fix(console): D1 security re-check fixes for the dev-owner start (#1873)

The agent-run refusal on a dev-local-owner start now checks the starting
process's own environment as well as the environment given for the
Console, so a caller cannot pass a clean env to get past it.

The start lock now names its holder (pid and start time). A detached
start that times out and kills its supervisor's process group clears the
lock only when it names that supervisor and the supervisor is gone, so a
forced kill no longer blocks every later start.

docs/TOOLS.md

@@ -333,7 +333,7 @@ For local development on a named instance only, `mosaic bus start BUSINESS --dev
 
 `mosaic console start --dev-local-owner --dev-insecure-bind ADDR --dev-allow-host NAME[,NAME...]` opens that dev owner Console to other machines, for UI testing from another device on a trusted LAN. It has no sign-in: anyone who can reach ADDR:PORT acts as the owner. Start, and `mosaic console status` while that Console runs, print `WARNING: dev owner Console listens on ADDR:PORT with no sign-in; anyone who can reach it acts as the owner.` on stderr. ADDR is `0.0.0.0` or an IPv4 address, and a specific address must also be one of the names. NAME is a host name or IPv4 address without a port; give several as one comma-separated value. The Console answers only `NAME:<bound port>` and the loopback names, and takes a write only when Origin is `http://` plus that same Host. A browser on such an http origin sends no Fetch Metadata, so for an allowed NAME a write with that exact Origin and no `Sec-Fetch-Site` counts as same-origin; a `Sec-Fetch-Site` the browser did send is still checked. Both flags are refused without `--dev-local-owner`, and `--dev-insecure-bind` is refused without a `--dev-allow-host`. SSO, preview and tester Consoles stay on loopback. The Console state file records `bind` and `allowHosts`. `mosaic console status` prints the state as JSON (`bind` is `127.0.0.1` for a loopback Console) and never moves a stale state file.
 
-`mosaic console restart` and `mosaic org reload ORG` are the Console's Settings controls (F3) from a terminal (U1, #1873). Each sends one request to the running Console named in the instance's console state, as the page sends it (the Console's own Host and Origin, `Sec-Fetch-Site: same-origin`, a JSON `{}` body), so the Console applies the same owner, `--send` and origin checks and writes the same action trail, `<dataRoot>/console/actions.jsonl`. Both refuse inside an agent run (exit 3) before reading anything, and refuse with exit 3 when no Console runs. `console restart` works only for a Console started in the foreground by `mosaic console start`, or by `mosaic console start --dev-local-owner --detach`, which leaves a detached supervisor (`packages/cli/src/console-supervisor.mjs`, its own session) running the Console as the foreground start does; its receipt adds the supervisor's `supervisor` pid, and its later warnings go to `console/supervisor-*.log` under the data root. `mosaic console stop` stops that Console, and its supervisor then exits. `mosaic console start --dev-local-owner` refuses inside an agent run (exit 3), with or without `--detach`; a start without `--dev-local-owner` is unchanged. Any other Console answers `not-supervised`. `console restart` waits up to 30 seconds for the relaunched Console to answer and prints `{"outcome":"restarted","from":OLD,"boot":NEW}`. `org reload ORG` restarts a running bus or starts a stopped one and prints the Console's done receipt. A request that left but got no answer is reported as outcome unknown (exit 1) and is never retried; check `mosaic console status`, `mosaic bus status ORG` and the trail. `mosaic bus restart ORG` remains the trail-less host verb.
+`mosaic console restart` and `mosaic org reload ORG` are the Console's Settings controls (F3) from a terminal (U1, #1873). Each sends one request to the running Console named in the instance's console state, as the page sends it (the Console's own Host and Origin, `Sec-Fetch-Site: same-origin`, a JSON `{}` body), so the Console applies the same owner, `--send` and origin checks and writes the same action trail, `<dataRoot>/console/actions.jsonl`. Both refuse inside an agent run (exit 3) before reading anything, and refuse with exit 3 when no Console runs. `console restart` works only for a Console started in the foreground by `mosaic console start`, or by `mosaic console start --dev-local-owner --detach`, which leaves a detached supervisor (`packages/cli/src/console-supervisor.mjs`, its own session) running the Console as the foreground start does; its receipt adds the supervisor's `supervisor` pid, and its later warnings go to `console/supervisor-*.log` under the data root. `mosaic console stop` stops that Console, and its supervisor then exits. `mosaic console start --dev-local-owner` refuses inside an agent run (exit 3), with or without `--detach`, checking the starting process's own environment as well as the one given for the Console; a start without `--dev-local-owner` is unchanged. A detached start whose supervisor does not reply within 60 seconds signals the supervisor's process group, then kills it, and clears `console/start.lock` only when the lock names that supervisor (pid and start time) and it is gone. Any other Console answers `not-supervised`. `console restart` waits up to 30 seconds for the relaunched Console to answer and prints `{"outcome":"restarted","from":OLD,"boot":NEW}`. `org reload ORG` restarts a running bus or starts a stopped one and prints the Console's done receipt. A request that left but got no answer is reported as outcome unknown (exit 1) and is never retried; check `mosaic console status`, `mosaic bus status ORG` and the trail. `mosaic bus restart ORG` remains the trail-less host verb.
 
