Fix fapolicyd for kiosk

diskless.pkr.hcl

@@ -531,14 +531,17 @@ build {
     destination = "${var.artifact_dir}/${var.image_version}/stig-arf.xml"
   }
 
-  # --- 2c-STIG-postfix. Keep the key-only APPLIANCE ACCOUNTS usable after STIG.
+  # --- 2c-STIG-postfix. Undo the STIG settings that lock out this architecture.
-  # STIG turns on pam_faillock (account phase) + password aging/inactivity
+  # Runs AFTER remediation so STIG can't re-break it. Two fixes:
-  # (account_disable_post_pw_expiration, accounts_*_age_*). The kiosk autologin
+  #  (a) APPLIANCE ACCOUNTS: STIG enables pam_faillock (account phase) + password
-  # user and the fleet admin are PASSWORDLESS / key-only, so the hardened account
+  #      aging/inactivity. The kiosk + fleet-admin accounts are PASSWORDLESS /
-  # phase rejects them: the kiosk PAM session never opens (getty keeps tty1), and
+  #      key-only, so the hardened account phase rejects them (kiosk session never
-  # sshd denies the otherwise-valid key as "Permission denied (publickey)" (key
+  #      opens; sshd account phase denies). Exempt just those two accounts.
-  # auth succeeds, then pam_acct_mgmt fails). Runs AFTER remediation so STIG can't
+  #  (b) fapolicyd: STIG's app-allowlisting blocks non-RPM executables, killing
-  # re-break it. Justified in docs/stig-deviations.md (#10, appliance accounts).
+  #      the kiosk (ExecStart=/usr/local/bin/kiosk-session). Keep fapolicyd (it's
+  #      the exec-guard for the writable overlay that dm-verity can't cover) and
+  #      trust the custom kiosk binary instead of disabling it.
+  # Both justified in docs/stig-deviations.md (#10).
   provisioner "shell" {
     environment_vars = [
       "ENABLE_STIG=${var.enable_stig}",
@@ -565,6 +568,17 @@ build {
       "  restorecon /etc/pam.d/sshd",
       "  echo \"==> /etc/pam.d/sshd: account-stack bypass ensured for $ADMIN_USER\"",
       "fi",
+      # fapolicyd (STIG app-allowlisting) blocks execution of any non-RPM file.
+      # The kiosk's ExecStart=/usr/local/bin/kiosk-session is exactly that, so
+      # fapolicyd kills the kiosk. KEEP fapolicyd (it's the real exec-guard for the
+      # writable RAM overlay, which dm-verity does NOT cover) and instead TRUST the
+      # custom kiosk files: add them to fapolicyd's trust DB and rebuild it. weston
+      # /chromium are RPM-installed so already trusted. (See docs/stig-deviations.md.)
+      "if command -v fapolicyd-cli >/dev/null 2>&1; then",
+      "  for f in /usr/local/bin/kiosk-session; do { [ -f \"$f\" ] && fapolicyd-cli --file add \"$f\"; } || true; done",
+      "  fapolicyd-cli --update 2>/dev/null || true",
+      "  echo '==> fapolicyd: trusted custom kiosk binaries + rebuilt trust db'",
+      "fi",
     ]
   }
 

docs/TODO.md

@@ -106,6 +106,24 @@ full session for seat); sshd gets an `account sufficient pam_succeed_if.so user
 keeps full STIG. Justified as register deviation #10. **Bench-verify on rebuild:
 kiosk renders + admin SSH works with hardening on.**
 
+**UPDATE 2026-06-19 — offline erofs inspection of the hardened image (mounted
+`rootfs.erofs` read-only) pinned the real causes; the PAM-account theory above
+was incomplete:**
+- **Kiosk = fapolicyd (CONFIRMED).** fapolicyd is `enabled` and `kiosk-session`
+  was not in its trust → it blocks `/usr/local/bin/kiosk-session`. Fix: KEEP
+  fapolicyd (it's the exec-guard for the writable overlay dm-verity can't cover)
+  and add the kiosk binary to its trust DB at build (2c-STIG-postfix). Not waived.
+- **SSH = NOT a STIG/PAM/SELinux issue.** `enforcing=0` didn't help; pam_namespace
+  is inert (empty namespace.conf); account bypass present; key file + perms +
+  sshd config all correct. Remaining cause: **baked key mismatch** — the
+  `SSH_AUTHORIZED_KEY` CI secret bakes `audiobuzz@s3`
+  (`SHA256:30p4vOQdWk0TdRQIR3VMo95yc6I7cRmCwIBRVjIHaV4`); if the client uses a
+  different private key → `Permission denied (publickey)`. AWAITING client
+  fingerprint / `ssh -vvv` to confirm; fix is the secret, not code.
+- Side note: SMBIOS cmdline-extra injection works only after a Proxmox VM
+  **stop/start** (not `reset`); and the debug-shell-on-tty2 attempt did not come
+  up — diagnosis ended up done via offline erofs mount instead.
+
 Remaining (future hardening, currently waived — see the register): FIPS,
 firewall, per-host/rotatable admin access. **Remote audit offload — WIRED both
 ends** (2026-06-19): receiver `serving/audit-collector/` (rsyslog TCP/514,

docs/stig-deviations.md

@@ -37,6 +37,12 @@ Posture decisions (2026-06-18, lab phase):
 ## Things that are NOT deviations (kept / enforced)
 
 - **SELinux enforcing** (`selinux --enforcing`; labels preserved into the erofs).
+- **fapolicyd** kept **enabled** (NOT waived) — it's the real exec-allowlist guard
+  for the writable RAM overlay, which dm-verity does **not** cover. The only
+  adjustment: the one non-RPM executable, `/usr/local/bin/kiosk-session`, is added
+  to fapolicyd's trust DB at build (provisioner 2c-STIG-postfix) so the kiosk runs;
+  everything else stays RPM-trust-only. (fapolicyd blocking that binary was the
+  original kiosk-won't-start cause.)
 - **auditd** enabled; STIG audit rules applied; logs locally.
 - **sshd hardening**: root login off, password & keyboard-interactive auth off,
   key-only (drop-in `50-fleet-hardening.conf`); STIG ciphers/MACs/KEX applied.