v in reversed(h['values']):
  688. if (this_ts <= newest):
  689. agt1 = now - this_ts
  690. agt2 = interval
  691. agetmp1 = (agt1.microseconds + (agt1.seconds + agt1.days * 24 * 3600)
  692. * 10**6) / 10**6
  693. agetmp2 = (agt2.microseconds + (agt2.seconds + agt2.days * 24 * 3600)
  694. * 10**6) / 10**6
  695. generic_history.append(
  696. { 'age': agetmp1,
  697. 'length': agetmp2,
  698. 'value': v
  699. })
  700. newest = this_ts
  701. this_ts -= interval
  702. if (this_ts + interval != parse_ts(h['first'])):
  703. logging.warning('Inconsistent time information in %s document for %s'
  704. %(p, which))
  705. #print json.dumps(generic_history, sort_keys=True,
  706. # indent=4, separators=(',', ': '))
  707. return generic_history
  708. @staticmethod
  709. def _avg_generic_history(generic_history):
  710. a = []
  711. for i in generic_history:
  712. if i['age'] > (ADDRESS_AND_PORT_STABLE_DAYS * 24 * 3600):
  713. continue
  714. if (i['length'] is not None
  715. and i['age'] is not None
  716. and i['value'] is not None):
  717. w = i['length'] * math.pow(AGE_ALPHA, i['age']/(3600*24))
  718. a.append( (i['value'] * w, w) )
  719. sv = math.fsum(map(lambda x: x[0], a))
  720. sw = math.fsum(map(lambda x: x[1], a))
  721. if sw == 0.0:
  722. svw = 0.0
  723. else:
  724. svw = sv/sw
  725. return svw
  726. def _add_generic_history(self, history):
  727. periods = r['read_history'].keys()
  728. periods.sort(key = lambda x: r['read_history'][x]['interval'] )
  729. print periods
  730. def add_running_history(self, history):
  731. pass
  732. def add_uptime(self, uptime):
  733. logging.debug('Adding uptime %s.'%(self._fpr,))
  734. # flags we care about: Running, V2Dir, Guard
  735. if not 'flags' in uptime:
  736. logging.debug('No flags in document for %s.'%(self._fpr,))
  737. return
  738. for f in ['Running', 'Guard', 'V2Dir']:
  739. if not f in uptime['flags']:
  740. logging.debug('No %s in flags for %s.'%(f, self._fpr,))
  741. return
  742. running = self._extract_generic_history(uptime['flags']['Running'],
  743. '%s-Running'%(self._fpr))
  744. guard = self._extract_generic_history(uptime['flags']['Guard'],
  745. '%s-Guard'%(self._fpr))
  746. v2dir = self._extract_generic_history(uptime['flags']['V2Dir'],
  747. '%s-V2Dir'%(self._fpr))
  748. if 'BadExit' in uptime['flags']:
  749. badexit = self._extract_generic_history(uptime['flags']['BadExit'],
  750. '%s-BadExit'%(self._fpr))
  751. self._running = self._avg_generic_history(running) / ONIONOO_SCALE_ONE
  752. self._guard = self._avg_generic_history(guard) / ONIONOO_SCALE_ONE
  753. self._v2dir = self._avg_generic_history(v2dir) / ONIONOO_SCALE_ONE
  754. self._badexit = None
  755. if 'BadExit' in uptime['flags']:
  756. self._badexit = self._avg_generic_history(badexit) / ONIONOO_SCALE_ONE
  757. def is_candidate(self):
  758. try:
  759. if (MUST_BE_RUNNING_NOW and not self.is_running()):
  760. log_excluded('%s not a candidate: not running now, unable to check ' +
  761. 'DirPort consensus download', self._fpr)
  762. return False
  763. if (self._data['last_changed_address_or_port'] >
  764. self.CUTOFF_ADDRESS_AND_PORT_STABLE):
  765. log_excluded('%s not a candidate: changed address/port recently (%s)',
  766. self._fpr, self._data['last_changed_address_or_port'])
  767. return False
  768. if self._running < CUTOFF_RUNNING:
  769. log_excluded('%s not a candidate: running avg too low (%lf)',
  770. self._fpr, self._running)
  771. return False
  772. if self._v2dir < CUTOFF_V2DIR:
  773. log_excluded('%s not a candidate: v2dir avg too low (%lf)',
  774. self._fpr, self._v2dir)
  775. return False
  776. if self._badexit is not None and self._badexit > PERMITTED_BADEXIT:
  777. log_excluded('%s not a candidate: badexit avg too high (%lf)',
  778. self._fpr, self._badexit)
  779. return False
  780. # this function logs a message depending on which check fails
  781. if not self.is_valid_version():
  782. return False
  783. if self._guard < CUTOFF_GUARD:
  784. log_excluded('%s not a candidate: guard avg too low (%lf)',
  785. self._fpr, self._guard)
  786. return False
  787. if (not self._data.has_key('consensus_weight')
  788. or self._data['consensus_weight'] < 1):
  789. log_excluded('%s not a candidate: consensus weight invalid', self._fpr)
  790. return False
  791. except BaseException as e:
  792. logging.warning("Exception %s when checking if fallback is a candidate",
  793. str(e))
  794. return False
  795. return True
  796. def is_in_whitelist(self, relaylist):
  797. """ A fallback matches if each key in the whitelist line matches:
  798. ipv4
  799. dirport
  800. orport
  801. id
  802. ipv6 address and port (if present)
  803. If the fallback has an ipv6 key, the whitelist line must also have
  804. it, and vice versa, otherwise they don't match. """
  805. ipv6 = None
  806. if self.has_ipv6():
  807. ipv6 = '%s:%d'%(self.ipv6addr, self.ipv6orport)
  808. for entry in relaylist:
  809. if entry['id'] != self._fpr:
  810. # can't log here unless we match an IP and port, because every relay's
  811. # fingerprint is compared to every entry's fingerprint
  812. if entry['ipv4'] == self.dirip and int(entry['orport']) == self.orport:
  813. logging.warning('%s excluded: has OR %s:%d changed fingerprint to ' +
  814. '%s?', entry['id'], self.dirip, self.orport,
  815. self._fpr)
  816. if self.has_ipv6() and entry.has_key('ipv6') and entry['ipv6'] == ipv6:
  817. logging.warning('%s excluded: has OR %s changed fingerprint to ' +
  818. '%s?', entry['id'], ipv6, self._fpr)
  819. continue
  820. if entry['ipv4'] != self.dirip:
  821. logging.warning('%s excluded: has it changed IPv4 from %s to %s?',
  822. self._fpr, entry['ipv4'], self.dirip)
  823. continue
  824. if int(entry['dirport']) != self.dirport:
  825. logging.warning('%s excluded: has it changed DirPort from %s:%d to ' +
  826. '%s:%d?', self._fpr, self.dirip, int(entry['dirport']),
  827. self.dirip, self.dirport)
  828. continue
  829. if int(entry['orport']) != self.orport:
  830. logging.warning('%s excluded: has it changed ORPort from %s:%d to ' +
  831. '%s:%d?', self._fpr, self.dirip, int(entry['orport']),
  832. self.dirip, self.orport)
  833. continue
  834. if entry.has_key('ipv6') and self.has_ipv6():
  835. # if both entry and fallback have an ipv6 address, compare them
  836. if entry['ipv6'] != ipv6:
  837. logging.warning('%s excluded: has it changed IPv6 ORPort from %s ' +
  838. 'to %s?', self._fpr, entry['ipv6'], ipv6)
  839. continue
  840. # if the fallback has an IPv6 address but the whitelist entry
  841. # doesn't, or vice versa, the whitelist entry doesn't match
  842. elif entry.has_key('ipv6') and not self.has_ipv6():
  843. logging.warning('%s excluded: has it lost its former IPv6 address %s?',
  844. self._fpr, entry['ipv6'])
  845. continue
  846. elif not entry.has_key('ipv6') and self.has_ipv6():
  847. logging.warning('%s excluded: has it gained an IPv6 address %s?',
  848. self._fpr, ipv6)
  849. continue
  850. return True
  851. return False
  852. def is_in_blacklist(self, relaylist):
  853. """ A fallback matches a blacklist line if a sufficiently specific group
  854. of attributes matches:
  855. ipv4 & dirport
  856. ipv4 & orport
  857. id
  858. ipv6 & dirport
  859. ipv6 & ipv6 orport
  860. If the fallback and the blacklist line both have an ipv6 key,
  861. their values will be compared, otherwise, they will be ignored.