 The instance processing switch is separate from per-business launches off/on.
 `mosaic halt` durably stops managed processing across every business in the

packages/cli/src/cli.mjs

@@ -596,7 +596,7 @@ async function consoleCommand(rest, io, deps) {
   // --dev-local-owner. ADDR is 0.0.0.0 or an IPv4 address; NAME[,NAME...] has no ports.
   const devOwner = parsed.flags.has("--dev-local-owner"), bind = parsed.values["--dev-insecure-bind"] ?? null;
   // D1: an agent run never starts a dev-local-owner Console (org-console.mjs consoleStart checks again).
-  if (devOwner) refuseInsideAgent(io.env);
+  if (devOwner) { refuseInsideAgent(process.env); refuseInsideAgent(io.env); }
   const names = Object.hasOwn(parsed.values, "--dev-allow-host") ? [parsed.values["--dev-allow-host"]] : [];
   let lan;
   try { lan = devLanOptions({ devOwner, bind, names }); } catch (err) { throw usage(err.message); }

packages/cli/src/org-console.mjs

@@ -3,7 +3,7 @@ import { randomBytes } from 'node:crypto';
 import { processState } from '../../seat/src/process-state.mjs';
 import { spawn } from 'node:child_process';
 import { request as httpRequest } from 'node:http';
-import { constants, closeSync, fstatSync, lstatSync, mkdirSync, openSync, readFileSync, rmSync, renameSync, writeFileSync } from 'node:fs';
+import { constants, closeSync, fstatSync, lstatSync, mkdirSync, openSync, readFileSync, rmSync, renameSync, writeFileSync, writeSync } from 'node:fs';
 import { basename, join } from 'node:path';
 import { fileURLToPath } from 'node:url';
 import { listOrgs, resolveOrg } from '../../organization/src/org-registry.mjs';
@@ -64,8 +64,25 @@ function takeLock(ctx) {
   checkDir(ctx);
   const path = join(directory(ctx), 'start.lock');
   let fd; try { fd = openSync(path, 'wx', 0o600); } catch { fail('console start lock exists'); }
+  // D1: the lock names its holder, so a detached start that timed out can clear the lock of the
+  // supervisor it killed (clearDeadLock) and no other.
+  try { writeSync(fd, JSON.stringify({ pid: process.pid, startTime: startTimeOf(process.pid) })); } catch { /* an unnamed lock is only never cleared */ }
   return () => { closeSync(fd); rmSync(path, { force: true }); };
 }
+// Removes start.lock only when it names this exact process (pid and start time) and that process
+// is gone. A lock that is unreadable, unnamed or another holder's stays.
+function clearDeadLock(ctx, pid, startTime) {
+  if (!startTime) return;
+  const path = join(directory(ctx), 'start.lock');
+  let owner;
+  try {
+    const st = lstatSync(path);
+    if (!st.isFile() || st.uid !== process.getuid()) return;
+    owner = JSON.parse(readFileSync(path, 'utf8'));
+  } catch { return; }
+  if (owner?.pid !== pid || owner.startTime !== startTime || startTimeOf(pid) === startTime) return;
+  rmSync(path, { force: true });
+}
