summaryrefslogtreecommitdiffstats
diff options
context:
space:
mode:
authorJagadeesh Krishnanjanappa <jkrishnanjanappa@mvista.com>2018-08-22 17:11:51 +0530
committerRichard Purdie <richard.purdie@linuxfoundation.org>2018-08-29 15:23:51 +0100
commit52fc5763c6cfc06fe7ca533814815c07a6e213f6 (patch)
treeb0815e7019e60800c1ee47d9982cc654ff59757b
parentdfe237a93b921ed7dd19222b4f6526498c2e7133 (diff)
downloadpoky-52fc5763c6cfc06fe7ca533814815c07a6e213f6.tar.gz
git: CVE-2018-11235
submodule-config: verify submodule names as paths Submodule "names" come from the untrusted .gitmodules file, but we blindly append them to $GIT_DIR/modules to create our on-disk repo paths. This means you can do bad things by putting "../" into the name (among other things). Let's sanity-check these names to avoid building a path that can be exploited. There are two main decisions: 1. What should the allowed syntax be? It's tempting to reuse verify_path(), since submodule names typically come from in-repo paths. But there are two reasons not to: a. It's technically more strict than what we need, as we really care only about breaking out of the $GIT_DIR/modules/ hierarchy. E.g., having a submodule named "foo/.git" isn't actually dangerous, and it's possible that somebody has manually given such a funny name. b. Since we'll eventually use this checking logic in fsck to prevent downstream repositories, it should be consistent across platforms. Because verify_path() relies on is_dir_sep(), it wouldn't block "foo\..\bar" on a non-Windows machine. 2. Where should we enforce it? These days most of the .gitmodules reads go through submodule-config.c, so I've put it there in the reading step. That should cover all of the C code. We also construct the name for "git submodule add" inside the git-submodule.sh script. This is probably not a big deal for security since the name is coming from the user anyway, but it would be polite to remind them if the name they pick is invalid (and we need to expose the name-checker to the shell anyway for our test scripts). This patch issues a warning when reading .gitmodules and just ignores the related config entry completely. This will generally end up producing a sensible error, as it works the same as a .gitmodules file which is missing a submodule entry (so "submodule update" will barf, but "git clone --recurse-submodules" will print an error but not abort the clone. There is one minor oddity, which is that we print the warning once per malformed config key (since that's how the config subsystem gives us the entries). So in the new test, for example, the user would see three warnings. That's OK, since the intent is that this case should never come up outside of malicious repositories (and then it might even benefit the user to see the message multiple times). Credit for finding this vulnerability and the proof of concept from which the test script was adapted goes to Etienne Stalmans. Affects: git < 2.13.7 and git < 2.14.4 and git < 2.15.2 and git < 2.16.4 and git < 2.17.1 (From OE-Core rev: 229bb7cd70c79944d54696d50f4f34df85a5804a) Signed-off-by: Jagadeesh Krishnanjanappa <jkrishnanjanappa@mvista.com> Signed-off-by: Armin Kuster <akuster808@gmail.com> Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
-rw-r--r--meta/recipes-devtools/git/files/CVE-2018-11235.patch288
-rw-r--r--meta/recipes-devtools/git/git.inc3
2 files changed, 290 insertions, 1 deletions
diff --git a/meta/recipes-devtools/git/files/CVE-2018-11235.patch b/meta/recipes-devtools/git/files/CVE-2018-11235.patch
new file mode 100644
index 0000000000..c272eac8d3
--- /dev/null
+++ b/meta/recipes-devtools/git/files/CVE-2018-11235.patch
@@ -0,0 +1,288 @@
1From 0383bbb9015898cbc79abd7b64316484d7713b44 Mon Sep 17 00:00:00 2001
2From: Jeff King <peff@peff.net>
3Date: Mon, 30 Apr 2018 03:25:25 -0400
4Subject: [PATCH] submodule-config: verify submodule names as paths
5
6Submodule "names" come from the untrusted .gitmodules file,
7but we blindly append them to $GIT_DIR/modules to create our
8on-disk repo paths. This means you can do bad things by
9putting "../" into the name (among other things).
