Unverified Commit 3b43f05a authored by Minseong Choi's avatar Minseong Choi 💬
Browse files

refactor(api): drop dead login concurrency limiter and reconcile passwordless comments

The passwordless migration (b330d77) removed the password-login route, leaving
concurrencyLimiter — its bcrypt concurrency cap — with no caller, and scattered
stale "local-password" / "change-password" references through the surviving auth
code's comments.

- Remove the dead concurrencyLimiter (type + newConcurrencyLimiter + acquire):
  no caller, no struct field, no test. Reword the one streamLimiter doc that
  contrasted against it.
- Realign comments in repo.go, pgrepo.go, session.go, util.go to the passwordless
  reality: staff lookups feed email-OTP / passkey / setup redeem, not a password
  compare; RevokeUserSessionsExcept and DeleteAllPasskeyCredentialsForUser are
  retained (uncalled) for the P5 account-remediation path (#78); "local sessions"
  no longer implies a password.

Comments and dead code only; no behavior change. Full WSL test tree green.
parent 0c1cc598
Loading
Loading
Loading
Loading
+2 −42
Changes for internal/api/api.go: 2 added lines, 42 removed lines.
Original line number Diff line number Diff line
@@ -622,54 +622,14 @@ func (c *cooldownLimiter) release(name string, reservedAt time.Time) {
	}
}

// ---- login concurrency cap ----

// concurrencyLimiter bounds how many holders may run a guarded section at once. It
// backs the public login route's bcrypt cap (handleLogin): a buffered channel of n
// tokens; acquire takes one WITHOUT blocking (returning ok=false when the section
// is already full), release returns it. Unlike cooldownLimiter — a per-key time
// window — this bounds simultaneity, not frequency, which is the right shape for a
// CPU-costly section a flood would otherwise pin every core running. A non-positive
// cap disables it (acquire always admits, release is a no-op), mirroring the "zero
// disables" idiom of WakeCooldown and MaxRunningServers.
type concurrencyLimiter struct {
	slots chan struct{}
}

// newConcurrencyLimiter builds a limiter admitting at most n concurrent holders. A
// non-positive n yields a disabled limiter (nil slots) that admits everyone.
func newConcurrencyLimiter(n int) *concurrencyLimiter {
	if n <= 0 {
		return &concurrencyLimiter{}
	}
	return &concurrencyLimiter{slots: make(chan struct{}, n)}
}

// acquire tries to take a slot without blocking. It returns a release func and true
// on success, or nil and false when the section is already at capacity. The disabled
// limiter (nil slots) always admits and returns a no-op release. release MUST be
// called exactly once on the success path, so it reads naturally as `release, ok :=
// l.acquire(); if !ok { shed }; defer/inline release()`.
func (l *concurrencyLimiter) acquire() (release func(), ok bool) {
	if l.slots == nil {
		return func() {}, true
	}
	select {
	case l.slots <- struct{}{}:
		return func() { <-l.slots }, true
	default:
		return nil, false
	}
}

// ---- per-principal stream cap ----

// streamLimiter bounds how many concurrent guarded sections a single KEY may hold at
// once. It backs the per-principal SSE stream cap (console + build-log relays): each
// relay blocks for the life of a client's attachment and, under a stalled reader,
// pins a goroutine plus a kube-apiserver follow connection, so an unbounded number of
// them from one principal is a control-plane connection-exhaustion vector. Unlike the
// login concurrencyLimiter (a single global semaphore), this counts per key. A
// them from one principal is a control-plane connection-exhaustion vector. Unlike a
// single global semaphore, this counts per key. A
// non-positive max disables it (acquire always admits, release is a no-op), the same
// "zero disables" idiom as the other levers.
type streamLimiter struct {
+9 −6
Changes for internal/api/pgrepo.go: 9 added lines, 6 removed lines.
Original line number Diff line number Diff line
@@ -722,7 +722,7 @@ func (p *PGRepo) IsProtectedAdminLink(ctx context.Context, mcUUID string) (bool,
	return ok, err
}

// ---- local-password auth (spec §B) ----
// ---- staff account lookups (spec §B, passwordless) ----

// UserByUsername loads a staff login projection by username, or ErrNotFound.
// The account is passwordless — staff authenticate via email-OTP / passkey, so
@@ -836,9 +836,10 @@ func (p *PGRepo) RevokeSession(ctx context.Context, tokenHash string) error {
	return err
}

