perf(session): cut yaegi allocations on authenticated requests (#164)

Under yaegi, the authenticated cookie path allocated about 3.8 MB per request. A yaegi harness measures bytes per request with a browser-like cookie jar. With it, these changes cut bytes by 83%, allocations by 69% and time by about 93%. - Replace range loops over tokens with index loops. Under yaegi, ranging over a string allocates O(n^2) bytes. - Pool gzip readers for combined session cookies. - Memoize the validated ID token within one request. - Cache content-only token validation across requests. The key is the token type and SHA-256, and the cache holds at most 4096 entries. The entry keeps the parsed exp and iat claims. The expiry and freshness checks still run on every request. New tests pin the validator results, the memo and the cache behavior.
2026-10-20 04:57:02 +00:00 · 2026-09-29 11:59:31 +01:00
parent 74decb21c2
commit 94a8664403
6 changed files with 409 additions and 46 deletions
@@ -10,6 +10,7 @@ import (
 	"sync"
 	"sync/atomic"
 	"time"
+	"unicode/utf8"

 	"github.com/gorilla/sessions"
 	"github.com/lukaszraczylo/traefikoidc/internal/pool"
@@ -89,9 +90,14 @@ type TokenRetrievalResult struct {
 // and error handling to ensure data integrity and prevent security vulnerabilities
 // throughout the process.
 type ChunkManager struct {
-	lastCleanup    time.Time
-	ctx            context.Context
-	mutex          *sync.RWMutex
+	lastCleanup time.Time
+	ctx         context.Context
+	mutex       *sync.RWMutex
+	// contentValid holds keys of tokens that passed every content-only check.
+	// Browsers resend the same cookie, and re-running those checks on every
+	// request dominated per-request cost under yaegi. Guarded by contentMu.
+	contentValid   map[string]tokenTimes
+	contentMu      sync.Mutex
 	cancel         context.CancelFunc
 	sessionMap     map[string]*SessionEntry
 	logger         *Logger
@@ -127,14 +133,15 @@ func NewChunkManager(logger *Logger) *ChunkManager {
 	ctx, cancel := context.WithCancel(context.Background())

 	cm := &ChunkManager{
-		logger:      logger,
-		mutex:       &sync.RWMutex{},
-		ctx:         ctx,
-		cancel:      cancel,
-		sessionMap:  make(map[string]*SessionEntry),
-		maxSessions: 200,              // CRITICAL FIX: Reduced from 1000 to 200 per instance
-		sessionTTL:  15 * time.Minute, // CRITICAL FIX: Reduced from 24h to 15 minutes
-		lastCleanup: time.Now(),
+		logger:       logger,
+		mutex:        &sync.RWMutex{},
+		ctx:          ctx,
+		cancel:       cancel,
+		sessionMap:   make(map[string]*SessionEntry),
+		contentValid: make(map[string]tokenTimes),
+		maxSessions:  200,              // CRITICAL FIX: Reduced from 1000 to 200 per instance
+		sessionTTL:   15 * time.Minute, // CRITICAL FIX: Reduced from 24h to 15 minutes
+		lastCleanup:  time.Now(),
 	}

 	// Start background cleanup routine
@@ -293,26 +300,38 @@ func (cm *ChunkManager) processSingleToken(token string, compressed bool, config
 // Returns:
 //   - TokenRetrievalResult with the validated token or validation error.
 func (cm *ChunkManager) validateToken(token string, config TokenConfig) TokenRetrievalResult {
-	if sizeErr := cm.validateTokenSize(token, config); sizeErr != nil {
-		return TokenRetrievalResult{Token: "", Error: sizeErr}
+	// Content-only checks depend on the token string alone, so a token that
+	// passed them once skips them. Expiry and freshness always run.
+	key := contentValidKey(token, config)
+	times, cached := cm.cachedTokenTimes(key)
+
+	if !cached {
+		if sizeErr := cm.validateTokenSize(token, config); sizeErr != nil {
+			return TokenRetrievalResult{Token: "", Error: sizeErr}
+		}
+
+		cm.validateChunkingEfficiency(token, config)
+
+		if contentErr := cm.validateTokenContent(token, config); contentErr != nil {
+			return TokenRetrievalResult{Token: "", Error: contentErr}
+		}
+		times = cm.tokenTimesOf(token)
 	}

