fix(firecracker): keep the host key canonical instead of clobbering the host token file

The previous firecracker fix let the VM generate its own signing key on the
guest volume and had the host overwrite the single host-wide control-plane-token
with it. That breaks a co-running Docker/macOS control plane: their orchestrators
still verify with the old key while new host clients start signing `cli` tokens
with the guest key, so supervise/teardown/policy calls against those backends
begin returning 401 (cross-backend operation is supported).

Keep the host token file the single source of truth. The launcher now pushes the
host-canonical key into the freshly booted infra VM over SSH (atomic write,
mirroring persist_env_var_secret); the VM's init waits for it and refuses to
start the control plane — rather than run open — if it never arrives. Nothing
ever writes back to the host file, so other backends are untouched, and the
best-effort silent path is gone (both the push and the guest fail loudly).

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>

bot_bottle/backend/firecracker/infra_vm.py

@@ -33,15 +33,17 @@ from pathlib import Path
 from typing import Generator
 
 from ...log import die, info
-from ...paths import CONTROL_PLANE_TOKEN_FILENAME, bot_bottle_root
+from ...paths import host_control_plane_token
 from .. import util as backend_util
 from ..docker import util as docker_mod
 from ..docker.gateway_provision import GatewayProvisionError
 from . import firecracker_vm, infra_artifact, netpool, util
 
-# Where the infra VM keeps its control-plane signing key (generated on the
+# The infra VM's control-plane signing-key path on its persistent /dev/vdb
-# persistent /dev/vdb volume mounted at BOT_BOTTLE_ROOT). The host mirrors it
+# volume (mounted at BOT_BOTTLE_ROOT=/var/lib/bot-bottle). The launcher seeds
-# back so the CLI signs `cli` tokens the VM verifies (issue #469 review).
+# this file with the host-canonical key BEFORE boot, so the VM verifies tokens
+# with the same key the host CLI signs with — the host token file stays the
+# single source of truth, never clobbered per-backend (issue #469 review).
 _GUEST_SIGNING_KEY_PATH = "/var/lib/bot-bottle/control-plane-token"
 
 # The single infra-VM image: gateway data plane + baked control-plane source
@@ -70,6 +72,10 @@ _INFRA_RESOLVER = "1.1.1.1"
 _HEALTH_TIMEOUT_SECONDS = 45.0
 _HEALTH_POLL_SECONDS = 0.5
 _CA_TIMEOUT_SECONDS = 30.0
+# How long the launcher retries pushing the signing key while the guest's SSH
+# comes up. Below the init's own wait window, so a failed push dies here first.
+_SIGNING_KEY_PUSH_TIMEOUT_SECONDS = 30.0
+_SIGNING_KEY_PUSH_POLL_SECONDS = 0.5
 
 
 @dataclass
@@ -173,44 +179,20 @@ def ensure_running() -> InfraVm:
     want = _expected_version()
     if _adoptable(key, url, want):
         info(f"adopting running infra VM at {url}")
-        return _with_signing_key(InfraVm(guest_ip=slot.guest_ip, private_key=key))
+        return InfraVm(guest_ip=slot.guest_ip, private_key=key)
 
     with _singleton_lock():
         # Re-check under the lock: another launcher may have booted it while
         # we waited for the lock (double-checked, so we adopt not re-boot).
         if _adoptable(key, url, want):
             info(f"adopting running infra VM at {url}")
-            return _with_signing_key(InfraVm(guest_ip=slot.guest_ip, private_key=key))
+            return InfraVm(guest_ip=slot.guest_ip, private_key=key)
         # Clear a stale/hung/OUTDATED VM holding the link before booting fresh.
         stop()
         ensure_built()
         infra = boot()
         wait_for_health(infra)
         _record_booted_version(want)
-        return _with_signing_key(infra)
-
-
-def _with_signing_key(infra: InfraVm) -> InfraVm:
-    """Mirror the infra VM's control-plane signing key (generated on its
-    persistent volume) into the host's control-plane-token file, so the host CLI
-    signs `cli` tokens the VM verifies (issue #469 review). Best-effort: an
-    unreadable key is logged, not fatal — the VM still enforces auth, but the CLI
-    may then be rejected until the key is readable. Returns `infra` for chaining."""
-    proc = subprocess.run(
-        util.ssh_base_argv(infra.private_key, infra.guest_ip)
-        + [f"cat {_GUEST_SIGNING_KEY_PATH}"],
-        capture_output=True, text=True, check=False,
-    )
-    signing_key = proc.stdout.strip()
-    if proc.returncode != 0 or not signing_key:
-        info("infra signing key not yet readable; control-plane auth may fail")
-        return infra
-    path = bot_bottle_root() / CONTROL_PLANE_TOKEN_FILENAME
-    path.parent.mkdir(parents=True, exist_ok=True)
-    fd = os.open(path, os.O_WRONLY | os.O_CREAT | os.O_TRUNC, 0o600)
-    with os.fdopen(fd, "w") as f:
-        f.write(signing_key)
-    os.chmod(path, stat.S_IRUSR | stat.S_IWUSR)
         return infra
 
