bootc status fails with EBUSY when the ESP is already mounted (e.g. via systemd.mount-extra)

[dustinkirkland]

On composefs-based bootc hosts where the ESP is auto-mounted in the root
mount namespace before bootc status runs, the status command fails
consistently with:

error: Status: Mounting /dev/vda1: Device or resource busy (os error 16)

Reproduction

Any bootc composefs deployment where the ESP is pre-mounted. One common
path: bootc install to-filesystem writes
systemd.mount-extra=UUID=<ESP>:/boot:auto:ro into the deployment kernel
cmdline (visible in /proc/cmdline). At boot, systemd's fstab-generator
processes the cmdline arg and emits a boot.mount unit that mounts the
ESP read-only at /boot before user services start.

bootc status at any later point then hits:

crates/lib/src/store/mod.rs BootedStorage::new (ComposefsBooted arm)
→ crates/lib/src/bootc_composefs/boot.rs mount_esp(...)
→ crates/mount/src/tempmount.rs TempMount::mount_dev(...)
→ rustix::mount::mount(dev, tmpdir, "vfat", ...)

which returns EBUSY because the same block device is already mounted
elsewhere in the same mount namespace.

Environment where reproduced

• bootc v1.16.6 (also reproduced on main at 3bfa9dd)
• Linux 6.18-series kernel (mainline)
• systemd 261.2
• QEMU x86_64, single-deployment composefs bootc host
• /boot mounted ro via systemd.mount-extra=UUID=...:/boot:auto:ro

Impact

bootc status, bootc status --json, and any consumer scripting
around them are broken on these hosts.

Suggested fix

Split the "mount ESP" primitive into mount_esp (unchanged, fresh
mount(2)) and a new mount_esp_readonly that gracefully reuses an
existing ESP mount via open_tree(OPEN_TREE_CLONE) + move_mount.
Wire the read-only status path to the latter; leave the write-path
callers (bootc install, gc) using mount_esp and failing loudly
on EBUSY — silently writing to a ro clone would be worse.

Fix + tests: (PR link forthcoming)