apparmor/utils/test/test-regex_matches.py

518 lines
22 KiB
Python
Raw Normal View History

utils: add simple parsing of multi-line rules [v3] D-Bus rules in particular seem to get written as multi-line rules. This patch adds very simple hackish support for multiple lines. Essentially, what it does is if the parsing of a line doesn't match anything and falls all the way through, it saves the line and prepends it to the next line that occurs in the profile, but *only* if the line does not have a trailing comma to indicate the end of a rule. If the trailing comma exists, then it assumes that it's a rule that it doesn't understand and aborts. With this patch, the simpler tools (aa-enforce, aa-complain, etc.) can parse policies containing multi-line rules to an extent and continue to function correctly. Again, aa-logprof and aa-genprof may have issues on the writing back of profiles, so some assistance testing here would be appreciated. Some testcases are added to exercise the regex that looks for a rule with a trailing comma but can still handle rules that have (,) or {,} in them. Patch history: v1 - initial version v2 - simplify and rearrange rule-ending comma search regex, since we only care about the trailing comma - add a new regex to search for trailing comments to filter out - simplify reset of lastline variable - restructure tests into a new script, and add more tests v3 - add additional testcases, most of which are problematic and thus commented out :( Signed-off-by: Steve Beattie <steve@nxnw.org> Acked-by: Seth Arnold <seth.arnold@canonical.com> Acked-by: Christian Boltz <apparmor@cboltz.de>
2014-03-07 10:04:57 -08:00
#! /usr/bin/env python
# ------------------------------------------------------------------
#
# Copyright (C) 2014 Canonical Ltd.
#
# This program is free software; you can redistribute it and/or
# modify it under the terms of version 2 of the GNU General Public
# License published by the Free Software Foundation.
#
# ------------------------------------------------------------------
import apparmor.aa as aa
import unittest
class AARegexHasComma(unittest.TestCase):
'''Tests for apparmor.aa.RE_RULE_HAS_COMMA'''
def _check(self, line, expected=True):
result = aa.RE_RULE_HAS_COMMA.search(line)
if expected:
self.assertTrue(result, 'Couldn\'t find a comma in "%s"' % line)
else:
self.assertEqual(None, result, 'Found an unexpected comma in "%s"' % line)
regex_has_comma_testcases = [
('dbus send%s', 'simple'),
('dbus (r, w, bind, eavesdrop)%s', 'embedded parens 01'),
('dbus (r, w,, bind, eavesdrop) %s', 'embedded parens 02'),
('dbus (r, w,, ) %s', 'embedded parens 03'),
('dbus () %s', 'empty parens'),
('member={Hello,AddMatch,RemoveMatch,GetNameOwner,NameHasOwner,StartServiceByName} %s ', 'embedded curly braces 01'),
('member={Hello,,,,,,AddMatch,,,NameHasOwner,StartServiceByName} %s ', 'embedded curly braces 02'),
('member={,Hello,,,,,,AddMatch,,,NameHasOwner,} %s ', 'embedded curly braces 03'),
('member={} %s ', 'empty curly braces'),
('dbus send%s# this is a comment', 'comment 01'),
('dbus send%s# this is a comment,', 'comment 02'),
('audit "/tmp/foo, bar" rw%s', 'quotes 01'),
('audit "/tmp/foo, bar" rw%s # comment', 'quotes 02'),
('audit "/tmp/foo, # bar" rw%s', 'comment embedded in quote 01'),
('audit "/tmp/foo, # bar" rw%s # comment', 'comment embedded in quote 02'),
# lifted from parser/tst/simple_tests/vars/vars_alternation_3.sd
('/does/not/@{BAR},exist,notexist} r%s', 'partial alternation'),
('signal%s', 'bare signal'),
('signal receive%s', 'simple signal'),
('signal (send, receive)%s', 'embedded parens signal 01'),
