Cap concurrent password checks and count attempts before checking

Every argon2 run takes 64 MiB and nothing limited how many ran at once, so parallel sign-in attempts could run the server out of memory (8 at once used about 600 MB). At most two now run at once; at most 16 sign-ins wait for one, more get HTTP 429. 30 parallel sign-ins peaked at 275 MB. A sign-in attempt now counts toward the lockout before its password is checked, so parallel attempts cannot get past it; a right password takes its own attempt back. IPv6 addresses are locked out by /64.
2026-10-05 00:23:05 +03:00
parent a59a095691
commit b54ff1b002
4 changed files with 136 additions and 34 deletions
@@ -232,7 +232,7 @@ func (a *App) login(w http.ResponseWriter, r *http.Request) {
 	if err != nil {
 		slog.Warn("login failed", "user", in.Username, "remote", ip, "reason", err.Error())
 		code := http.StatusUnauthorized
-		if errors.Is(err, errLocked) {
+		if errors.Is(err, errLocked) || errors.Is(err, errBusy) {
 			code = http.StatusTooManyRequests
 		}
 		writeJSON(w, code, map[string]string{"error": err.Error()})
@@ -10,6 +10,7 @@ import (
 	"fmt"
 	"net"
 	"net/http"
+	"net/netip"
 	"strings"
 	"sync"
 	"time"
@@ -26,12 +27,23 @@ const (
 	argonKeyLen  = 32
 )

+// Every argon2 run takes argonMemory (64 MiB). argonSlots caps how many run
+// at once, so a burst of sign-ins cannot run the server out of memory: two
+// slots are 128 MiB at most.
+var argonSlots = make(chan struct{}, 2)
+
+func argonKey(pw, salt []byte, t, m uint32, p uint8, n uint32) []byte {
+	argonSlots <- struct{}{}
+	defer func() { <-argonSlots }()
+	return argon2.IDKey(pw, salt, t, m, p, n)
+}
+
 func hashPassword(pw string) (string, error) {
 	salt := make([]byte, 16)
 	if _, err := rand.Read(salt); err != nil {
 		return "", err
 	}
-	key := argon2.IDKey([]byte(pw), salt, argonTime, argonMemory, argonThreads, argonKeyLen)
+	key := argonKey([]byte(pw), salt, argonTime, argonMemory, argonThreads, argonKeyLen)
 	b64 := base64.RawStdEncoding
 	return fmt.Sprintf("$argon2id$v=%d$m=%d,t=%d,p=%d$%s$%s",
 		argon2.Version, argonMemory, argonTime, argonThreads, b64.EncodeToString(salt), b64.EncodeToString(key)), nil
@@ -54,7 +66,7 @@ func verifyPassword(encoded, pw string) bool {
 	if err1 != nil || err2 != nil {
 		return false
 	}
-	got := argon2.IDKey([]byte(pw), salt, t, m, p, uint32(len(want)))
+	got := argonKey([]byte(pw), salt, t, m, p, uint32(len(want)))
 	return subtle.ConstantTimeCompare(got, want) == 1
 }

@@ -139,14 +151,18 @@ type Auth struct {
 	mu       sync.Mutex
 	sessions map[string]*session
 	used     map[string]tokenUse
-	logins   map[string]tokenUse // last sign-in per user ID
-	fails    map[string]*failState
+	logins   map[string]tokenUse   // last sign-in per user ID
+	fails    map[string]*failState // by lockKey
+	waiting  int                   // sign-ins waiting for or running a password check
 	mfa      mfaState
 }

 const (
 	maxFailures = 5
 	lockoutTime = 15 * time.Minute
+	// maxWaiting sign-ins may wait for a password check; more are turned
+	// away until the queue is shorter.
+	maxWaiting = 16
 )

 func newAuth(s *Store) *Auth {
@@ -155,23 +171,51 @@ func newAuth(s *Store) *Auth {

 func cookieName() string { return appName + "_session" }

-var errLocked = errors.New("too many failed attempts, try again later")
+var (
+	errLocked = errors.New("too many failed attempts, try again later")
+	errBusy   = errors.New("too many sign-ins at once, try again in a moment")
+)
+
+// lockKey is what failed sign-ins are counted by: the IPv4 address, or the
+// /64 network of an IPv6 address, since one device can pick any address in
+// its /64.
+func lockKey(ip string) string {
+	a, err := netip.ParseAddr(ip)
+	if err != nil || a.Unmap().Is4() {
+		return ip
+	}
+	p, _ := a.Prefix(64)
+	return p.String()
+}

