|
|
923a60 |
From ddad59f50e67a5e36cd9c40e774d28240a6a7c0c Mon Sep 17 00:00:00 2001
|
|
|
923a60 |
From: Filipe Brandenburger <filbranden@google.com>
|
|
|
923a60 |
Date: Wed, 11 Nov 2015 09:24:34 -0800
|
|
|
923a60 |
Subject: [PATCH] test-execute: Clarify interaction of PassEnvironment= and
|
|
|
923a60 |
MANAGER_USER
|
|
|
923a60 |
|
|
|
923a60 |
@evverx brought up that test-execute runs under MANAGER_USER which
|
|
|
923a60 |
forwards all its environment variables to the services. It turns out it
|
|
|
923a60 |
only forwards those that were in the environment at the time of manager
|
|
|
923a60 |
creation, so this test was still working.
|
|
|
923a60 |
|
|
|
923a60 |
It was still possible to attack it by running something like:
|
|
|
923a60 |
$ sudo VAR1=a VAR2=b VAR3=c ./test-execute
|
|
|
923a60 |
|
|
|
923a60 |
Prevent that attack by unsetting the three variables explicitly before
|
|
|
923a60 |
creating the manager for the test case.
|
|
|
923a60 |
|
|
|
923a60 |
Also add comments explaining the interactions with MANAGER_USER and,
|
|
|
923a60 |
while it has some caveats, this tests are still valid in that context.
|
|
|
923a60 |
|
|
|
923a60 |
Tested by checking that the test running with the variables set from the
|
|
|
923a60 |
external environment will still pass.
|
|
|
923a60 |
|
|
|
923a60 |
(cherry picked from commit e1abca2ee42e5938ee1f2542c3eba9e70edb0be2)
|
|
|
923a60 |
|
|
|
923a60 |
Related: #1426214
|
|
|
923a60 |
---
|
|
|
923a60 |
src/test/test-execute.c | 21 +++++++++++++++++++++
|
|
|
923a60 |
1 file changed, 21 insertions(+)
|
|
|
923a60 |
|
|
|
923a60 |
diff --git a/src/test/test-execute.c b/src/test/test-execute.c
|
|
|
923a60 |
index 8def1946dc..6e5567c3ed 100644
|
|
|
923a60 |
--- a/src/test/test-execute.c
|
|
|
923a60 |
+++ b/src/test/test-execute.c
|
|
|
923a60 |
@@ -143,6 +143,17 @@ static void test_exec_environment(Manager *m) {
|
|
|
923a60 |
}
|
|
|
923a60 |
|
|
|
923a60 |
static void test_exec_passenvironment(Manager *m) {
|
|
|
923a60 |
+ /* test-execute runs under MANAGER_USER which, by default, forwards all
|
|
|
923a60 |
+ * variables present in the environment, but only those that are
|
|
|
923a60 |
+ * present _at the time it is created_!
|
|
|
923a60 |
+ *
|
|
|
923a60 |
+ * So these PassEnvironment checks are still expected to work, since we
|
|
|
923a60 |
+ * are ensuring the variables are not present at manager creation (they
|
|
|
923a60 |
+ * are unset explicitly in main) and are only set here.
|
|
|
923a60 |
+ *
|
|
|
923a60 |
+ * This is still a good approximation of how a test for MANAGER_SYSTEM
|
|
|
923a60 |
+ * would work.
|
|
|
923a60 |
+ */
|
|
|
923a60 |
assert_se(setenv("VAR1", "word1 word2", 1) == 0);
|
|
|
923a60 |
assert_se(setenv("VAR2", "word3", 1) == 0);
|
|
|
923a60 |
assert_se(setenv("VAR3", "$word 5 6", 1) == 0);
|
|
|
923a60 |
@@ -199,6 +210,16 @@ int main(int argc, char *argv[]) {
|
|
|
923a60 |
assert_se(setenv("XDG_RUNTIME_DIR", "/tmp/", 1) == 0);
|
|
|
923a60 |
assert_se(set_unit_path(TEST_DIR ":") >= 0);
|
|
|
923a60 |
|
|
|
923a60 |
+ /* Unset VAR1, VAR2 and VAR3 which are used in the PassEnvironment test
|
|
|
923a60 |
+ * cases, otherwise (and if they are present in the environment),
|
|
|
923a60 |
+ * `manager_default_environment` will copy them into the default
|
|
|
923a60 |
+ * environment which is passed to each created job, which will make the
|
|
|
923a60 |
+ * tests that expect those not to be present to fail.
|
|
|
923a60 |
+ */
|
|
|
923a60 |
+ assert_se(unsetenv("VAR1") == 0);
|
|
|
923a60 |
+ assert_se(unsetenv("VAR2") == 0);
|
|
|
923a60 |
+ assert_se(unsetenv("VAR3") == 0);
|
|
|
923a60 |
+
|
|
|
923a60 |
r = manager_new(SYSTEMD_USER, true, &m);
|
|
|
923a60 |
if (IN_SET(r, -EPERM, -EACCES, -EADDRINUSE, -EHOSTDOWN, -ENOENT)) {
|
|
|
923a60 |
printf("Skipping test: manager_new: %s", strerror(-r));
|