chiark / gitweb /
Add SELinux support to systemd-nspawn
[elogind.git] / src / nspawn / nspawn.c
index caf1aa9bedfee1373d5aca396cbf26a1dc50d37f..c91f6cce21faf264a2f1680b0e337d1a53ba26a6 100644 (file)
@@ -41,6 +41,9 @@
 #include <sys/socket.h>
 #include <linux/netlink.h>
 #include <sys/eventfd.h>
+#if HAVE_SELINUX
+#include <selinux/selinux.h>
+#endif
 
 #include "sd-daemon.h"
 #include "sd-bus.h"
 #include "ptyfwd.h"
 #include "bus-kernel.h"
 #include "env-util.h"
-
-#ifndef TTY_GID
-#define TTY_GID 5
-#endif
+#include "def.h"
 
 typedef enum LinkJournal {
         LINK_NO,
@@ -80,6 +80,8 @@ static char *arg_directory = NULL;
 static char *arg_user = NULL;
 static sd_id128_t arg_uuid = {};
 static char *arg_machine = NULL;
+static char *process_label = NULL;
+static char *file_label = NULL;
 static const char *arg_slice = NULL;
 static bool arg_private_network = false;
 static bool arg_read_only = false;
@@ -110,7 +112,8 @@ static uint64_t arg_retain =
         (1ULL << CAP_SYS_RESOURCE) |
         (1ULL << CAP_SYS_BOOT) |
         (1ULL << CAP_AUDIT_WRITE) |
-        (1ULL << CAP_AUDIT_CONTROL);
+        (1ULL << CAP_AUDIT_CONTROL) |
+        (1ULL << CAP_MKNOD);
 static char **arg_bind = NULL;
 static char **arg_bind_ro = NULL;
 static char **arg_setenv = NULL;
@@ -119,25 +122,27 @@ static int help(void) {
 
         printf("%s [OPTIONS...] [PATH] [ARGUMENTS...]\n\n"
                "Spawn a minimal namespace container for debugging, testing and building.\n\n"
-               "  -h --help                Show this help\n"
-               "     --version             Print version string\n"
-               "  -D --directory=NAME      Root directory for the container\n"
-               "  -b --boot                Boot up full system (i.e. invoke init)\n"
-               "  -u --user=USER           Run the command under specified user or uid\n"
-               "     --uuid=UUID           Set a specific machine UUID for the container\n"
-               "  -M --machine=NAME        Set the machine name for the container\n"
-               "  -S --slice=SLICE         Place the container in the specified slice\n"
-               "     --private-network     Disable network in container\n"
-               "     --read-only           Mount the root directory read-only\n"
-               "     --capability=CAP      In addition to the default, retain specified\n"
-               "                           capability\n"
-               "     --drop-capability=CAP Drop the specified capability from the default set\n"
-               "     --link-journal=MODE   Link up guest journal, one of no, auto, guest, host\n"
-               "  -j                       Equivalent to --link-journal=host\n"
-               "     --bind=PATH[:PATH]    Bind mount a file or directory from the host into\n"
-               "                           the container\n"
-               "     --bind-ro=PATH[:PATH] Similar, but creates a read-only bind mount\n"
-               "     --setenv=NAME=VALUE   Pass an environment variable to PID 1\n",
+               "  -h --help                 Show this help\n"
+               "     --version              Print version string\n"
+               "  -D --directory=NAME       Root directory for the container\n"
+               "  -b --boot                 Boot up full system (i.e. invoke init)\n"
+               "  -u --user=USER            Run the command under specified user or uid\n"
+               "     --uuid=UUID            Set a specific machine UUID for the container\n"
+               "  -M --machine=NAME         Set the machine name for the container\n"
+               "  -S --slice=SLICE          Place the container in the specified slice\n"
+               "  -L --file-label=LABEL     Set the MAC file label to be used by tmpfs file systems in container\n"
+               "  -Z --process-label=LABEL  Set the MAC label to be used by processes in container\n"
+               "     --private-network      Disable network in container\n"
+               "     --read-only            Mount the root directory read-only\n"
+               "     --capability=CAP       In addition to the default, retain specified\n"
+               "                            capability\n"
+               "     --drop-capability=CAP  Drop the specified capability from the default set\n"
+               "     --link-journal=MODE    Link up guest journal, one of no, auto, guest, host\n"
+               "  -j                        Equivalent to --link-journal=host\n"
+               "     --bind=PATH[:PATH]     Bind mount a file or directory from the host into\n"
+               "                            the container\n"
+               "     --bind-ro=PATH[:PATH]  Similar, but creates a read-only bind mount\n"
+               "     --setenv=NAME=VALUE    Pass an environment variable to PID 1\n",
                program_invocation_short_name);
 
