SymbX/systemd
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

user units: implicitly enable PrivateUsers= when sandboxing options are set

Browse Source 
Enabling these options when not running as root requires a user
namespace, so implicitly enable PrivateUsers=.
This has a side effect as it changes which users are visible to the unit.
However until now these options did not work at all for user units, and
in practice just a handful of user units in Fedora, Debian and Ubuntu
mistakenly used them (and they have been all fixed since).

This fixes the long-standing confusing issue that the user and system
units take the same options but the behaviour is wildly (and sometimes
silently) different depending on which is which, with user units
requiring manually specifiying PrivateUsers= in order for sandboxing
options to actually work and not be silently ignored.
This commit is contained in:
Luca Boccassi
2023-04-13 21:33:48 +01:00
committed by Luca Boccassi
parent ce963a747f
commit 6ef721cbc7
4 changed files with 98 additions and 52 deletions
Download Patch File Download Diff File Expand all files Collapse all files
src/core/execute.c
+46 -3
View File
@@ -4400,6 +4400,44 @@ static void log_command_line(Unit *unit, const char *msg, const char *executable
LOG_UNIT_INVOCATION_ID(unit));
}
static bool exec_context_need_unprivileged_private_users(const ExecContext *context, const Manager *manager) {
assert(context);
assert(manager);
/* These options require PrivateUsers= when used in user units, as we need to be in a user namespace
* to have permission to enable them when not running as root. If we have effective CAP_SYS_ADMIN
* (system manager) then we have privileges and don't need this. */
if (MANAGER_IS_SYSTEM(manager))
return false;
return context->private_users ||
context->private_tmp ||
context->private_devices ||
context->private_network ||
context->network_namespace_path ||
context->private_ipc ||
context->ipc_namespace_path ||
context->private_mounts ||
context->mount_apivfs ||
context->n_bind_mounts > 0 ||
context->n_temporary_filesystems > 0 ||
context->root_directory ||
!strv_isempty(context->extension_directories) ||
context->protect_system != PROTECT_SYSTEM_NO ||
context->protect_home != PROTECT_HOME_NO ||
context->protect_kernel_tunables ||
context->protect_kernel_modules ||
context->protect_kernel_logs ||
context->protect_control_groups ||
context->protect_clock ||
context->protect_hostname ||
!strv_isempty(context->read_write_paths) ||
!strv_isempty(context->read_only_paths) ||
!strv_isempty(context->inaccessible_paths) ||
!strv_isempty(context->exec_paths) ||
!strv_isempty(context->no_exec_paths);
}
static int exec_child(
Unit *unit,
const ExecCommand *command,
@@ -5032,17 +5070,22 @@ static int exec_child(
}
}
if (needs_sandboxing && context->private_users && have_effective_cap(CAP_SYS_ADMIN) <= 0) {
if (needs_sandboxing && exec_context_need_unprivileged_private_users(context, unit->manager)) {
/* If we're unprivileged, set up the user namespace first to enable use of the other namespaces.
* Users with CAP_SYS_ADMIN can set up user namespaces last because they will be able to
* set up the all of the other namespaces (i.e. network, mount, UTS) without a user namespace. */
userns_set_up = true;
