#include <sys/mman.h>
#include <sys/personality.h>
#include <sys/prctl.h>
+#include <sys/shm.h>
#include <sys/socket.h>
#include <sys/stat.h>
+#include <sys/types.h>
#include <sys/un.h>
#include <unistd.h>
#include <utmpx.h>
#include "exit-status.h"
#include "fd-util.h"
#include "fileio.h"
-#include "formats-util.h"
+#include "format-util.h"
#include "fs-util.h"
#include "glob-util.h"
#include "io-util.h"
return 0;
}
-static int setup_confirm_stdio(int *_saved_stdin, int *_saved_stdout) {
+static int setup_confirm_stdio(const char *vc, int *_saved_stdin, int *_saved_stdout) {
_cleanup_close_ int fd = -1, saved_stdin = -1, saved_stdout = -1;
int r;
if (saved_stdout < 0)
return -errno;
- fd = acquire_terminal(
- "/dev/console",
- false,
- false,
- false,
- DEFAULT_CONFIRM_USEC);
+ fd = acquire_terminal(vc, false, false, false, DEFAULT_CONFIRM_USEC);
if (fd < 0)
return fd;
return 0;
}
-_printf_(1, 2) static int write_confirm_message(const char *format, ...) {
+static void write_confirm_error_fd(int err, int fd) {
+ assert(err < 0);
+
+ if (err == -ETIMEDOUT)
+ dprintf(fd, "Confirmation question timed out, assuming positive response.\n");
+ else {
+ errno = -err;
+ dprintf(fd, "Couldn't ask confirmation: %m, assuming positive response.\n");
+ }
+}
+
+static void write_confirm_error(int err, const char *vc) {
_cleanup_close_ int fd = -1;
- va_list ap;
- assert(format);
+ assert(vc);
- fd = open_terminal("/dev/console", O_WRONLY|O_NOCTTY|O_CLOEXEC);
+ fd = open_terminal(vc, O_WRONLY|O_NOCTTY|O_CLOEXEC);
if (fd < 0)
- return fd;
-
- va_start(ap, format);
- vdprintf(fd, format, ap);
- va_end(ap);
+ return;
- return 0;
+ write_confirm_error_fd(err, fd);
}
static int restore_confirm_stdio(int *saved_stdin, int *saved_stdout) {
return r;
}
-static int ask_for_confirmation(char *response, char **argv) {
+enum {
+ CONFIRM_PRETEND_FAILURE = -1,
+ CONFIRM_PRETEND_SUCCESS = 0,
+ CONFIRM_EXECUTE = 1,
+};
+
+static int ask_for_confirmation(const char *vc, Unit *u, const char *cmdline) {
int saved_stdout = -1, saved_stdin = -1, r;
- _cleanup_free_ char *line = NULL;
+ _cleanup_free_ char *e = NULL;
+ char c;
- r = setup_confirm_stdio(&saved_stdin, &saved_stdout);
- if (r < 0)
- return r;
+ /* For any internal errors, assume a positive response. */
+ r = setup_confirm_stdio(vc, &saved_stdin, &saved_stdout);
+ if (r < 0) {
+ write_confirm_error(r, vc);
+ return CONFIRM_EXECUTE;
+ }
- line = exec_command_line(argv);
- if (!line)
- return -ENOMEM;
+ e = ellipsize(cmdline, 60, 100);
+ if (!e) {
+ log_oom();
+ r = CONFIRM_EXECUTE;
+ goto restore_stdio;
+ }
- r = ask_char(response, "yns", "Execute %s? [Yes, No, Skip] ", line);
+ for (;;) {
+ r = ask_char(&c, "yfshiDj", "Execute %s? [y, f, s – h for help] ", e);
+ if (r < 0) {
+ write_confirm_error_fd(r, STDOUT_FILENO);
+ r = CONFIRM_EXECUTE;
+ goto restore_stdio;
+ }
- restore_confirm_stdio(&saved_stdin, &saved_stdout);
+ switch (c) {
+ case 'D':
+ unit_dump(u, stdout, " ");
+ continue; /* ask again */
+ case 'f':
+ printf("Failing execution.\n");
+ r = CONFIRM_PRETEND_FAILURE;
+ break;
+ case 'h':
+ printf(" D - dump, show the state of the unit\n"