-	cm.validateChunkingEfficiency(token, config)
-
-	if contentErr := cm.validateTokenContent(token, config); contentErr != nil {
-		return TokenRetrievalResult{Token: "", Error: contentErr}
-	}
-
-	if expErr := cm.validateTokenExpiration(token, config); expErr != nil {
+	if expErr := cm.validateTokenExpiration(times, config); expErr != nil {
 		return TokenRetrievalResult{Token: "", Error: expErr}
 	}

-	if freshnessErr := cm.validateTokenFreshness(token, config); freshnessErr != nil {
+	if freshnessErr := cm.validateTokenFreshness(times, config); freshnessErr != nil {
 		return TokenRetrievalResult{Token: "", Error: freshnessErr}
 	}

 	// Determine if token is opaque or JWT based on format
 	// JWT tokens have exactly 2 dots (3 parts: header.payload.signature)
+	if cached {
+		return TokenRetrievalResult{Token: token, Error: nil}
+	}
+
 	dotCount := strings.Count(token, ".")
 	isJWT := dotCount == 2

@@ -338,6 +357,7 @@ func (cm *ChunkManager) validateToken(token string, config TokenConfig) TokenRet
 		}
 	}

+	cm.markContentValid(key, times)
 	return TokenRetrievalResult{Token: token, Error: nil}
 }

@@ -350,6 +370,53 @@ func (cm *ChunkManager) validateToken(token string, config TokenConfig) TokenRet
 //
 // Returns:
 //   - TokenRetrievalResult with the reassembled token or error.
+//
+// tokenTimes holds a token's parsed exp and iat claims. They are part of the
+// token content, so a content-cache entry keeps them to avoid re-parsing the
+// payload on every request. The checks that use them still run each time.
+type tokenTimes struct {
+	exp, iat       *time.Time
+	expErr, iatErr error
+	jwt            bool
+}
+
+// tokenTimesOf parses exp and iat from a JWT-shaped token.
+func (cm *ChunkManager) tokenTimesOf(token string) tokenTimes {
+	if !strings.Contains(token, ".") {
+		return tokenTimes{}
+	}
+	t := tokenTimes{jwt: true}
+	t.exp, t.expErr = cm.extractJWTExpiration(token)
+	t.iat, t.iatErr = cm.extractJWTIssuedAt(token)
+	return t
+}
+
+// contentValidCacheMax bounds contentValid; the map is cleared when full.
+const contentValidCacheMax = 4096
+
+// contentValidKey keys a token by type and SHA-256 so the cache does not
+// retain token values.
+func contentValidKey(token string, config TokenConfig) string {
+	sum := sha256.Sum256([]byte(token))
+	return config.Type + ":" + string(sum[:])
+}
+
+func (cm *ChunkManager) cachedTokenTimes(key string) (tokenTimes, bool) {
+	cm.contentMu.Lock()
+	t, ok := cm.contentValid[key]
+	cm.contentMu.Unlock()
+	return t, ok
+}
+
+func (cm *ChunkManager) markContentValid(key string, t tokenTimes) {
+	cm.contentMu.Lock()
+	if len(cm.contentValid) >= contentValidCacheMax {
+		cm.contentValid = make(map[string]tokenTimes)
+	}
+	cm.contentValid[key] = t
+	cm.contentMu.Unlock()
+}
+
 func (cm *ChunkManager) processChunkedToken(chunks map[int]*sessions.Session, config TokenConfig) TokenRetrievalResult {
 	if len(chunks) > config.MaxChunks {
 		err := fmt.Errorf("too many %s token chunks (%d, max: %d)", config.Type, len(chunks), config.MaxChunks)
@@ -504,7 +571,9 @@ func (cm *ChunkManager) validateJWTFormat(token string, tokenType string) error
 			return err
 		}

-		for _, char := range part {
+		// Byte loop: a multi-byte rune fails on its first byte either way.
+		for j := 0; j < len(part); j++ {
+			char := part[j]
 			if !((char >= 'A' && char <= 'Z') ||
 				(char >= 'a' && char <= 'z') ||
 				(char >= '0' && char <= '9') ||
@@ -717,9 +786,10 @@ func (cm *ChunkManager) validateTokenSanitization(token string, config TokenConf
 		return err
 	}