('signal (send, receive) set=(hup, quit)%s', 'embedded parens signal 02'),
utils: add simple parsing of multi-line rules [v3] D-Bus rules in particular seem to get written as multi-line rules. This patch adds very simple hackish support for multiple lines. Essentially, what it does is if the parsing of a line doesn't match anything and falls all the way through, it saves the line and prepends it to the next line that occurs in the profile, but *only* if the line does not have a trailing comma to indicate the end of a rule. If the trailing comma exists, then it assumes that it's a rule that it doesn't understand and aborts. With this patch, the simpler tools (aa-enforce, aa-complain, etc.) can parse policies containing multi-line rules to an extent and continue to function correctly. Again, aa-logprof and aa-genprof may have issues on the writing back of profiles, so some assistance testing here would be appreciated. Some testcases are added to exercise the regex that looks for a rule with a trailing comma but can still handle rules that have (,) or {,} in them. Patch history: v1 - initial version v2 - simplify and rearrange rule-ending comma search regex, since we only care about the trailing comma - add a new regex to search for trailing comments to filter out - simplify reset of lastline variable - restructure tests into a new script, and add more tests v3 - add additional testcases, most of which are problematic and thus commented out :( Signed-off-by: Steve Beattie <steve@nxnw.org> Acked-by: Seth Arnold <seth.arnold@canonical.com> Acked-by: Christian Boltz <apparmor@cboltz.de>
2014-03-07 10:04:57 -08:00
('ptrace%s', 'bare ptrace'),
('ptrace trace%s', 'simple ptrace'),
('ptrace (tracedby, readby)%s', 'embedded parens ptrace 01'),
('ptrace (trace) peer=/usr/bin/foo%s', 'embedded parens ptrace 02'),
('pivot_root%s', 'bare pivot_root'),
('pivot_root /old%s', 'pivot_root with old'),
('pivot_root /old new%s', 'pivot_root with new'),
('pivot_root /old /new -> child%s', 'pivot_root with child'),
utils: add simple parsing of multi-line rules [v3] D-Bus rules in particular seem to get written as multi-line rules. This patch adds very simple hackish support for multiple lines. Essentially, what it does is if the parsing of a line doesn't match anything and falls all the way through, it saves the line and prepends it to the next line that occurs in the profile, but *only* if the line does not have a trailing comma to indicate the end of a rule. If the trailing comma exists, then it assumes that it's a rule that it doesn't understand and aborts. With this patch, the simpler tools (aa-enforce, aa-complain, etc.) can parse policies containing multi-line rules to an extent and continue to function correctly. Again, aa-logprof and aa-genprof may have issues on the writing back of profiles, so some assistance testing here would be appreciated. Some testcases are added to exercise the regex that looks for a rule with a trailing comma but can still handle rules that have (,) or {,} in them. Patch history: v1 - initial version v2 - simplify and rearrange rule-ending comma search regex, since we only care about the trailing comma - add a new regex to search for trailing comments to filter out - simplify reset of lastline variable - restructure tests into a new script, and add more tests v3 - add additional testcases, most of which are problematic and thus commented out :( Signed-off-by: Steve Beattie <steve@nxnw.org> Acked-by: Seth Arnold <seth.arnold@canonical.com> Acked-by: Christian Boltz <apparmor@cboltz.de>