+ " f - fail, don't execute the command and pretend it failed\n"
+ " h - help\n"
+ " i - info, show a short summary of the unit\n"
+ " j - jobs, show jobs that are in progress\n"
+ " s - skip, don't execute the command and pretend it succeeded\n"
+ " y - yes, execute the command\n");
+ continue; /* ask again */
+ case 'i':
+ printf(" Description: %s\n"
+ " Unit: %s\n"
+ " Command: %s\n",
+ u->id, u->description, cmdline);
+ continue; /* ask again */
+ case 'j':
+ manager_dump_jobs(u->manager, stdout, " ");
+ continue; /* ask again */
+ case 's':
+ printf("Skipping execution.\n");
+ r = CONFIRM_PRETEND_SUCCESS;
+ break;
+ case 'y':
+ r = CONFIRM_EXECUTE;
+ break;
+ default:
+ assert_not_reached("Unhandled choice");
+ }
+ break;
+ }
+restore_stdio:
+ restore_confirm_stdio(&saved_stdin, &saved_stdout);
return r;
}
return 0;
}
-static int get_fixed_supplementary_groups(const ExecContext *c,
- const char *user,
- const char *group,
- gid_t gid,
- gid_t **supplementary_gids, int *ngids) {
+static int get_supplementary_groups(const ExecContext *c, const char *user,
+ const char *group, gid_t gid,
+ gid_t **supplementary_gids, int *ngids) {
char **i;
int r, k = 0;
int ngroups_max;
assert(c);
- if (!c->supplementary_groups)
- return 0;
-
/*
- * If user is given, then lookup GID and supplementary group list.
- * We avoid NSS lookups for gid=0.
+ * If user is given, then lookup GID and supplementary groups list.
+ * We avoid NSS lookups for gid=0. Also we have to initialize groups
+ * here and as early as possible so we keep the list of supplementary
+ * groups of the caller.
*/
if (user && gid_is_valid(gid) && gid != 0) {
/* First step, initialize groups from /etc/groups */
keep_groups = true;
}
- assert_se((ngroups_max = (int) sysconf(_SC_NGROUPS_MAX)) > 0);
+ if (!c->supplementary_groups)
+ return 0;
+
+ /*
+ * If SupplementaryGroups= was passed then NGROUPS_MAX has to
+ * be positive, otherwise fail.
+ */
+ errno = 0;
+ ngroups_max = (int) sysconf(_SC_NGROUPS_MAX);
+ if (ngroups_max <= 0) {
+ if (errno > 0)
+ return -errno;
+ else
+ return -EOPNOTSUPP; /* For all other values */
+ }
l_gids = new(gid_t, ngroups_max);
if (!l_gids)
#ifdef HAVE_SECCOMP
static bool skip_seccomp_unavailable(const Unit* u, const char* msg) {
- if (!is_seccomp_available()) {
- log_open();
- log_unit_debug(u, "SECCOMP features not detected in the kernel, skipping %s", msg);
- log_close();
- return true;
- }
- return false;
+
+ if (is_seccomp_available())
+ return false;
+
+ log_open();
+ log_unit_debug(u, "SECCOMP features not detected in the kernel, skipping %s", msg);
+ log_close();
+ return true;
}
static int apply_seccomp(const Unit* u, const ExecContext *c) {
uint32_t negative_action, action;
- scmp_filter_ctx *seccomp;
+ scmp_filter_ctx seccomp;
Iterator i;
void *id;
int r;
}
static int apply_address_families(const Unit* u, const ExecContext *c) {
- scmp_filter_ctx *seccomp;
+ scmp_filter_ctx seccomp;
Iterator i;
int r;
if (skip_seccomp_unavailable(u, "RestrictAddressFamilies="))
return 0;
- seccomp = seccomp_init(SCMP_ACT_ALLOW);
- if (!seccomp)