// RevokeUserSessionsExcept revokes every live session of a user except
// keepTokenHash — the change-password flow logs out the account's other devices
// while keeping the current one.
// RevokeUserSessionsExcept revokes every live session of a user except keepTokenHash
// — logs out an account's other devices while keeping the current one. Its original
// caller (the change-password flow) was removed in the passwordless migration; it is
// retained for the account-remediation path (P5, #78) and currently has no caller.
func (p *PGRepo) RevokeUserSessionsExcept(ctx context.Context, userID, keepTokenHash string) error {
	_, err := p.db.ExecContext(ctx,
		`UPDATE sessions SET revoked_at = now()
@@ -1030,8 +1031,10 @@ func (p *PGRepo) DeletePasskeyCredential(ctx context.Context, userID, id string)
// DeleteAllPasskeyCredentialsForUser unbinds every passkey a user holds. Unlike the
// single-credential delete this does NOT report ErrNotFound on zero rows: removing all of
// a user's passkeys when they have none is a successful no-op, since "the user holds no
// passkeys" is exactly the intended post-condition. The change-password flow calls it so a
// passkey planted through a transiently-hijacked session cannot survive the remediation.
// passkeys" is exactly the intended post-condition. It is the remediation that stops a
// passkey planted through a transiently-hijacked session from surviving; its original
// caller (the change-password flow) was removed in the passwordless migration, so it is
// currently uncalled, retained for the account-remediation/reset path (P5, #78).
func (p *PGRepo) DeleteAllPasskeyCredentialsForUser(ctx context.Context, userID string) error {
	_, err := p.db.ExecContext(ctx,
		`DELETE FROM webauthn_credentials WHERE user_id = $1`, userID)
+22 −20
Changes for internal/api/repo.go: 22 added lines, 20 removed lines.
Original line number Diff line number Diff line
@@ -92,9 +92,9 @@ type StaffUser struct {
// be public (unlike a session token), so it is safe at rest. CredentialID is the
// authenticator's globally-unique handle (base64url) and PublicKey the COSE key
// (base64); SignCount is the uint32 signature counter captured at registration.
// LastUsedAt is nil until an assertion is verified — the login/step-up path that
// would stamp it is out of scope for this enrollment-only slice (deferred), so it
// stays nil through the flow this type backs.
// LastUsedAt is nil until an assertion stamps it. The passkey login door now exists
// (Public /auth/passkey/login/{begin,finish}, #72), but no path yet writes
// last_used_at, so in practice it stays nil; wiring the stamp is a follow-up there.
type PasskeyCredential struct {
	ID           string
	UserID       string
@@ -352,11 +352,13 @@ type Repo interface {
	// can only unbind their OWN credential. No matching (user, id) row → ErrNotFound,
	// so a stale or cross-user id cannot silently no-op as success.
	DeletePasskeyCredential(ctx context.Context, userID, id string) error
	// DeleteAllPasskeyCredentialsForUser unbinds every passkey a user holds. The
	// change-password flow calls it so a passkey planted via a transiently-hijacked
	// session does not survive the remediation (password reset + session revoke) as a
	// standing login foothold. Removing zero rows is success, not an error — an account
	// with no passkeys is the intended post-condition either way.
	// DeleteAllPasskeyCredentialsForUser unbinds every passkey a user holds — the
	// remediation that stops a passkey planted via a transiently-hijacked session from
	// surviving as a standing login foothold. Its original caller, the change-password
	// flow, was removed in the passwordless migration, so it currently has no production
	// caller; it is retained for the account-remediation/reset path (P5, #78). Removing
	// zero rows is success, not an error — an account with no passkeys is the intended
	// post-condition either way.
	DeleteAllPasskeyCredentialsForUser(ctx context.Context, userID string) error

	// ---- player game-login: username-collision reclaim (spec §B3) ----
@@ -398,18 +400,18 @@ type Repo interface {
	// name to protect). Keyed by UUID — the only identity velocity knows.
	IsProtectedAdminLink(ctx context.Context, mcUUID string) (bool, error)

	// ---- local-password auth (spec §B) ----
	// ---- staff account lookups (spec §B, passwordless) ----