r = setup_private_users(saved_uid, saved_gid, uid, gid);
if (r < 0) {
/* If it was requested explicitly and we can't set it up, fail early. Otherwise, continue and let
* the actual requested operations fail (or silently continue). */
if (r < 0 && context->private_users) {
*exit_status = EXIT_USER;
return log_unit_error_errno(unit, r, "Failed to set up user namespacing for unprivileged user: %m");
}
if (r < 0)
log_unit_info_errno(unit, r, "Failed to set up user namespacing for unprivileged user, ignoring: %m");
else
userns_set_up = true;
}
if (exec_needs_network_namespace(context) && runtime && runtime->shared && runtime->shared->netns_storage_socket[0] >= 0) {
src/test/test-execute.c
+29 -30
View File
@@ -401,9 +401,9 @@ static void test_exec_ignoresigpipe(Manager *m) {
static void test_exec_privatetmp(Manager *m) {
assert_se(touch("/tmp/test-exec_privatetmp") >= 0);
test(m, "exec-privatetmp-yes.service", can_unshare ? 0 : EXIT_FAILURE, CLD_EXITED);
test(m, "exec-privatetmp-yes.service", can_unshare ? 0 : MANAGER_IS_SYSTEM(m) ? EXIT_FAILURE : EXIT_NAMESPACE, CLD_EXITED);
test(m, "exec-privatetmp-no.service", 0, CLD_EXITED);
test(m, "exec-privatetmp-disabled-by-prefix.service", can_unshare ? 0 : EXIT_FAILURE, CLD_EXITED);
test(m, "exec-privatetmp-disabled-by-prefix.service", can_unshare ? 0 : MANAGER_IS_SYSTEM(m) ? EXIT_FAILURE : EXIT_NAMESPACE, CLD_EXITED);
unlink("/tmp/test-exec_privatetmp");
}
@@ -420,10 +420,10 @@ static void test_exec_privatedevices(Manager *m) {
return;
}
test(m, "exec-privatedevices-yes.service", can_unshare ? 0 : EXIT_FAILURE, CLD_EXITED);
test(m, "exec-privatedevices-yes.service", can_unshare ? 0 : MANAGER_IS_SYSTEM(m) ? EXIT_FAILURE : EXIT_NAMESPACE, CLD_EXITED);
test(m, "exec-privatedevices-no.service", 0, CLD_EXITED);
test(m, "exec-privatedevices-disabled-by-prefix.service", can_unshare ? 0 : EXIT_FAILURE, CLD_EXITED);
test(m, "exec-privatedevices-yes-with-group.service", can_unshare ? 0 : MANAGER_IS_SYSTEM(m) ? EXIT_FAILURE : EXIT_GROUP, CLD_EXITED);
test(m, "exec-privatedevices-disabled-by-prefix.service", can_unshare ? 0 : MANAGER_IS_SYSTEM(m) ? EXIT_FAILURE : EXIT_NAMESPACE, CLD_EXITED);
test(m, "exec-privatedevices-yes-with-group.service", can_unshare ? 0 : MANAGER_IS_SYSTEM(m) ? EXIT_FAILURE : EXIT_NAMESPACE, CLD_EXITED);
/* We use capsh to test if the capabilities are
* properly set, so be sure that it exists */
@@ -433,10 +433,10 @@ static void test_exec_privatedevices(Manager *m) {
return;
}
test(m, "exec-privatedevices-yes-capability-mknod.service", 0, CLD_EXITED);
test(m, "exec-privatedevices-no-capability-mknod.service", MANAGER_IS_SYSTEM(m) ? 0 : EXIT_FAILURE, CLD_EXITED);
test(m, "exec-privatedevices-yes-capability-sys-rawio.service", 0, CLD_EXITED);
test(m, "exec-privatedevices-no-capability-sys-rawio.service", MANAGER_IS_SYSTEM(m) ? 0 : EXIT_FAILURE, CLD_EXITED);
test(m, "exec-privatedevices-yes-capability-mknod.service", can_unshare || MANAGER_IS_SYSTEM(m) ? 0 : EXIT_NAMESPACE, CLD_EXITED);
test(m, "exec-privatedevices-no-capability-mknod.service", 0, CLD_EXITED);
test(m, "exec-privatedevices-yes-capability-sys-rawio.service", MANAGER_IS_SYSTEM(m) ? 0 : EXIT_NAMESPACE, CLD_EXITED);
test(m, "exec-privatedevices-no-capability-sys-rawio.service", 0, CLD_EXITED);
}
static void test_exec_protecthome(Manager *m) {