  862. If there is no dirport and no orport, the entry matches all relays on
  863. that ip. """
  864. for entry in relaylist:
  865. for key in entry:
  866. value = entry[key]
  867. if key == 'id' and value == self._fpr:
  868. log_excluded('%s is in the blacklist: fingerprint matches',
  869. self._fpr)
  870. return True
  871. if key == 'ipv4' and value == self.dirip:
  872. # if the dirport is present, check it too
  873. if entry.has_key('dirport'):
  874. if int(entry['dirport']) == self.dirport:
  875. log_excluded('%s is in the blacklist: IPv4 (%s) and ' +
  876. 'DirPort (%d) match', self._fpr, self.dirip,
  877. self.dirport)
  878. return True
  879. # if the orport is present, check it too
  880. elif entry.has_key('orport'):
  881. if int(entry['orport']) == self.orport:
  882. log_excluded('%s is in the blacklist: IPv4 (%s) and ' +
  883. 'ORPort (%d) match', self._fpr, self.dirip,
  884. self.orport)
  885. return True
  886. else:
  887. log_excluded('%s is in the blacklist: IPv4 (%s) matches, and ' +
  888. 'entry has no DirPort or ORPort', self._fpr,
  889. self.dirip)
  890. return True
  891. ipv6 = None
  892. if self.has_ipv6():
  893. ipv6 = '%s:%d'%(self.ipv6addr, self.ipv6orport)
  894. if (key == 'ipv6' and self.has_ipv6()):
  895. # if both entry and fallback have an ipv6 address, compare them,
  896. # otherwise, disregard ipv6 addresses
  897. if value == ipv6:
  898. # if the dirport is present, check it too
  899. if entry.has_key('dirport'):
  900. if int(entry['dirport']) == self.dirport:
  901. log_excluded('%s is in the blacklist: IPv6 (%s) and ' +
  902. 'DirPort (%d) match', self._fpr, ipv6,
  903. self.dirport)
  904. return True
  905. # we've already checked the ORPort, it's part of entry['ipv6']
  906. else:
  907. log_excluded('%s is in the blacklist: IPv6 (%s) matches, and' +
  908. 'entry has no DirPort', self._fpr, ipv6)
  909. return True
  910. elif (key == 'ipv6' or self.has_ipv6()):
  911. # only log if the fingerprint matches but the IPv6 doesn't
  912. if entry.has_key('id') and entry['id'] == self._fpr:
  913. log_excluded('%s skipping IPv6 blacklist comparison: relay ' +
  914. 'has%s IPv6%s, but entry has%s IPv6%s', self._fpr,
  915. '' if self.has_ipv6() else ' no',
  916. (' (' + ipv6 + ')') if self.has_ipv6() else '',
  917. '' if key == 'ipv6' else ' no',
  918. (' (' + value + ')') if key == 'ipv6' else '')
  919. logging.warning('Has %s %s IPv6 address %s?', self._fpr,
  920. 'gained an' if self.has_ipv6() else 'lost its former',
  921. ipv6 if self.has_ipv6() else value)
  922. return False
  923. def cw_to_bw_factor(self):
  924. # any relays with a missing or zero consensus weight are not candidates
  925. # any relays with a missing advertised bandwidth have it set to zero
  926. return self._data['advertised_bandwidth'] / self._data['consensus_weight']
  927. # since advertised_bandwidth is reported by the relay, it can be gamed
  928. # to avoid this, use the median consensus weight to bandwidth factor to
  929. # estimate this relay's measured bandwidth, and make that the upper limit
  930. def measured_bandwidth(self, median_cw_to_bw_factor):
  931. cw_to_bw= median_cw_to_bw_factor
  932. # Reduce exit bandwidth to make sure we're not overloading them
  933. if self.is_exit():
  934. cw_to_bw *= EXIT_BANDWIDTH_FRACTION
  935. measured_bandwidth = self._data['consensus_weight'] * cw_to_bw
  936. if self._data['advertised_bandwidth'] != 0:
  937. # limit advertised bandwidth (if available) to measured bandwidth
  938. return min(measured_bandwidth, self._data['advertised_bandwidth'])
  939. else:
  940. return measured_bandwidth
  941. def set_measured_bandwidth(self, median_cw_to_bw_factor):
  942. self._data['measured_bandwidth'] = self.measured_bandwidth(
  943. median_cw_to_bw_factor)
  944. def is_exit(self):
  945. return 'Exit' in self._data['flags']
  946. def is_guard(self):
  947. return 'Guard' in self._data['flags']
  948. def is_running(self):
  949. return 'Running' in self._data['flags']
  950. # does this fallback have an IPv6 address and orport?
  951. def has_ipv6(self):
  952. return self.ipv6addr is not None and self.ipv6orport is not None
  953. # strip leading and trailing brackets from an IPv6 address
  954. # safe to use on non-bracketed IPv6 and on IPv4 addresses
  955. # also convert to unicode, and make None appear as ''
  956. @staticmethod
  957. def strip_ipv6_brackets(ip):
  958. if ip is None:
  959. return unicode('')
  960. if len(ip) < 2:
  961. return unicode(ip)
  962. if ip[0] == '[' and ip[-1] == ']':
  963. return unicode(ip[1:-1])
  964. return unicode(ip)
  965. # are ip_a and ip_b in the same netblock?
  966. # mask_bits is the size of the netblock
  967. # takes both IPv4 and IPv6 addresses
  968. # the versions of ip_a and ip_b must be the same
  969. # the mask must be valid for the IP version
  970. @staticmethod
  971. def netblocks_equal(ip_a, ip_b, mask_bits):
  972. if ip_a is None or ip_b is None:
  973. return False
  974. ip_a = Candidate.strip_ipv6_brackets(ip_a)
  975. ip_b = Candidate.strip_ipv6_brackets(ip_b)
  976. a = ipaddress.ip_address(ip_a)
  977. b = ipaddress.ip_address(ip_b)
  978. if a.version != b.version:
  979. raise Exception('Mismatching IP versions in %s and %s'%(ip_a, ip_b))
  980. if mask_bits > a.max_prefixlen:
  981. logging.error('Bad IP mask %d for %s and %s'%(mask_bits, ip_a, ip_b))
  982. mask_bits = a.max_prefixlen
  983. if mask_bits < 0:
  984. logging.error('Bad IP mask %d for %s and %s'%(mask_bits, ip_a, ip_b))
  985. mask_bits = 0
  986. a_net = ipaddress.ip_network('%s/%d'%(ip_a, mask_bits), strict=False)
  987. return b in a_net
  988. # is this fallback's IPv4 address (dirip) in the same netblock as other's
  989. # IPv4 address?
  990. # mask_bits is the size of the netblock
  991. def ipv4_netblocks_equal(self, other, mask_bits):
  992. return Candidate.netblocks_equal(self.dirip, other.dirip, mask_bits)
  993. # is this fallback's IPv6 address (ipv6addr) in the same netblock as
  994. # other's IPv6 address?
  995. # Returns False if either fallback has no IPv6 address
  996. # mask_bits is the size of the netblock
  997. def ipv6_netblocks_equal(self, other, mask_bits):
  998. if not self.has_ipv6() or not other.has_ipv6():
  999. return False
  1000. return Candidate.netblocks_equal(self.ipv6addr, other.ipv6addr, mask_bits)
  1001. # is this fallback's IPv4 DirPort the same as other's IPv4 DirPort?
  1002. def dirport_equal(self, other):
  1003. return self.dirport == other.dirport
  1004. # is this fallback's IPv4 ORPort the same as other's IPv4 ORPort?
  1005. def ipv4_orport_equal(self, other):
  1006. return self.orport == other.orport
  1007. # is this fallback's IPv6 ORPort the same as other's IPv6 ORPort?
  1008. # Returns False if either fallback has no IPv6 address
  1009. def ipv6_orport_equal(self, other):
  1010. if not self.has_ipv6() or not other.has_ipv6():
  1011. return False
  1012. return self.ipv6orport == other.ipv6orport
  1013. # does this fallback have the same DirPort, IPv4 ORPort, or
  1014. # IPv6 ORPort as other?
  1015. # Ignores IPv6 ORPort if either fallback has no IPv6 address
  1016. def port_equal(self, other):
  1017. return (self.dirport_equal(other) or self.ipv4_orport_equal(other)
  1018. or self.ipv6_orport_equal(other))
  1019. # return a list containing IPv4 ORPort, DirPort, and IPv6 ORPort (if present)
  1020. def port_list(self):
  1021. ports = [self.dirport, self.orport]
  1022. if self.has_ipv6() and not self.ipv6orport in ports:
  1023. ports.append(self.ipv6orport)
  1024. return ports
  1025. # does this fallback share a port with other, regardless of whether the
  1026. # port types match?
