Compare commits

...

2 Commits

Author SHA1 Message Date
lukaszraczylo 7c6f09fb20 feat(middleware): RS-aware token validators (kill ~21 RLocks/request)
Adds token_validation_rs.go with requestState-aware variants of the
token validation path:

  isUserAuthenticatedRS(rs) -> dispatches by provider
    validateStandardTokensRS(rs) -> standard path (eliminates 17 RLocks)
    validateAzureTokensRS(rs)    -> Azure path     (eliminates 10 RLocks)
    validateGoogleTokensRS(rs)   -> delegates to standard
  validateTokenExpiryRS(rs, tok) -> shared expiry check (eliminates 4 RLocks)

middleware.ServeHTTP now calls isUserAuthenticatedRS(rs) on the hot
path. The pre-v1.0.20 non-RS variants are kept untouched for tests
and any future caller that doesn't have a captured snapshot.

Why
---
The standard validation path read SessionData via session.GetX() 17
times, with GetRefreshToken alone called 11 times (every "return
'needs refresh'" branch re-reads it). Each call acquires
sd.sessionMutex.RLock(). Under Yaegi each RLock costs ~1-5ms of
interpreter dispatch. The captured snapshot already lives on rs, so
the RS variants substitute direct struct field reads.

Per-request cost on the hot authenticated path
----------------------------------------------
  ServeHTTP enters:
    + 1 RLock to populate rs (was 0)
  Validation path:
    Standard: was 17 RLocks, now 0
    Azure:    was 10 RLocks, now 0
  processAuthorizedRequestRS:
    was 4-6 GetX calls, now 0 (already in v1.0.19)

Net: ~22-27 fewer Yaegi-dispatched RLock acquisitions per authenticated
request on the hot path.

Caveats
-------
* Refresh / expired / callback paths still use the non-RS validators
  because they can mutate session state between validation and use.
* The RS variants are by-design line-for-line equivalents of the
  originals. If logic in the originals changes, the RS variants need
  matching updates. This is acceptable for now; a future refactor
  could collapse them once the non-RS callers are gone.

All tests pass with -race; golangci-lint clean.
2026-05-23 12:38:42 +01:00
lukaszraczylo 68e1c4319c feat(middleware): per-request context object (requestState)
Adds requeststate.go and threads a *requestState through the
ServeHTTP -> processAuthorizedRequestRS -> forwardAuthorized path.
rs is allocated once at the top of ServeHTTP, populates SessionData
field snapshots under a SINGLE sd.sessionMutex.RLock, and caches the
MetadataSnapshot. Downstream handlers read the cached fields instead
of calling session.GetX() / t.metadataSnap() repeatedly.

Why
---
Under Yaegi each method dispatch (including RWMutex.RLock) costs
~1-5ms of interpreter overhead. SessionData getters each take an
RLock on sd.sessionMutex; the previous hot path called 5-7 of them
per request (GetAuthenticated, GetAccessToken, GetIDToken,
GetRefreshToken, GetUserIdentifier, plus the same set again inside
processAuthorizedRequest). With one batched RLock + cached fields,
that drops to a single RLock for the whole handler chain.

This is scoped — not a wholesale architectural refactor:

* requestState is per-request (alloc at ServeHTTP entry, dropped on
  return). It is NOT a shared cache and never escapes the request.
* The original processAuthorizedRequest is kept unchanged for any
  callers we don't migrate this round (bearer path, callback
  handlers, expired-token handlers). New code path is the RS-aware
  processAuthorizedRequestRS, which middleware.ServeHTTP now uses for
  the happy authenticated-and-not-needing-refresh case.
* Cross-request caches (tokenCache, JWKCache, sessionEntries,
  sessionInvalidationCache) are unchanged. rs is additive, not a
  replacement.

What this does NOT change
-------------------------
* The refresh path still calls session.GetX() in middleware.go
  (handleExpiredToken, refreshToken, defaultInitiateAuthentication)
  because those flows can mutate session state and a stale rs would
  be wrong.
* validateStandardTokens still has its own session.GetX() calls.
  Deep plumbing into the token-verification path is a follow-up.
* No semantic changes to authentication, refresh, or session
  lifecycle — only the read path is optimised.