- return -ENOMEM;
-
- r = seccomp_add_secondary_archs(seccomp);
+ r = seccomp_init_conservative(&seccomp, SCMP_ACT_ALLOW);
if (r < 0)
- goto finish;
+ return r;
if (c->address_families_whitelist) {
int af, first = 0, last = 0;
}
}
- r = seccomp_attr_set(seccomp, SCMP_FLTATR_CTL_NNP, 0);
- if (r < 0)
- goto finish;
-
r = seccomp_load(seccomp);
finish:
}
static int apply_memory_deny_write_execute(const Unit* u, const ExecContext *c) {
- scmp_filter_ctx *seccomp;
+ scmp_filter_ctx seccomp;
int r;
assert(c);
if (skip_seccomp_unavailable(u, "MemoryDenyWriteExecute="))
return 0;
- seccomp = seccomp_init(SCMP_ACT_ALLOW);
- if (!seccomp)
- return -ENOMEM;
-
- r = seccomp_add_secondary_archs(seccomp);
+ r = seccomp_init_conservative(&seccomp, SCMP_ACT_ALLOW);
if (r < 0)
- goto finish;
+ return r;
r = seccomp_rule_add(
seccomp,
if (r < 0)
goto finish;
- r = seccomp_attr_set(seccomp, SCMP_FLTATR_CTL_NNP, 0);
+ r = seccomp_rule_add(
+ seccomp,
+ SCMP_ACT_ERRNO(EPERM),
+ SCMP_SYS(shmat),
+ 1,
+ SCMP_A2(SCMP_CMP_MASKED_EQ, SHM_EXEC, SHM_EXEC));
if (r < 0)
goto finish;
SCHED_IDLE,
};
- scmp_filter_ctx *seccomp;
+ scmp_filter_ctx seccomp;
unsigned i;
int r, p, max_policy = 0;
if (skip_seccomp_unavailable(u, "RestrictRealtime="))
return 0;
- seccomp = seccomp_init(SCMP_ACT_ALLOW);
- if (!seccomp)
- return -ENOMEM;
-
- r = seccomp_add_secondary_archs(seccomp);
+ r = seccomp_init_conservative(&seccomp, SCMP_ACT_ALLOW);
if (r < 0)
- goto finish;
+ return r;
/* Determine the highest policy constant we want to allow */
for (i = 0; i < ELEMENTSOF(permitted_policies); i++)
if (r < 0)
goto finish;
- r = seccomp_attr_set(seccomp, SCMP_FLTATR_CTL_NNP, 0);
- if (r < 0)
- goto finish;
-
r = seccomp_load(seccomp);
finish:
return r;
}
-static int apply_protect_sysctl(Unit *u, const ExecContext *c) {
- scmp_filter_ctx *seccomp;
+static int apply_protect_sysctl(const Unit *u, const ExecContext *c) {
+ scmp_filter_ctx seccomp;
int r;
assert(c);
if (skip_seccomp_unavailable(u, "ProtectKernelTunables="))
return 0;
- seccomp = seccomp_init(SCMP_ACT_ALLOW);
- if (!seccomp)
- return -ENOMEM;
-
- r = seccomp_add_secondary_archs(seccomp);
+ r = seccomp_init_conservative(&seccomp, SCMP_ACT_ALLOW);
if (r < 0)
- goto finish;
+ return r;
r = seccomp_rule_add(
seccomp,
if (r < 0)
goto finish;
- r = seccomp_attr_set(seccomp, SCMP_FLTATR_CTL_NNP, 0);
- if (r < 0)
- goto finish;
-
r = seccomp_load(seccomp);
finish:
return r;
}
-static int apply_protect_kernel_modules(Unit *u, const ExecContext *c) {
- static const int module_syscalls[] = {
- SCMP_SYS(delete_module),
- SCMP_SYS(finit_module),
- SCMP_SYS(init_module),
- };
-
- scmp_filter_ctx *seccomp;
- unsigned i;
- int r;
-
+static int apply_protect_kernel_modules(const Unit *u, const ExecContext *c) {
assert(c);
- /* Turn of module syscalls on ProtectKernelModules=yes */
+ /* Turn off module syscalls on ProtectKernelModules=yes */
if (skip_seccomp_unavailable(u, "ProtectKernelModules="))
return 0;
- seccomp = seccomp_init(SCMP_ACT_ALLOW);
- if (!seccomp)
- return -ENOMEM;
-
- r = seccomp_add_secondary_archs(seccomp);
- if (r < 0)
- goto finish;
-