  1027. # For example, if self's IPv4 ORPort is 80 and other's DirPort is 80,
  1028. # return True
  1029. def port_shared(self, other):
  1030. for p in self.port_list():
  1031. if p in other.port_list():
  1032. return True
  1033. return False
  1034. # log how long it takes to download a consensus from dirip:dirport
  1035. # returns True if the download failed, False if it succeeded within max_time
  1036. @staticmethod
  1037. def fallback_consensus_download_speed(dirip, dirport, nickname, fingerprint,
  1038. max_time):
  1039. download_failed = False
  1040. # some directory mirrors respond to requests in ways that hang python
  1041. # sockets, which is why we log this line here
  1042. logging.info('Initiating %sconsensus download from %s (%s:%d) %s.',
  1043. 'microdesc ' if DOWNLOAD_MICRODESC_CONSENSUS else '',
  1044. nickname, dirip, dirport, fingerprint)
  1045. # there appears to be about 1 second of overhead when comparing stem's
  1046. # internal trace time and the elapsed time calculated here
  1047. TIMEOUT_SLOP = 1.0
  1048. start = datetime.datetime.utcnow()
  1049. try:
  1050. consensus = get_consensus(
  1051. endpoints = [(dirip, dirport)],
  1052. timeout = (max_time + TIMEOUT_SLOP),
  1053. validate = True,
  1054. retries = 0,
  1055. fall_back_to_authority = False,
  1056. document_handler = DocumentHandler.BARE_DOCUMENT,
  1057. microdescriptor = DOWNLOAD_MICRODESC_CONSENSUS
  1058. ).run()[0]
  1059. end = datetime.datetime.utcnow()
  1060. time_since_expiry = (end - consensus.valid_until).total_seconds()
  1061. except Exception, stem_error:
  1062. end = datetime.datetime.utcnow()
  1063. log_excluded('Unable to retrieve a consensus from %s: %s', nickname,
  1064. stem_error)
  1065. status = 'error: "%s"' % (stem_error)
  1066. level = logging.WARNING
  1067. download_failed = True
  1068. elapsed = (end - start).total_seconds()
  1069. if download_failed:
  1070. # keep the error failure status, and avoid using the variables
  1071. pass
  1072. elif elapsed > max_time:
  1073. status = 'too slow'
  1074. level = logging.WARNING
  1075. download_failed = True
  1076. elif (time_since_expiry > 0):
  1077. status = 'outdated consensus, expired %ds ago'%(int(time_since_expiry))
  1078. if time_since_expiry <= CONSENSUS_EXPIRY_TOLERANCE:
  1079. status += ', tolerating up to %ds'%(CONSENSUS_EXPIRY_TOLERANCE)
  1080. level = logging.INFO
  1081. else:
  1082. status += ', invalid'
  1083. level = logging.WARNING
  1084. download_failed = True
  1085. else:
  1086. status = 'ok'
  1087. level = logging.DEBUG
  1088. logging.log(level, 'Consensus download: %0.1fs %s from %s (%s:%d) %s, ' +
  1089. 'max download time %0.1fs.', elapsed, status, nickname,
  1090. dirip, dirport, fingerprint, max_time)
  1091. return download_failed
  1092. # does this fallback download the consensus fast enough?
  1093. def check_fallback_download_consensus(self):
  1094. # include the relay if we're not doing a check, or we can't check (IPv6)
  1095. ipv4_failed = False
  1096. ipv6_failed = False
  1097. if PERFORM_IPV4_DIRPORT_CHECKS:
  1098. ipv4_failed = Candidate.fallback_consensus_download_speed(self.dirip,
  1099. self.dirport,
  1100. self._data['nickname'],
  1101. self._fpr,
  1102. CONSENSUS_DOWNLOAD_SPEED_MAX)
  1103. if self.has_ipv6() and PERFORM_IPV6_DIRPORT_CHECKS:
  1104. # Clients assume the IPv6 DirPort is the same as the IPv4 DirPort
  1105. ipv6_failed = Candidate.fallback_consensus_download_speed(self.ipv6addr,
  1106. self.dirport,
  1107. self._data['nickname'],
  1108. self._fpr,
  1109. CONSENSUS_DOWNLOAD_SPEED_MAX)
  1110. return ((not ipv4_failed) and (not ipv6_failed))
  1111. # if this fallback has not passed a download check, try it again,
  1112. # and record the result, available in get_fallback_download_consensus
  1113. def try_fallback_download_consensus(self):
  1114. if not self.get_fallback_download_consensus():
  1115. self._data['download_check'] = self.check_fallback_download_consensus()
  1116. # did this fallback pass the download check?
  1117. def get_fallback_download_consensus(self):
  1118. # if we're not performing checks, return True
  1119. if not PERFORM_IPV4_DIRPORT_CHECKS and not PERFORM_IPV6_DIRPORT_CHECKS:
  1120. return True
  1121. # if we are performing checks, but haven't done one, return False
  1122. if not self._data.has_key('download_check'):
  1123. return False
  1124. return self._data['download_check']
  1125. # output an optional header comment and info for this fallback
  1126. # try_fallback_download_consensus before calling this
  1127. def fallbackdir_line(self, fallbacks, prefilter_fallbacks):
  1128. s = ''
  1129. if OUTPUT_COMMENTS:
  1130. s += self.fallbackdir_comment(fallbacks, prefilter_fallbacks)
  1131. # if the download speed is ok, output a C string
  1132. # if it's not, but we OUTPUT_COMMENTS, output a commented-out C string
  1133. if self.get_fallback_download_consensus() or OUTPUT_COMMENTS:
  1134. s += self.fallbackdir_info(self.get_fallback_download_consensus())
  1135. return s
  1136. # output a header comment for this fallback
  1137. def fallbackdir_comment(self, fallbacks, prefilter_fallbacks):
  1138. # /*
  1139. # nickname
  1140. # flags
  1141. # adjusted bandwidth, consensus weight
  1142. # [contact]
  1143. # [identical contact counts]
  1144. # */
  1145. # Multiline C comment
  1146. s = '/*'
  1147. s += '\n'
  1148. s += cleanse_c_multiline_comment(self._data['nickname'])
  1149. s += '\n'
  1150. s += 'Flags: '
  1151. s += cleanse_c_multiline_comment(' '.join(sorted(self._data['flags'])))
  1152. s += '\n'
  1153. # this is an adjusted bandwidth, see calculate_measured_bandwidth()
  1154. bandwidth = self._data['measured_bandwidth']
  1155. weight = self._data['consensus_weight']
  1156. s += 'Bandwidth: %.1f MByte/s, Consensus Weight: %d'%(
  1157. bandwidth/(1024.0*1024.0),
  1158. weight)
  1159. s += '\n'
  1160. if self._data['contact'] is not None:
  1161. s += cleanse_c_multiline_comment(self._data['contact'])
  1162. if CONTACT_COUNT or CONTACT_BLACKLIST_COUNT:
  1163. fallback_count = len([f for f in fallbacks
  1164. if f._data['contact'] == self._data['contact']])
  1165. if fallback_count > 1:
  1166. s += '\n'
  1167. s += '%d identical contacts listed' % (fallback_count)
  1168. if CONTACT_BLACKLIST_COUNT:
  1169. prefilter_count = len([f for f in prefilter_fallbacks
  1170. if f._data['contact'] == self._data['contact']])
  1171. filter_count = prefilter_count - fallback_count
  1172. if filter_count > 0:
  1173. if fallback_count > 1:
  1174. s += ' '
  1175. else:
  1176. s += '\n'
  1177. s += '%d blacklisted' % (filter_count)
  1178. s += '\n'
  1179. s += '*/'
  1180. s += '\n'
  1181. return s
  1182. # output the fallback info C string for this fallback
  1183. # this is the text that would go after FallbackDir in a torrc
  1184. # if this relay failed the download test and we OUTPUT_COMMENTS,
  1185. # comment-out the returned string
  1186. def fallbackdir_info(self, dl_speed_ok):
  1187. # "address:dirport orport=port id=fingerprint"
  1188. # "[ipv6=addr:orport]"
  1189. # "weight=FALLBACK_OUTPUT_WEIGHT",
  1190. #
  1191. # Do we want a C string, or a commented-out string?