@@ -466,23 +466,23 @@ static void test_exec_protectkernelmodules(Manager *m) {
return;
}
test(m, "exec-protectkernelmodules-no-capabilities.service", MANAGER_IS_SYSTEM(m) ? 0 : EXIT_FAILURE, CLD_EXITED);
test(m, "exec-protectkernelmodules-yes-capabilities.service", 0, CLD_EXITED);
test(m, "exec-protectkernelmodules-yes-mount-propagation.service", can_unshare ? 0 : EXIT_FAILURE, CLD_EXITED);
test(m, "exec-protectkernelmodules-no-capabilities.service", 0, CLD_EXITED);
test(m, "exec-protectkernelmodules-yes-capabilities.service", MANAGER_IS_SYSTEM(m) ? 0 : EXIT_NAMESPACE, CLD_EXITED);
test(m, "exec-protectkernelmodules-yes-mount-propagation.service", can_unshare ? 0 : MANAGER_IS_SYSTEM(m) ? EXIT_FAILURE : EXIT_NAMESPACE, CLD_EXITED);
}
static void test_exec_readonlypaths(Manager *m) {
test(m, "exec-readonlypaths-simple.service", can_unshare ? 0 : EXIT_FAILURE, CLD_EXITED);
test(m, "exec-readonlypaths-simple.service", can_unshare ? 0 : MANAGER_IS_SYSTEM(m) ? EXIT_FAILURE : EXIT_NAMESPACE, CLD_EXITED);
if (path_is_read_only_fs("/var") > 0) {
log_notice("Directory /var is readonly, skipping remaining tests in %s", __func__);
return;
}
test(m, "exec-readonlypaths.service", can_unshare ? 0 : EXIT_FAILURE, CLD_EXITED);
test(m, "exec-readonlypaths.service", can_unshare ? 0 : MANAGER_IS_SYSTEM(m) ? EXIT_FAILURE : EXIT_NAMESPACE, CLD_EXITED);
test(m, "exec-readonlypaths-with-bindpaths.service", can_unshare ? 0 : EXIT_NAMESPACE, CLD_EXITED);
test(m, "exec-readonlypaths-mount-propagation.service", can_unshare ? 0 : EXIT_FAILURE, CLD_EXITED);
test(m, "exec-readonlypaths-mount-propagation.service", can_unshare ? 0 : MANAGER_IS_SYSTEM(m) ? EXIT_FAILURE : EXIT_NAMESPACE, CLD_EXITED);
}
static void test_exec_readwritepaths(Manager *m) {
@@ -492,7 +492,7 @@ static void test_exec_readwritepaths(Manager *m) {
return;
}
test(m, "exec-readwritepaths-mount-propagation.service", can_unshare ? 0 : EXIT_FAILURE, CLD_EXITED);
test(m, "exec-readwritepaths-mount-propagation.service", can_unshare ? 0 : MANAGER_IS_SYSTEM(m) ? EXIT_FAILURE : EXIT_NAMESPACE, CLD_EXITED);
}
static void test_exec_inaccessiblepaths(Manager *m) {
@@ -502,14 +502,14 @@ static void test_exec_inaccessiblepaths(Manager *m) {
return;
}
test(m, "exec-inaccessiblepaths-sys.service", can_unshare ? 0 : EXIT_FAILURE, CLD_EXITED);
test(m, "exec-inaccessiblepaths-sys.service", can_unshare ? 0 : MANAGER_IS_SYSTEM(m) ? EXIT_FAILURE : EXIT_NAMESPACE, CLD_EXITED);
if (path_is_read_only_fs("/") > 0) {
log_notice("Root directory is readonly, skipping remaining tests in %s", __func__);
return;
}
test(m, "exec-inaccessiblepaths-mount-propagation.service", can_unshare ? 0 : EXIT_FAILURE, CLD_EXITED);
test(m, "exec-inaccessiblepaths-mount-propagation.service", can_unshare ? 0 : MANAGER_IS_SYSTEM(m) ? EXIT_FAILURE : EXIT_NAMESPACE, CLD_EXITED);
}
static int on_spawn_io(sd_event_source *s, int fd, uint32_t revents, void *userdata) {
@@ -687,14 +687,14 @@ static void test_exec_mount_apivfs(Manager *m) {
assert_se(mkdir_p("/tmp/test-exec-mount-apivfs-no/root", 0755) >= 0);
test(m, "exec-mount-apivfs-no.service", can_unshare ? 0 : EXIT_NAMESPACE, CLD_EXITED);
test(m, "exec-mount-apivfs-no.service", can_unshare || !MANAGER_IS_SYSTEM(m) ? 0 : EXIT_NAMESPACE, CLD_EXITED);