 // Login checks the credentials and returns a new session id, or, for a user
 // with two-step sign-in, a ticket for the second step.
 func (a *Auth) Login(user, pw, ip string) (sessionID, ticket string, err error) {
-	a.mu.Lock()
-	f := a.fails[ip]
-	if f != nil && time.Now().Before(f.until) {
-		a.mu.Unlock()
-		return "", "", errLocked
-	}
-	a.mu.Unlock()
-
 	cfg := a.store.Get()
 	if !cfg.passwordSet() {
 		return "", "", errors.New("no password is set; run: " + appName + " passwd")
 	}
+	// The attempt counts as failed before the password is checked, so
+	// parallel attempts cannot get past the lockout; a right password takes
+	// it back.
+	a.mu.Lock()
+	if a.lockedLocked(ip) {
+		a.mu.Unlock()
+		return "", "", errLocked
+	}
+	if a.waiting >= maxWaiting {
+		a.mu.Unlock()
+		return "", "", errBusy
+	}
+	a.waiting++
+	undo := a.failLocked(ip)
+	a.mu.Unlock()
+	defer func() {
+		a.mu.Lock()
+		a.waiting--
+		a.mu.Unlock()
+	}()
+
 	// An unknown username costs as much time as a wrong password, so the
 	// answer time does not tell which usernames exist.
 	u := cfg.userByName(strings.TrimSpace(user))
@@ -185,21 +229,13 @@ func (a *Auth) Login(user, pw, ip string) (sessionID, ticket string, err error)
 	a.mu.Lock()
 	defer a.mu.Unlock()
 	if !okUser || !okPw {
-		if f == nil {
-			f = &failState{}
-			a.fails[ip] = f
-		}
-		f.count++
-		if f.count >= maxFailures {
-			f.count = 0
-			f.until = time.Now().Add(lockoutTime)
-		}
 		return "", "", errors.New("wrong username or password")
 	}
+	undo()
 	if u.hasMFA() {
 		return "", a.newTicketLocked(u, ip), nil
 	}
-	delete(a.fails, ip)
+	delete(a.fails, lockKey(ip))
 	a.logins[u.ID] = tokenUse{At: time.Now(), IP: ip}
 	return a.newSessionLocked(cfg, u, sessionInfo{Started: time.Now(), IP: ip}), "", nil
 }
@@ -321,9 +357,9 @@ func (a *Auth) sweep() {
 			delete(a.sessions, id)
 		}
 	}
-	for ip, f := range a.fails {
+	for key, f := range a.fails {
 		if now.After(f.until) && f.count == 0 {
-			delete(a.fails, ip)
+			delete(a.fails, key)
 		}
 	}
 	for id, t := range a.mfa.tickets {
@@ -14,6 +14,7 @@ import (
 	"path/filepath"
 	"slices"
 	"strings"
+	"sync"
 	"testing"
 	"time"
 )
@@ -409,6 +410,60 @@ func TestSysctlConf(t *testing.T) {
 	}
 }