  1192. c_string = dl_speed_ok
  1193. comment_string = not dl_speed_ok and OUTPUT_COMMENTS
  1194. # If we don't want either kind of string, bail
  1195. if not c_string and not comment_string:
  1196. return ''
  1197. s = ''
  1198. # Comment out the fallback directory entry if it's too slow
  1199. # See the debug output for which address and port is failing
  1200. if comment_string:
  1201. s += '/* Consensus download failed or was too slow:\n'
  1202. # Multi-Line C string with trailing comma (part of a string list)
  1203. # This makes it easier to diff the file, and remove IPv6 lines using grep
  1204. # Integers don't need escaping
  1205. s += '"%s orport=%d id=%s"'%(
  1206. cleanse_c_string(self._data['dir_address']),
  1207. self.orport,
  1208. cleanse_c_string(self._fpr))
  1209. s += '\n'
  1210. if self.has_ipv6():
  1211. s += '" ipv6=%s:%d"'%(cleanse_c_string(self.ipv6addr), self.ipv6orport)
  1212. s += '\n'
  1213. s += '" weight=%d",'%(FALLBACK_OUTPUT_WEIGHT)
  1214. if comment_string:
  1215. s += '\n'
  1216. s += '*/'
  1217. return s
  1218. ## Fallback Candidate List Class
  1219. class CandidateList(dict):
  1220. def __init__(self):
  1221. pass
  1222. def _add_relay(self, details):
  1223. if not 'dir_address' in details: return
  1224. c = Candidate(details)
  1225. self[ c.get_fingerprint() ] = c
  1226. def _add_uptime(self, uptime):
  1227. try:
  1228. fpr = uptime['fingerprint']
  1229. except KeyError:
  1230. raise Exception("Document has no fingerprint field.")
  1231. try:
  1232. c = self[fpr]
  1233. except KeyError:
  1234. logging.debug('Got unknown relay %s in uptime document.'%(fpr,))
  1235. return
  1236. c.add_uptime(uptime)
  1237. def _add_details(self):
  1238. logging.debug('Loading details document.')
  1239. d = fetch('details',
  1240. fields=('fingerprint,nickname,contact,last_changed_address_or_port,' +
  1241. 'consensus_weight,advertised_bandwidth,or_addresses,' +
  1242. 'dir_address,recommended_version,flags,effective_family,' +
  1243. 'platform'))
  1244. logging.debug('Loading details document done.')
  1245. if not 'relays' in d: raise Exception("No relays found in document.")
  1246. for r in d['relays']: self._add_relay(r)
  1247. def _add_uptimes(self):
  1248. logging.debug('Loading uptime document.')
  1249. d = fetch('uptime')
  1250. logging.debug('Loading uptime document done.')
  1251. if not 'relays' in d: raise Exception("No relays found in document.")
  1252. for r in d['relays']: self._add_uptime(r)
  1253. def add_relays(self):
  1254. self._add_details()
  1255. self._add_uptimes()
  1256. def count_guards(self):
  1257. guard_count = 0
  1258. for fpr in self.keys():
  1259. if self[fpr].is_guard():
  1260. guard_count += 1
  1261. return guard_count
  1262. # Find fallbacks that fit the uptime, stability, and flags criteria,
  1263. # and make an array of them in self.fallbacks
  1264. def compute_fallbacks(self):
  1265. self.fallbacks = map(lambda x: self[x],
  1266. filter(lambda x: self[x].is_candidate(),
  1267. self.keys()))
  1268. # sort fallbacks by their consensus weight to advertised bandwidth factor,
  1269. # lowest to highest
  1270. # used to find the median cw_to_bw_factor()
  1271. def sort_fallbacks_by_cw_to_bw_factor(self):
  1272. self.fallbacks.sort(key=lambda f: f.cw_to_bw_factor())
  1273. # sort fallbacks by their measured bandwidth, highest to lowest
  1274. # calculate_measured_bandwidth before calling this
  1275. # this is useful for reviewing candidates in priority order
  1276. def sort_fallbacks_by_measured_bandwidth(self):
  1277. self.fallbacks.sort(key=lambda f: f._data['measured_bandwidth'],
  1278. reverse=True)
  1279. # sort fallbacks by the data field data_field, lowest to highest
  1280. def sort_fallbacks_by(self, data_field):
  1281. self.fallbacks.sort(key=lambda f: f._data[data_field])
  1282. @staticmethod
  1283. def load_relaylist(file_obj):
  1284. """ Read each line in the file, and parse it like a FallbackDir line:
  1285. an IPv4 address and optional port:
  1286. <IPv4 address>:<port>
  1287. which are parsed into dictionary entries:
  1288. ipv4=<IPv4 address>
  1289. dirport=<port>
  1290. followed by a series of key=value entries:
  1291. orport=<port>
  1292. id=<fingerprint>
  1293. ipv6=<IPv6 address>:<IPv6 orport>
  1294. each line's key/value pairs are placed in a dictonary,
  1295. (of string -> string key/value pairs),
  1296. and these dictionaries are placed in an array.
  1297. comments start with # and are ignored """
  1298. file_data = file_obj['data']
  1299. file_name = file_obj['name']
  1300. relaylist = []
  1301. if file_data is None:
  1302. return relaylist
  1303. for line in file_data.split('\n'):
  1304. relay_entry = {}
  1305. # ignore comments
  1306. line_comment_split = line.split('#')
  1307. line = line_comment_split[0]
  1308. # cleanup whitespace
  1309. line = cleanse_whitespace(line)
  1310. line = line.strip()
  1311. if len(line) == 0:
  1312. continue
  1313. for item in line.split(' '):
  1314. item = item.strip()
  1315. if len(item) == 0:
  1316. continue
  1317. key_value_split = item.split('=')
  1318. kvl = len(key_value_split)
  1319. if kvl < 1 or kvl > 2:
  1320. print '#error Bad %s item: %s, format is key=value.'%(
  1321. file_name, item)
  1322. if kvl == 1:
  1323. # assume that entries without a key are the ipv4 address,
  1324. # perhaps with a dirport
  1325. ipv4_maybe_dirport = key_value_split[0]
  1326. ipv4_maybe_dirport_split = ipv4_maybe_dirport.split(':')
  1327. dirl = len(ipv4_maybe_dirport_split)
  1328. if dirl < 1 or dirl > 2:
  1329. print '#error Bad %s IPv4 item: %s, format is ipv4:port.'%(
  1330. file_name, item)
  1331. if dirl >= 1:
  1332. relay_entry['ipv4'] = ipv4_maybe_dirport_split[0]
  1333. if dirl == 2:
  1334. relay_entry['dirport'] = ipv4_maybe_dirport_split[1]
  1335. elif kvl == 2:
  1336. relay_entry[key_value_split[0]] = key_value_split[1]
  1337. relaylist.append(relay_entry)
  1338. return relaylist
  1339. # apply the fallback whitelist and blacklist
  1340. def apply_filter_lists(self, whitelist_obj, blacklist_obj):
  1341. excluded_count = 0
  1342. logging.debug('Applying whitelist and blacklist.')