2014-03-07 10:04:57 -08:00
# the following fail due to inadequacies in the regex
# ('dbus (r, w, %s', 'incomplete dbus action'),
# ('member="{Hello,AddMatch,RemoveMatch, %s', 'incomplete {} regex'), # also invalid policy
# ('member={Hello,AddMatch,RemoveMatch, %s', 'incomplete {} regex'), # also invalid policy when trailing comma exists
# the following passes the tests, but variable declarations are
# odd in that they *don't* allow trailing commas; commas at the end
# of the line need to be quoted.
# ('@{BAR}={bar,baz,blort %s', 'tricksy variable declaration')
# ('@{BAR}="{bar,baz,blort," %s', 'tricksy variable declaration')
# The following fails the no comma test, but is invalid
# ('@{BAR}={bar,baz,blort, %s', 'tricksy variable declaration')
# The following fails the comma test, because it's really a no comma situation
# ('@{BAR}="{bar,baz,blort%s" ', 'tricksy variable declaration')
]
def setup_has_comma_testcases():
i = 0
for (test_string, description) in regex_has_comma_testcases:
i += 1
def stub_test_comma(self, test_string=test_string):
self._check(test_string % ',')
def stub_test_no_comma(self, test_string=test_string):
self._check(test_string % ' ', False)
stub_test_comma.__doc__ = "test %s (w/comma)" % (description)
stub_test_no_comma.__doc__ = "test %s (no comma)" % (description)
setattr(AARegexHasComma, 'test_comma_%d' % (i), stub_test_comma)
setattr(AARegexHasComma, 'test_no_comma_%d' % (i), stub_test_no_comma)
class AARegexSplitComment(unittest.TestCase):
'''Tests for RE_HAS_COMMENT_SPLIT'''
def _check(self, line, expected, comment=None, not_comment=None):
result = aa.RE_HAS_COMMENT_SPLIT.search(line)
if expected:
self.assertTrue(result, 'Couldn\'t find a comment in "%s"' % line)
self.assertEqual(result.group('comment'), comment, 'Expected comment "%s", got "%s"'
% (comment, result.group('comment')))
self.assertEqual(result.group('not_comment'), not_comment, 'Expected not comment "%s", got "%s"'
% (not_comment, result.group('not_comment')))
else:
self.assertEqual(None, result, 'Found an unexpected comment "%s" in "%s"'
% ("" if result is None else result.group('comment'), line ))
# Tuples of (string, expected result), where expected result is False if
# the string should not be considered as having a comment, or a second
# tuple of the not comment and comment sections split apart
regex_split_comment_testcases = [
('dbus send # this is a comment', ('dbus send ', '# this is a comment')),
('dbus send member=no_comment', False),
('dbus send member=no_comment, ', False),
('audit "/tmp/foo, # bar" rw', False),
('audit "/tmp/foo, # bar" rw # comment', ('audit "/tmp/foo, # bar" rw ', '# comment')),
('file,', False),
('file, # bare', ('file, ', '# bare')),
('file /tmp/foo rw, # read-write', ('file /tmp/foo rw, ', '# read-write')),
('signal, # comment', ('signal, ', '# comment')),
('signal receive set=(usr1 usr2) peer=foo,', False),
('ptrace, # comment', ('ptrace, ', '# comment')),
('ptrace (trace read) peer=/usr/bin/foo,', False),
('pivot_root, # comment', ('pivot_root, ', '# comment')),
('pivot_root /old /new -> child,', False),
