##// END OF EJS Templates
pathauditor: change parts verification order to be root first...
Durham Goode -
r28087:0b7ce0b1 default
parent child Browse files
Show More
@@ -1,213 +1,214 b''
1 from __future__ import absolute_import
1 from __future__ import absolute_import
2
2
3 import errno
3 import errno
4 import os
4 import os
5 import posixpath
5 import posixpath
6 import stat
6 import stat
7
7
8 from .i18n import _
8 from .i18n import _
9 from . import (
9 from . import (
10 encoding,
10 encoding,
11 error,
11 error,
12 util,
12 util,
13 )
13 )
14
14
15 def _lowerclean(s):
15 def _lowerclean(s):
16 return encoding.hfsignoreclean(s.lower())
16 return encoding.hfsignoreclean(s.lower())
17
17
18 class pathauditor(object):
18 class pathauditor(object):
19 '''ensure that a filesystem path contains no banned components.
19 '''ensure that a filesystem path contains no banned components.
20 the following properties of a path are checked:
20 the following properties of a path are checked:
21
21
22 - ends with a directory separator
22 - ends with a directory separator
23 - under top-level .hg
23 - under top-level .hg
24 - starts at the root of a windows drive
24 - starts at the root of a windows drive
25 - contains ".."
25 - contains ".."
26
26
27 More check are also done about the file system states:
27 More check are also done about the file system states:
28 - traverses a symlink (e.g. a/symlink_here/b)
28 - traverses a symlink (e.g. a/symlink_here/b)
29 - inside a nested repository (a callback can be used to approve
29 - inside a nested repository (a callback can be used to approve
30 some nested repositories, e.g., subrepositories)
30 some nested repositories, e.g., subrepositories)
31
31
32 The file system checks are only done when 'realfs' is set to True (the
32 The file system checks are only done when 'realfs' is set to True (the
33 default). They should be disable then we are auditing path for operation on
33 default). They should be disable then we are auditing path for operation on
34 stored history.
34 stored history.
35 '''
35 '''
36
36
37 def __init__(self, root, callback=None, realfs=True):
37 def __init__(self, root, callback=None, realfs=True):
38 self.audited = set()
38 self.audited = set()
39 self.auditeddir = set()
39 self.auditeddir = set()
40 self.root = root
40 self.root = root
41 self._realfs = realfs
41 self._realfs = realfs
42 self.callback = callback
42 self.callback = callback
43 if os.path.lexists(root) and not util.checkcase(root):
43 if os.path.lexists(root) and not util.checkcase(root):
44 self.normcase = util.normcase
44 self.normcase = util.normcase
45 else:
45 else:
46 self.normcase = lambda x: x
46 self.normcase = lambda x: x
47
47
48 def __call__(self, path):
48 def __call__(self, path):
49 '''Check the relative path.
49 '''Check the relative path.
50 path may contain a pattern (e.g. foodir/**.txt)'''
50 path may contain a pattern (e.g. foodir/**.txt)'''
51
51
52 path = util.localpath(path)
52 path = util.localpath(path)
53 normpath = self.normcase(path)
53 normpath = self.normcase(path)
54 if normpath in self.audited:
54 if normpath in self.audited:
55 return
55 return
56 # AIX ignores "/" at end of path, others raise EISDIR.
56 # AIX ignores "/" at end of path, others raise EISDIR.
57 if util.endswithsep(path):
57 if util.endswithsep(path):
58 raise error.Abort(_("path ends in directory separator: %s") % path)
58 raise error.Abort(_("path ends in directory separator: %s") % path)
59 parts = util.splitpath(path)
59 parts = util.splitpath(path)
60 if (os.path.splitdrive(path)[0]
60 if (os.path.splitdrive(path)[0]
61 or _lowerclean(parts[0]) in ('.hg', '.hg.', '')
61 or _lowerclean(parts[0]) in ('.hg', '.hg.', '')
62 or os.pardir in parts):
62 or os.pardir in parts):
63 raise error.Abort(_("path contains illegal component: %s") % path)
63 raise error.Abort(_("path contains illegal component: %s") % path)
64 # Windows shortname aliases
64 # Windows shortname aliases
65 for p in parts:
65 for p in parts:
66 if "~" in p:
66 if "~" in p:
67 first, last = p.split("~", 1)
67 first, last = p.split("~", 1)
68 if last.isdigit() and first.upper() in ["HG", "HG8B6C"]:
68 if last.isdigit() and first.upper() in ["HG", "HG8B6C"]:
69 raise error.Abort(_("path contains illegal component: %s")
69 raise error.Abort(_("path contains illegal component: %s")
70 % path)
70 % path)
71 if '.hg' in _lowerclean(path):
71 if '.hg' in _lowerclean(path):
72 lparts = [_lowerclean(p.lower()) for p in parts]
72 lparts = [_lowerclean(p.lower()) for p in parts]
73 for p in '.hg', '.hg.':
73 for p in '.hg', '.hg.':
74 if p in lparts[1:]:
74 if p in lparts[1:]:
75 pos = lparts.index(p)
75 pos = lparts.index(p)
76 base = os.path.join(*parts[:pos])
76 base = os.path.join(*parts[:pos])
77 raise error.Abort(_("path '%s' is inside nested repo %r")
77 raise error.Abort(_("path '%s' is inside nested repo %r")
78 % (path, base))
78 % (path, base))
79
79
80 normparts = util.splitpath(normpath)
80 normparts = util.splitpath(normpath)
81 assert len(parts) == len(normparts)
81 assert len(parts) == len(normparts)
82
82
83 parts.pop()
83 parts.pop()
84 normparts.pop()
84 normparts.pop()
85 prefixes = []
85 prefixes = []
86 while parts:
86 # It's important that we check the path parts starting from the root.