All tests pass with -race; golangci-lint clean.
2026-05-23 12:22:51 +01:00
3 changed files with 459 additions and 4 deletions
+109 -4
View File
@@ -311,6 +311,19 @@ func (t *TraefikOidc) ServeHTTP(rw http.ResponseWriter, req *http.Request) {
host := utils.DetermineHost(req)
redirectURL := buildFullURL(scheme, host, t.redirURLPath)
// Capture per-request state: one RLock on sd.sessionMutex covers all the
// getter values the handler chain needs (instead of 5-7 separate
// session.GetX() calls each acquiring their own RLock under Yaegi).
// metadataSnap is also stored once so downstream handlers don't repeat
// the atomic.Value.Load.
rs := (&requestState{
scheme: scheme,
host: host,
redirectURL: redirectURL,
next: t.next,
metadata: t.metadataSnap(),
}).captureSession(session)
// Check if the current request is the OIDC callback
t.logger.Debugf("Checking callback URL match: request_path=%q, configured_callback=%q", req.URL.Path, t.redirURLPath)
if req.URL.Path == t.redirURLPath {
@@ -320,7 +333,10 @@ func (t *TraefikOidc) ServeHTTP(rw http.ResponseWriter, req *http.Request) {
}
t.logger.Debugf("Callback URL did not match (request_path=%q != configured=%q), continuing auth flow", req.URL.Path, t.redirURLPath)
authenticated, needsRefresh, expired := t.isUserAuthenticated(session)
// Token validation reads session via the captured snapshot — saves ~21
// sd.sessionMutex.RLock acquisitions (Yaegi-dispatched, ~1-5ms each)
// across the validation path.
authenticated, needsRefresh, expired := t.isUserAuthenticatedRS(rs)
if expired {
t.logger.Debug("Session token is definitively expired or invalid, initiating re-auth")
@@ -328,7 +344,7 @@ func (t *TraefikOidc) ServeHTTP(rw http.ResponseWriter, req *http.Request) {
return
}
userIdentifier := session.GetUserIdentifier()
userIdentifier := rs.userIdentifier
// User authorization check
if authenticated && userIdentifier != "" {
if !t.isAllowedUser(userIdentifier) {
@@ -345,11 +361,11 @@ func (t *TraefikOidc) ServeHTTP(rw http.ResponseWriter, req *http.Request) {
// methods (validateAzureTokens/validateStandardTokens) before reaching this point.
// Redundant validation here was causing issues with Azure AD tokens that have
// JWT format but unverifiable signatures. See issue #89.
t.processAuthorizedRequest(rw, req, session, redirectURL)
t.processAuthorizedRequestRS(rw, req, rs)
return
}
refreshTokenPresent := session.GetRefreshToken() != ""
refreshTokenPresent := rs.refreshToken != ""
// Decide whether to answer with 401 instead of a redirect. AJAX requests
// cannot follow a 302 into an IdP, and sub-resource loads (script/image/
@@ -456,6 +472,95 @@ func (t *TraefikOidc) ServeHTTP(rw http.ResponseWriter, req *http.Request) {
// - req: The HTTP request to process.
// - session: The user's session data containing tokens and claims.
// - redirectURL: The callback URL for re-authentication if needed.
// processAuthorizedRequestRS is the requestState-aware variant of
// processAuthorizedRequest. It reads SessionData fields from the captured
// snapshot in rs instead of calling session.GetX() (each of which acquires
// sd.sessionMutex.RLock — under Yaegi every RLock pays ~1-5ms of interpreter
// dispatch). Only session-mutating operations (Save, ResetRedirectCount,
// Clear, IsDirty) still go through the session pointer because those write
// state and have no snapshot.
func (t *TraefikOidc) processAuthorizedRequestRS(rw http.ResponseWriter, req *http.Request, rs *requestState) {
session := rs.session
redirectURL := rs.redirectURL
userIdentifier := rs.userIdentifier
if userIdentifier == "" {
t.logger.Info("No user identifier found in session during final processing, initiating re-auth")
session.ResetRedirectCount()
t.defaultInitiateAuthentication(rw, req, session, redirectURL)
return
}
// Check if session has been invalidated via backchannel or front-channel logout
idToken := rs.idToken
if t.enableBackchannelLogout || t.enableFrontchannelLogout {
if idToken != "" {
sid, sub, createdAt := t.extractSessionInfo(idToken)