         return 0;
@@ -175,6 +180,8 @@ static int parse_argv(int argc, char *argv[]) {
                 { "machine",         required_argument, NULL, 'M'                 },
                 { "slice",           required_argument, NULL, 'S'                 },
                 { "setenv",          required_argument, NULL, ARG_SETENV          },
+                { "process-label",   required_argument, NULL, 'Z'                 },
+                { "file-label",      required_argument, NULL, 'L'                 },
                 {}
         };
 
@@ -183,7 +190,7 @@ static int parse_argv(int argc, char *argv[]) {
         assert(argc >= 0);
         assert(argv);
 
-        while ((c = getopt_long(argc, argv, "+hD:u:bM:jS:", options, NULL)) >= 0) {
+        while ((c = getopt_long(argc, argv, "+hD:u:bL:M:jS:Z:", options, NULL)) >= 0) {
 
                 switch (c) {
 
@@ -249,6 +256,20 @@ static int parse_argv(int argc, char *argv[]) {
 
                         break;
 
+                case 'L':
+                        file_label = strdup(optarg);
+                        if (!file_label)
+                                return log_oom();
+
+                        break;
+
+                case 'Z':
+                        process_label = strdup(optarg);
+                        if (!process_label)
+                                return log_oom();
+
+                        break;
+
                 case ARG_READ_ONLY:
                         arg_read_only = true;
                         break;
@@ -398,6 +419,7 @@ static int mount_all(const char *dest) {
 
         for (k = 0; k < ELEMENTSOF(mount_table); k++) {
                 _cleanup_free_ char *where = NULL;
+                _cleanup_free_ char *options = NULL;
                 int t;
 
                 where = strjoin(dest, "/", mount_table[k].where, NULL);
@@ -420,11 +442,22 @@ static int mount_all(const char *dest) {
 
                 mkdir_p(where, 0755);
 
+#ifdef HAVE_SELINUX
+                if (file_label && (streq_ptr(mount_table[k].what, "tmpfs") ||
+                              streq_ptr(mount_table[k].what, "devpts")))
+                        options = strjoin(mount_table[k].options, ",context=\"", file_label, "\"", NULL);
+                else
+#endif
+                        options = strjoin(mount_table[k].options, NULL);
+
+                if (!options)
+                        return log_oom();
+
                 if (mount(mount_table[k].what,
                           where,
                           mount_table[k].type,
                           mount_table[k].flags,
-                          mount_table[k].options) < 0 &&
+                          options) < 0 &&
                     mount_table[k].fatal) {
 
                         log_error("mount(%s) failed: %m", where);
@@ -639,40 +672,30 @@ static int copy_devnodes(const char *dest) {
         u = umask(0000);
 
         NULSTR_FOREACH(d, devnodes) {
-                struct stat st;
                 _cleanup_free_ char *from = NULL, *to = NULL;
+                struct stat st;
 
-                asprintf(&from, "/dev/%s", d);
-                asprintf(&to, "%s/dev/%s", dest, d);
-
-                if (!from || !to) {
-                        log_oom();
-
-                        if (r == 0)
-                                r = -ENOMEM;
-
-                        break;
-                }
+                from = strappend("/dev/", d);
+                to = strjoin(dest, "/dev/", d, NULL);
+                if (!from || !to)
+                        return log_oom();
 
                 if (stat(from, &st) < 0) {
 
                         if (errno != ENOENT) {
                                 log_error("Failed to stat %s: %m", from);
-                                if (r == 0)
-                                        r = -errno;
+                                return -errno;
                         }
 
