core: split out setup_private_users_child()

Drop support for kernels older than 3.19, as this is where
/proc/<pid>/setgroups was added.

9cc46516dd
This commit is contained in:
Matteo Croce
2024-11-26 11:54:29 +01:00
parent 9554ac3052
commit 2c7dabff50

View File

@@ -2234,6 +2234,42 @@ static int build_pass_environment(const ExecContext *c, char ***ret) {
return 0;
}
static int setup_private_users_child(int unshare_ready_fd, const char *uid_map, const char *gid_map, bool allow_setgroups) {
int r;
/* Child process, running in the original user namespace. Let's update the parent's UID/GID map from
* here, after the parent opened its own user namespace. */
pid_t ppid = getppid();
/* Wait until the parent unshared the user namespace */
uint64_t c;
if (read(unshare_ready_fd, &c, sizeof(c)) < 0)
return log_debug_errno(errno, "Failed to read from signaling eventfd: %m");
/* Disable the setgroups() system call in the child user namespace, for good, unless PrivateUsers=full
* and using the system service manager. */
const char *a = procfs_file_alloca(ppid, "setgroups");
const char *setgroups = allow_setgroups ? "allow" : "deny";
r = write_string_file(a, setgroups, WRITE_STRING_FILE_DISABLE_BUFFER);
if (r < 0)
return log_debug_errno(r, "Failed to write '%s' to %s: %m", setgroups, a);
/* First write the GID map */
a = procfs_file_alloca(ppid, "gid_map");
r = write_string_file(a, gid_map, WRITE_STRING_FILE_DISABLE_BUFFER);
if (r < 0)
return log_debug_errno(r, "Failed to write GID map to %s: %m", a);
/* Then write the UID map */
a = procfs_file_alloca(ppid, "uid_map");
r = write_string_file(a, uid_map, WRITE_STRING_FILE_DISABLE_BUFFER);
if (r < 0)
return log_debug_errno(r, "Failed to write UID map to %s: %m", a);
return 0;
}
static int setup_private_users(PrivateUsers private_users, uid_t ouid, gid_t ogid, uid_t uid, gid_t gid, bool allow_setgroups) {
_cleanup_free_ char *uid_map = NULL, *gid_map = NULL;
_cleanup_close_pair_ int errno_pipe[2] = EBADF_PAIR;
@@ -2339,69 +2375,10 @@ static int setup_private_users(PrivateUsers private_users, uid_t ouid, gid_t ogi
if (r < 0)
return r;
if (r == 0) {
_cleanup_close_ int fd = -EBADF;
const char *a;
pid_t ppid;
/* Child process, running in the original user namespace. Let's update the parent's UID/GID map from
* here, after the parent opened its own user namespace. */
ppid = getppid();
errno_pipe[0] = safe_close(errno_pipe[0]);
/* Wait until the parent unshared the user namespace */
if (read(unshare_ready_fd, &c, sizeof(c)) < 0)
report_errno_and_exit(errno_pipe[1], -errno);
/* Disable the setgroups() system call in the child user namespace, for good, unless PrivateUsers=full
* and using the system service manager. */
a = procfs_file_alloca(ppid, "setgroups");
fd = open(a, O_WRONLY|O_CLOEXEC);
if (fd < 0) {
if (errno != ENOENT) {
r = log_debug_errno(errno, "Failed to open %s: %m", a);
report_errno_and_exit(errno_pipe[1], r);
}
/* If the file is missing the kernel is too old, let's continue anyway. */
} else {
const char *setgroups = allow_setgroups ? "allow\n" : "deny\n";
if (write(fd, setgroups, strlen(setgroups)) < 0) {
r = log_debug_errno(errno, "Failed to write '%s' to %s: %m", setgroups, a);
report_errno_and_exit(errno_pipe[1], r);
}
fd = safe_close(fd);
}
/* First write the GID map */
a = procfs_file_alloca(ppid, "gid_map");
fd = open(a, O_WRONLY|O_CLOEXEC);
if (fd < 0) {
r = log_debug_errno(errno, "Failed to open %s: %m", a);
r = setup_private_users_child(unshare_ready_fd, uid_map, gid_map, allow_setgroups);
if (r < 0)
report_errno_and_exit(errno_pipe[1], r);
}
if (write(fd, gid_map, strlen(gid_map)) < 0) {
r = log_debug_errno(errno, "Failed to write GID map to %s: %m", a);
report_errno_and_exit(errno_pipe[1], r);
}
fd = safe_close(fd);
/* The write the UID map */
a = procfs_file_alloca(ppid, "uid_map");
fd = open(a, O_WRONLY|O_CLOEXEC);
if (fd < 0) {
r = log_debug_errno(errno, "Failed to open %s: %m", a);
report_errno_and_exit(errno_pipe[1], r);
}
if (write(fd, uid_map, strlen(uid_map)) < 0) {
r = log_debug_errno(errno, "Failed to write UID map to %s: %m", a);
report_errno_and_exit(errno_pipe[1], r);
}
_exit(EXIT_SUCCESS);
}