if t.isSessionInvalidated(sid, sub, createdAt) {
t.logger.Infof("Session for user %s has been invalidated via IdP-initiated logout", userIdentifier)
if err := session.Clear(req, rw); err != nil {
t.logger.Errorf("Error clearing invalidated session: %v", err)
}
session.ResetRedirectCount()
t.defaultInitiateAuthentication(rw, req, session, redirectURL)
return
}
}
}
// Resolve ID-token claims at most once per request. SessionData caches
// the parsed claims keyed on the raw ID token.
var (
idClaims map[string]interface{}
idClaimsErr error
)
if idToken != "" {
idClaims, idClaimsErr = session.GetIDTokenClaims(t.extractClaimsFunc)
}
var (
groupClaims map[string]interface{}
groupClaimsErr error
)
if idToken != "" {
groupClaims, groupClaimsErr = idClaims, idClaimsErr
} else if rs.accessToken != "" {
groupClaims, groupClaimsErr = t.extractClaimsFunc(rs.accessToken)
} else if len(t.allowedRolesAndGroups) > 0 {
t.logger.Error("No token available but roles/groups checks are required")
session.ResetRedirectCount()
t.defaultInitiateAuthentication(rw, req, session, redirectURL)
return
}
if groupClaimsErr != nil && len(t.allowedRolesAndGroups) > 0 {
t.logger.Errorf("Failed to extract claims for roles/groups check: %v", groupClaimsErr)
session.ResetRedirectCount()
t.defaultInitiateAuthentication(rw, req, session, redirectURL)
return
}
// Persist any dirty session state BEFORE forwardAuthorized writes the
// response.
if session.IsDirty() {
if err := session.Save(req, rw); err != nil {
t.logger.Errorf("Failed to save session after processing headers: %v", err)
}
} else {
t.logger.Debug("Session not dirty, skipping save in processAuthorizedRequest")
}
p := &principal{
Source: sourceSession,
Identifier: userIdentifier,
AccessToken: rs.accessToken,
IDToken: idToken,
RefreshToken: rs.refreshToken,
Claims: groupClaims,
}
t.forwardAuthorized(rw, req, p)
}
func (t *TraefikOidc) processAuthorizedRequest(rw http.ResponseWriter, req *http.Request, session *SessionData, redirectURL string) {
userIdentifier := session.GetUserIdentifier()
if userIdentifier == "" {
+71
View File
@@ -0,0 +1,71 @@
// Package traefikoidc provides OIDC authentication middleware for Traefik.
// requestState bundles read-mostly fields for a single ServeHTTP call.
package traefikoidc
import "net/http"
// requestState is a per-request context object allocated at the top of
// ServeHTTP and threaded through to downstream handlers. It caches values
// that would otherwise require a Yaegi-dispatched lock acquisition each time
// they're read:
//
// - The metadata snapshot (atomic.Value.Load once, not per-handler).
// - SessionData getter results (one RLock on sd.sessionMutex covers all
// fields, instead of 5-7 separate RLock/RUnlock pairs scattered through
// the handler chain).
//
// The struct is alloc'd at request entry, populated under at most one RLock
// of sd.sessionMutex, and discarded at request exit. It is NOT shared across
// requests and never written from another goroutine, so no synchronization
// on its fields is required.
//
// Cross-request global caches (tokenCache, JWKCache, sessionEntries,
// sessionInvalidationCache) remain — they're orthogonal. requestState's job
// is to eliminate redundant per-handler reads of values that don't change
// within a single request.
type requestState struct {
// Globals snapshotted once.
metadata *MetadataSnapshot
// SessionData fields snapshotted under one RLock. The pointer to the
// SessionData is retained so handlers that genuinely need to mutate
// (Save, Clear, etc.) still have access.
session *SessionData
authenticated bool
accessToken string
idToken string
refreshToken string
userIdentifier string
createdAtUnixSec int64
// Output: scheme/host/redirect path determined at top of ServeHTTP.
scheme string
host string
redirectURL string
// Carry the next handler so forwardAuthorized doesn't need to close over t.
next http.Handler
}
// captureSession populates requestState's SessionData-derived fields under a
// single RLock of sd.sessionMutex. Returns the populated rs for chaining.