10
11Let's sanity-check these names to avoid building a path that
12can be exploited. There are two main decisions:
13
14 1. What should the allowed syntax be?
15
16 It's tempting to reuse verify_path(), since submodule
17 names typically come from in-repo paths. But there are
18 two reasons not to:
19
20 a. It's technically more strict than what we need, as
21 we really care only about breaking out of the
22 $GIT_DIR/modules/ hierarchy. E.g., having a
23 submodule named "foo/.git" isn't actually
24 dangerous, and it's possible that somebody has
25 manually given such a funny name.
26
27 b. Since we'll eventually use this checking logic in
28 fsck to prevent downstream repositories, it should
29 be consistent across platforms. Because
30 verify_path() relies on is_dir_sep(), it wouldn't
31 block "foo\..\bar" on a non-Windows machine.
32
33 2. Where should we enforce it? These days most of the
34 .gitmodules reads go through submodule-config.c, so
35 I've put it there in the reading step. That should
36 cover all of the C code.
37
38 We also construct the name for "git submodule add"
39 inside the git-submodule.sh script. This is probably
40 not a big deal for security since the name is coming
41 from the user anyway, but it would be polite to remind
42 them if the name they pick is invalid (and we need to
43 expose the name-checker to the shell anyway for our
44 test scripts).
45
46 This patch issues a warning when reading .gitmodules
47 and just ignores the related config entry completely.
48 This will generally end up producing a sensible error,
49 as it works the same as a .gitmodules file which is
50 missing a submodule entry (so "submodule update" will
51 barf, but "git clone --recurse-submodules" will print
52 an error but not abort the clone.
53
54 There is one minor oddity, which is that we print the
55 warning once per malformed config key (since that's how
56 the config subsystem gives us the entries). So in the
57 new test, for example, the user would see three
58 warnings. That's OK, since the intent is that this case
59 should never come up outside of malicious repositories
60 (and then it might even benefit the user to see the
61 message multiple times).
62
63Credit for finding this vulnerability and the proof of
64concept from which the test script was adapted goes to
65Etienne Stalmans.
66
67CVE: CVE-2018-11235
68Upstream-Status: Backport [https://github.com/gitster/git/commit/0383bbb9015898cbc79abd7b64316484d7713b44#diff-1772b951776d1647ca31a2256f7fe88f]
69
70Signed-off-by: Jeff King <peff@peff.net>
71Signed-off-by: Jagadeesh Krishnanjanappa <jkrishnanjanappa@mvista.com>
72---
73 builtin/submodule--helper.c | 24 ++++++++++++++
74 git-submodule.sh | 5 +++
75 submodule-config.c | 31 ++++++++++++++++++
76 submodule-config.h | 7 +++++
77 t/t7415-submodule-names.sh | 76 +++++++++++++++++++++++++++++++++++++++++++++
78 5 files changed, 143 insertions(+)
79 create mode 100755 t/t7415-submodule-names.sh
80
81diff --git a/builtin/submodule--helper.c b/builtin/submodule--helper.c
82index cbb17a902..b4b4d29d8 100644
83--- a/builtin/submodule--helper.c
84+++ b/builtin/submodule--helper.c
85@@ -1480,6 +1480,29 @@ static int is_active(int argc, const cha
86 return !is_submodule_active(the_repository, argv[1]);
87 }
88
89+/*
90+ * Exit non-zero if any of the submodule names given on the command line is
91+ * invalid. If no names are given, filter stdin to print only valid names
92+ * (which is primarily intended for testing).