	// UserByUsername loads the login projection of a staff account by its unique
	// username, or ErrNotFound. The caller compares PasswordHash itself so the
	// anti-enumeration dummy-hash compare runs even on a miss; a player row (NULL
	// password_hash → empty PasswordHash) is returned too and is rejected by the
	// caller's hash compare, never by leaking "no such user".
	// username, or ErrNotFound. Its caller is the `felis breakGlass` recovery TUI,
	// which resolves an Owner/Operator username before sending an email-OTP — there is
	// no password compare (the account is passwordless). A non-staff (role='user') row
	// resolves too; callers that require staff enforce the role themselves.
	UserByUsername(ctx context.Context, username string) (*StaffUser, error)
	// UserByID loads the same staff projection by user id, or ErrNotFound. The
	// change-password flow uses it to re-verify the caller's current password: the
	// session yields a user id, not a username, so this is the id-keyed counterpart
	// of UserByUsername.
	// UserByID loads the same staff projection by user id, or ErrNotFound. Callers hold
	// a session (which yields a user id, not a username) and need the account behind it
	// — e.g. setup redeem/status resolving the lockdown session's owner. It is the
	// id-keyed counterpart of UserByUsername.
	UserByID(ctx context.Context, id string) (*StaffUser, error)
	// UserByEmail resolves a VERIFIED email address to its login projection,
	// case-insensitively, or ErrNotFound (spec §B email-first login). It is the
@@ -419,9 +421,9 @@ type Repo interface {
	// else's login by typing their email. Matching is on lower(email) to align with
	// the users_verified_email_unique partial index (migration 0010), which
	// guarantees at most one verified row per normalized address, so the result is
	// unambiguous. A player row (empty PasswordHash) resolves too — email-first
	// login is passwordless and does not consult the hash — unlike the password
	// path, which this deliberately does not gate on.
	// unambiguous. A player (role='user') row resolves too — email-first login is
	// passwordless and role-agnostic here; the door that consumes this result decides
	// what each role may do.
	UserByEmail(ctx context.Context, email string) (*StaffUser, error)
	// UpsertOwner creates or resets the single Owner account direct-to-Postgres
	// (the `felis setup` / `felis breakGlass` recovery path). role is forced to
+9 −9
Changes for internal/api/session.go: 9 added lines, 9 removed lines.
Original line number Diff line number Diff line
@@ -15,23 +15,23 @@ import (
	"time"
)

// Local-password sessions (spec §B). The remote face authenticates statelessly
// with a Cloudflare-Access JWT and sets no cookie; local-password auth, used on
// op.console when Zero Trust is not configured (and as the demo's primary web
// login), needs a server-minted session. We store only the sha-256 of the opaque
// cookie value, mirroring how service tokens are stored, so a database read never
// yields a usable cookie.
// Local sessions (spec §B, passwordless). The remote face authenticates statelessly
// with a Cloudflare-Access JWT and sets no cookie; the passwordless console login
// (email-OTP / passkey / setup redeem), used on op.console when Zero Trust is not
// configured (and as the demo's primary web login), needs a server-minted session.
// We store only the sha-256 of the opaque cookie value, mirroring how service tokens
// are stored, so a database read never yields a usable cookie.

const (
	// sessionCookieName is the host-only session cookie. It carries no Domain
	// attribute, so an op.console session is never sent to the player console.
	sessionCookieName = "felis_session"
	// sessionTTL bounds a local-password session. Staff re-authenticate after it.
	// sessionTTL bounds a local session. Staff re-authenticate after it.
	sessionTTL = 12 * time.Hour
)

// LocalAuthEnabledKey is the platform_settings key that gates whether
// local-password sessions are honored. It is flipped on by `felis breakGlass`
// local sessions are honored. It is flipped on by `felis breakGlass`
// direct-to-Postgres at first-run and read live per-request, so enabling local
// auth needs no pod roll. Exported so the break-glass writer and this
// per-request reader share one source of truth instead of drifting copies.
@@ -109,7 +109,7 @@ func hostIsAdminConsole(r *http.Request, rootDomain, adminHostname string) bool
}

// SessionAuth is the composite ExternalAuth for the web face. It prefers a
// local-password session cookie and otherwise delegates to the remote JWT
// local session cookie and otherwise delegates to the remote JWT
// verifier, so both auth models coexist on one face:
//
//   - No cookie  → delegate to Delegate (the Cloudflare-Access JWT path).
+2 −1
Changes for internal/api/util.go: 2 added lines, 1 removed line.
Original line number Diff line number Diff line
@@ -12,7 +12,8 @@ const maxBodyBytes = 1 << 20 // 1 MiB

// requireJSONContentType rejects a request whose body is not declared
// application/json, returning 415 before any decode. It guards the credential-bearing
// auth writes (login, change-password) against a cross-site forgery: an HTML form can
// auth writes (email-OTP, passkey, op-login, setup redeem) against a cross-site
// forgery: an HTML form can
// only POST as application/x-www-form-urlencoded, multipart/form-data, or text/plain
// — never JSON — and a cross-site fetch that forces application/json triggers a CORS
// preflight this API never answers, so neither form can be forged off-origin. The