utils: add simple parsing of multi-line rules [v3] D-Bus rules in particular seem to get written as multi-line rules. This patch adds very simple hackish support for multiple lines. Essentially, what it does is if the parsing of a line doesn't match anything and falls all the way through, it saves the line and prepends it to the next line that occurs in the profile, but *only* if the line does not have a trailing comma to indicate the end of a rule. If the trailing comma exists, then it assumes that it's a rule that it doesn't understand and aborts. With this patch, the simpler tools (aa-enforce, aa-complain, etc.) can parse policies containing multi-line rules to an extent and continue to function correctly. Again, aa-logprof and aa-genprof may have issues on the writing back of profiles, so some assistance testing here would be appreciated. Some testcases are added to exercise the regex that looks for a rule with a trailing comma but can still handle rules that have (,) or {,} in them. Patch history: v1 - initial version v2 - simplify and rearrange rule-ending comma search regex, since we only care about the trailing comma - add a new regex to search for trailing comments to filter out - simplify reset of lastline variable - restructure tests into a new script, and add more tests v3 - add additional testcases, most of which are problematic and thus commented out :( Signed-off-by: Steve Beattie <steve@nxnw.org> Acked-by: Seth Arnold <seth.arnold@canonical.com> Acked-by: Christian Boltz <apparmor@cboltz.de>
2014-03-07 10:04:57 -08:00
]
def setup_split_comment_testcases():
i = 0
for (test_string, result) in regex_split_comment_testcases:
i += 1
def stub_test(self, test_string=test_string, result=result):
if result is False:
self._check(test_string, False)
else:
self._check(test_string, True, not_comment=result[0], comment=result[1])
stub_test.__doc__ = "test '%s'" % (test_string)
setattr(AARegexSplitComment, 'test_split_comment_%d' % (i), stub_test)
class AARegexCapability(unittest.TestCase):
'''Tests for RE_PROFILE_CAP'''
def test_simple_capability_01(self):
'''test ' capability net_raw,' '''
line = ' capability net_raw,'
result = aa.RE_PROFILE_CAP.search(line)
self.assertTrue(result, 'Couldn\'t find capability rule in "%s"' % line)
cap = result.groups()[2].strip()
self.assertEqual(cap, 'net_raw', 'Expected capability "%s", got "%s"'
% ('net_raw', cap))
def test_simple_capability_02(self):
'''test ' capability net_raw , ' '''
line = 'capability net_raw , '
result = aa.RE_PROFILE_CAP.search(line)
self.assertTrue(result, 'Couldn\'t find capability rule in "%s"' % line)
cap = result.groups()[2].strip()
self.assertEqual(cap, 'net_raw', 'Expected capability "%s", got "%s"'
% ('net_raw', cap))
def test_capability_all_01(self):
'''test ' capability,' '''
line = ' capability,'
result = aa.RE_PROFILE_CAP.search(line)
self.assertTrue(result, 'Couldn\'t find capability rule in "%s"' % line)
def test_capability_all_02(self):
'''test ' capability , ' '''
line = ' capability , '
result = aa.RE_PROFILE_CAP.search(line)
self.assertTrue(result, 'Couldn\'t find capability rule in "%s"' % line)
def test_simple_bad_capability_01(self):
'''test ' capabilitynet_raw,' '''
line = ' capabilitynet_raw,'
result = aa.RE_PROFILE_CAP.search(line)
self.assertFalse(result, 'Found unexpected capability rule in "%s"' % line)
class AARegexPath(unittest.TestCase):
'''Tests for RE_PROFILE_PATH_ENTRY'''
def test_simple_path_01(self):
'''test ' /tmp/foo r,' '''
line = ' /tmp/foo r,'
result = aa.RE_PROFILE_PATH_ENTRY.search(line)
self.assertTrue(result, 'Couldn\'t find file rule in "%s"' % line)
mode = result.groups()[5].strip()
self.assertEqual(mode, 'r', 'Expected mode "r", got "%s"' % (mode))
def test_simple_path_02(self):
'''test ' audit /tmp/foo rw,' '''
line = ' audit /tmp/foo rw,'
result = aa.RE_PROFILE_PATH_ENTRY.search(line)
self.assertTrue(result, 'Couldn\'t find file rule in "%s"' % line)
audit = result.groups()[0].strip()