-	// Check for control characters (ASCII 0-31 and 127)
-	for i, char := range token {
-		if char < 32 || char == 127 {
+	// Check for control characters (ASCII 0-31 and 127). Byte loop: bytes of
+	// a multi-byte rune are >= 0x80, and i is a byte offset either way.
+	for i := 0; i < len(token); i++ {
+		if char := token[i]; char < 32 || char == 127 {
 			err := fmt.Errorf("%s token contains control character at position %d", config.Type, i)
 			return err
 		}
@@ -921,7 +991,13 @@ func (cm *ChunkManager) detectRepeatedCharacters(token string, config TokenConfi
 	currentRepeated := 1
 	var lastChar rune

-	for i, char := range token {
+	// Index loops, not range: under yaegi, ranging over a string allocates
+	// O(n²) bytes, which dominated per-request cost for 1-2 KB tokens.
+	for i := 0; i < len(token); {
+		char, size := rune(token[i]), 1
+		if char >= utf8.RuneSelf {
+			char, size = utf8.DecodeRuneInString(token[i:])
+		}
 		if i > 0 && char == lastChar {
 			currentRepeated++
 			if currentRepeated > maxRepeated {
@@ -931,6 +1007,7 @@ func (cm *ChunkManager) detectRepeatedCharacters(token string, config TokenConfi
 			currentRepeated = 1
 		}
 		lastChar = char
+		i += size
 	}

 	threshold := 20
@@ -941,8 +1018,13 @@ func (cm *ChunkManager) detectRepeatedCharacters(token string, config TokenConfi
 	}

 	charFreq := make(map[rune]int)
-	for _, char := range token {
+	for i := 0; i < len(token); {
+		char, size := rune(token[i]), 1
+		if char >= utf8.RuneSelf {
+			char, size = utf8.DecodeRuneInString(token[i:])
+		}
 		charFreq[char]++
+		i += size
 	}

 	tokenLen := len(token)
@@ -963,21 +1045,21 @@ func (cm *ChunkManager) detectRepeatedCharacters(token string, config TokenConfi
 // It extracts and checks JWT expiration claims to ensure tokens are not expired
 // and detects tokens with suspicious expiration times.
 // Parameters:
-//   - token: The token to check expiration for.
+//   - times: The token's parsed claims (see tokenTimesOf).
 //   - config: Token configuration for error context.
 //
 // Returns:
 //   - An error if the token is expired or has invalid expiration, nil if valid.
 //
 //nolint:unparam // error return kept for API consistency and future use
-func (cm *ChunkManager) validateTokenExpiration(token string, config TokenConfig) error {
-	if !strings.Contains(token, ".") {
+func (cm *ChunkManager) validateTokenExpiration(times tokenTimes, config TokenConfig) error {
+	if !times.jwt {
 		return nil
 	}

-	expiration, err := cm.extractJWTExpiration(token)
-	if err != nil {
-		cm.logger.Debugf("Could not extract expiration from %s token: %v", config.Type, err)
+	expiration := times.exp
+	if times.expErr != nil {
+		cm.logger.Debugf("Could not extract expiration from %s token: %v", config.Type, times.expErr)
 		return nil
 	}

@@ -1062,19 +1144,19 @@ func (cm *ChunkManager) extractJWTExpiration(token string) (*time.Time, error) {
 // It examines the 'iat' (issued at) claim to detect tokens issued too far
 // in the future or suspiciously old tokens that might indicate replay attacks.
 // Parameters:
-//   - token: The token to check freshness for.
+//   - times: The token's parsed claims (see tokenTimesOf).
 //   - config: Token configuration for error context.
 //
 // Returns:
 //   - An error if the token freshness is suspicious, nil if acceptable.
-func (cm *ChunkManager) validateTokenFreshness(token string, config TokenConfig) error {
-	if !strings.Contains(token, ".") {
+func (cm *ChunkManager) validateTokenFreshness(times tokenTimes, config TokenConfig) error {
+	if !times.jwt {
 		return nil
 	}

-	issuedAt, err := cm.extractJWTIssuedAt(token)
-	if err != nil {
-		cm.logger.Debugf("Could not extract issued time from %s token: %v", config.Type, err)
+	issuedAt := times.iat
+	if times.iatErr != nil {
+		cm.logger.Debugf("Could not extract issued time from %s token: %v", config.Type, times.iatErr)
 		return nil
 	}