93+ */
94+static int check_name(int argc, const char **argv, const char *prefix)
95+{
96+ if (argc > 1) {
97+ while (*++argv) {
98+ if (check_submodule_name(*argv) < 0)
99+ return 1;
100+ }
101+ } else {
102+ struct strbuf buf = STRBUF_INIT;
103+ while (strbuf_getline(&buf, stdin) != EOF) {
104+ if (!check_submodule_name(buf.buf))
105+ printf("%s\n", buf.buf);
106+ }
107+ strbuf_release(&buf);
108+ }
109+ return 0;
110+}
111+
112 #define SUPPORT_SUPER_PREFIX (1<<0)
113
114 struct cmd_struct {
115@@ -1502,6 +1525,7 @@ static struct cmd_struct commands[] = {
116 {"push-check", push_check, 0},
117 {"absorb-git-dirs", absorb_git_dirs, SUPPORT_SUPER_PREFIX},
118 {"is-active", is_active, 0},
119+ {"check-name", check_name, 0},
120 };
121
122 int cmd_submodule__helper(int argc, const char **argv, const char *prefix)
123diff --git a/git-submodule.sh b/git-submodule.sh
124index c0d0e9a4c..92750b9e2 100755
125--- a/git-submodule.sh
126+++ b/git-submodule.sh
127@@ -229,6 +229,11 @@ Use -f if you really want to add it." >&
128 sm_name="$sm_path"
129 fi
130
131+ if ! git submodule--helper check-name "$sm_name"
132+ then
133+ die "$(eval_gettext "'$sm_name' is not a valid submodule name")"
134+ fi
135+
136 # perhaps the path exists and is already a git repo, else clone it
137 if test -e "$sm_path"
138 then
139diff --git a/submodule-config.c b/submodule-config.c
140index 4f58491dd..de54351c6 100644
141--- a/submodule-config.c
142+++ b/submodule-config.c
143@@ -190,6 +190,31 @@ static struct submodule *cache_lookup_na
144 return NULL;
145 }
146
147+int check_submodule_name(const char *name)
148+{
149+ /* Disallow empty names */
150+ if (!*name)
151+ return -1;
152+
153+ /*
154+ * Look for '..' as a path component. Check both '/' and '\\' as
155+ * separators rather than is_dir_sep(), because we want the name rules
156+ * to be consistent across platforms.
157+ */
158+ goto in_component; /* always start inside component */
159+ while (*name) {
160+ char c = *name++;
161+ if (c == '/' || c == '\\') {
162+in_component:
163+ if (name[0] == '.' && name[1] == '.' &&
164+ (!name[2] || name[2] == '/' || name[2] == '\\'))
165+ return -1;
166+ }
167+ }
168+
169+ return 0;
170+}
171+
172 static int name_and_item_from_var(const char *var, struct strbuf *name,
173 struct strbuf *item)
174 {
175@@ -201,6 +226,12 @@ static int name_and_item_from_var(const
176 return 0;
177
178 strbuf_add(name, subsection, subsection_len);
179+ if (check_submodule_name(name->buf) < 0) {
180+ warning(_("ignoring suspicious submodule name: %s"), name->buf);
181+ strbuf_release(name);
182+ return 0;
183+ }
184+
185 strbuf_addstr(item, key);
186
187 return 1;
188diff --git a/submodule-config.h b/submodule-config.h
189index d434ecdb4..103cc79dd 100644
190--- a/submodule-config.h
191+++ b/submodule-config.h
192@@ -48,4 +48,11 @@ extern const struct submodule *submodule
193 const char *key);
194 extern void submodule_free(void);
195
196+/*
197+ * Returns 0 if the name is syntactically acceptable as a submodule "name"
198+ * (e.g., that may be found in the subsection of a .gitmodules file) and -1
199+ * otherwise.
