selftests/seccomp: Add test for unknown SECCOMP_RET kill behavior
authorKees Cook <keescook@chromium.org>
Tue, 8 Sep 2020 19:28:27 +0000 (12:28 -0700)
committerKees Cook <keescook@chromium.org>
Tue, 8 Sep 2020 23:26:07 +0000 (16:26 -0700)
While we were testing for the behavior of unknown seccomp filter return
values, there was no test for how it acted in a thread group. Add a test
in the thread group tests for this.

Reviewed-by: Shuah Khan <skhan@linuxfoundation.org>
Acked-by: Christian Brauner <christian.brauner@ubuntu.com>
Signed-off-by: Kees Cook <keescook@chromium.org>
tools/testing/selftests/seccomp/seccomp_bpf.c

index 7a6d402..bfb3825 100644 (file)
@@ -774,8 +774,15 @@ void *kill_thread(void *data)
        return (void *)SIBLING_EXIT_UNKILLED;
 }
 
+enum kill_t {
+       KILL_THREAD,
+       KILL_PROCESS,
+       RET_UNKNOWN
+};
+
 /* Prepare a thread that will kill itself or both of us. */
-void kill_thread_or_group(struct __test_metadata *_metadata, bool kill_process)
+void kill_thread_or_group(struct __test_metadata *_metadata,
+                         enum kill_t kill_how)
 {
        pthread_t thread;
        void *status;
@@ -791,11 +798,12 @@ void kill_thread_or_group(struct __test_metadata *_metadata, bool kill_process)
                .len = (unsigned short)ARRAY_SIZE(filter_thread),
                .filter = filter_thread,
        };
+       int kill = kill_how == KILL_PROCESS ? SECCOMP_RET_KILL_PROCESS : 0xAAAAAAAAA;
        struct sock_filter filter_process[] = {
                BPF_STMT(BPF_LD|BPF_W|BPF_ABS,
                        offsetof(struct seccomp_data, nr)),
                BPF_JUMP(BPF_JMP|BPF_JEQ|BPF_K, __NR_prctl, 0, 1),
-               BPF_STMT(BPF_RET|BPF_K, SECCOMP_RET_KILL_PROCESS),
+               BPF_STMT(BPF_RET|BPF_K, kill),
                BPF_STMT(BPF_RET|BPF_K, SECCOMP_RET_ALLOW),
        };
        struct sock_fprog prog_process = {
@@ -808,13 +816,15 @@ void kill_thread_or_group(struct __test_metadata *_metadata, bool kill_process)
        }
 
        ASSERT_EQ(0, seccomp(SECCOMP_SET_MODE_FILTER, 0,
-                            kill_process ? &prog_process : &prog_thread));
+                            kill_how == KILL_THREAD ? &prog_thread
+                                                    : &prog_process));
 
        /*
         * Add the KILL_THREAD rule again to make sure that the KILL_PROCESS
         * flag cannot be downgraded by a new filter.
         */
-       ASSERT_EQ(0, seccomp(SECCOMP_SET_MODE_FILTER, 0, &prog_thread));
+       if (kill_how == KILL_PROCESS)
+               ASSERT_EQ(0, seccomp(SECCOMP_SET_MODE_FILTER, 0, &prog_thread));
 
        /* Start a thread that will exit immediately. */
        ASSERT_EQ(0, pthread_create(&thread, NULL, kill_thread, (void *)false));
@@ -842,7 +852,7 @@ TEST(KILL_thread)
        child_pid = fork();
        ASSERT_LE(0, child_pid);
        if (child_pid == 0) {
-               kill_thread_or_group(_metadata, false);
+               kill_thread_or_group(_metadata, KILL_THREAD);
                _exit(38);
        }
 
@@ -861,7 +871,7 @@ TEST(KILL_process)
        child_pid = fork();
        ASSERT_LE(0, child_pid);
        if (child_pid == 0) {
-               kill_thread_or_group(_metadata, true);
+               kill_thread_or_group(_metadata, KILL_PROCESS);
                _exit(38);
        }
 
@@ -872,6 +882,27 @@ TEST(KILL_process)
        ASSERT_EQ(SIGSYS, WTERMSIG(status));
 }
 
+TEST(KILL_unknown)
+{
+       int status;
+       pid_t child_pid;
+
+       child_pid = fork();
+       ASSERT_LE(0, child_pid);
+       if (child_pid == 0) {
+               kill_thread_or_group(_metadata, RET_UNKNOWN);
+               _exit(38);
+       }
+
+       ASSERT_EQ(child_pid, waitpid(child_pid, &status, 0));
+
+       /* If the entire process was killed, we'll see SIGSYS. */
+       EXPECT_TRUE(WIFSIGNALED(status)) {
+               TH_LOG("Unknown SECCOMP_RET is only killing the thread?");
+       }
+       ASSERT_EQ(SIGSYS, WTERMSIG(status));
+}
+
 /* TODO(wad) add 64-bit versus 32-bit arg tests. */
 TEST(arg_out_of_range)
 {