- for (i = 0; i < ELEMENTSOF(module_syscalls); i++) {
- r = seccomp_rule_add(seccomp, SCMP_ACT_ERRNO(EPERM),
- module_syscalls[i], 0);
- if (r < 0)
- goto finish;
- }
-
- r = seccomp_attr_set(seccomp, SCMP_FLTATR_CTL_NNP, 0);
- if (r < 0)
- goto finish;
-
- r = seccomp_load(seccomp);
-
-finish:
- seccomp_release(seccomp);
- return r;
+ return seccomp_load_filter_set(SCMP_ACT_ALLOW, syscall_filter_sets + SYSCALL_FILTER_SET_MODULE, SCMP_ACT_ERRNO(EPERM));
}
-static int apply_private_devices(Unit *u, const ExecContext *c) {
- const SystemCallFilterSet *set;
- scmp_filter_ctx *seccomp;
- const char *sys;
- bool syscalls_found = false;
- int r;
-
+static int apply_private_devices(const Unit *u, const ExecContext *c) {
assert(c);
/* If PrivateDevices= is set, also turn off iopl and all @raw-io syscalls. */
if (skip_seccomp_unavailable(u, "PrivateDevices="))
return 0;
- seccomp = seccomp_init(SCMP_ACT_ALLOW);
- if (!seccomp)
- return -ENOMEM;
-
- r = seccomp_add_secondary_archs(seccomp);
- if (r < 0)
- goto finish;
-
- for (set = syscall_filter_sets; set->set_name; set++)
- if (streq(set->set_name, "@raw-io")) {
- syscalls_found = true;
- break;
- }
-
- /* We should never fail here */
- if (!syscalls_found) {
- r = -EOPNOTSUPP;
- goto finish;
- }
-
- NULSTR_FOREACH(sys, set->value) {
- int id;
- bool add = true;
-
-#ifndef __NR_s390_pci_mmio_read
- if (streq(sys, "s390_pci_mmio_read"))
- add = false;
-#endif
-#ifndef __NR_s390_pci_mmio_write
- if (streq(sys, "s390_pci_mmio_write"))
- add = false;
-#endif
-
- if (!add)
- continue;
-
- id = seccomp_syscall_resolve_name(sys);
+ return seccomp_load_filter_set(SCMP_ACT_ALLOW, syscall_filter_sets + SYSCALL_FILTER_SET_RAW_IO, SCMP_ACT_ERRNO(EPERM));
+}
- r = seccomp_rule_add(
- seccomp,
- SCMP_ACT_ERRNO(EPERM),
- id, 0);
- if (r < 0)
- goto finish;
- }
+static int apply_restrict_namespaces(Unit *u, const ExecContext *c) {
+ assert(c);
- r = seccomp_attr_set(seccomp, SCMP_FLTATR_CTL_NNP, 0);
- if (r < 0)
- goto finish;
+ if (!exec_context_restrict_namespaces_set(c))
+ return 0;
- r = seccomp_load(seccomp);
+ if (skip_seccomp_unavailable(u, "RestrictNamespaces="))
+ return 0;
-finish:
- seccomp_release(seccomp);
- return r;
+ return seccomp_restrict_namespaces(c->restrict_namespaces);
}
#endif
if (!joined)
return -ENOMEM;
- x = strjoin("LISTEN_FDNAMES=", joined, NULL);
+ x = strjoin("LISTEN_FDNAMES=", joined);
if (!x)
return -ENOMEM;
our_env[n_env++] = x;
v = getenv(*i);
if (!v)
continue;
- x = strjoin(*i, "=", v, NULL);
+ x = strjoin(*i, "=", v);
if (!x)
return -ENOMEM;
if (!GREEDY_REALLOC(pass_env, n_bufsize, n_env + 2))
asprintf(&uid_map,
"0 0 1\n" /* Map root → root */
UID_FMT " " UID_FMT " 1\n", /* Map $UID → $UID */
- uid, uid); /* The case where the above is the same */
+ uid, uid);
else
- uid_map = strdup("0 0 1\n");
+ uid_map = strdup("0 0 1\n"); /* The case where the above is the same */
if (!uid_map)
return -ENOMEM;
STRV_FOREACH(rt, context->runtime_directory) {
_cleanup_free_ char *p;
- p = strjoin(params->runtime_prefix, "/", *rt, NULL);
+ p = strjoin(params->runtime_prefix, "/", *rt);
if (!p)
return -ENOMEM;
