fix(hub): a per-account sign-in lockout, and a reviewed unauthenticated surface

Passphrase sign-in locks per username: after `login.max_failures` wrong
passphrases (default 4) the name is refused with `429 account_locked` and a
`Retry-After` for `login.lockout_minutes` (default 60), without the passphrase
being checked. Both numbers are instance policy an admin sets from the panel;
zero failures turns it off. The per-IP limit bounds one address, and IPv6
gives every subscriber a /64 of them — an online guess targets an account, so
the account is what is counted.

- Counted by the name as typed, existing or not, so `login` stays uniform (M1).
  The key is a hash: people type passphrases into the username field.
- The attempt is taken before the check in one `INSERT … ON CONFLICT DO UPDATE
  … WHERE … RETURNING`, so a concurrent burst gets no more than the limit.
- Sign-in, passphrase change and account deletion count on the same row; the
  last had no rate limit at all.
- A lockout refuses passphrase sign-in and nothing else: sessions, renewal and
  device sign-in continue, and a reset code clears it (AV26). A session learns
  its own lockout from `/v1/users/me`, and the passphrase change checks it
  before re-wrapping any node's bundle — the hub accepts the new passphrase
  only after the nodes have it.

The SPA now shows what the hub said. `loginAndRecover` threw "Login failed:
{json}", so `email_verification_required` never matched and was never shown;
the passphrase-change form rendered no error at all in its first phase.

The unauthenticated surface, reviewed route by route:

- No `/docs`, `/redoc` or `/openapi.json`, in the code. The Caddyfile hid them
  on meshbay.org only; a packaged hub behind any other proxy published all three.
- The node socket's first message must arrive within ten seconds. It is
  accepted before anyone is known, and an unbounded read is a connection any
  stranger holds for free.
- `/v1/relays` answers 503 behind `relay.RELAYS_ENABLED`, as federation does:
  nothing in the tree calls it and two of its routes take no account.
- `test_unauthenticated_surface.py` walks every route and fails on one without
  an authentication dependency that is not listed with its reason.

Verified in Chrome against a local hub: the lockout and wrong-passphrase
messages, the admin section saving both lockout and mail limits, and the
passphrase change refused while locked. Not verified in Firefox (a running
instance blocks the headless one), nor the upsert's concurrency on PostgreSQL.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01LcF3QKWii7uQ2kSyXErzCt