 // The exit code a supervised Console uses to ask for a relaunch (EX_TEMPFAIL). Only the Console's
 // own restart route (POST /api/console/restart, a human action) exits with it.
 export const RESTART_EXIT = 75;
@@ -169,6 +186,7 @@ async function detachedSupervisor(ctx, { port, bind, allowHosts, warn }) {
     child = spawn(process.execPath, [SUPERVISOR, JSON.stringify({ configDir: ctx.configDir, dataRoot: ctx.dataRoot, instance: ctx.instance, port, bind, allowHosts })],
       { env: ctx.env ?? process.env, detached: true, stdio: ['ignore', 'pipe', fd] });
   } finally { closeSync(fd); }
+  const supervisorStart = child.pid ? startTimeOf(child.pid) : null;
   const done = new Promise(resolve => { child.once('error', () => resolve(1)); child.once('exit', code => resolve(code ?? 1)); });
   // The log is named under the data root, never by host path (PR #1916).
   const unclear = `console failed to start; see console/${name} under the data root`;
@@ -184,13 +202,15 @@ async function detachedSupervisor(ctx, { port, bind, allowHosts, warn }) {
   clearTimeout(timer);
   child.stdout.destroy();
   // The supervisor leads its own process group, with the Console it may have started: a timed-out
-  // start signals that group, then kills it if it is still there after 10 seconds.
+  // start signals that group, then kills it if it is still there after 10 seconds. A supervisor
+  // killed while it held the start lock never released it, so the lock it named is cleared.
   if (line === undefined) {
     const group = signal => { try { process.kill(-child.pid, signal); } catch { /* already gone */ } };
     group('SIGTERM');
     let grace; const exited = await Promise.race([done.then(() => true), new Promise(r => { grace = setTimeout(() => r(false), 10000); })]);
     clearTimeout(grace);
     if (!exited) { group('SIGKILL'); await done; }
+    clearDeadLock(ctx, child.pid, supervisorStart);
     fail(unclear);
   }
   let reply = null;
@@ -217,8 +237,9 @@ export async function consoleStart(ctx, { devLocalOwner = false, detach = false,
   let lan;
   try { lan = devLanOptions({ devOwner: devLocalOwner, bind, names: allowHosts }); } catch (e) { throw new CliError(e.message, 4); }
   // D1 fix round (security review P1): a dev-local-owner Console acts as the owner with no sign-in,
-  // so an agent run never starts one, detached or not. Other starts are unchanged.
-  if (devLocalOwner) refuseInsideAgent(ctx.env ?? process.env);
+  // so an agent run never starts one, detached or not. Other starts are unchanged. This process's own
+  // environment is checked as well as the one given for the Console (security re-check P1).
+  if (devLocalOwner) { refuseInsideAgent(process.env); if (ctx.env) refuseInsideAgent(ctx.env); }
   if (detach && devLocalOwner) return detachedSupervisor(ctx, { port, bind, allowHosts, warn });
   const release = takeLock(ctx);
   let run, args;