+func TestLoginLockout(t *testing.T) {
+	store, err := openStore(filepath.Join(t.TempDir(), "config.json"))
+	if err != nil {
+		t.Fatal(err)
+	}
+	hash, _ := hashPassword("a long test password")
+	_ = store.Update(func(c *Config) error { c.Users[0].PasswordHash = hash; return nil })
+	a := newAuth(store)
+	const right, wrong = "a long test password", "a wrong password"
+
+	// Ten wrong attempts at once from one /64: five are checked, the others
+	// are locked out before any password check.
+	var wg sync.WaitGroup
+	var mu sync.Mutex
+	got := map[string]int{}
+	for i := range 10 {
+		wg.Add(1)
+		go func() {
+			defer wg.Done()
+			_, _, err := a.Login("admin", wrong, fmt.Sprintf("2001:db8::%x", i+1))
+			mu.Lock()
+			got[err.Error()]++
+			mu.Unlock()
+		}()
+	}
+	wg.Wait()
+	if got["wrong username or password"] != 5 || got[errLocked.Error()] != 5 {
+		t.Fatalf("parallel attempts: %v", got)
+	}
+	if _, _, err := a.Login("admin", right, "2001:db8::ffff"); !errors.Is(err, errLocked) {
+		t.Fatalf("same /64: %v, want locked", err)
+	}
+	if _, _, err := a.Login("admin", right, "2001:db8:0:1::1"); err != nil {
+		t.Fatalf("other /64: %v", err)
+	}
+
+	// A right password takes its own attempt back. With two-step sign-in
+	// the earlier failures stay, so wrong codes still lead to the lockout.
+	_ = store.Update(func(c *Config) error { c.Users[0].MFA = &UserMFA{TOTPSecret: newTOTPSecret()}; return nil })
+	ip := "192.0.2.7"
+	for range maxFailures - 1 {
+		_, _, _ = a.Login("admin", wrong, ip)
+	}
+	if _, tk, err := a.Login("admin", right, ip); err != nil || tk == "" {
+		t.Fatalf("5th attempt, right password: ticket %q, %v", tk, err)
+	}
+	if _, _, err := a.Login("admin", wrong, ip); err == nil || errors.Is(err, errLocked) {
+		t.Fatalf("6th attempt: %v, want wrong password", err)
+	}
+	if _, _, err := a.Login("admin", right, ip); !errors.Is(err, errLocked) {
+		t.Fatalf("7th attempt: %v, want locked", err)
+	}
+}
+
 func TestWriteIfChanged(t *testing.T) {
 	p := filepath.Join(t.TempDir(), "x.conf")
 	if ch, err := writeIfChanged(p, "a\n", 0o644); !ch || err != nil {
@@ -236,23 +236,34 @@ func newMFAState() mfaState {

 var errBadTicket = errors.New("the sign-in expired; enter your password again")

-// failLocked counts a failed attempt from ip toward the lockout. a.mu must
-// be held.
-func (a *Auth) failLocked(ip string) {
-	f := a.fails[ip]
+// failLocked counts a failed attempt from ip toward the lockout and returns
+// a function that takes it back, for an attempt counted before it was
+// checked. a.mu must be held, also when calling undo.
+func (a *Auth) failLocked(ip string) (undo func()) {
+	key := lockKey(ip)
+	f := a.fails[key]
 	if f == nil {
 		f = &failState{}
-		a.fails[ip] = f
+		a.fails[key] = f
 	}
 	f.count++
-	if f.count >= maxFailures {
+	locked := f.count >= maxFailures
+	if locked {
 		f.count = 0
 		f.until = time.Now().Add(lockoutTime)
 	}
+	return func() {
+		switch {
+		case locked:
+			f.count, f.until = maxFailures-1, time.Time{}
+		case f.count > 0:
+			f.count--
+		}
+	}
 }

 func (a *Auth) lockedLocked(ip string) bool {
-	f := a.fails[ip]
+	f := a.fails[lockKey(ip)]
 	return f != nil && time.Now().Before(f.until)
 }

@@ -302,7 +313,7 @@ func (a *Auth) finishSignIn(u *User, ip string) string {
 	cfg := a.store.Get()
 	a.mu.Lock()
 	defer a.mu.Unlock()
-	delete(a.fails, ip)
+	delete(a.fails, lockKey(ip))
 	a.logins[u.ID] = tokenUse{At: time.Now(), IP: ip}
 	return a.newSessionLocked(cfg, u, sessionInfo{Started: time.Now(), IP: ip})
 }