STRV_FOREACH(rt, context->runtime_directory) {
char *s;
- s = strjoin(params->runtime_prefix, "/", *rt, NULL);
+ s = strjoin(params->runtime_prefix, "/", *rt);
if (!s)
return -ENOMEM;
return 0;
}
+static int apply_mount_namespace(Unit *u, const ExecContext *context,
+ const ExecParameters *params,
+ ExecRuntime *runtime) {
+ int r;
+ _cleanup_free_ char **rw = NULL;
+ char *tmp = NULL, *var = NULL;
+ const char *root_dir = NULL;
+ NameSpaceInfo ns_info = {
+ .ignore_protect_paths = false,
+ .private_dev = context->private_devices,
+ .protect_control_groups = context->protect_control_groups,
+ .protect_kernel_tunables = context->protect_kernel_tunables,
+ .protect_kernel_modules = context->protect_kernel_modules,
+ };
+
+ assert(context);
+
+ /* The runtime struct only contains the parent of the private /tmp,
+ * which is non-accessible to world users. Inside of it there's a /tmp
+ * that is sticky, and that's the one we want to use here. */
+
+ if (context->private_tmp && runtime) {
+ if (runtime->tmp_dir)
+ tmp = strjoina(runtime->tmp_dir, "/tmp");
+ if (runtime->var_tmp_dir)
+ var = strjoina(runtime->var_tmp_dir, "/tmp");
+ }
+
+ r = compile_read_write_paths(context, params, &rw);
+ if (r < 0)
+ return r;
+
+ if (params->flags & EXEC_APPLY_CHROOT)
+ root_dir = context->root_directory;
+
+ /*
+ * If DynamicUser=no and RootDirectory= is set then lets pass a relaxed
+ * sandbox info, otherwise enforce it, don't ignore protected paths and
+ * fail if we are enable to apply the sandbox inside the mount namespace.
+ */
+ if (!context->dynamic_user && root_dir)
+ ns_info.ignore_protect_paths = true;
+
+ r = setup_namespace(root_dir, &ns_info, rw,
+ context->read_only_paths,
+ context->inaccessible_paths,
+ tmp,
+ var,
+ context->protect_home,
+ context->protect_system,
+ context->mount_flags);
+
+ /* If we couldn't set up the namespace this is probably due to a
+ * missing capability. In this case, silently proceeed. */
+ if (IN_SET(r, -EPERM, -EACCES)) {
+ log_open();
+ log_unit_debug_errno(u, r, "Failed to set up namespace, assuming containerized execution, ignoring: %m");
+ log_close();
+ r = 0;
+ }
+
+ return r;
+}
+
+static int apply_working_directory(const ExecContext *context,
+ const ExecParameters *params,
+ const char *home,
+ const bool needs_mount_ns) {
+ const char *d;
+ const char *wd;
+
+ assert(context);
+
+ if (context->working_directory_home)
+ wd = home;
+ else if (context->working_directory)
+ wd = context->working_directory;
+ else
+ wd = "/";
+
+ if (params->flags & EXEC_APPLY_CHROOT) {
+ if (!needs_mount_ns && context->root_directory)
+ if (chroot(context->root_directory) < 0)
+ return -errno;
+
+ d = wd;
+ } else
+ d = strjoina(strempty(context->root_directory), "/", strempty(wd));
+
+ if (chdir(d) < 0 && !context->working_directory_missing_ok)
+ return -errno;
+
+ return 0;
+}
+
static void append_socket_pair(int *array, unsigned *n, int pair[2]) {
assert(array);
assert(n);
return context_has_address_families(c) || /* we need NNP if we have any form of seccomp and are unprivileged */
c->memory_deny_write_execute ||
c->restrict_realtime ||
