Merge pull request #21094 from poettering/userns-split

util-lib: split out userns allocation into new helper function

src/basic/namespace-util.c

@@ -5,6 +5,7 @@
 #include <sys/mount.h>
 
 #include "fd-util.h"
+#include "fileio.h"
 #include "missing_fs.h"
 #include "missing_magic.h"
 #include "namespace-util.h"
@@ -181,3 +182,41 @@ int detach_mount_namespace(void) {
 
         return 0;
 }
+
+int userns_acquire(const char *uid_map, const char *gid_map) {
+        char path[STRLEN("/proc//uid_map") + DECIMAL_STR_MAX(pid_t) + 1];
+        _cleanup_(sigkill_waitp) pid_t pid = 0;
+        _cleanup_close_ int userns_fd = -1;
+        int r;
+
+        assert(uid_map);
+        assert(gid_map);
+
+        /* Forks off a process in a new userns, configures the specified uidmap/gidmap, acquires an fd to it,
+         * and then kills the process again. This way we have a userns fd that is not bound to any
+         * process. We can use that for file system mounts and similar. */
+
+        r = safe_fork("(sd-mkuserns)", FORK_CLOSE_ALL_FDS|FORK_DEATHSIG|FORK_NEW_USERNS, &pid);
+        if (r < 0)
+                return r;
+        if (r == 0)
+                /* Child. We do nothing here, just freeze until somebody kills us. */
+                freeze();
+
+        xsprintf(path, "/proc/" PID_FMT "/uid_map", pid);
+        r = write_string_file(path, uid_map, WRITE_STRING_FILE_DISABLE_BUFFER);
+        if (r < 0)
+                return log_error_errno(r, "Failed to write UID map: %m");
+
+        xsprintf(path, "/proc/" PID_FMT "/gid_map", pid);
+        r = write_string_file(path, gid_map, WRITE_STRING_FILE_DISABLE_BUFFER);
+        if (r < 0)
+                return log_error_errno(r, "Failed to write GID map: %m");
+
+        r = namespace_open(pid, NULL, NULL, NULL, &userns_fd, NULL);
+        if (r < 0)
+                return log_error_errno(r, "Failed to open netns fd: %m");
+
+        return TAKE_FD(userns_fd);
+
+}

src/basic/namespace-util.h

@@ -24,3 +24,5 @@ static inline bool userns_shift_range_valid(uid_t shift, uid_t range) {
 
         return true;
 }
+
+int userns_acquire(const char *uid_map, const char *gid_map);

src/basic/process-util.c

@@ -1618,6 +1618,25 @@ bool invoked_as(char *argv[], const char *token) {
         return strstr(last_path_component(argv[0]), token);
 }
 
+_noreturn_ void freeze(void) {
+        log_close();
+
+        /* Make sure nobody waits for us on a socket anymore */
+        (void) close_all_fds_full(NULL, 0, false);
+
+        /* Let's not freeze right away, but keep reaping zombies. */
+        for (;;) {
+                siginfo_t si = {};
+
+                if (waitid(P_ALL, 0, &si, WEXITED) < 0 && errno != EINTR)
+                        break;
+        }
+
+        /* waitid() failed with an unexpected error, things are really borked. Freeze now! */
+        for (;;)
+                pause();
+}
+
 static const char *const ioprio_class_table[] = {
         [IOPRIO_CLASS_NONE] = "none",
         [IOPRIO_CLASS_RT] = "realtime",

src/basic/process-util.h

@@ -202,3 +202,5 @@ int pidfd_get_pid(int fd, pid_t *ret);
 int setpriority_closest(int priority);
 
 bool invoked_as(char *argv[], const char *token);
+
+_noreturn_ void freeze(void);