self.assertEqual(audit, 'audit', 'Couldn\t find audit modifier')
mode = result.groups()[5].strip()
self.assertEqual(mode, 'rw', 'Expected mode "rw", got "%s"' % (mode))
def test_simple_path_03(self):
'''test ' audit deny /tmp/foo rw,' '''
line = ' audit deny /tmp/foo rw,'
result = aa.RE_PROFILE_PATH_ENTRY.search(line)
self.assertTrue(result, 'Couldn\'t find file rule in "%s"' % line)
audit = result.groups()[0].strip()
self.assertEqual(audit, 'audit', 'Couldn\t find audit modifier')
deny = result.groups()[1].strip()
self.assertEqual(deny, 'deny', 'Couldn\t find deny modifier')
mode = result.groups()[5].strip()
self.assertEqual(mode, 'rw', 'Expected mode "rw", got "%s"' % (mode))
def test_simple_path_04(self):
'''test ' file /tmp/foo rw,' '''
line = ' file /tmp/foo rw,'
result = aa.RE_PROFILE_PATH_ENTRY.search(line)
self.assertTrue(result, 'Couldn\'t find file rule in "%s"' % line)
path = result.groups()[4].strip()
self.assertEqual(path, "/tmp/foo", 'Couldn\'t find path')
mode = result.groups()[5].strip()
self.assertEqual(mode, 'rw', 'Expected mode "rw", got "%s"' % (mode))
def test_simple_bad_path_01(self):
'''test ' file,' '''
line = ' file,'
result = aa.RE_PROFILE_PATH_ENTRY.search(line)
self.assertFalse(result, 'RE_PROFILE_PATH_ENTRY unexpectedly matched "%s"' % line)
class AARegexBareFile(unittest.TestCase):
'''Tests for RE_PROFILE_BARE_FILE_ENTRY'''
def _assertEqualStrings(self, str1, str2):
self.assertEqual(str1, str2, 'Expected %s, got "%s"' % (str1, str2))
def test_bare_file_01(self):
'''test ' file,' '''
line = ' file,'
result = aa.RE_PROFILE_BARE_FILE_ENTRY.search(line)
self.assertTrue(result, 'Couldn\'t find file rule in "%s"' % line)
def test_simple_bad_file_01(self):
'''test ' dbus,' '''
line = ' dbus,'
result = aa.RE_PROFILE_BARE_FILE_ENTRY.search(line)
self.assertFalse(result, 'RE_PROFILE_BARE_FILE_ENTRY unexpectedly matched "%s"' % line)
def test_simple_bad_file_02(self):
'''test ' file /tmp/foo rw,' '''
line = ' file /tmp/foo rw,'
result = aa.RE_PROFILE_BARE_FILE_ENTRY.search(line)
self.assertFalse(result, 'RE_PROFILE_BARE_FILE_ENTRY unexpectedly matched "%s"' % line)
def test_simple_bad_file_03(self):
'''test ' file /tmp/foo,' '''
line = ' file /tmp/foo,'
result = aa.RE_PROFILE_BARE_FILE_ENTRY.search(line)
self.assertFalse(result, 'RE_PROFILE_BARE_FILE_ENTRY unexpectedly matched "%s"' % line)
def test_simple_bad_file_04(self):
'''test ' file r,' '''
line = ' file r,'
result = aa.RE_PROFILE_BARE_FILE_ENTRY.search(line)
self.assertFalse(result, 'RE_PROFILE_BARE_FILE_ENTRY unexpectedly matched "%s"' % line)
class AARegexSignal(unittest.TestCase):
'''Tests for RE_PROFILE_SIGNAL'''
def test_bare_signal_01(self):
'''test ' signal,' '''
rule = 'signal,'
line = ' %s' % rule
result = aa.RE_PROFILE_SIGNAL.search(line)
self.assertTrue(result, 'Couldn\'t find signal rule in "%s"' % line)
parsed = result.groups()[2].strip()
self.assertEqual(parsed, rule, 'Expected signal rule "%s", got "%s"'
% (rule, parsed))
def test_bare_signal_02(self):
'''test ' audit signal,' '''
rule = 'signal,'
line = ' audit %s' % rule
result = aa.RE_PROFILE_SIGNAL.search(line)
self.assertTrue(result, 'Couldn\'t find signal rule in "%s"' % line)
self.assertTrue(result.groups()[0], 'Couldn\'t find audit modifier in "%s"' % line)
parsed = result.groups()[2].strip()
self.assertEqual(parsed, rule, 'Expected signal rule "%s", got "%s"'
% (rule, parsed))
def test_simple_signal_01(self):
'''test ' signal receive,' '''
rule = 'signal receive,'
line = ' %s' % rule
result = aa.RE_PROFILE_SIGNAL.search(line)