//
// Replaces a sequence of SessionData.GetX() calls each of which acquires
// sd.sessionMutex.RLock(). Under Yaegi each RLock costs ~1-5ms of
// interpreter dispatch; batching saves the rest.
func (rs *requestState) captureSession(sd *SessionData) *requestState {
if sd == nil {
return rs
}
rs.session = sd
sd.sessionMutex.RLock()
rs.authenticated = sd.getAuthenticatedUnsafe()
rs.accessToken = sd.getAccessTokenUnsafe()
rs.idToken = sd.getIDTokenUnsafe()
rs.refreshToken = sd.getRefreshTokenUnsafe()
rs.userIdentifier = sd.getUserIdentifierUnsafe()
rs.createdAtUnixSec = sd.getCreatedAtUnsafe()
sd.sessionMutex.RUnlock()
return rs
}
+279
View File
@@ -0,0 +1,279 @@
// Package traefikoidc provides OIDC authentication middleware for Traefik.
// This file contains requestState-aware variants of the token validation
// functions. They read session field values from the captured snapshot in
// *requestState instead of calling session.GetX(), eliminating ~21 RLock
// acquisitions on sd.sessionMutex per request through the validation path
// (validateStandardTokens reads 17, validateAzureTokens reads 10,
// validateTokenExpiry reads 4 — and many are the SAME field). Under Yaegi
// each RLock costs ~1-5ms of interpreter dispatch.
//
// The non-RS variants are retained for paths that don't have a captured
// snapshot (tests that drive the validators directly, the Azure/Google path
// when reached without rs threading, etc).
package traefikoidc
import (
"encoding/base64"
"encoding/json"
"strings"
"time"
)
// isUserAuthenticatedRS is the requestState-aware variant of
// isUserAuthenticated. Dispatches to the right per-provider validator based
// on the configured provider, all of which read from rs instead of session.
func (t *TraefikOidc) isUserAuthenticatedRS(rs *requestState) (bool, bool, bool) {
if t.isAzureProvider() {
return t.validateAzureTokensRS(rs)
} else if t.isGoogleProvider() {
return t.validateStandardTokensRS(rs)
}
return t.validateStandardTokensRS(rs)
}
// validateTokenExpiryRS is the requestState-aware variant of validateTokenExpiry.
// Reads rs.refreshToken instead of session.GetRefreshToken() (4 RLocks avoided).
func (t *TraefikOidc) validateTokenExpiryRS(rs *requestState, token string) (bool, bool, bool) {
cachedClaims, found := t.tokenCache.Get(token)
if !found {
t.logger.Debug("Claims not found in cache after successful token verification")
if rs.refreshToken != "" {
return false, true, false
}
return false, false, true
}
expClaim, ok := cachedClaims["exp"].(float64)
if !ok {
t.logger.Error("Failed to get expiration time ('exp' claim) from verified token")
if rs.refreshToken != "" {
return false, true, false
}
return false, false, true
}
expTimeObj := time.Unix(int64(expClaim), 0)
nowObj := time.Now()
if expTimeObj.Before(nowObj) {
if rs.refreshToken != "" {
return false, true, false
}
return false, false, true
}
refreshThreshold := nowObj.Add(t.refreshGracePeriod)
if expTimeObj.Before(refreshThreshold) {
if rs.refreshToken != "" {
return true, true, false
}
return true, false, false
}
return true, false, false
}
// validateStandardTokensRS is the requestState-aware variant of
// validateStandardTokens. Replaces all session.GetX() calls (17 of them in
// the non-RS variant, dominated by GetRefreshToken called 11 times) with
// rs field reads. Same control flow.
//
//nolint:gocognit,gocyclo // Mirrors validateStandardTokens complexity by design.
func (t *TraefikOidc) validateStandardTokensRS(rs *requestState) (bool, bool, bool) {
if !rs.authenticated {
if rs.refreshToken != "" {
return false, true, false
}
return false, false, false
}
if rs.accessToken == "" {
if rs.refreshToken != "" {
// ID-token grace-period check (only when accessToken is absent).
if rs.idToken != "" {
parts := strings.Split(rs.idToken, ".")
if len(parts) == 3 {
if claimsData, err := base64.RawURLEncoding.DecodeString(parts[1]); err == nil {
var claims map[string]interface{}
if err := json.Unmarshal(claimsData, &claims); err == nil {
if expClaim, ok := claims["exp"].(float64); ok {