  1343. # parse the whitelist and blacklist
  1344. whitelist = self.load_relaylist(whitelist_obj)
  1345. blacklist = self.load_relaylist(blacklist_obj)
  1346. filtered_fallbacks = []
  1347. for f in self.fallbacks:
  1348. in_whitelist = f.is_in_whitelist(whitelist)
  1349. in_blacklist = f.is_in_blacklist(blacklist)
  1350. if in_whitelist and in_blacklist:
  1351. if BLACKLIST_EXCLUDES_WHITELIST_ENTRIES:
  1352. # exclude
  1353. excluded_count += 1
  1354. logging.warning('Excluding %s: in both blacklist and whitelist.',
  1355. f._fpr)
  1356. else:
  1357. # include
  1358. filtered_fallbacks.append(f)
  1359. elif in_whitelist:
  1360. # include
  1361. filtered_fallbacks.append(f)
  1362. elif in_blacklist:
  1363. # exclude
  1364. excluded_count += 1
  1365. log_excluded('Excluding %s: in blacklist.', f._fpr)
  1366. else:
  1367. if INCLUDE_UNLISTED_ENTRIES:
  1368. # include
  1369. filtered_fallbacks.append(f)
  1370. else:
  1371. # exclude
  1372. excluded_count += 1
  1373. log_excluded('Excluding %s: in neither blacklist nor whitelist.',
  1374. f._fpr)
  1375. self.fallbacks = filtered_fallbacks
  1376. return excluded_count
  1377. @staticmethod
  1378. def summarise_filters(initial_count, excluded_count):
  1379. return '/* Whitelist & blacklist excluded %d of %d candidates. */'%(
  1380. excluded_count, initial_count)
  1381. # calculate each fallback's measured bandwidth based on the median
  1382. # consensus weight to advertised bandwdith ratio
  1383. def calculate_measured_bandwidth(self):
  1384. self.sort_fallbacks_by_cw_to_bw_factor()
  1385. median_fallback = self.fallback_median(True)
  1386. if median_fallback is not None:
  1387. median_cw_to_bw_factor = median_fallback.cw_to_bw_factor()
  1388. else:
  1389. # this will never be used, because there are no fallbacks
  1390. median_cw_to_bw_factor = None
  1391. for f in self.fallbacks:
  1392. f.set_measured_bandwidth(median_cw_to_bw_factor)
  1393. # remove relays with low measured bandwidth from the fallback list
  1394. # calculate_measured_bandwidth for each relay before calling this
  1395. def remove_low_bandwidth_relays(self):
  1396. if MIN_BANDWIDTH is None:
  1397. return
  1398. above_min_bw_fallbacks = []
  1399. for f in self.fallbacks:
  1400. if f._data['measured_bandwidth'] >= MIN_BANDWIDTH:
  1401. above_min_bw_fallbacks.append(f)
  1402. else:
  1403. # the bandwidth we log here is limited by the relay's consensus weight
  1404. # as well as its adverttised bandwidth. See set_measured_bandwidth
  1405. # for details
  1406. log_excluded('%s not a candidate: bandwidth %.1fMByte/s too low, ' +
  1407. 'must be at least %.1fMByte/s', f._fpr,
  1408. f._data['measured_bandwidth']/(1024.0*1024.0),
  1409. MIN_BANDWIDTH/(1024.0*1024.0))
  1410. self.fallbacks = above_min_bw_fallbacks
  1411. # the minimum fallback in the list
  1412. # call one of the sort_fallbacks_* functions before calling this
  1413. def fallback_min(self):
  1414. if len(self.fallbacks) > 0:
  1415. return self.fallbacks[-1]
  1416. else:
  1417. return None
  1418. # the median fallback in the list
  1419. # call one of the sort_fallbacks_* functions before calling this
  1420. def fallback_median(self, require_advertised_bandwidth):
  1421. # use the low-median when there are an evan number of fallbacks,
  1422. # for consistency with the bandwidth authorities
  1423. if len(self.fallbacks) > 0:
  1424. median_position = (len(self.fallbacks) - 1) / 2
  1425. if not require_advertised_bandwidth:
  1426. return self.fallbacks[median_position]
  1427. # if we need advertised_bandwidth but this relay doesn't have it,
  1428. # move to a fallback with greater consensus weight until we find one
  1429. while not self.fallbacks[median_position]._data['advertised_bandwidth']:
  1430. median_position += 1
  1431. if median_position >= len(self.fallbacks):
  1432. return None
  1433. return self.fallbacks[median_position]
  1434. else:
  1435. return None
  1436. # the maximum fallback in the list
  1437. # call one of the sort_fallbacks_* functions before calling this
  1438. def fallback_max(self):
  1439. if len(self.fallbacks) > 0:
  1440. return self.fallbacks[0]
  1441. else:
  1442. return None
  1443. # return a new bag suitable for storing attributes
  1444. @staticmethod
  1445. def attribute_new():
  1446. return dict()
  1447. # get the count of attribute in attribute_bag
  1448. # if attribute is None or the empty string, return 0
  1449. @staticmethod
  1450. def attribute_count(attribute, attribute_bag):
  1451. if attribute is None or attribute == '':
  1452. return 0
  1453. if attribute not in attribute_bag:
  1454. return 0
  1455. return attribute_bag[attribute]
  1456. # does attribute_bag contain more than max_count instances of attribute?
  1457. # if so, return False
  1458. # if not, return True
  1459. # if attribute is None or the empty string, or max_count is invalid,
  1460. # always return True
  1461. @staticmethod
  1462. def attribute_allow(attribute, attribute_bag, max_count=1):
  1463. if attribute is None or attribute == '' or max_count <= 0:
  1464. return True
  1465. elif CandidateList.attribute_count(attribute, attribute_bag) >= max_count:
  1466. return False
  1467. else:
  1468. return True
  1469. # add attribute to attribute_bag, incrementing the count if it is already
  1470. # present
  1471. # if attribute is None or the empty string, or count is invalid,
  1472. # do nothing
  1473. @staticmethod
  1474. def attribute_add(attribute, attribute_bag, count=1):
  1475. if attribute is None or attribute == '' or count <= 0:
  1476. pass
  1477. attribute_bag.setdefault(attribute, 0)
  1478. attribute_bag[attribute] += count
  1479. # make sure there are only MAX_FALLBACKS_PER_IP fallbacks per IPv4 address,
  1480. # and per IPv6 address
  1481. # there is only one IPv4 address on each fallback: the IPv4 DirPort address
  1482. # (we choose the IPv4 ORPort which is on the same IPv4 as the DirPort)
  1483. # there is at most one IPv6 address on each fallback: the IPv6 ORPort address
  1484. # we try to match the IPv4 ORPort, but will use any IPv6 address if needed
  1485. # (clients only use the IPv6 ORPort)
  1486. # if there is no IPv6 address, only the IPv4 address is checked
  1487. # return the number of candidates we excluded
  1488. def limit_fallbacks_same_ip(self):
  1489. ip_limit_fallbacks = []
  1490. ip_list = CandidateList.attribute_new()
  1491. for f in self.fallbacks:
  1492. if (CandidateList.attribute_allow(f.dirip, ip_list,
  1493. MAX_FALLBACKS_PER_IPV4)
  1494. and CandidateList.attribute_allow(f.ipv6addr, ip_list,
  1495. MAX_FALLBACKS_PER_IPV6)):
  1496. ip_limit_fallbacks.append(f)
  1497. CandidateList.attribute_add(f.dirip, ip_list)
  1498. if f.has_ipv6():
  1499. CandidateList.attribute_add(f.ipv6addr, ip_list)
  1500. elif not CandidateList.attribute_allow(f.dirip, ip_list,
  1501. MAX_FALLBACKS_PER_IPV4):
  1502. log_excluded('Eliminated %s: already have %d fallback(s) on IPv4 %s'
  1503. %(f._fpr, CandidateList.attribute_count(f.dirip, ip_list),
  1504. f.dirip))
  1505. elif (f.has_ipv6() and
  1506. not CandidateList.attribute_allow(f.ipv6addr, ip_list,
  1507. MAX_FALLBACKS_PER_IPV6)):
  1508. log_excluded('Eliminated %s: already have %d fallback(s) on IPv6 %s'
  1509. %(f._fpr, CandidateList.attribute_count(f.ipv6addr,
  1510. ip_list),
  1511. f.ipv6addr))
  1512. original_count = len(self.fallbacks)
  1513. self.fallbacks = ip_limit_fallbacks
  1514. return original_count - len(self.fallbacks)
  1515. # make sure there are only MAX_FALLBACKS_PER_CONTACT fallbacks for each
  1516. # ContactInfo
  1517. # if there is no ContactInfo, allow the fallback
  1518. # this check can be gamed by providing no ContactInfo, or by setting the
  1519. # ContactInfo to match another fallback
  1520. # However, given the likelihood that relays with the same ContactInfo will
  1521. # go down at similar times, its usefulness outweighs the risk
  1522. def limit_fallbacks_same_contact(self):
