Inside a user namespace, why am I not allowed to remount a filesystem I have mounted?

Question

https://github.com/systemd/systemd/issues/9914#issuecomment-416387637

$ uname -r
4.17.18-200.fc28.x86_64

$ unshare -U -r -m
# mkdir TMP
# mount -t tmpfs tmpfs TMP/
# mount -o remount,ro TMP/
mount: ./TMP: permission denied.

---

# grep TMP /proc/self/mountinfo
834 831 0:74 / /home/alan/TMP rw,relatime - tmpfs tmpfs rw,seclabel,uid=1001,gid=1001
# strace -f mount -o remount TMP/
...
mount("tmpfs", "/home/alan/TMP", 0x557c3cec9600, MS_REMOUNT|MS_RELATIME, "seclabel,uid=1001,gid=1001") = -1 EPERM (Operation not permitted)
...

Bind-remounting works ok.

 # strace -f mount -o remount,bind,ro TMP/
 mount("tmpfs", "/home/alan/TMP", 0x5615b7ebc130, MS_RDONLY|MS_REMOUNT|MS_BIND|MS_RELATIME, "seclabel,uid=1001,gid=1001") = 0

Answer 1

It was not implemented. It is in the next version though!

v4.18 includes the following commit. This does not provide any more specific reason for it not being supported previously.

https://github.com/torvalds/linux/commit/bc6155d13260

fs: Allow superblock owner to access do_remount_sb()

Superblock level remounts are currently restricted to global CAP_SYS_ADMIN, as is the path for changing the root mount to read only on umount. Loosen both of these permission checks to also allow CAP_SYS_ADMIN in any namespace which is privileged towards the userns which originally mounted the filesystem.