87 prefix = os.sep.join(parts)
87 # This means we won't accidentaly traverse a symlink into some other
88 normprefix = os.sep.join(normparts)
88 # filesystem (which is potentially expensive to access).
89 for i in range(len(parts)):
90 prefix = os.sep.join(parts[:i + 1])
91 normprefix = os.sep.join(normparts[:i + 1])
89 if normprefix in self.auditeddir:
92 if normprefix in self.auditeddir:
90 break
93 continue
91 if self._realfs:
94 if self._realfs:
92 self._checkfs(prefix, path)
95 self._checkfs(prefix, path)
93 prefixes.append(normprefix)
96 prefixes.append(normprefix)
94 parts.pop()
95 normparts.pop()
96
97
97 self.audited.add(normpath)
98 self.audited.add(normpath)
98 # only add prefixes to the cache after checking everything: we don't
99 # only add prefixes to the cache after checking everything: we don't
99 # want to add "foo/bar/baz" before checking if there's a "foo/.hg"
100 # want to add "foo/bar/baz" before checking if there's a "foo/.hg"
100 self.auditeddir.update(prefixes)
101 self.auditeddir.update(prefixes)
101
102
102 def _checkfs(self, prefix, path):
103 def _checkfs(self, prefix, path):
103 """raise exception if a file system backed check fails"""
104 """raise exception if a file system backed check fails"""
104 curpath = os.path.join(self.root, prefix)
105 curpath = os.path.join(self.root, prefix)
105 try:
106 try:
106 st = os.lstat(curpath)
107 st = os.lstat(curpath)
107 except OSError as err:
108 except OSError as err:
108 # EINVAL can be raised as invalid path syntax under win32.
109 # EINVAL can be raised as invalid path syntax under win32.
109 # They must be ignored for patterns can be checked too.
110 # They must be ignored for patterns can be checked too.
110 if err.errno not in (errno.ENOENT, errno.ENOTDIR, errno.EINVAL):
111 if err.errno not in (errno.ENOENT, errno.ENOTDIR, errno.EINVAL):
111 raise
112 raise
112 else:
113 else:
113 if stat.S_ISLNK(st.st_mode):
114 if stat.S_ISLNK(st.st_mode):
114 msg = _('path %r traverses symbolic link %r') % (path, prefix)
115 msg = _('path %r traverses symbolic link %r') % (path, prefix)
115 raise error.Abort(msg)
116 raise error.Abort(msg)
116 elif (stat.S_ISDIR(st.st_mode) and
117 elif (stat.S_ISDIR(st.st_mode) and
117 os.path.isdir(os.path.join(curpath, '.hg'))):
118 os.path.isdir(os.path.join(curpath, '.hg'))):
118 if not self.callback or not self.callback(curpath):
119 if not self.callback or not self.callback(curpath):
119 msg = _("path '%s' is inside nested repo %r")
120 msg = _("path '%s' is inside nested repo %r")
120 raise error.Abort(msg % (path, prefix))
121 raise error.Abort(msg % (path, prefix))
121
122
122 def check(self, path):
123 def check(self, path):
123 try:
124 try:
124 self(path)
125 self(path)
125 return True
126 return True
126 except (OSError, error.Abort):
127 except (OSError, error.Abort):
127 return False
128 return False
128
129
129 def canonpath(root, cwd, myname, auditor=None):
130 def canonpath(root, cwd, myname, auditor=None):
130 '''return the canonical path of myname, given cwd and root'''
131 '''return the canonical path of myname, given cwd and root'''
131 if util.endswithsep(root):
132 if util.endswithsep(root):
132 rootsep = root
133 rootsep = root
133 else:
134 else:
134 rootsep = root + os.sep
135 rootsep = root + os.sep
135 name = myname
136 name = myname
136 if not os.path.isabs(name):
137 if not os.path.isabs(name):
137 name = os.path.join(root, cwd, name)
138 name = os.path.join(root, cwd, name)
138 name = os.path.normpath(name)
139 name = os.path.normpath(name)
139 if auditor is None:
140 if auditor is None:
140 auditor = pathauditor(root)
141 auditor = pathauditor(root)
141 if name != rootsep and name.startswith(rootsep):
142 if name != rootsep and name.startswith(rootsep):
142 name = name[len(rootsep):]
143 name = name[len(rootsep):]
143 auditor(name)
144 auditor(name)
144 return util.pconvert(name)
145 return util.pconvert(name)
145 elif name == root:
146 elif name == root:
146 return ''
147 return ''
147 else:
148 else:
148 # Determine whether `name' is in the hierarchy at or beneath `root',
149 # Determine whether `name' is in the hierarchy at or beneath `root',
149 # by iterating name=dirname(name) until that causes no change (can't
150 # by iterating name=dirname(name) until that causes no change (can't
150 # check name == '/', because that doesn't work on windows). The list
151 # check name == '/', because that doesn't work on windows). The list
151 # `rel' holds the reversed list of components making up the relative
152 # `rel' holds the reversed list of components making up the relative
152 # file name we want.