200+ */
201+int check_submodule_name(const char *name);
202+
203 #endif /* SUBMODULE_CONFIG_H */
204diff --git a/t/t7415-submodule-names.sh b/t/t7415-submodule-names.sh
205new file mode 100755
206index 000000000..75fa071c6
207--- /dev/null
208+++ b/t/t7415-submodule-names.sh
209@@ -0,0 +1,76 @@
210+#!/bin/sh
211+
212+test_description='check handling of .. in submodule names
213+
214+Exercise the name-checking function on a variety of names, and then give a
215+real-world setup that confirms we catch this in practice.
216+'
217+. ./test-lib.sh
218+
219+test_expect_success 'check names' '
220+ cat >expect <<-\EOF &&
221+ valid
222+ valid/with/paths
223+ EOF
224+
225+ git submodule--helper check-name >actual <<-\EOF &&
226+ valid
227+ valid/with/paths
228+
229+ ../foo
230+ /../foo
231+ ..\foo
232+ \..\foo
233+ foo/..
234+ foo/../
235+ foo\..
236+ foo\..\
237+ foo/../bar
238+ EOF
239+
240+ test_cmp expect actual
241+'
242+
243+test_expect_success 'create innocent subrepo' '
244+ git init innocent &&
245+ git -C innocent commit --allow-empty -m foo
246+'
247+
248+test_expect_success 'submodule add refuses invalid names' '
249+ test_must_fail \
250+ git submodule add --name ../../modules/evil "$PWD/innocent" evil
251+'
252+
253+test_expect_success 'add evil submodule' '
254+ git submodule add "$PWD/innocent" evil &&
255+
256+ mkdir modules &&
257+ cp -r .git/modules/evil modules &&
258+ write_script modules/evil/hooks/post-checkout <<-\EOF &&
259+ echo >&2 "RUNNING POST CHECKOUT"
260+ EOF
261+
262+ git config -f .gitmodules submodule.evil.update checkout &&
263+ git config -f .gitmodules --rename-section \
264+ submodule.evil submodule.../../modules/evil &&
265+ git add modules &&
266+ git commit -am evil
267+'
268+
269+# This step seems like it shouldn't be necessary, since the payload is
270+# contained entirely in the evil submodule. But due to the vagaries of the
271+# submodule code, checking out the evil module will fail unless ".git/modules"
272+# exists. Adding another submodule (with a name that sorts before "evil") is an
273+# easy way to make sure this is the case in the victim clone.
274+test_expect_success 'add other submodule' '
275+ git submodule add "$PWD/innocent" another-module &&
276+ git add another-module &&
277+ git commit -am another
278+'
279+
280+test_expect_success 'clone evil superproject' '
281+ git clone --recurse-submodules . victim >output 2>&1 &&
282+ ! grep "RUNNING POST CHECKOUT" output
283+'
284+
285+test_done
286--
2872.13.3
288
diff --git a/meta/recipes-devtools/git/git.inc b/meta/recipes-devtools/git/git.inc
index dd9d792c5c..bea23ec783 100644
--- a/meta/recipes-devtools/git/git.inc
+++ b/meta/recipes-devtools/git/git.inc
@@ -7,7 +7,8 @@ DEPENDS = "openssl curl zlib expat"
7PROVIDES_append_class-native = " git-replacement-native" 7PROVIDES_append_class-native = " git-replacement-native"
8 8
9SRC_URI = "${KERNELORG_MIRROR}/software/scm/git/git-${PV}.tar.gz;name=tarball \ 9SRC_URI = "${KERNELORG_MIRROR}/software/scm/git/git-${PV}.tar.gz;name=tarball \
10 ${KERNELORG_MIRROR}/software/scm/git/git-manpages-${PV}.tar.gz;name=manpages" 10 ${KERNELORG_MIRROR}/software/scm/git/git-manpages-${PV}.tar.gz;name=manpages \
11 file://CVE-2018-11235.patch"
11 12
12S = "${WORKDIR}/git-${PV}" 13S = "${WORKDIR}/git-${PV}"
13 14