  1523. contact_limit_fallbacks = []
  1524. contact_list = CandidateList.attribute_new()
  1525. for f in self.fallbacks:
  1526. if CandidateList.attribute_allow(f._data['contact'], contact_list,
  1527. MAX_FALLBACKS_PER_CONTACT):
  1528. contact_limit_fallbacks.append(f)
  1529. CandidateList.attribute_add(f._data['contact'], contact_list)
  1530. else:
  1531. log_excluded(
  1532. 'Eliminated %s: already have %d fallback(s) on ContactInfo %s'
  1533. %(f._fpr, CandidateList.attribute_count(f._data['contact'],
  1534. contact_list),
  1535. f._data['contact']))
  1536. original_count = len(self.fallbacks)
  1537. self.fallbacks = contact_limit_fallbacks
  1538. return original_count - len(self.fallbacks)
  1539. # make sure there are only MAX_FALLBACKS_PER_FAMILY fallbacks per effective
  1540. # family
  1541. # if there is no family, allow the fallback
  1542. # we use effective family, which ensures mutual family declarations
  1543. # but the check can be gamed by not declaring a family at all
  1544. # if any indirect families exist, the result depends on the order in which
  1545. # fallbacks are sorted in the list
  1546. def limit_fallbacks_same_family(self):
  1547. family_limit_fallbacks = []
  1548. fingerprint_list = CandidateList.attribute_new()
  1549. for f in self.fallbacks:
  1550. if CandidateList.attribute_allow(f._fpr, fingerprint_list,
  1551. MAX_FALLBACKS_PER_FAMILY):
  1552. family_limit_fallbacks.append(f)
  1553. CandidateList.attribute_add(f._fpr, fingerprint_list)
  1554. for family_fingerprint in f._data['effective_family']:
  1555. CandidateList.attribute_add(family_fingerprint, fingerprint_list)
  1556. else:
  1557. # we already have a fallback with this fallback in its effective
  1558. # family
  1559. log_excluded(
  1560. 'Eliminated %s: already have %d fallback(s) in effective family'
  1561. %(f._fpr, CandidateList.attribute_count(f._fpr, fingerprint_list)))
  1562. original_count = len(self.fallbacks)
  1563. self.fallbacks = family_limit_fallbacks
  1564. return original_count - len(self.fallbacks)
  1565. # try a download check on each fallback candidate in order
  1566. # stop after max_count successful downloads
  1567. # but don't remove any candidates from the array
  1568. def try_download_consensus_checks(self, max_count):
  1569. dl_ok_count = 0
  1570. for f in self.fallbacks:
  1571. f.try_fallback_download_consensus()
  1572. if f.get_fallback_download_consensus():
  1573. # this fallback downloaded a consensus ok
  1574. dl_ok_count += 1
  1575. if dl_ok_count >= max_count:
  1576. # we have enough fallbacks
  1577. return
  1578. # put max_count successful candidates in the fallbacks array:
  1579. # - perform download checks on each fallback candidate
  1580. # - retry failed candidates if CONSENSUS_DOWNLOAD_RETRY is set
  1581. # - eliminate failed candidates
  1582. # - if there are more than max_count candidates, eliminate lowest bandwidth
  1583. # - if there are fewer than max_count candidates, leave only successful
  1584. # Return the number of fallbacks that failed the consensus check
  1585. def perform_download_consensus_checks(self, max_count):
  1586. self.sort_fallbacks_by_measured_bandwidth()
  1587. self.try_download_consensus_checks(max_count)
  1588. if CONSENSUS_DOWNLOAD_RETRY:
  1589. # try unsuccessful candidates again
  1590. # we could end up with more than max_count successful candidates here
  1591. self.try_download_consensus_checks(max_count)
  1592. # now we have at least max_count successful candidates,
  1593. # or we've tried them all
  1594. original_count = len(self.fallbacks)
  1595. self.fallbacks = filter(lambda x: x.get_fallback_download_consensus(),
  1596. self.fallbacks)
  1597. # some of these failed the check, others skipped the check,
  1598. # if we already had enough successful downloads
  1599. failed_count = original_count - len(self.fallbacks)
  1600. self.fallbacks = self.fallbacks[:max_count]
  1601. return failed_count
  1602. # return a string that describes a/b as a percentage
  1603. @staticmethod
  1604. def describe_percentage(a, b):
  1605. if b != 0:
  1606. return '%d/%d = %.0f%%'%(a, b, (a*100.0)/b)
  1607. else:
  1608. # technically, 0/0 is undefined, but 0.0% is a sensible result
  1609. return '%d/%d = %.0f%%'%(a, b, 0.0)
  1610. # return a dictionary of lists of fallbacks by IPv4 netblock
  1611. # the dictionary is keyed by the fingerprint of an arbitrary fallback
  1612. # in each netblock
  1613. # mask_bits is the size of the netblock
  1614. def fallbacks_by_ipv4_netblock(self, mask_bits):
  1615. netblocks = {}
  1616. for f in self.fallbacks:
  1617. found_netblock = False
  1618. for b in netblocks.keys():
  1619. # we found an existing netblock containing this fallback
  1620. if f.ipv4_netblocks_equal(self[b], mask_bits):
  1621. # add it to the list
  1622. netblocks[b].append(f)
  1623. found_netblock = True
  1624. break
  1625. # make a new netblock based on this fallback's fingerprint
  1626. if not found_netblock:
  1627. netblocks[f._fpr] = [f]
  1628. return netblocks
  1629. # return a dictionary of lists of fallbacks by IPv6 netblock
  1630. # where mask_bits is the size of the netblock
  1631. def fallbacks_by_ipv6_netblock(self, mask_bits):
  1632. netblocks = {}
  1633. for f in self.fallbacks:
  1634. # skip fallbacks without IPv6 addresses
  1635. if not f.has_ipv6():
  1636. continue
  1637. found_netblock = False
  1638. for b in netblocks.keys():
  1639. # we found an existing netblock containing this fallback
  1640. if f.ipv6_netblocks_equal(self[b], mask_bits):
  1641. # add it to the list
  1642. netblocks[b].append(f)
  1643. found_netblock = True
  1644. break
  1645. # make a new netblock based on this fallback's fingerprint
  1646. if not found_netblock:
  1647. netblocks[f._fpr] = [f]
  1648. return netblocks
  1649. # log a message about the proportion of fallbacks in each IPv4 netblock,
  1650. # where mask_bits is the size of the netblock
  1651. def describe_fallback_ipv4_netblock_mask(self, mask_bits):
  1652. fallback_count = len(self.fallbacks)
  1653. shared_netblock_fallback_count = 0
  1654. most_frequent_netblock = None
  1655. netblocks = self.fallbacks_by_ipv4_netblock(mask_bits)
  1656. for b in netblocks.keys():
  1657. if len(netblocks[b]) > 1:
  1658. # how many fallbacks are in a netblock with other fallbacks?
  1659. shared_netblock_fallback_count += len(netblocks[b])
  1660. # what's the netblock with the most fallbacks?
  1661. if (most_frequent_netblock is None
  1662. or len(netblocks[b]) > len(netblocks[most_frequent_netblock])):
  1663. most_frequent_netblock = b
  1664. logging.debug('Fallback IPv4 addresses in the same /%d:'%(mask_bits))
  1665. for f in netblocks[b]:
  1666. logging.debug('%s - %s', f.dirip, f._fpr)
  1667. if most_frequent_netblock is not None:
  1668. logging.warning('There are %s fallbacks in the IPv4 /%d containing %s'%(
  1669. CandidateList.describe_percentage(
  1670. len(netblocks[most_frequent_netblock]),
  1671. fallback_count),
  1672. mask_bits,
  1673. self[most_frequent_netblock].dirip))
  1674. if shared_netblock_fallback_count > 0:
  1675. logging.warning(('%s of fallbacks are in an IPv4 /%d with other ' +
  1676. 'fallbacks')%(CandidateList.describe_percentage(
  1677. shared_netblock_fallback_count,
  1678. fallback_count),
  1679. mask_bits))
  1680. # log a message about the proportion of fallbacks in each IPv6 netblock,
  1681. # where mask_bits is the size of the netblock
  1682. def describe_fallback_ipv6_netblock_mask(self, mask_bits):
  1683. fallback_count = len(self.fallbacks_with_ipv6())
  1684. shared_netblock_fallback_count = 0
  1685. most_frequent_netblock = None
  1686. netblocks = self.fallbacks_by_ipv6_netblock(mask_bits)
  1687. for b in netblocks.keys():
  1688. if len(netblocks[b]) > 1:
  1689. # how many fallbacks are in a netblock with other fallbacks?
  1690. shared_netblock_fallback_count += len(netblocks[b])
  1691. # what's the netblock with the most fallbacks?
