stripping ssh host keys from build

diskless.pkr.hcl

@@ -278,6 +278,11 @@ build {
       "printf 'PermitRootLogin no\\nPasswordAuthentication no\\nKbdInteractiveAuthentication no\\n' > /etc/ssh/sshd_config.d/50-fleet-hardening.conf",
       "chmod 0644 /etc/ssh/sshd_config.d/50-fleet-hardening.conf",
       "restorecon /etc/ssh/sshd_config.d/50-fleet-hardening.conf",
+      # Guarantee sshd is enabled (it's our ONLY way in once root is locked and
+      # the admin user is key-only). Don't rely on the distro preset. Validate
+      # the full sshd config so a bad drop-in fails the BUILD, not the boot.
+      "systemctl enable sshd.service",
+      "sshd -t",
       # Lock root in the image. Done last; admin access is via the key user.
       "passwd -l root",
       "echo \"root locked; admin user $ADMIN_USER created (wheel + NOPASSWD sudo, key-only).\"",

docs/TODO.md

@@ -64,6 +64,14 @@ and cleanup.
   but document/deviate), and the admin key/user are **fleet-wide baked** — for
   per-host/rotatable access without rebuilds, layer the signed HTTP/SMBIOS
   config-agent or an SSH CA (`TrustedUserCAKeys`) discussed in session.
+- **SSH host keys (done):** `extract-artifacts.sh` strips `/etc/ssh/ssh_host_*`
+  so they're not baked (baked keys = shared fleet identity + the build's
+  perms/group don't survive the tar→erofs round-trip, so sshd rejected them
+  "Permissions 640 too open" and refused to start). RHEL's `sshd-keygen@.service`
+  regenerates them into the overlay before sshd starts. **⚠ Caveat:** the
+  ephemeral overlay means they regenerate **every boot** → the host key changes
+  per reboot (clients warn). Follow-up for a stable host identity: persist the
+  host keys or derive them from the TPM (ties to the hostnqn/identity model).
 
 ## [ ] 3. Roll the errata into design doc v3
 

scripts/extract-artifacts.sh

@@ -68,6 +68,17 @@ tar -C "$WORK/root" --numeric-owner --xattrs --xattrs-include='*' --acls --selin
     -xpf "$WORK/rootfs.tar"
 rm -rf "$WORK/root/boot/"* "$WORK/rootfs.tar"   # diskless hosts don't need /boot
 
+# Strip SSH host keys. They are PER-HOST identity (invariant #1) and must not be
+# baked: a baked key means every host shares one SSH identity (MITM risk), and
+# the build VM's key perms/group don't survive the tar->erofs round-trip, so
+# sshd rejects them ("Permissions 640 ... too open") and refuses to start.
+# RHEL's sshd-keygen@.service regenerates them with correct perms before sshd
+# starts, into the writable volatile overlay. NOTE: with the ephemeral overlay
+# they regenerate EVERY boot, so the host key changes each reboot (SSH clients
+# will warn) — for a stable identity, persist them or derive from the TPM
+# (future; ties to the hostnqn/identity model).
+rm -f "$WORK/root/etc/ssh/"ssh_host_*
+
 # Fail loudly if labels didn't make it — shipping an unlabeled enforcing image
 # boots but denies all logins (the symptom we just chased down).
 if ! getfattr -h -n security.selinux "$WORK/root/usr/bin/bash" >/dev/null 2>&1; then