self.assertTrue(result, 'Couldn\'t find signal rule in "%s"' % line)
parsed = result.groups()[2].strip()
self.assertEqual(parsed, rule, 'Expected signal rule "%s", got "%s"'
% (rule, parsed))
def test_simple_signal_02(self):
'''test ' signal (send, receive),' '''
rule = 'signal (send, receive),'
line = ' %s' % rule
result = aa.RE_PROFILE_SIGNAL.search(line)
self.assertTrue(result, 'Couldn\'t find signal rule in "%s"' % line)
parsed = result.groups()[2].strip()
self.assertEqual(parsed, rule, 'Expected signal rule "%s", got "%s"'
% (rule, parsed))
def test_simple_signal_03(self):
'''test ' audit signal (receive),' '''
rule = 'signal (receive),'
line = ' audit %s' % rule
result = aa.RE_PROFILE_SIGNAL.search(line)
self.assertTrue(result, 'Couldn\'t find signal rule in "%s"' % line)
self.assertTrue(result.groups()[0], 'Couldn\'t find audit modifier in "%s"' % line)
parsed = result.groups()[2].strip()
self.assertEqual(parsed, rule, 'Expected signal rule "%s", got "%s"'
% (rule, parsed))
def test_set_signal_01(self):
'''test ' signal (send, receive) set=(usr1 usr2),' '''
rule = 'signal (send, receive) set=(usr1 usr2),'
line = ' %s' % rule
result = aa.RE_PROFILE_SIGNAL.search(line)
self.assertTrue(result, 'Couldn\'t find signal rule in "%s"' % line)
parsed = result.groups()[2].strip()
self.assertEqual(parsed, rule, 'Expected signal rule "%s", got "%s"'
% (rule, parsed))
def test_peer_signal_01(self):
'''test ' signal send set=(hup, quit) peer=/usr/sbin/daemon,' '''
rule = 'signal send set=(hup, quit) peer=/usr/sbin/daemon,'
line = ' %s' % rule
result = aa.RE_PROFILE_SIGNAL.search(line)
self.assertTrue(result, 'Couldn\'t find signal rule in "%s"' % line)
parsed = result.groups()[2].strip()
self.assertEqual(parsed, rule, 'Expected signal rule "%s", got "%s"'
% (rule, parsed))
class AARegexPtrace(unittest.TestCase):
'''Tests for RE_PROFILE_PTRACE'''
def test_bare_ptrace_01(self):
'''test ' ptrace,' '''
rule = 'ptrace,'
line = ' %s' % rule
result = aa.RE_PROFILE_PTRACE.search(line)
self.assertTrue(result, 'Couldn\'t find ptrace rule in "%s"' % line)
parsed = result.groups()[2].strip()
self.assertEqual(parsed, rule, 'Expected ptrace rule "%s", got "%s"'
% (rule, parsed))
def test_bare_ptrace_02(self):
'''test ' audit ptrace,' '''
rule = 'ptrace,'
line = ' audit %s' % rule
result = aa.RE_PROFILE_PTRACE.search(line)
self.assertTrue(result, 'Couldn\'t find ptrace rule in "%s"' % line)
self.assertTrue(result.groups()[0], 'Couldn\'t find audit modifier in "%s"' % line)
parsed = result.groups()[2].strip()
self.assertEqual(parsed, rule, 'Expected ptrace rule "%s", got "%s"'
% (rule, parsed))
def test_simple_ptrace_01(self):
'''test ' ptrace trace,' '''
rule = 'ptrace trace,'
line = ' %s' % rule
result = aa.RE_PROFILE_PTRACE.search(line)
self.assertTrue(result, 'Couldn\'t find ptrace rule in "%s"' % line)
parsed = result.groups()[2].strip()
self.assertEqual(parsed, rule, 'Expected ptrace rule "%s", got "%s"'
% (rule, parsed))
def test_simple_ptrace_02(self):
'''test ' ptrace (tracedby, readby),' '''
rule = 'ptrace (tracedby, readby),'
line = ' %s' % rule
result = aa.RE_PROFILE_PTRACE.search(line)
self.assertTrue(result, 'Couldn\'t find ptrace rule in "%s"' % line)
parsed = result.groups()[2].strip()
self.assertEqual(parsed, rule, 'Expected ptrace rule "%s", got "%s"'
% (rule, parsed))
def test_simple_ptrace_03(self):
'''test ' audit ptrace (read),' '''
rule = 'ptrace (read),'
line = ' audit %s' % rule
result = aa.RE_PROFILE_PTRACE.search(line)
self.assertTrue(result, 'Couldn\'t find ptrace rule in "%s"' % line)