 
@@ -267,7 +249,13 @@ def boot() -> InfraVm:
         data_drive=_ensure_registry_volume(),
     )
     _pid_file().write_text(str(vm.process.pid))
-    return InfraVm(guest_ip=slot.guest_ip, private_key=private_key, vm=vm)
+    infra = InfraVm(guest_ip=slot.guest_ip, private_key=private_key, vm=vm)
+    # Push the host-canonical control-plane signing key into the guest over SSH
+    # (the init waits for it before starting the control plane). The host token
+    # file stays the single source of truth, so a co-running Docker/macOS control
+    # plane keeps working (issue #469 review).
+    _push_signing_key(infra)
+    return infra
 
 
 def _infra_dir() -> Path:
@@ -339,6 +327,33 @@ def _ensure_registry_volume() -> Path:
     return vol
 
 
+def _push_signing_key(infra: InfraVm) -> None:
+    """Push the host-canonical control-plane signing key into the freshly booted
+    infra VM over SSH (atomic write), so its orchestrator verifies tokens with
+    the same key the host CLI signs from. Mirrors the existing
+    `persist_env_var_secret` transport. Retries until the guest's SSH is up (it
+    comes up before the init waits for this file); dies if it never lands, since
+    the VM then refuses to start its control plane rather than run OPEN."""
+    key = host_control_plane_token()
+    push = (
+        f"umask 077; cat > {_GUEST_SIGNING_KEY_PATH}.tmp "
+        f"&& mv {_GUEST_SIGNING_KEY_PATH}.tmp {_GUEST_SIGNING_KEY_PATH}"
+    )
+    deadline = time.monotonic() + _SIGNING_KEY_PUSH_TIMEOUT_SECONDS
+    last = ""
+    while time.monotonic() < deadline:
+        proc = subprocess.run(
+            util.ssh_base_argv(infra.private_key, infra.guest_ip) + [push],
+            input=key, capture_output=True, text=True, check=False,
+        )
+        if proc.returncode == 0:
+            return
+        last = proc.stderr.strip()
+        time.sleep(_SIGNING_KEY_PUSH_POLL_SECONDS)
+    die("could not push the control-plane signing key to the infra VM "
+        f"(its control plane will not start): {last or '<no stderr>'}")
+
+
 def _stable_keypair() -> tuple[Path, str]:
     """The infra VM's SSH keypair — generated once and reused, so any later
     launcher can SSH in (fetch CA / provision) even though a different process
@@ -517,32 +532,43 @@ mount -t ext4 /dev/vdb /var/lib/bot-bottle 2>/dev/null || true
 
