Merge pull request #4483 from poettering/exec-order

more seccomp fixes, and change of order of selinux/aa/smack and seccomp application on exec

4 files changed, 88 insertions, 35 deletions

diff --git a/src/core/execute.c b/src/core/execute.c
index a1bd0c1238..3f053602b5 100644
--- a/src/core/execute.c
+++ b/src/core/execute.c
@@ -2540,12 +2540,6 @@ static int exec_child(
         (void) umask(context->umask);
 
         if ((params->flags & EXEC_APPLY_PERMISSIONS) && !command->privileged) {
-                r = setup_smack(context, command);
-                if (r < 0) {
-                        *exit_status = EXIT_SMACK_PROCESS_LABEL;
-                        return r;
-                }
-
                 if (context->pam_name && username) {
                         r = setup_pam(context->pam_name, username, uid, gid, context->tty_path, &accum_env, fds, n_fds);
                         if (r < 0) {
@@ -2695,6 +2689,41 @@ static int exec_child(
                         }
                 }
 
+                /* Apply the MAC contexts late, but before seccomp syscall filtering, as those should really be last to
+                 * influence our own codepaths as little as possible. Moreover, applying MAC contexts usually requires
+                 * syscalls that are subject to seccomp filtering, hence should probably be applied before the syscalls
+                 * are restricted. */
+
+#ifdef HAVE_SELINUX
+                if (mac_selinux_use()) {
+                        char *exec_context = mac_selinux_context_net ?: context->selinux_context;
+
+                        if (exec_context) {
+                                r = setexeccon(exec_context);
+                                if (r < 0) {
+                                        *exit_status = EXIT_SELINUX_CONTEXT;
+                                        return r;
+                                }
+                        }
+                }
+#endif
+
+                r = setup_smack(context, command);
+                if (r < 0) {
+                        *exit_status = EXIT_SMACK_PROCESS_LABEL;
+                        return r;
+                }
+
+#ifdef HAVE_APPARMOR
+                if (context->apparmor_profile && mac_apparmor_use()) {
+                        r = aa_change_onexec(context->apparmor_profile);
+                        if (r < 0 && !context->apparmor_profile_ignore) {
+                                *exit_status = EXIT_APPARMOR_PROFILE;
+                                return -errno;
+                        }
+                }
+#endif
+
                 /* PR_GET_SECUREBITS is not privileged, while
                  * PR_SET_SECUREBITS is. So to suppress
                  * potential EPERMs we'll try not to call
@@ -2760,6 +2789,8 @@ static int exec_child(
                         }
                 }
 
+                /* This really should remain the last step before the execve(), to make sure our own code is unaffected
+                 * by the filter as little as possible. */
                 if (context_has_syscall_filters(context)) {
                         r = apply_seccomp(unit, context);
                         if (r < 0) {
@@ -2768,30 +2799,6 @@ static int exec_child(
                         }
                 }
 #endif
-
-#ifdef HAVE_SELINUX
-                if (mac_selinux_use()) {
-                        char *exec_context = mac_selinux_context_net ?: context->selinux_context;
-
-                        if (exec_context) {
-                                r = setexeccon(exec_context);
-                                if (r < 0) {
-                                        *exit_status = EXIT_SELINUX_CONTEXT;
-                                        return r;
-                                }
-                        }
-                }
-#endif
-
-#ifdef HAVE_APPARMOR
-                if (context->apparmor_profile && mac_apparmor_use()) {
-                        r = aa_change_onexec(context->apparmor_profile);
-                        if (r < 0 && !context->apparmor_profile_ignore) {
-                                *exit_status = EXIT_APPARMOR_PROFILE;
-                                return -errno;
-                        }
-                }
-#endif
         }
 
         final_argv = replace_env_argv(argv, accum_env);
@@ -3613,7 +3620,8 @@ char *exec_command_line(char **argv) {
         STRV_FOREACH(a, argv)
                 k += strlen(*a)+3;
 
-        if (!(n = new(char, k)))
+        n = new(char, k);
+        if (!n)
                 return NULL;
 
         p = n;
diff --git a/src/shared/seccomp-util.c b/src/shared/seccomp-util.c
index 1cbbb9d757..c9b24f1065 100644
--- a/src/shared/seccomp-util.c
+++ b/src/shared/seccomp-util.c
@@ -217,6 +217,24 @@ bool is_seccomp_available(void) {
 }
 