self.assertTrue(result.groups()[0], 'Couldn\'t find audit modifier in "%s"' % line)
parsed = result.groups()[2].strip()
self.assertEqual(parsed, rule, 'Expected ptrace rule "%s", got "%s"'
% (rule, parsed))
def test_peer_ptrace_01(self):
'''test ' ptrace trace peer=/usr/sbin/daemon,' '''
rule = 'ptrace trace peer=/usr/sbin/daemon,'
line = ' %s' % rule
result = aa.RE_PROFILE_PTRACE.search(line)
self.assertTrue(result, 'Couldn\'t find ptrace rule in "%s"' % line)
parsed = result.groups()[2].strip()
self.assertEqual(parsed, rule, 'Expected ptrace rule "%s", got "%s"'
% (rule, parsed))
class AARegexPivotRoot(unittest.TestCase):
'''Tests for RE_PROFILE_PIVOT_ROOT'''
def test_bare_pivot_root_01(self):
'''test ' pivot_root,' '''
rule = 'pivot_root,'
line = ' %s' % rule
result = aa.RE_PROFILE_PIVOT_ROOT.search(line)
self.assertTrue(result, 'Couldn\'t find pivot_root rule in "%s"' % line)
parsed = result.groups()[2].strip()
self.assertEqual(parsed, rule, 'Expected pivot_root rule "%s", got "%s"'
% (rule, parsed))
def test_bare_pivot_root_02(self):
'''test ' audit pivot_root,' '''
rule = 'pivot_root,'
line = ' audit %s' % rule
result = aa.RE_PROFILE_PIVOT_ROOT.search(line)
self.assertTrue(result, 'Couldn\'t find pivot_root rule in "%s"' % line)
self.assertTrue(result.groups()[0], 'Couldn\'t find audit modifier in "%s"' % line)
parsed = result.groups()[2].strip()
self.assertEqual(parsed, rule, 'Expected pivot_root rule "%s", got "%s"'
% (rule, parsed))
def test_old_pivot_root_01(self):
'''test ' pivot_root /old,' '''
rule = 'pivot_root /old,'
line = ' %s' % rule
result = aa.RE_PROFILE_PIVOT_ROOT.search(line)
self.assertTrue(result, 'Couldn\'t find pivot_root rule in "%s"' % line)
parsed = result.groups()[2].strip()
self.assertEqual(parsed, rule, 'Expected pivot_root rule "%s", got "%s"'
% (rule, parsed))
def test_new_pivot_root_01(self):
'''test ' pivot_root /old /new,' '''
rule = 'pivot_root /old /new,'
line = ' %s' % rule
result = aa.RE_PROFILE_PIVOT_ROOT.search(line)
self.assertTrue(result, 'Couldn\'t find pivot_root rule in "%s"' % line)
parsed = result.groups()[2].strip()
self.assertEqual(parsed, rule, 'Expected pivot_root rule "%s", got "%s"'
% (rule, parsed))
def test_child_pivot_root_01(self):
'''test ' pivot_root /old /new -> child,' '''
rule = 'pivot_root /old /new -> child,'
line = ' %s' % rule
result = aa.RE_PROFILE_PIVOT_ROOT.search(line)
self.assertTrue(result, 'Couldn\'t find pivot_root rule in "%s"' % line)
parsed = result.groups()[2].strip()
self.assertEqual(parsed, rule, 'Expected pivot_root rule "%s", got "%s"'
% (rule, parsed))
def test_child_pivot_root_02(self):
'''test ' audit pivot_root /old /new -> child,' '''
rule = 'pivot_root /old /new -> child,'
line = ' audit %s' % rule
result = aa.RE_PROFILE_PIVOT_ROOT.search(line)
self.assertTrue(result, 'Couldn\'t find pivot_root rule in "%s"' % line)
self.assertTrue(result.groups()[0], 'Couldn\'t find audit modifier in "%s"' % line)
parsed = result.groups()[2].strip()
self.assertEqual(parsed, rule, 'Expected pivot_root rule "%s", got "%s"'
% (rule, parsed))