packages/cli/tests/org-console.test.mjs

@@ -3,7 +3,8 @@ import assert from 'node:assert/strict';
 import { spawn, spawnSync } from 'node:child_process';
 import { mkdtempSync, mkdirSync, writeFileSync, readdirSync, readFileSync, rmSync, existsSync, statSync } from 'node:fs';
 import { get, request } from 'node:http';
-import { join } from 'node:path';
+import { dirname, join } from 'node:path';
+import { pathToFileURL } from 'node:url';
 import { tmpdir } from 'node:os';
 import { reserveOrg, activateOrg } from '../../business/src/org-registry.mjs';
 import { humanEnv } from './helpers.mjs';
@@ -343,6 +344,13 @@ test('an agent run cannot start a dev-local-owner console; a start without it is
   for (const detach of [true, false]) {
     await assert.rejects(consoleStart(agent, { devLocalOwner: true, detach, port: 'auto', warn: () => {} }), e => e.exitCode === 3 && /only from a human shell/.test(e.message));
   }
+  // A clean environment given for the Console does not hide this process's own marker (re-check P1).
+  process.env.MOSAIC_RUN_ID = 'run-2';
+  try {
+    for (const detach of [true, false]) {
+      await assert.rejects(consoleStart(ctx, { devLocalOwner: true, detach, port: 'auto', warn: () => {} }), e => e.exitCode === 3 && /found MOSAIC_RUN_ID/.test(e.message));
+    }
+  } finally { delete process.env.MOSAIC_RUN_ID; }
   const agentCli = args => spawnSync(join(REPO, 'scripts/mosaic'), args, { env: { ...ctx.env, MOSAIC_AGENT_RUN: '1' }, encoding: 'utf8', timeout: 30000 });
   const refused = agentCli(['console', 'start', '--dev-local-owner', '--detach', '--port', 'auto']);
   assert.equal(refused.status, 3); assert.equal(refused.stdout, ''); assert.match(refused.stderr, /only from a human shell/);
@@ -355,6 +363,31 @@ test('an agent run cannot start a dev-local-owner console; a start without it is
   assert.equal((await consoleStop(ctx)).stopped, true);
 });
 
+// D1 security re-check P2: a supervisor that hangs holding the start lock is killed with its group
+// after the 60 s start timeout and 10 s grace. Its lock names it, so the start clears it and the
+// next start proceeds. The hook blocks the supervisor's event loop once it holds the lock (at its
+// console state read), so its stop handler never runs and only the group SIGKILL ends it.
+test('a detached start that times out clears the start lock of the supervisor it killed', { timeout: 150000 }, async t => {
+  const { ctx } = await instance(t);
+  const hook = join(dirname(ctx.dataRoot), 'hang-after-lock.mjs');
+  writeFileSync(hook, `import fs from 'node:fs'; import { syncBuiltinESMExports } from 'node:module';
+    if (process.argv[1]?.endsWith('console-supervisor.mjs')) { const real = fs.openSync;
+      fs.openSync = (...args) => { if (String(args[0]).endsWith('/console/console.json')) Atomics.wait(new Int32Array(new SharedArrayBuffer(4)), 0, 0); return real(...args); };
+      syncBuiltinESMExports(); }\n`);
+  const hung = { ...ctx, env: { ...ctx.env, NODE_OPTIONS: `--import=${pathToFileURL(hook).href}` } };
+  const lock = join(ctx.dataRoot, 'console/start.lock');
+  const began = Date.now();
+  await assert.rejects(consoleStart(hung, { devLocalOwner: true, detach: true, port: 'auto', warn: () => {} }),
+    e => e.exitCode === 3 && /^console failed to start; see console\/supervisor-\d+-\d+\.log under the data root$/.test(e.message));
+  assert.ok(Date.now() - began >= 69000, 'the start waited out its timeout and grace');
+  assert.equal(existsSync(lock), false);
+  assert.deepEqual(consoleStatus(ctx), { running: false });
+  // The next start is not blocked by the killed supervisor's lock.
+  const result = await consoleStart(ctx, { devLocalOwner: true, detach: true, port: 'auto', warn: () => {} });
+  assert.equal((await consoleStop(ctx)).stopped, true);
+  assert.equal(await exitOf(result.done), 0);
+});
+
 // Without --dev-local-owner a detached start is what it was before D1: no supervisor, restart refused.
 test('a detached console without the dev owner has no supervisor, so it refuses a restart and stays up', { timeout: 60000 }, async t => {
   const { ctx, mosaic } = await instance(t);