1 files changed, 256 insertions, 0 deletions

diff --git a/packages/meshbay-hub/tests/test_login_lockout.py b/packages/meshbay-hub/tests/test_login_lockout.py
new file mode 100644
index 0000000..43300d0
--- /dev/null
+++ b/packages/meshbay-hub/tests/test_login_lockout.py
@@ -0,0 +1,256 @@
+"""
+Per-account sign-in lockout (`login_throttle.py`).
+
+The per-IP limit bounds one address, and IPv6 gives every subscriber a /64 of
+them; an online guess targets an account, so the account is what is counted.
+The properties pinned here are the ones that make that safe to ship:
+
+  * the Nth wrong passphrase locks, and a locked account is refused **before**
+    the passphrase is checked — the right one is refused too
+  * an unknown username locks exactly like a real one, so the 429 says nothing
+    the 401 did not (M1)
+  * a right passphrase clears the count, and failures age out of the window
+  * `change_password` checks the same passphrase, so it counts on the same row
+  * the two numbers are the admin's to change, and zero turns it off
+
+What one account costs another through this — a stranger locking your name —
+is `test_availability_between_members.py`, because it takes two accounts.
+"""
+
+import asyncio
+from datetime import datetime, timedelta, timezone
+
+import pytest
+from sqlalchemy import select, update
+
+from meshbay_hub.api.deps import set_admin_usernames
+from meshbay_hub.db.models import LoginThrottle
+from meshbay_hub.login_throttle import _key
+
+RIGHT = "r" * 44
+WRONG = "w" * 44
+
+
+async def _register(client, username, auth_key=RIGHT):
+    r = await client.post("/v1/users/register", json={
+        "username": username, "email": f"{username}@example.com",
+        "auth_key": auth_key})
+    assert r.status_code == 201, r.text
+
+
+async def _login(client, username, auth_key):
+    return await client.post("/v1/users/login", json={
+        "username": username, "auth_key": auth_key})
+
+
+async def _fail(client, username, times):
+    for _ in range(times):
+        r = await _login(client, username, WRONG)
+        assert r.status_code == 401, r.text
+
+
+@pytest.mark.asyncio
+async def test_the_fourth_failure_locks_and_the_right_passphrase_is_refused(client):
+    await _register(client, "alice")
+    await _fail(client, "alice", 4)
+
+    r = await _login(client, "alice", RIGHT)
+    assert r.status_code == 429, r.text
+    assert r.json()["detail"] == "account_locked"
+    # An hour, give or take the time the four failures took.
+    assert 3500 <= int(r.headers["retry-after"]) <= 3600
+
+
+@pytest.mark.asyncio
+async def test_an_unknown_name_locks_exactly_like_a_real_one(client):
+    """M1: the lockout must not become the enumeration oracle `login` avoids."""
+    await _register(client, "bob")
+    await _fail(client, "bob", 4)
+    await _fail(client, "nobody-by-this-name", 4)
+
+    real = await _login(client, "bob", WRONG)
+    ghost = await _login(client, "nobody-by-this-name", WRONG)
+    assert (real.status_code, real.json()) == (ghost.status_code, ghost.json())
+    assert real.status_code == 429
+
+
+@pytest.mark.asyncio
+async def test_the_right_passphrase_clears_the_count(client):
+    await _register(client, "carol")
+    await _fail(client, "carol", 3)
+    r = await _login(client, "carol", RIGHT)
+    assert r.status_code == 200, r.text
+
+    # Three more would have been seven in a row without the reset.
+    await _fail(client, "carol", 3)
+    assert (await _login(client, "carol", RIGHT)).status_code == 200
+
+
+@pytest.mark.asyncio
+async def test_a_lockout_ends_when_its_window_does(client, db_session):
+    await _register(client, "dave")
+    await _fail(client, "dave", 4)
+    assert (await _login(client, "dave", RIGHT)).status_code == 429
+
+    await db_session.execute(
+        update(LoginThrottle).where(LoginThrottle.key == _key("dave"))
+        .values(last_failure_at=datetime.now(timezone.utc) - timedelta(minutes=61)))
+    await db_session.commit()
+
+    assert (await _login(client, "dave", RIGHT)).status_code == 200
+
+
+@pytest.mark.asyncio
+async def test_old_failures_do_not_carry_into_a_new_window(client, db_session):
+    await _register(client, "erin")
+    await _fail(client, "erin", 3)
+    await db_session.execute(
+        update(LoginThrottle).where(LoginThrottle.key == _key("erin"))
+        .values(last_failure_at=datetime.now(timezone.utc) - timedelta(minutes=61)))
+    await db_session.commit()
+
+    # One stale window of three, then one fresh failure: a count of one, not four.
+    await _fail(client, "erin", 1)
+    assert (await _login(client, "erin", RIGHT)).status_code == 200
+
+
+@pytest.mark.asyncio
+async def test_a_burst_of_concurrent_guesses_gets_no_more_than_the_limit(client):
+    """The attempt is taken before the check, in one statement.
+
+    Read-then-write would let every request in a burst read "no failures yet"
+    and be checked. On SQLite writes serialise anyway, so this pins the
+    behaviour rather than proving the statement under PostgreSQL's concurrency;
+    the statement is an `ON CONFLICT DO UPDATE … WHERE`, which both evaluate
+    against the row as locked.
+    """
+    await _register(client, "frank")
+    results = await asyncio.gather(*[_login(client, "frank", WRONG) for _ in range(10)])
+    codes = sorted(r.status_code for r in results)
+    assert codes.count(401) == 4, codes