utils: add simple parsing of multi-line rules [v3] D-Bus rules in particular seem to get written as multi-line rules. This patch adds very simple hackish support for multiple lines. Essentially, what it does is if the parsing of a line doesn't match anything and falls all the way through, it saves the line and prepends it to the next line that occurs in the profile, but *only* if the line does not have a trailing comma to indicate the end of a rule. If the trailing comma exists, then it assumes that it's a rule that it doesn't understand and aborts. With this patch, the simpler tools (aa-enforce, aa-complain, etc.) can parse policies containing multi-line rules to an extent and continue to function correctly. Again, aa-logprof and aa-genprof may have issues on the writing back of profiles, so some assistance testing here would be appreciated. Some testcases are added to exercise the regex that looks for a rule with a trailing comma but can still handle rules that have (,) or {,} in them. Patch history: v1 - initial version v2 - simplify and rearrange rule-ending comma search regex, since we only care about the trailing comma - add a new regex to search for trailing comments to filter out - simplify reset of lastline variable - restructure tests into a new script, and add more tests v3 - add additional testcases, most of which are problematic and thus commented out :( Signed-off-by: Steve Beattie <steve@nxnw.org> Acked-by: Seth Arnold <seth.arnold@canonical.com> Acked-by: Christian Boltz <apparmor@cboltz.de>
2014-03-07 10:04:57 -08:00
if __name__ == '__main__':
verbosity = 2
setup_has_comma_testcases()
setup_split_comment_testcases()
test_suite = unittest.TestSuite()
test_suite.addTest(unittest.TestLoader().loadTestsFromTestCase(AARegexHasComma))
test_suite.addTest(unittest.TestLoader().loadTestsFromTestCase(AARegexSplitComment))
test_suite.addTest(unittest.TestLoader().loadTestsFromTestCase(AARegexCapability))
test_suite.addTest(unittest.TestLoader().loadTestsFromTestCase(AARegexPath))
test_suite.addTest(unittest.TestLoader().loadTestsFromTestCase(AARegexBareFile))
test_suite.addTest(unittest.TestLoader().loadTestsFromTestCase(AARegexSignal))
test_suite.addTest(unittest.TestLoader().loadTestsFromTestCase(AARegexPtrace))
test_suite.addTest(unittest.TestLoader().loadTestsFromTestCase(AARegexPivotRoot))
utils: add simple parsing of multi-line rules [v3] D-Bus rules in particular seem to get written as multi-line rules. This patch adds very simple hackish support for multiple lines. Essentially, what it does is if the parsing of a line doesn't match anything and falls all the way through, it saves the line and prepends it to the next line that occurs in the profile, but *only* if the line does not have a trailing comma to indicate the end of a rule. If the trailing comma exists, then it assumes that it's a rule that it doesn't understand and aborts. With this patch, the simpler tools (aa-enforce, aa-complain, etc.) can parse policies containing multi-line rules to an extent and continue to function correctly. Again, aa-logprof and aa-genprof may have issues on the writing back of profiles, so some assistance testing here would be appreciated. Some testcases are added to exercise the regex that looks for a rule with a trailing comma but can still handle rules that have (,) or {,} in them. Patch history: v1 - initial version v2 - simplify and rearrange rule-ending comma search regex, since we only care about the trailing comma - add a new regex to search for trailing comments to filter out - simplify reset of lastline variable - restructure tests into a new script, and add more tests v3 - add additional testcases, most of which are problematic and thus commented out :( Signed-off-by: Steve Beattie <steve@nxnw.org> Acked-by: Seth Arnold <seth.arnold@canonical.com> Acked-by: Christian Boltz <apparmor@cboltz.de>
2014-03-07 10:04:57 -08:00
result = unittest.TextTestRunner(verbosity=verbosity).run(test_suite)
if not result.wasSuccessful():
exit(1)