(void) rm_rf("/tmp/test-exec-mount-apivfs-no/root", REMOVE_ROOT|REMOVE_PHYSICAL);
}
static void test_exec_noexecpaths(Manager *m) {
test(m, "exec-noexecpaths-simple.service", can_unshare ? 0 : EXIT_FAILURE, CLD_EXITED);
test(m, "exec-noexecpaths-simple.service", can_unshare ? 0 : MANAGER_IS_SYSTEM(m) ? EXIT_FAILURE : EXIT_NAMESPACE, CLD_EXITED);
}
static void test_exec_temporaryfilesystem(Manager *m) {
@@ -964,8 +964,8 @@ static void test_exec_passenvironment(Manager *m) {
}
static void test_exec_umask(Manager *m) {
test(m, "exec-umask-default.service", 0, CLD_EXITED);
test(m, "exec-umask-0177.service", 0, CLD_EXITED);
test(m, "exec-umask-default.service", can_unshare || MANAGER_IS_SYSTEM(m) ? 0 : EXIT_NAMESPACE, CLD_EXITED);
test(m, "exec-umask-0177.service", can_unshare || MANAGER_IS_SYSTEM(m) ? 0 : EXIT_NAMESPACE, CLD_EXITED);
}
static void test_exec_runtimedirectory(Manager *m) {
@@ -1012,7 +1012,7 @@ static void test_exec_capabilityboundingset(Manager *m) {
}
static void test_exec_basic(Manager *m) {
test(m, "exec-basic.service", 0, CLD_EXITED);
test(m, "exec-basic.service", can_unshare || MANAGER_IS_SYSTEM(m) ? 0 : EXIT_NAMESPACE, CLD_EXITED);
}
static void test_exec_ambientcapabilities(Manager *m) {
@@ -1052,7 +1052,7 @@ static void test_exec_ambientcapabilities(Manager *m) {
}
static void test_exec_privatenetwork(Manager *m) {
int r, status;
int r;
r = find_executable("ip", NULL);
if (r < 0) {
@@ -1060,9 +1060,8 @@ static void test_exec_privatenetwork(Manager *m) {
return;
}
status = can_unshare ? 0 : MANAGER_IS_SYSTEM(m) ? EXIT_NETWORK : EXIT_FAILURE;
test(m, "exec-privatenetwork-yes-privatemounts-no.service", status, CLD_EXITED);
test(m, "exec-privatenetwork-yes-privatemounts-yes.service", status, CLD_EXITED);
test(m, "exec-privatenetwork-yes-privatemounts-no.service", can_unshare ? 0 : MANAGER_IS_SYSTEM(m) ? EXIT_NETWORK : EXIT_FAILURE, CLD_EXITED);
test(m, "exec-privatenetwork-yes-privatemounts-yes.service", can_unshare ? 0 : MANAGER_IS_SYSTEM(m) ? EXIT_NETWORK : EXIT_NAMESPACE, CLD_EXITED);
}
static void test_exec_networknamespacepath(Manager *m) {
@@ -1075,7 +1074,7 @@ static void test_exec_networknamespacepath(Manager *m) {
}
test(m, "exec-networknamespacepath-privatemounts-no.service", MANAGER_IS_SYSTEM(m) ? EXIT_SUCCESS : EXIT_FAILURE, CLD_EXITED);
test(m, "exec-networknamespacepath-privatemounts-yes.service", can_unshare ? EXIT_SUCCESS : EXIT_FAILURE, CLD_EXITED);
test(m, "exec-networknamespacepath-privatemounts-yes.service", can_unshare ? EXIT_SUCCESS : MANAGER_IS_SYSTEM(m) ? EXIT_FAILURE : EXIT_NAMESPACE, CLD_EXITED);
}
static void test_exec_oomscoreadjust(Manager *m) {
@@ -1105,12 +1104,12 @@ static void test_exec_unsetenvironment(Manager *m) {
}
static void test_exec_specifier(Manager *m) {
test(m, "exec-specifier.service", 0, CLD_EXITED);
test(m, "exec-specifier.service", can_unshare || MANAGER_IS_SYSTEM(m) ? 0 : EXIT_FAILURE, CLD_EXITED);
if (MANAGER_IS_SYSTEM(m))
test(m, "exec-specifier-system.service", 0, CLD_EXITED);
else
test(m, "exec-specifier-user.service", 0, CLD_EXITED);
test(m, "exec-specifier@foo-bar.service", 0, CLD_EXITED);
test(m, "exec-specifier@foo-bar.service", can_unshare || MANAGER_IS_SYSTEM(m) ? 0 : EXIT_FAILURE, CLD_EXITED);
test(m, "exec-specifier-interpolation.service", 0, CLD_EXITED);
}