+    assert codes.count(429) == 6, codes
+
+
+@pytest.mark.asyncio
+async def test_change_password_counts_on_the_same_row(client):
+    """It checks the same passphrase, so it is the same oracle."""
+    await _register(client, "grace")
+    token = (await _login(client, "grace", RIGHT)).json()["access_token"]
+    auth = {"Authorization": f"Bearer {token}"}
+
+    for _ in range(4):
+        r = await client.post("/v1/users/password", headers=auth, json={
+            "old_auth_key": WRONG, "new_auth_key": "n" * 44})
+        assert r.status_code == 403, r.text
+
+    r = await client.post("/v1/users/password", headers=auth, json={
+        "old_auth_key": RIGHT, "new_auth_key": "n" * 44})
+    assert r.status_code == 429, r.text
+    assert (await _login(client, "grace", RIGHT)).status_code == 429
+
+
+@pytest.mark.asyncio
+async def test_a_signed_in_session_is_told_its_own_lockout(client):
+    """A passphrase change re-wraps every node's bundle before the hub accepts
+    it, so the client must know not to start one the hub would then refuse."""
+    await _register(client, "olivia")
+    token = (await _login(client, "olivia", RIGHT)).json()["access_token"]
+    auth = {"Authorization": f"Bearer {token}"}
+    assert (await client.get("/v1/users/me", headers=auth)).json()["passphrase_locked_for"] == 0
+
+    await _fail(client, "olivia", 4)
+    left = (await client.get("/v1/users/me", headers=auth)).json()["passphrase_locked_for"]
+    assert 3500 <= left <= 3600
+
+
+@pytest.mark.asyncio
+async def test_an_attempt_that_checks_no_passphrase_is_not_counted(client, db_session):
+    """A legacy account asked to upgrade has been told nothing about its passphrase."""
+    from meshbay_hub.db.models import User
+
+    await _register(client, "heidi")
+    await db_session.execute(
+        update(User).where(User.username == "heidi").values(pw_version=2))
+    await db_session.commit()
+
+    for _ in range(6):
+        r = await _login(client, "heidi", RIGHT)
+        assert r.status_code == 401 and r.json()["detail"] == "auth_upgrade_required"
+    failures = await db_session.scalar(
+        select(LoginThrottle.failures).where(LoginThrottle.key == _key("heidi")))
+    assert not failures
+
+
+@pytest.mark.asyncio
+async def test_the_table_never_holds_what_was_typed(client, db_session):
+    """People type passphrases into the username field."""
+    await _login(client, "my-secret-passphrase-typed-in-the-wrong-box", WRONG)
+    keys = (await db_session.execute(select(LoginThrottle.key))).scalars().all()
+    assert keys and all("secret" not in k for k in keys)
+
+
+# ── The admin's two numbers ──────────────────────────────────────────────────
+
+async def _admin_headers(client, username="root"):
+    await _register(client, username)
+    set_admin_usernames([username])
+    token = (await _login(client, username, RIGHT)).json()["access_token"]
+    return {"Authorization": f"Bearer {token}"}
+
+
+@pytest.mark.asyncio
+async def test_the_admin_sets_the_limit_and_the_hub_applies_it(client):
+    admin = await _admin_headers(client)
+
+    r = await client.get("/v1/admin/settings", headers=admin)
+    assert r.json()["login"] == {"max_failures": 4, "lockout_minutes": 60}
+    assert r.json()["login_defaults"] == {"max_failures": 4, "lockout_minutes": 60}
+
+    r = await client.patch("/v1/admin/settings", headers=admin,
+                           json={"login": {"max_failures": 2, "lockout_minutes": 5}})
+    assert r.status_code == 200, r.text
+    assert r.json()["login"] == {"max_failures": 2, "lockout_minutes": 5}
+
+    await _register(client, "ivan")
+    await _fail(client, "ivan", 2)
+    r = await _login(client, "ivan", RIGHT)
+    assert r.status_code == 429
+    assert int(r.headers["retry-after"]) <= 300
+
+
+@pytest.mark.asyncio
+async def test_zero_failures_turns_the_lockout_off(client):
+    admin = await _admin_headers(client)
+    await client.patch("/v1/admin/settings", headers=admin,
+                       json={"login": {"max_failures": 0}})
+
+    await _register(client, "judy")
+    await _fail(client, "judy", 8)
+    assert (await _login(client, "judy", RIGHT)).status_code == 200
+
+
+@pytest.mark.asyncio
+async def test_values_are_clamped_and_unknown_keys_refused(client):
+    admin = await _admin_headers(client)
+
+    r = await client.patch("/v1/admin/settings", headers=admin,
+                           json={"login": {"max_failures": -3, "lockout_minutes": 10**9}})
+    assert r.status_code == 200
+    low, _ = r.json()["login_bounds"]["max_failures"]
+    _, high = r.json()["login_bounds"]["lockout_minutes"]
+    assert r.json()["login"] == {"max_failures": low, "lockout_minutes": high}
+
+    r = await client.patch("/v1/admin/settings", headers=admin,
+                           json={"login": {"lockout_hours": 1}})
+    assert r.status_code == 422
+
+
+@pytest.mark.asyncio
+async def test_only_an_admin_changes_them(client):
+    await _admin_headers(client)          # an admin exists; this is someone else
+    await _register(client, "mallory")
+    token = (await _login(client, "mallory", RIGHT)).json()["access_token"]
+    r = await client.patch("/v1/admin/settings",
+                           headers={"Authorization": f"Bearer {token}"},
+                           json={"login": {"max_failures": 0}})
+    assert r.status_code == 403