  1692. if (most_frequent_netblock is None
  1693. or len(netblocks[b]) > len(netblocks[most_frequent_netblock])):
  1694. most_frequent_netblock = b
  1695. logging.debug('Fallback IPv6 addresses in the same /%d:'%(mask_bits))
  1696. for f in netblocks[b]:
  1697. logging.debug('%s - %s', f.ipv6addr, f._fpr)
  1698. if most_frequent_netblock is not None:
  1699. logging.warning('There are %s fallbacks in the IPv6 /%d containing %s'%(
  1700. CandidateList.describe_percentage(
  1701. len(netblocks[most_frequent_netblock]),
  1702. fallback_count),
  1703. mask_bits,
  1704. self[most_frequent_netblock].ipv6addr))
  1705. if shared_netblock_fallback_count > 0:
  1706. logging.warning(('%s of fallbacks are in an IPv6 /%d with other ' +
  1707. 'fallbacks')%(CandidateList.describe_percentage(
  1708. shared_netblock_fallback_count,
  1709. fallback_count),
  1710. mask_bits))
  1711. # log a message about the proportion of fallbacks in each IPv4 /8, /16,
  1712. # and /24
  1713. def describe_fallback_ipv4_netblocks(self):
  1714. # this doesn't actually tell us anything useful
  1715. #self.describe_fallback_ipv4_netblock_mask(8)
  1716. self.describe_fallback_ipv4_netblock_mask(16)
  1717. self.describe_fallback_ipv4_netblock_mask(24)
  1718. # log a message about the proportion of fallbacks in each IPv6 /12 (RIR),
  1719. # /23 (smaller RIR blocks), /32 (LIR), /48 (Customer), and /64 (Host)
  1720. # https://www.iana.org/assignments/ipv6-unicast-address-assignments/
  1721. def describe_fallback_ipv6_netblocks(self):
  1722. # these don't actually tell us anything useful
  1723. #self.describe_fallback_ipv6_netblock_mask(12)
  1724. #self.describe_fallback_ipv6_netblock_mask(23)
  1725. self.describe_fallback_ipv6_netblock_mask(32)
  1726. self.describe_fallback_ipv6_netblock_mask(48)
  1727. self.describe_fallback_ipv6_netblock_mask(64)
  1728. # log a message about the proportion of fallbacks in each IPv4 and IPv6
  1729. # netblock
  1730. def describe_fallback_netblocks(self):
  1731. self.describe_fallback_ipv4_netblocks()
  1732. self.describe_fallback_ipv6_netblocks()
  1733. # return a list of fallbacks which are on the IPv4 ORPort port
  1734. def fallbacks_on_ipv4_orport(self, port):
  1735. return filter(lambda x: x.orport == port, self.fallbacks)
  1736. # return a list of fallbacks which are on the IPv6 ORPort port
  1737. def fallbacks_on_ipv6_orport(self, port):
  1738. return filter(lambda x: x.ipv6orport == port, self.fallbacks_with_ipv6())
  1739. # return a list of fallbacks which are on the DirPort port
  1740. def fallbacks_on_dirport(self, port):
  1741. return filter(lambda x: x.dirport == port, self.fallbacks)
  1742. # log a message about the proportion of fallbacks on IPv4 ORPort port
  1743. # and return that count
  1744. def describe_fallback_ipv4_orport(self, port):
  1745. port_count = len(self.fallbacks_on_ipv4_orport(port))
  1746. fallback_count = len(self.fallbacks)
  1747. logging.warning('%s of fallbacks are on IPv4 ORPort %d'%(
  1748. CandidateList.describe_percentage(port_count,
  1749. fallback_count),
  1750. port))
  1751. return port_count
  1752. # log a message about the proportion of IPv6 fallbacks on IPv6 ORPort port
  1753. # and return that count
  1754. def describe_fallback_ipv6_orport(self, port):
  1755. port_count = len(self.fallbacks_on_ipv6_orport(port))
  1756. fallback_count = len(self.fallbacks_with_ipv6())
  1757. logging.warning('%s of IPv6 fallbacks are on IPv6 ORPort %d'%(
  1758. CandidateList.describe_percentage(port_count,
  1759. fallback_count),
  1760. port))
  1761. return port_count
  1762. # log a message about the proportion of fallbacks on DirPort port
  1763. # and return that count
  1764. def describe_fallback_dirport(self, port):
  1765. port_count = len(self.fallbacks_on_dirport(port))
  1766. fallback_count = len(self.fallbacks)
  1767. logging.warning('%s of fallbacks are on DirPort %d'%(
  1768. CandidateList.describe_percentage(port_count,
  1769. fallback_count),
  1770. port))
  1771. return port_count
  1772. # log a message about the proportion of fallbacks on each dirport,
  1773. # each IPv4 orport, and each IPv6 orport
  1774. def describe_fallback_ports(self):
  1775. fallback_count = len(self.fallbacks)
  1776. ipv4_or_count = fallback_count
  1777. ipv4_or_count -= self.describe_fallback_ipv4_orport(443)
  1778. ipv4_or_count -= self.describe_fallback_ipv4_orport(9001)
  1779. logging.warning('%s of fallbacks are on other IPv4 ORPorts'%(
  1780. CandidateList.describe_percentage(ipv4_or_count,
  1781. fallback_count)))
  1782. ipv6_fallback_count = len(self.fallbacks_with_ipv6())
  1783. ipv6_or_count = ipv6_fallback_count
  1784. ipv6_or_count -= self.describe_fallback_ipv6_orport(443)
  1785. ipv6_or_count -= self.describe_fallback_ipv6_orport(9001)
  1786. logging.warning('%s of IPv6 fallbacks are on other IPv6 ORPorts'%(
  1787. CandidateList.describe_percentage(ipv6_or_count,
  1788. ipv6_fallback_count)))
  1789. dir_count = fallback_count
  1790. dir_count -= self.describe_fallback_dirport(80)
  1791. dir_count -= self.describe_fallback_dirport(9030)
  1792. logging.warning('%s of fallbacks are on other DirPorts'%(
  1793. CandidateList.describe_percentage(dir_count,
  1794. fallback_count)))
  1795. # return a list of fallbacks which have the Exit flag
  1796. def fallbacks_with_exit(self):
  1797. return filter(lambda x: x.is_exit(), self.fallbacks)
  1798. # log a message about the proportion of fallbacks with an Exit flag
  1799. def describe_fallback_exit_flag(self):
  1800. exit_falback_count = len(self.fallbacks_with_exit())
  1801. fallback_count = len(self.fallbacks)
  1802. logging.warning('%s of fallbacks have the Exit flag'%(
  1803. CandidateList.describe_percentage(exit_falback_count,
  1804. fallback_count)))
  1805. # return a list of fallbacks which have an IPv6 address
  1806. def fallbacks_with_ipv6(self):
  1807. return filter(lambda x: x.has_ipv6(), self.fallbacks)
  1808. # log a message about the proportion of fallbacks on IPv6
  1809. def describe_fallback_ip_family(self):
  1810. ipv6_falback_count = len(self.fallbacks_with_ipv6())
  1811. fallback_count = len(self.fallbacks)
  1812. logging.warning('%s of fallbacks are on IPv6'%(
  1813. CandidateList.describe_percentage(ipv6_falback_count,
  1814. fallback_count)))
  1815. def summarise_fallbacks(self, eligible_count, operator_count, failed_count,
  1816. guard_count, target_count):
  1817. s = ''
  1818. s += '/* To comment-out entries in this file, use C comments, and add *'
  1819. s += ' to the start of each line. (stem finds fallback entries using "'
  1820. s += ' at the start of a line.) */'
  1821. s += '\n'
  1822. # Report:
  1823. # whether we checked consensus download times
  1824. # the number of fallback directories (and limits/exclusions, if relevant)
  1825. # min & max fallback bandwidths
  1826. # #error if below minimum count
  1827. if PERFORM_IPV4_DIRPORT_CHECKS or PERFORM_IPV6_DIRPORT_CHECKS:
  1828. s += '/* Checked %s%s%s DirPorts served a consensus within %.1fs. */'%(
  1829. 'IPv4' if PERFORM_IPV4_DIRPORT_CHECKS else '',
  1830. ' and ' if (PERFORM_IPV4_DIRPORT_CHECKS
  1831. and PERFORM_IPV6_DIRPORT_CHECKS) else '',
  1832. 'IPv6' if PERFORM_IPV6_DIRPORT_CHECKS else '',
  1833. CONSENSUS_DOWNLOAD_SPEED_MAX)
  1834. else:
  1835. s += '/* Did not check IPv4 or IPv6 DirPort consensus downloads. */'
  1836. s += '\n'
  1837. # Multiline C comment with #error if things go bad
  1838. s += '/*'
  1839. s += '\n'
  1840. # Integers don't need escaping in C comments
  1841. fallback_count = len(self.fallbacks)
  1842. if FALLBACK_PROPORTION_OF_GUARDS is None:
  1843. fallback_proportion = ''
  1844. else:
  1845. fallback_proportion = ', Target %d (%d * %.2f)'%(target_count,
  1846. guard_count,
  1847. FALLBACK_PROPORTION_OF_GUARDS)
  1848. s += 'Final Count: %d (Eligible %d%s'%(fallback_count, eligible_count,
  1849. fallback_proportion)
  1850. if MAX_FALLBACK_COUNT is not None:
  1851. s += ', Max %d'%(MAX_FALLBACK_COUNT)
  1852. s += ')\n'
  1853. if eligible_count != fallback_count:
  1854. removed_count = eligible_count - fallback_count
  1855. excess_to_target_or_max = (eligible_count - operator_count - failed_count
  1856. - fallback_count)
  1857. # some 'Failed' failed the check, others 'Skipped' the check,
  1858. # if we already had enough successful downloads
  1859. s += ('Excluded: %d (Same Operator %d, Failed/Skipped Download %d, ' +
  1860. 'Excess %d)')%(removed_count, operator_count, failed_count,
  1861. excess_to_target_or_max)
  1862. s += '\n'
  1863. min_fb = self.fallback_min()
  1864. min_bw = min_fb._data['measured_bandwidth']
  1865. max_fb = self.fallback_max()
  1866. max_bw = max_fb._data['measured_bandwidth']
  1867. s += 'Bandwidth Range: %.1f - %.1f MByte/s'%(min_bw/(1024.0*1024.0),
  1868. max_bw/(1024.0*1024.0))
  1869. s += '\n'
  1870. s += '*/'
  1871. if fallback_count < MIN_FALLBACK_COUNT:
  1872. # We must have a minimum number of fallbacks so they are always
  1873. # reachable, and are in diverse locations
  1874. s += '\n'
  1875. s += '#error Fallback Count %d is too low. '%(fallback_count)
  1876. s += 'Must be at least %d for diversity. '%(MIN_FALLBACK_COUNT)
  1877. s += 'Try adding entries to the whitelist, '
  1878. s += 'or setting INCLUDE_UNLISTED_ENTRIES = True.'
