systemd,nspawn: use extended attributes to store metadata

All attributes are stored as text, since root_directory is already
text, and it seems easier to have all of them in text format.

Attributes are written in the trusted. namespace, because the kernel
currently does not allow user. attributes on cgroups. This is a PITA,
and CAP_SYS_ADMIN is required to *read* the attributes. Alas.

A second pipe is opened for the child to signal the parent that the
cgroup hierarchy has been set up.

3 files changed, 67 insertions, 6 deletions

diff --git a/TODO b/TODO
index 786275143..9adec5ea9 100644
--- a/TODO
+++ b/TODO
@@ -55,8 +55,6 @@ Features:
 
 * timedatctl, localectl: possibly make some commands work without the daemon, for chroot situations...
 
-* nspawn: expose init PID, UUID, root directory as cgroup EAs. names should be: 'user.machine_id', 'user.init_pid', 'user.root_directory'.
-
 * logind: consider making suspend inhibitor locks non-session specific, but keep suspend-key inhibitor locks session specific
 
 * logind: add Suspend() bus calls which take timestamps to fix double suspend issues when somebody hits suspend and closes laptop quickly.
@@ -405,7 +403,6 @@ Features:
     interfaces to the container
   - nspawn: maybe add a way to drop additional caps, in addition to add additional caps
   - nspawn: maybe explicitly reset loginuid?
-  - nspawn: move all containers to name=systemd:/containers/, and introduce externally visible names for containers
   - nspawn: make it work for dwalsh and shared /usr containers -- tmpfs mounts as command line parameters, selinux exec context
 
 * cryptsetup:
diff --git a/src/core/mount-setup.c b/src/core/mount-setup.c
index e45a6bc1c..56d358b09 100644
--- a/src/core/mount-setup.c
+++ b/src/core/mount-setup.c
@@ -68,6 +68,12 @@ typedef struct MountPoint {
  * other ones we can delay until SELinux and IMA are loaded. */
 #define N_EARLY_MOUNT 5
 
+#ifdef HAVE_XATTR
+#  define FS_XATTR_OPT ",xattr"
+#else
+#  define FS_XATTR_OPT ""
+#endif
+
 static const MountPoint mount_table[] = {
         { "proc",       "/proc",                     "proc",       NULL, MS_NOSUID|MS_NOEXEC|MS_NODEV,
           NULL,       MNT_FATAL|MNT_IN_CONTAINER },
@@ -87,7 +93,7 @@ static const MountPoint mount_table[] = {
           NULL,       MNT_FATAL|MNT_IN_CONTAINER },
         { "tmpfs",      "/sys/fs/cgroup",            "tmpfs",      "mode=755", MS_NOSUID|MS_NOEXEC|MS_NODEV|MS_STRICTATIME,
           NULL,       MNT_IN_CONTAINER },
-        { "cgroup",     "/sys/fs/cgroup/systemd",    "cgroup",     "none,name=systemd", MS_NOSUID|MS_NOEXEC|MS_NODEV,
+        { "cgroup",     "/sys/fs/cgroup/systemd",    "cgroup",     "none,name=systemd" FS_XATTR_OPT, MS_NOSUID|MS_NOEXEC|MS_NODEV,
           NULL,       MNT_IN_CONTAINER },
         { "pstore",     "/sys/fs/pstore",            "pstore",     NULL, MS_NOSUID|MS_NOEXEC|MS_NODEV,
           NULL,       MNT_NONE },
diff --git a/src/nspawn/nspawn.c b/src/nspawn/nspawn.c
index 5a43d5ed1..b59b2673c 100644
--- a/src/nspawn/nspawn.c
+++ b/src/nspawn/nspawn.c
@@ -22,6 +22,7 @@
 #include <signal.h>
 #include <sched.h>
 #include <unistd.h>
+#include <attr/xattr.h>
 #include <sys/types.h>
 #include <sys/syscall.h>
 #include <sys/mount.h>
@@ -921,6 +922,46 @@ static int setup_cgroup(const char *path) {
         return 0;
 }
 