+ exec_context_restrict_namespaces_set(c) ||
c->protect_kernel_tunables ||
c->protect_kernel_modules ||
c->private_devices ||
_cleanup_free_ char *mac_selinux_context_net = NULL;
_cleanup_free_ gid_t *supplementary_gids = NULL;
const char *username = NULL, *groupname = NULL;
- const char *home = NULL, *shell = NULL, *wd;
+ const char *home = NULL, *shell = NULL;
dev_t journal_stream_dev = 0;
ino_t journal_stream_ino = 0;
bool needs_mount_namespace;
exec_context_tty_reset(context, params);
- if (params->flags & EXEC_CONFIRM_SPAWN) {
- char response;
+ if (params->confirm_spawn) {
+ const char *vc = params->confirm_spawn;
+ _cleanup_free_ char *cmdline = NULL;
- r = ask_for_confirmation(&response, argv);
- if (r == -ETIMEDOUT)
- write_confirm_message("Confirmation question timed out, assuming positive response.\n");
- else if (r < 0)
- write_confirm_message("Couldn't ask confirmation question, assuming positive response: %s\n", strerror(-r));
- else if (response == 's') {
- write_confirm_message("Skipping execution.\n");
+ cmdline = exec_command_line(argv);
+ if (!cmdline) {
+ *exit_status = EXIT_CONFIRM;
+ return -ENOMEM;
+ }
+
+ r = ask_for_confirmation(vc, unit, cmdline);
+ if (r != CONFIRM_EXECUTE) {
+ if (r == CONFIRM_PRETEND_SUCCESS) {
+ *exit_status = EXIT_SUCCESS;
+ return 0;
+ }
*exit_status = EXIT_CONFIRM;
return -ECANCELED;
- } else if (response == 'n') {
- write_confirm_message("Failing execution.\n");
- *exit_status = 0;
- return 0;
}
}
*exit_status = EXIT_GROUP;
return r;
}
+ }
- r = get_fixed_supplementary_groups(context, username, groupname,
- gid, &supplementary_gids, &ngids);
- if (r < 0) {
- *exit_status = EXIT_GROUP;
- return r;
- }
+ /* Initialize user supplementary groups and get SupplementaryGroups= ones */
+ r = get_supplementary_groups(context, username, groupname, gid,
+ &supplementary_gids, &ngids);
+ if (r < 0) {
+ *exit_status = EXIT_GROUP;
+ return r;
}
r = send_user_lookup(unit, user_lookup_fd, uid, gid);
(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) {
needs_mount_namespace = exec_needs_mount_namespace(context, params, runtime);
if (needs_mount_namespace) {
- _cleanup_free_ char **rw = NULL;
- char *tmp = NULL, *var = NULL;
- NameSpaceInfo ns_info = {
- .private_dev = context->private_devices,
- .protect_control_groups = context->protect_control_groups,
- .protect_kernel_tunables = context->protect_kernel_tunables,
- .protect_kernel_modules = context->protect_kernel_modules,
- };
-
- /* The runtime struct only contains the parent
- * of the private /tmp, which is
- * non-accessible to world users. Inside of it
- * there's a /tmp that is sticky, and that's
- * the one we want to use here. */
-
- if (context->private_tmp && runtime) {
- if (runtime->tmp_dir)
- tmp = strjoina(runtime->tmp_dir, "/tmp");
- if (runtime->var_tmp_dir)
- var = strjoina(runtime->var_tmp_dir, "/tmp");
- }
-
- r = compile_read_write_paths(context, params, &rw);
+ r = apply_mount_namespace(unit, context, params, runtime);
if (r < 0) {
*exit_status = EXIT_NAMESPACE;
return r;
}
-
- r = setup_namespace(
- (params->flags & EXEC_APPLY_CHROOT) ? context->root_directory : NULL,