expTime := time.Unix(int64(expClaim), 0)
if time.Now().After(expTime) {
expiredDuration := time.Since(expTime)
if expiredDuration > t.refreshGracePeriod {
return false, false, true
}
}
}
}
}
}
}
return false, true, false
}
return false, false, true
}
dotCount := strings.Count(rs.accessToken, ".")
isOpaqueToken := dotCount != 2
if isOpaqueToken {
if t.allowOpaqueTokens {
if err := t.validateOpaqueToken(rs.accessToken); err != nil {
errMsg := err.Error()
isTokenInvalid := strings.Contains(errMsg, "token is not active") ||
strings.Contains(errMsg, "revoked") ||
strings.Contains(errMsg, "token has expired")
if isTokenInvalid {
if rs.refreshToken != "" {
return false, true, false
}
return false, false, true
}
if t.requireTokenIntrospection {
if rs.refreshToken != "" {
return false, true, false
}
return false, false, true
}
// Transient introspection error: fall through to ID-token validation.
} else {
// Introspection succeeded.
if rs.idToken != "" {
return t.validateTokenExpiryRS(rs, rs.idToken)
}
return true, false, false
}
}
// Fall back to ID-token validation when opaque + no successful introspection.
if rs.idToken == "" {
if rs.refreshToken != "" {
return false, true, false
}
return true, false, false
}
if err := t.verifyToken(rs.idToken); err != nil {
if strings.Contains(err.Error(), "token has expired") {
if rs.refreshToken != "" {
return false, true, false
}
return false, false, true
}
if rs.refreshToken != "" {
return false, true, false
}
return false, false, true
}
return t.validateTokenExpiryRS(rs, rs.idToken)
}
// JWT access token present.
accessTokenValid := false
if err := t.verifyToken(rs.accessToken); err != nil {
errMsg := err.Error()
if strings.Contains(errMsg, "invalid audience") || strings.Contains(errMsg, "audience") {
if t.strictAudienceValidation {
if rs.refreshToken != "" {
return false, true, false
}
return false, false, true
}
// Fall through to ID-token validation.
}
} else {
accessTokenValid = true
}
if rs.idToken == "" {
if accessTokenValid {
return t.validateTokenExpiryRS(rs, rs.accessToken)
}
if rs.refreshToken != "" {
return true, true, false
}
return true, false, false
}
if err := t.verifyToken(rs.idToken); err != nil {
if strings.Contains(err.Error(), "token has expired") {
if rs.refreshToken != "" {
return false, true, false
}
return false, false, true
}
if rs.refreshToken != "" {
return false, true, false
}
return false, false, true
}
if accessTokenValid {
return t.validateTokenExpiryRS(rs, rs.accessToken)
}
return t.validateTokenExpiryRS(rs, rs.idToken)
}
// validateAzureTokensRS is the requestState-aware variant of validateAzureTokens.
// Eliminates 10 session.GetX() RLocks per Azure-path request.
func (t *TraefikOidc) validateAzureTokensRS(rs *requestState) (bool, bool, bool) {
if !rs.authenticated {
if rs.refreshToken != "" {
return false, true, false
}
return false, true, false
}
if rs.accessToken != "" {
if strings.Count(rs.accessToken, ".") == 2 {
if t.isUnverifiableAzureAccessToken(rs.accessToken) {
if rs.idToken != "" {
if err := t.verifyToken(rs.idToken); err != nil {
if rs.refreshToken != "" {
return false, true, false
}
return false, false, true
}
return t.validateTokenExpiryRS(rs, rs.idToken)
}
return true, false, false
}
if err := t.verifyToken(rs.accessToken); err != nil {
if rs.idToken != "" {
if err := t.verifyToken(rs.idToken); err != nil {
if rs.refreshToken != "" {
return false, true, false
}
return false, false, true
}
return t.validateTokenExpiryRS(rs, rs.idToken)
}
if rs.refreshToken != "" {
return false, true, false
}
return false, false, true
}
return t.validateTokenExpiryRS(rs, rs.accessToken)
}
// Opaque access token.
if rs.idToken != "" {
return t.validateTokenExpiryRS(rs, rs.idToken)
}
return true, false, false
}
if rs.idToken != "" {
if err := t.verifyToken(rs.idToken); err != nil {
if rs.refreshToken != "" {
return false, true, false
}
return false, false, true
}
return t.validateTokenExpiryRS(rs, rs.idToken)
}
if rs.refreshToken != "" {
return false, true, false
}
return false, false, true
}