153 # file name we want.
153 rel = []
154 rel = []
154 while True:
155 while True:
155 try:
156 try:
156 s = util.samefile(name, root)
157 s = util.samefile(name, root)
157 except OSError:
158 except OSError:
158 s = False
159 s = False
159 if s:
160 if s:
160 if not rel:
161 if not rel:
161 # name was actually the same as root (maybe a symlink)
162 # name was actually the same as root (maybe a symlink)
162 return ''
163 return ''
163 rel.reverse()
164 rel.reverse()
164 name = os.path.join(*rel)
165 name = os.path.join(*rel)
165 auditor(name)
166 auditor(name)
166 return util.pconvert(name)
167 return util.pconvert(name)
167 dirname, basename = util.split(name)
168 dirname, basename = util.split(name)
168 rel.append(basename)
169 rel.append(basename)
169 if dirname == name:
170 if dirname == name:
170 break
171 break
171 name = dirname
172 name = dirname
172
173
173 # A common mistake is to use -R, but specify a file relative to the repo
174 # A common mistake is to use -R, but specify a file relative to the repo
174 # instead of cwd. Detect that case, and provide a hint to the user.
175 # instead of cwd. Detect that case, and provide a hint to the user.
175 hint = None
176 hint = None
176 try:
177 try:
177 if cwd != root:
178 if cwd != root:
178 canonpath(root, root, myname, auditor)
179 canonpath(root, root, myname, auditor)
179 hint = (_("consider using '--cwd %s'")
180 hint = (_("consider using '--cwd %s'")
180 % os.path.relpath(root, cwd))
181 % os.path.relpath(root, cwd))
181 except error.Abort:
182 except error.Abort:
182 pass
183 pass
183
184
184 raise error.Abort(_("%s not under root '%s'") % (myname, root),
185 raise error.Abort(_("%s not under root '%s'") % (myname, root),
185 hint=hint)
186 hint=hint)
186
187
187 def normasprefix(path):
188 def normasprefix(path):
188 '''normalize the specified path as path prefix
189 '''normalize the specified path as path prefix
189
190
190 Returned value can be used safely for "p.startswith(prefix)",
191 Returned value can be used safely for "p.startswith(prefix)",
191 "p[len(prefix):]", and so on.
192 "p[len(prefix):]", and so on.
192
193
193 For efficiency, this expects "path" argument to be already
194 For efficiency, this expects "path" argument to be already
194 normalized by "os.path.normpath", "os.path.realpath", and so on.
195 normalized by "os.path.normpath", "os.path.realpath", and so on.
195
196
196 See also issue3033 for detail about need of this function.
197 See also issue3033 for detail about need of this function.
197
198
198 >>> normasprefix('/foo/bar').replace(os.sep, '/')
199 >>> normasprefix('/foo/bar').replace(os.sep, '/')
199 '/foo/bar/'
200 '/foo/bar/'
200 >>> normasprefix('/').replace(os.sep, '/')
201 >>> normasprefix('/').replace(os.sep, '/')
201 '/'
202 '/'
202 '''
203 '''
203 d, p = os.path.splitdrive(path)
204 d, p = os.path.splitdrive(path)
204 if len(p) != len(os.sep):
205 if len(p) != len(os.sep):
205 return path + os.sep
206 return path + os.sep
206 else:
207 else:
207 return path
208 return path
208
209
209 # forward two methods from posixpath that do what we need, but we'd
210 # forward two methods from posixpath that do what we need, but we'd
210 # rather not let our internals know that we're thinking in posix terms
211 # rather not let our internals know that we're thinking in posix terms
211 # - instead we'll let them be oblivious.
212 # - instead we'll let them be oblivious.
212 join = posixpath.join
213 join = posixpath.join
213 dirname = posixpath.dirname
214 dirname = posixpath.dirname
General Comments 0
You need to be logged in to leave comments. Login now