# The root instance-control service (isolated-exec design §5). Socket-
# activated, so it runs only while there is work. Its whole job is
# systemctl start/stop/set-property on the three instance templates' units
# and their host slices, on the agent's request, plus running the readiness
# probe. Every unit name it passes to systemctl is built from a validated
# host by the platform's own naming; nothing from a request reaches
# systemctl as a string.
#
# Runs as ROOT deliberately, with no new-privileges lockout and no syscall
# filter, for the systemctl child (the same rationale as
# carlos-stage.service). It accepts a connection only from the edge unit's
# main process (SO_PEERCRED uid carlos, pid == MainPID of
# carlos-edge.service): a flat tenant is a child of that pid, never the pid
# itself, and a nested tenant cannot reach the socket at all.
#
# At start it also looks UNDER the two instance bind mounts (unshare -m in
# a private namespace, umount, list; design §8.2, "Recovery for a layout
# that was unmounted anyway"). Files left in the raw mountpoint directory
# while the layout was unmounted would otherwise be hidden by the mount and
# invisible to the layout pass's destination check; a held file is a
# failing line in every probe report and keeps /ready refusing nested wakes
# until an operator moves or removes it and restarts this service. Root is
# what makes that inspection possible: the private namespace and the umount
# need CAP_SYS_ADMIN. ProtectSystem=strict leaves /data and /run visible.
[Unit]
Description=carlos instance control (root: start/stop carlos-instance*@ units for the agent)
Requires=carlos-instance-control.socket data-instances.mount run-carlos-instances.mount run-carlos-logs.mount
After=data-instances.mount run-carlos-instances.mount run-carlos-logs.mount

[Service]
ExecStart=/opt/carlos/carlos instance-control
ProtectSystem=strict
# /run/systemd/system: the per-generation runtime drop-in (zero-downtime
# deploys, plan 2 D1); nothing else is written there.
ReadWritePaths=/run/carlos-control /run/systemd/system
ProtectHome=yes
PrivateTmp=yes
PrivateDevices=yes
ProtectKernelTunables=yes
ProtectKernelModules=yes
ProtectKernelLogs=yes
RestrictAddressFamilies=AF_UNIX
MemoryDenyWriteExecute=yes
MemoryMax=64M