  1879. return s
  1880. def process_existing():
  1881. logging.basicConfig(level=logging.INFO)
  1882. logging.getLogger('stem').setLevel(logging.INFO)
  1883. whitelist = {'data': parse_fallback_file(FALLBACK_FILE_NAME),
  1884. 'name': FALLBACK_FILE_NAME}
  1885. blacklist = {'data': read_from_file(BLACKLIST_FILE_NAME, MAX_LIST_FILE_SIZE),
  1886. 'name': BLACKLIST_FILE_NAME}
  1887. list_fallbacks(whitelist, blacklist)
  1888. def process_default():
  1889. logging.basicConfig(level=logging.WARNING)
  1890. logging.getLogger('stem').setLevel(logging.WARNING)
  1891. whitelist = {'data': read_from_file(WHITELIST_FILE_NAME, MAX_LIST_FILE_SIZE),
  1892. 'name': WHITELIST_FILE_NAME}
  1893. blacklist = {'data': read_from_file(BLACKLIST_FILE_NAME, MAX_LIST_FILE_SIZE),
  1894. 'name': BLACKLIST_FILE_NAME}
  1895. list_fallbacks(whitelist, blacklist)
  1896. ## Main Function
  1897. def main():
  1898. if get_command() == 'check_existing':
  1899. process_existing()
  1900. else:
  1901. process_default()
  1902. def get_command():
  1903. if len(sys.argv) == 2:
  1904. return sys.argv[1]
  1905. else:
  1906. return None
  1907. def log_excluded(msg, *args):
  1908. if get_command() == 'check_existing':
  1909. logging.warning(msg, *args)
  1910. else:
  1911. logging.info(msg, *args)
  1912. def list_fallbacks(whitelist, blacklist):
  1913. """ Fetches required onionoo documents and evaluates the
  1914. fallback directory criteria for each of the relays """
  1915. logging.warning('Downloading and parsing Onionoo data. ' +
  1916. 'This may take some time.')
  1917. # find relays that could be fallbacks
  1918. candidates = CandidateList()
  1919. candidates.add_relays()
  1920. # work out how many fallbacks we want
  1921. guard_count = candidates.count_guards()
  1922. if FALLBACK_PROPORTION_OF_GUARDS is None:
  1923. target_count = guard_count
  1924. else:
  1925. target_count = int(guard_count * FALLBACK_PROPORTION_OF_GUARDS)
  1926. # the maximum number of fallbacks is the least of:
  1927. # - the target fallback count (FALLBACK_PROPORTION_OF_GUARDS * guard count)
  1928. # - the maximum fallback count (MAX_FALLBACK_COUNT)
  1929. if MAX_FALLBACK_COUNT is None:
  1930. max_count = target_count
  1931. else:
  1932. max_count = min(target_count, MAX_FALLBACK_COUNT)
  1933. candidates.compute_fallbacks()
  1934. prefilter_fallbacks = copy.copy(candidates.fallbacks)
  1935. # filter with the whitelist and blacklist
  1936. # if a relay has changed IPv4 address or ports recently, it will be excluded
  1937. # as ineligible before we call apply_filter_lists, and so there will be no
  1938. # warning that the details have changed from those in the whitelist.
  1939. # instead, there will be an info-level log during the eligibility check.
  1940. initial_count = len(candidates.fallbacks)
  1941. excluded_count = candidates.apply_filter_lists(whitelist, blacklist)
  1942. print candidates.summarise_filters(initial_count, excluded_count)
  1943. eligible_count = len(candidates.fallbacks)
  1944. # calculate the measured bandwidth of each relay,
  1945. # then remove low-bandwidth relays
  1946. candidates.calculate_measured_bandwidth()
  1947. candidates.remove_low_bandwidth_relays()
  1948. # print the raw fallback list
  1949. #for x in candidates.fallbacks:
  1950. # print x.fallbackdir_line(True)
  1951. # print json.dumps(candidates[x]._data, sort_keys=True, indent=4,
  1952. # separators=(',', ': '), default=json_util.default)
  1953. # impose mandatory conditions here, like one per contact, family, IP
  1954. # in measured bandwidth order
  1955. candidates.sort_fallbacks_by_measured_bandwidth()
  1956. operator_count = 0
  1957. # only impose these limits on the final list - operators can nominate
  1958. # multiple candidate fallbacks, and then we choose the best set
  1959. if not OUTPUT_CANDIDATES:
  1960. operator_count += candidates.limit_fallbacks_same_ip()
  1961. operator_count += candidates.limit_fallbacks_same_contact()
  1962. operator_count += candidates.limit_fallbacks_same_family()
  1963. # check if each candidate can serve a consensus
  1964. # there's a small risk we've eliminated relays from the same operator that
  1965. # can serve a consensus, in favour of one that can't
  1966. # but given it takes up to 15 seconds to check each consensus download,
  1967. # the risk is worth it
  1968. if PERFORM_IPV4_DIRPORT_CHECKS or PERFORM_IPV6_DIRPORT_CHECKS:
  1969. logging.warning('Checking consensus download speeds. ' +
  1970. 'This may take some time.')
  1971. failed_count = candidates.perform_download_consensus_checks(max_count)
  1972. # analyse and log interesting diversity metrics
  1973. # like netblock, ports, exit, IPv4-only
  1974. # (we can't easily analyse AS, and it's hard to accurately analyse country)
  1975. candidates.describe_fallback_ip_family()
  1976. # if we can't import the ipaddress module, we can't do netblock analysis
  1977. if HAVE_IPADDRESS:
  1978. candidates.describe_fallback_netblocks()
  1979. candidates.describe_fallback_ports()
  1980. candidates.describe_fallback_exit_flag()
  1981. # output C comments summarising the fallback selection process
  1982. if len(candidates.fallbacks) > 0:
  1983. print candidates.summarise_fallbacks(eligible_count, operator_count,
  1984. failed_count, guard_count,
  1985. target_count)
  1986. else:
  1987. print '/* No Fallbacks met criteria */'
  1988. # output C comments specifying the OnionOO data used to create the list
  1989. for s in fetch_source_list():
  1990. print describe_fetch_source(s)
  1991. # sort the list differently depending on why we've created it:
  1992. # if we're outputting the final fallback list, sort by fingerprint
  1993. # this makes diffs much more stable
  1994. # otherwise, if we're trying to find a bandwidth cutoff, or we want to
  1995. # contact operators in priority order, sort by bandwidth (not yet
  1996. # implemented)
  1997. # otherwise, if we're contacting operators, sort by contact
  1998. candidates.sort_fallbacks_by(OUTPUT_SORT_FIELD)
  1999. for x in candidates.fallbacks:
  2000. print x.fallbackdir_line(candidates.fallbacks, prefilter_fallbacks)
  2001. if __name__ == "__main__":
  2002. main()