- &ns_info,
- rw,
- context->read_only_paths,
- context->inaccessible_paths,
- tmp,
- var,
- context->protect_home,
- context->protect_system,
- context->mount_flags);
-
- /* If we couldn't set up the namespace this is
- * probably due to a missing capability. In this case,
- * silently proceeed. */
- if (r == -EPERM || r == -EACCES) {
- log_open();
- log_unit_debug_errno(unit, r, "Failed to set up namespace, assuming containerized execution, ignoring: %m");
- log_close();
- } else if (r < 0) {
- *exit_status = EXIT_NAMESPACE;
- return r;
- }
}
- if (context->working_directory_home)
- wd = home;
- else if (context->working_directory)
- wd = context->working_directory;
- else
- wd = "/";
+ /* Apply just after mount namespace setup */
+ r = apply_working_directory(context, params, home, needs_mount_namespace);
+ if (r < 0) {
+ *exit_status = EXIT_CHROOT;
+ return r;
+ }
- /* Drop group as early as possbile */
+ /* Drop groups as early as possbile */
if ((params->flags & EXEC_APPLY_PERMISSIONS) && !command->privileged) {
r = enforce_groups(context, gid, supplementary_gids, ngids);
if (r < 0) {
}
}
- if (params->flags & EXEC_APPLY_CHROOT) {
- if (!needs_mount_namespace && context->root_directory)
- if (chroot(context->root_directory) < 0) {
- *exit_status = EXIT_CHROOT;
- return -errno;
- }
-
- if (chdir(wd) < 0 &&
- !context->working_directory_missing_ok) {
- *exit_status = EXIT_CHDIR;
- return -errno;
- }
- } else {
- const char *d;
-
- d = strjoina(strempty(context->root_directory), "/", strempty(wd));
- if (chdir(d) < 0 &&
- !context->working_directory_missing_ok) {
- *exit_status = EXIT_CHDIR;
- return -errno;
- }
- }
-
#ifdef HAVE_SELINUX
if ((params->flags & EXEC_APPLY_PERMISSIONS) &&
mac_selinux_use() &&
}
}
+ /* 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
}
}
+ r = apply_restrict_namespaces(unit, context);
+ if (r < 0) {
+ *exit_status = EXIT_SECCOMP;
+ return r;
+ }
+
if (context->protect_kernel_tunables) {
r = apply_protect_sysctl(unit, context);
if (r < 0) {
}
}
+ /* 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) {
}
}
#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);
c->personality = PERSONALITY_INVALID;
c->runtime_directory_mode = 0755;
c->capability_bounding_set = CAP_ALL;
+ c->restrict_namespaces = NAMESPACE_FLAGS_ALL;
}
void exec_context_done(ExecContext *c) {
STRV_FOREACH(i, c->runtime_directory) {
_cleanup_free_ char *p;
- p = strjoin(runtime_prefix, "/", *i, NULL);
+ p = strjoin(runtime_prefix, "/", *i);
if (!p)
return -ENOMEM;
void exec_context_dump(ExecContext *c, FILE* f, const char *prefix) {
char **e, **d;
unsigned i;
+ int r;
assert(c);
assert(f);
fputc('\n', f);
}
+ if (exec_context_restrict_namespaces_set(c)) {
+ _cleanup_free_ char *s = NULL;
+
+ r = namespace_flag_to_string_many(c->restrict_namespaces, &s);
+ if (r >= 0)
+ fprintf(f, "%sRestrictNamespaces: %s\n",
+ prefix, s);
+ }
+
if (c->syscall_errno > 0)
fprintf(f,
"%sSystemCallErrorNumber: %s\n",
STRV_FOREACH(a, argv)
k += strlen(*a)+3;
- if (!(n = new(char, k)))
+ n = new(char, k);
+ if (!n)
return NULL;
p = n;