 # Control plane. Source is baked at /app; the package is stdlib-only.
 cd /app
-# Control-plane signing key + a pre-minted `gateway` JWT, scoped per-process
+# Wait for the launcher to push the host-canonical control-plane signing key
-# (issue #469 review). The key is generated once on the persistent volume and
+# over SSH, then hand it ONLY to the orchestrator (to verify tokens); the
-# handed ONLY to the orchestrator (to verify tokens); the data-plane daemons get
+# data-plane daemons get a pre-minted `gateway` JWT they present, never the key.
-# the `gateway` JWT they present, never the key. Without this the control plane
+# If it never arrives, REFUSE to start the control plane rather than run OPEN —
-# would run OPEN and a compromised egress / supervise / git-http daemon in this
+# open mode would grant every unauthenticated caller the `cli` role, and a
-# same VM — reaching the orchestrator over 127.0.0.1, past the nft boundary that
+# compromised egress / supervise / git-http daemon in this same VM (reaching the
-# only fences off the separate agent VM — could drive the operator routes
+# orchestrator over 127.0.0.1, past the nft boundary that only fences off the
-# (approve its own supervise proposals, rewrite policy, read injected tokens).
+# separate agent VM) could then drive the operator routes (issue #469).
-CP_KEY=$(BOT_BOTTLE_ROOT=/var/lib/bot-bottle python3 -c 'from bot_bottle.paths import host_control_plane_token as t; print(t())')
+CP_KEY=""
-GW_JWT=$(BB_SIGNING_KEY="$CP_KEY" python3 -c 'import os; from bot_bottle.control_auth import mint, ROLE_GATEWAY; print(mint(ROLE_GATEWAY, os.environ["BB_SIGNING_KEY"]))')
+i=0
-
+while [ "$i" -lt 600 ]; do
-BOT_BOTTLE_ROOT=/var/lib/bot-bottle BOT_BOTTLE_CONTROL_PLANE_TOKEN="$CP_KEY" python3 -m bot_bottle.orchestrator \\
+    CP_KEY=$(cat /var/lib/bot-bottle/control-plane-token 2>/dev/null)
+    [ -n "$CP_KEY" ] && break
+    i=$((i + 1))
+    sleep 0.1
+done
+if [ -z "$CP_KEY" ]; then
+    echo "infra: control-plane signing key never arrived; refusing to start the control plane (would run OPEN)" >&2
+else
+    chmod 600 /var/lib/bot-bottle/control-plane-token 2>/dev/null || true
+    GW_JWT=$(BB_SIGNING_KEY="$CP_KEY" python3 -c 'import os; from bot_bottle.control_auth import mint, ROLE_GATEWAY; print(mint(ROLE_GATEWAY, os.environ["BB_SIGNING_KEY"]))')
+    BOT_BOTTLE_ROOT=/var/lib/bot-bottle BOT_BOTTLE_CONTROL_PLANE_TOKEN="$CP_KEY" python3 -m bot_bottle.orchestrator \\
         --host 0.0.0.0 --port {CONTROL_PLANE_PORT} --broker stub &
 
-# Gateway data plane, multi-tenant: each request resolves source-IP ->
+    # Gateway data plane, multi-tenant: each request resolves source-IP ->
-# policy against the local control plane. The VM backend reaches git over
+    # policy against the local control plane. The VM backend reaches git over
-# git-http (9420), so the git:// daemon (git-gate, needs a per-bottle
+    # git-http (9420), so the git:// daemon (git-gate, needs a per-bottle
-# entrypoint the consolidated model doesn't use) is left out. No
+    # entrypoint the consolidated model doesn't use) is left out. No
-# SUPERVISE_DB_PATH: the data plane reaches the supervise queue over the
+    # SUPERVISE_DB_PATH: the data plane reaches the supervise queue over the
-# control-plane RPC and never opens bot-bottle.db (PRD 0070 / #469). It presents
+    # control-plane RPC and never opens bot-bottle.db (PRD 0070 / #469). It
-# the pre-minted `gateway` JWT; gateway_init keeps the signing key out of the
+    # presents the pre-minted `gateway` JWT; gateway_init keeps the signing key
-# data-plane daemons' env.
+    # out of the data-plane daemons' env.
-BOT_BOTTLE_GATEWAY_DAEMONS=egress,git-http,supervise \\
+    BOT_BOTTLE_GATEWAY_DAEMONS=egress,git-http,supervise \\
-BOT_BOTTLE_ORCHESTRATOR_URL=http://127.0.0.1:{CONTROL_PLANE_PORT} \\
+    BOT_BOTTLE_ORCHESTRATOR_URL=http://127.0.0.1:{CONTROL_PLANE_PORT} \\
-BOT_BOTTLE_CONTROL_AUTH_JWT="$GW_JWT" \\
+    BOT_BOTTLE_CONTROL_AUTH_JWT="$GW_JWT" \\
         python3 -m bot_bottle.gateway_init &
+fi
 
 # Reap as PID 1; children are backgrounded, so `wait` blocks.
 while : ; do wait ; done

tests/unit/test_firecracker_infra_vm.py

@@ -8,7 +8,6 @@ decisions that must hold without a VM.
 from __future__ import annotations
 
 import os
-import tempfile
 import unittest
 from pathlib import Path
 from unittest.mock import MagicMock, patch
@@ -16,35 +15,36 @@ from unittest.mock import MagicMock, patch
 from bot_bottle.backend.firecracker import infra_vm
 
 
-class TestSigningKeySync(unittest.TestCase):
+class TestPushSigningKey(unittest.TestCase):
-    """`_with_signing_key` mirrors the VM's control-plane signing key into the
+    """`_push_signing_key` pushes the host-canonical key into the guest over SSH
-    host token file so the CLI signs `cli` tokens the VM verifies (issue #469)."""
+    (the init waits for it). The host token file stays the single source of
+    truth, never clobbered per-backend (issue #469)."""
 
     def _infra(self) -> infra_vm.InfraVm:
         return infra_vm.InfraVm(guest_ip="10.0.0.1", private_key=Path("/k"))
 