+static int save_attributes(const char *cgroup, pid_t pid, const char *uuid, const char *directory) {
+        char buf[DECIMAL_STR_MAX(pid_t)], path[PATH_MAX];
+        int r = 0, k;
+
+        assert(cgroup);
+        assert(pid >= 0);
+        assert(arg_directory);
+
+#ifdef HAVE_XATTR
+        assert_se(snprintf(buf, sizeof(buf), "%lu", (unsigned long) pid) < (int) sizeof(buf));
+
+        r = snprintf(path, sizeof(path), "/sys/fs/cgroup/systemd/%s", cgroup);
+        if (r >= (int) sizeof(path)) {
+                log_error("cgroup name too long");
+                return -EINVAL;
+        }
+
+        r = setxattr(path, "trusted.init_pid", buf, strlen(buf), XATTR_CREATE);
+        if (r < 0)
+                log_warning("Failed to set %s attribute on %s: %m", "trusted.init_pid", path);
+
+        if (uuid) {
+                k = setxattr(path, "trusted.machine_id", uuid, strlen(uuid), XATTR_CREATE);
+                if (k < 0) {
+                        log_warning("Failed to set %s attribute on %s: %m", "trusted.machine_id", path);
+                        if (r == 0)
+                                r = k;
+                }
+        }
+
+        k = setxattr(path, "trusted.root_directory", directory, strlen(directory), XATTR_CREATE);
+        if (k < 0) {
+                log_warning("Failed to set %s attribute on %s: %m", "trusted.machine_id", path);
+                if (r == 0)
+                        r = k;
+        }
+#endif
+        return r;
+}
+
 static int drop_capabilities(void) {
         return capability_bounding_set_drop(~arg_retain, false);
 }
@@ -1198,7 +1239,7 @@ int main(int argc, char *argv[]) {
                 arg_directory = get_current_dir_name();
 
         if (!arg_directory) {
-                log_error("Failed to determine path");
+                log_error("Failed to determine path, please use -D.");
                 goto finish;
         }
 
@@ -1313,13 +1354,19 @@ int main(int argc, char *argv[]) {
 
         for (;;) {
                 siginfo_t status;
-                int pipefd[2];
+                int pipefd[2], pipefd2[2];
 
                 if (pipe2(pipefd, O_NONBLOCK|O_CLOEXEC) < 0) {
                         log_error("pipe2(): %m");
                         goto finish;
                 }
 
+                if (pipe2(pipefd2, O_NONBLOCK|O_CLOEXEC) < 0) {
+                        log_error("pipe2(): %m");
+                        close_pipe(pipefd);
+                        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)
@@ -1353,6 +1400,7 @@ int main(int argc, char *argv[]) {
                         if (envp[n_env])
                                 n_env ++;
 
+                        /* Wait for the parent process to log our PID */
                         close_nointr_nofail(pipefd[1]);
                         fd_wait_for_event(pipefd[0], POLLHUP, -1);
                         close_nointr_nofail(pipefd[0]);
@@ -1409,6 +1457,9 @@ int main(int argc, char *argv[]) {
                         if (setup_cgroup(newcg) < 0)
                                 goto child_fail;
 
+                        close_nointr_nofail(pipefd2[1]);
+                        close_nointr_nofail(pipefd2[0]);
+
                         /* Mark everything as slave, so that we still
                          * receive mounts from the real root, but don't
                          * propagate mounts to the real root. */
@@ -1617,6 +1668,13 @@ int main(int argc, char *argv[]) {
                 close_nointr_nofail(pipefd[0]);
                 close_nointr_nofail(pipefd[1]);
 
+                /* Wait for the child process to establish cgroup hierarchy */
+                close_nointr_nofail(pipefd2[1]);
+                fd_wait_for_event(pipefd2[0], POLLHUP, -1);
+                close_nointr_nofail(pipefd2[0]);
+
+                save_attributes(newcg, pid, arg_uuid, arg_directory);
+
                 fdset_free(fds);
                 fds = NULL;