 const SyscallFilterSet syscall_filter_sets[_SYSCALL_FILTER_SET_MAX] = {
+        [SYSCALL_FILTER_SET_BASIC_IO] = {
+                /* Basic IO */
+                .name = "@basic-io",
+                .value =
+                "close\0"
+                "dup2\0"
+                "dup3\0"
+                "dup\0"
+                "lseek\0"
+                "pread64\0"
+                "preadv\0"
+                "pwrite64\0"
+                "pwritev\0"
+                "read\0"
+                "readv\0"
+                "write\0"
+                "writev\0"
+        },
         [SYSCALL_FILTER_SET_CLOCK] = {
                 /* Clock */
                 .name = "@clock",
@@ -253,15 +271,22 @@ const SyscallFilterSet syscall_filter_sets[_SYSCALL_FILTER_SET_MAX] = {
                 "sys_debug_setcontext\0"
         },
         [SYSCALL_FILTER_SET_DEFAULT] = {
-                /* Default list */
+                /* Default list: the most basic of operations */
                 .name = "@default",
                 .value =
+                "clock_getres\0"
+                "clock_gettime\0"
+                "clock_nanosleep\0"
                 "execve\0"
                 "exit\0"
                 "exit_group\0"
                 "getrlimit\0"      /* make sure processes can query stack size and such */
+                "gettimeofday\0"
+                "nanosleep\0"
+                "pause\0"
                 "rt_sigreturn\0"
                 "sigreturn\0"
+                "time\0"
         },
         [SYSCALL_FILTER_SET_IO_EVENT] = {
                 /* Event loop use */
@@ -283,9 +308,10 @@ const SyscallFilterSet syscall_filter_sets[_SYSCALL_FILTER_SET_MAX] = {
                 "select\0"
         },
         [SYSCALL_FILTER_SET_IPC] = {
-                /* Message queues, SYSV IPC or other IPC: unusual */
+                /* Message queues, SYSV IPC or other IPC */
                 .name = "@ipc",
                 .value = "ipc\0"
+                "memfd_create\0"
                 "mq_getsetattr\0"
                 "mq_notify\0"
                 "mq_open\0"
@@ -296,6 +322,8 @@ const SyscallFilterSet syscall_filter_sets[_SYSCALL_FILTER_SET_MAX] = {
                 "msgget\0"
                 "msgrcv\0"
                 "msgsnd\0"
+                "pipe2\0"
+                "pipe\0"
                 "process_vm_readv\0"
                 "process_vm_writev\0"
                 "semctl\0"
@@ -436,7 +464,6 @@ const SyscallFilterSet syscall_filter_sets[_SYSCALL_FILTER_SET_MAX] = {
                 .value =
                 "arch_prctl\0"
                 "clone\0"
-                "execve\0"
                 "execveat\0"
                 "fork\0"
                 "kill\0"
@@ -463,6 +490,22 @@ const SyscallFilterSet syscall_filter_sets[_SYSCALL_FILTER_SET_MAX] = {
                 "s390_pci_mmio_write\0"
 #endif
         },
+        [SYSCALL_FILTER_SET_RESOURCES] = {
+                /* Alter resource settings */
+                .name = "@resources",
+                .value =
+                "sched_setparam\0"
+                "sched_setscheduler\0"
+                "sched_setaffinity\0"
+                "setpriority\0"
+                "setrlimit\0"
+                "set_mempolicy\0"
+                "migrate_pages\0"
+                "move_pages\0"
+                "mbind\0"
+                "sched_setattr\0"
+                "prlimit64\0"
+        },
 };
 
 const SyscallFilterSet *syscall_filter_set_find(const char *name) {
diff --git a/src/shared/seccomp-util.h b/src/shared/seccomp-util.h
index 8050fc6fbf..8e209efef2 100644
--- a/src/shared/seccomp-util.h
+++ b/src/shared/seccomp-util.h
@@ -38,6 +38,7 @@ typedef struct SyscallFilterSet {
 } SyscallFilterSet;
 
 enum {
+        SYSCALL_FILTER_SET_BASIC_IO,
         SYSCALL_FILTER_SET_CLOCK,
         SYSCALL_FILTER_SET_CPU_EMULATION,
         SYSCALL_FILTER_SET_DEBUG,
@@ -52,6 +53,7 @@ enum {
         SYSCALL_FILTER_SET_PRIVILEGED,
         SYSCALL_FILTER_SET_PROCESS,
         SYSCALL_FILTER_SET_RAW_IO,
+        SYSCALL_FILTER_SET_RESOURCES,
         _SYSCALL_FILTER_SET_MAX
 };
 
diff --git a/src/test/test-unit-file.c b/src/test/test-unit-file.c
index 7ef087a2e3..12f48bf435 100644
--- a/src/test/test-unit-file.c
+++ b/src/test/test-unit-file.c
@@ -589,7 +589,7 @@ static void test_install_printf(void) {
         assert_se(specifier_machine_id('m', NULL, NULL, &mid) >= 0 && mid);
         assert_se(specifier_boot_id('b', NULL, NULL, &bid) >= 0 && bid);
         assert_se((host = gethostname_malloc()));
-        assert_se((user = getusername_malloc()));
+        assert_se((user = uid_to_name(getuid())));
         assert_se(asprintf(&uid, UID_FMT, getuid()) >= 0);
 
 #define expect(src, pattern, result)                                    \