-    def test_mirrors_guest_key_to_host_file(self):
+    def test_writes_host_key_over_ssh_atomically(self):
-        with tempfile.TemporaryDirectory() as d:
+        proc = MagicMock(returncode=0, stderr="")
-            root = Path(d)
+        with patch.object(infra_vm, "host_control_plane_token", return_value="host-key"), \
-            proc = MagicMock(returncode=0, stdout="the-signing-key\n")
+                patch.object(infra_vm.subprocess, "run", return_value=proc) as run:
-            with patch.object(infra_vm.subprocess, "run", return_value=proc) as run, \
+            infra_vm._push_signing_key(self._infra())
-                    patch.object(infra_vm, "bot_bottle_root", return_value=root):
+        # Piped the key to an atomic write of the guest token path.
-                out = infra_vm._with_signing_key(self._infra())
+        self.assertEqual("host-key", run.call_args.kwargs["input"])
-            self.assertIsInstance(out, infra_vm.InfraVm)  # returned for chaining
+        remote_cmd = run.call_args.args[0][-1]
-            token = root / infra_vm.CONTROL_PLANE_TOKEN_FILENAME
+        self.assertIn(f"cat > {infra_vm._GUEST_SIGNING_KEY_PATH}.tmp", remote_cmd)
-            self.assertEqual("the-signing-key", token.read_text())
+        self.assertIn(f"mv {infra_vm._GUEST_SIGNING_KEY_PATH}.tmp "
-            self.assertEqual(0o600, token.stat().st_mode & 0o777)
+                      f"{infra_vm._GUEST_SIGNING_KEY_PATH}", remote_cmd)
-            # It cat'd the guest volume path over SSH.
-            self.assertIn(infra_vm._GUEST_SIGNING_KEY_PATH, run.call_args.args[0][-1])
 
-    def test_unreadable_key_is_not_fatal(self):
+    def test_dies_when_push_never_succeeds(self):
-        with tempfile.TemporaryDirectory() as d:
+        proc = MagicMock(returncode=255, stderr="ssh: connect refused")
-            root = Path(d)
+        with patch.object(infra_vm, "host_control_plane_token", return_value="host-key"), \
-            proc = MagicMock(returncode=255, stdout="")
+                patch.object(infra_vm, "_SIGNING_KEY_PUSH_TIMEOUT_SECONDS", 0.05), \
-            with patch.object(infra_vm.subprocess, "run", return_value=proc), \
+                patch.object(infra_vm, "_SIGNING_KEY_PUSH_POLL_SECONDS", 0.0), \
-                    patch.object(infra_vm, "bot_bottle_root", return_value=root):
+                patch.object(infra_vm.subprocess, "run", return_value=proc), \
-                infra_vm._with_signing_key(self._infra())  # no raise
+                patch.object(infra_vm, "die", side_effect=SystemExit) as die:
-            self.assertFalse((root / infra_vm.CONTROL_PLANE_TOKEN_FILENAME).exists())
+            with self.assertRaises(SystemExit):
+                infra_vm._push_signing_key(self._infra())
+        die.assert_called_once()
 
 
 class TestControlPlaneUrl(unittest.TestCase):
@@ -79,10 +79,11 @@ class TestBuildInfraRootfs(unittest.TestCase):
         # VM backend uses git-http (9420); the git:// daemon is left out.
         self.assertIn("BOT_BOTTLE_GATEWAY_DAEMONS=egress,git-http,supervise", init)
         # Role-scoped control-plane auth (issue #469 review): the orchestrator
-        # gets the signing key, the gateway daemons get a pre-minted `gateway`
+        # gets the host-seeded signing key, the gateway daemons get a pre-minted
-        # JWT — never open mode in the infra VM.
+        # `gateway` JWT, and the VM refuses to run OPEN if the key is missing.
-        self.assertIn("host_control_plane_token", init)  # key generated on the volume
+        self.assertIn("cat /var/lib/bot-bottle/control-plane-token", init)  # host-seeded key
-        self.assertIn("mint, ROLE_GATEWAY", init)         # gateway JWT minted from it
+        self.assertIn("refusing to start the control plane", init)          # no open mode
+        self.assertIn("mint, ROLE_GATEWAY", init)                           # gateway JWT
         self.assertIn('BOT_BOTTLE_CONTROL_PLANE_TOKEN="$CP_KEY" python3 -m bot_bottle.orchestrator',
                       init)  # key -> orchestrator only
         self.assertIn('BOT_BOTTLE_CONTROL_AUTH_JWT="$GW_JWT"', init)  # JWT -> gateway daemons