                 } else if (!S_ISCHR(st.st_mode) && !S_ISBLK(st.st_mode)) {
 
                         log_error("%s is not a char or block device, cannot copy", from);
-                        if (r == 0)
-                                r = -EIO;
+                        return -EIO;
 
                 } else if (mknod(to, st.st_mode, st.st_rdev) < 0) {
 
                         log_error("mknod(%s) failed: %m", dest);
-                        if (r == 0)
-                                r = -errno;
+                        return  -errno;
                 }
         }
 
@@ -986,7 +1009,7 @@ static int setup_kdbus(const char *dest, const char *path) {
         }
 
         if (mount(path, p, "bind", MS_BIND, NULL) < 0) {
-                log_error("Failed to mount kdbus namespace path: %m");
+                log_error("Failed to mount kdbus domain path: %m");
                 return -errno;
         }
 
@@ -1104,7 +1127,7 @@ int main(int argc, char *argv[]) {
         sigset_t mask;
         _cleanup_close_pipe_ int kmsg_socket_pair[2] = { -1, -1 };
         _cleanup_fdset_free_ FDSet *fds = NULL;
-        _cleanup_free_ char *kdbus_namespace = NULL;
+        _cleanup_free_ char *kdbus_domain = NULL;
         const char *ns;
 
         log_parse_environment();
@@ -1207,23 +1230,17 @@ int main(int argc, char *argv[]) {
         }
 
         ns = strappenda("machine-", arg_machine);
-        kdbus_fd = bus_kernel_create_namespace(ns, &kdbus_namespace);
+        kdbus_fd = bus_kernel_create_domain(ns, &kdbus_domain);
         if (r < 0)
-                log_debug("Failed to create kdbus namespace: %s", strerror(-r));
+                log_debug("Failed to create kdbus domain: %s", strerror(-r));
         else
-                log_debug("Successfully created kdbus namespace as %s", kdbus_namespace);
+                log_debug("Successfully created kdbus domain as %s", kdbus_domain);
 
         if (socketpair(AF_UNIX, SOCK_DGRAM|SOCK_NONBLOCK|SOCK_CLOEXEC, 0, kmsg_socket_pair) < 0) {
                 log_error("Failed to create kmsg socket pair: %m");
                 goto finish;
         }
 
-        sync_fd = eventfd(0, EFD_CLOEXEC);
-        if (sync_fd < 0) {
-                log_error("Failed to create event fd: %m");
-                goto finish;
-        }
-
         sd_notify(0, "READY=1");
 
         assert_se(sigemptyset(&mask) == 0);
@@ -1233,6 +1250,12 @@ int main(int argc, char *argv[]) {
         for (;;) {
                 siginfo_t status;
 
+                sync_fd = eventfd(0, EFD_CLOEXEC);
+                if (sync_fd < 0) {
+                        log_error("Failed to create event fd: %m");
+                        goto finish;
+                }
+
                 pid = syscall(__NR_clone, SIGCHLD|CLONE_NEWIPC|CLONE_NEWNS|CLONE_NEWPID|CLONE_NEWUTS|(arg_private_network ? CLONE_NEWNET : 0), NULL);
                 if (pid < 0) {
                         if (errno == EINVAL)
@@ -1368,7 +1391,7 @@ int main(int argc, char *argv[]) {
                         if (mount_binds(arg_directory, arg_bind_ro, MS_RDONLY) < 0)
                                 goto child_fail;
 
-                        if (setup_kdbus(arg_directory, kdbus_namespace) < 0)
+                        if (setup_kdbus(arg_directory, kdbus_domain) < 0)
                                 goto child_fail;
 
                         if (chdir(arg_directory) < 0) {
@@ -1503,6 +1526,11 @@ int main(int argc, char *argv[]) {
                         } else
                                 env_use = (char**) envp;
 
+#if HAVE_SELINUX
+                        if (process_label)
+                                if (setexeccon(process_label) < 0)
+                                        log_error("setexeccon(\"%s\") failed: %m", process_label);
+#endif
                         if (arg_boot) {
                                 char **a;
                                 size_t l;