Catalyst 4500 Series Switch Software Configuration Guide, Release IOS XE 3.3(0)XO

Device Roles

With 802.1X port-based authentication, network devices have specific roles. Figure 41-1 shows the role of each device, which is described below.

Figure 41-1 802.1X Device Roles

• Client—The workstation that requests access to the LAN, and responds to requests from the switch. The workstation must be running 802.1X-compliant client software.
• Authenticator—Controls physical access to the network based on the authentication status of the client. The Catalyst 4500 series switch acts as an intermediary between the client and the authentication server, requesting identity information from the client, verifying that information with the authentication server, and relaying a response to the client. The switch encapsulates and decapsulates the Extensible Authentication Protocol (EAP) frames and interacts with the RADIUS authentication server.

When the switch receives EAPOL frames and relays them to the authentication server, the Ethernet header is stripped and the remaining EAP frame is reencapsulated in the RADIUS format. The EAP frames are not modified or examined during encapsulation, and the authentication server must support EAP within the native frame format. When the switch receives frames from the authentication server, the frame header is removed from the server, leaving the EAP frame, which is then encapsulated for Ethernet and sent to the client.

Note The Catalyst 4500 series switches must be running software that supports the RADIUS client and 802.1X.

• Authentication server—Performs the actual authentication of the client. The authentication server validates the identity of the client and notifies the switch that the client is authorized to access the LAN and switch services. (The only supported authentication server is the RADIUS authentication server with EAP extensions; it is available in Cisco Secure Access Control Server version 3.2 and later releases.)

802.1X and Network Access Control

Network Access Control is a feature that allows port access policies to be influenced by the antivirus posture of the authenticating device.

Antivirus posture includes such elements as the operating system running on the device, the operating system version, whether antivirus software is installed and what version of antivirus signatures is available. If the authenticating device has a NAC-aware 802.1X supplicant and the authentication server is configured to support NAC using 802.1X, antivirus posture information is automatically included as part of the 802.1X authentication exchange.

For information on NAC, refer to the URL:

http://www.cisco.com/en/US/products/ps6128/index.html

Authentication Initiation and Message Exchange

The switch or the client can initiate authentication. If you enable authentication on a port by using the
authentication port-control auto interface configuration command, the switch must initiate authentication when it determines that the port link state has changed. It then sends an EAP-request/identity frame to the client to request its identity (typically, the switch sends an initial identity/request frame followed by one or more requests for authentication information). Upon receipt of the frame, the client responds with an EAP-response/identity frame.

However, if during bootup, the client does not receive an EAP-request/identity frame from the switch, the client can initiate authentication by sending an EAPOL-start frame, which prompts the switch to request the client’s identity.

If 802.1X is not enabled or supported on the network access switch, any EAPOL frames from the client are dropped. If the client does not receive an EAP-request/identity frame after three attempts to start authentication, the client transmits frames as if the port is in the authorized state. A port in the authorized state means that the client was successfully authenticated. When the client supplies its identity, the switch begins its role as the intermediary, passing EAP frames between the client and the authentication server until authentication succeeds or fails. If the authentication succeeds, the switch port becomes authorized.

The specific exchange of EAP frames depends on the authentication method being used. Figure 41-2 shows a message exchange that is initiated by the client using the One-Time Password (OTP) authentication method with an authentication server.

Figure 41-2 Message Exchange

Ports in Authorized and Unauthorized States

The switch port state determines whether the client is granted access to the network. The port starts in the unauthorized state. While in this state, the port disallows all ingress and egress traffic except for 802.1X protocol packets. When a client is successfully authenticated, the port transitions to the authorized state, allowing all traffic for the client to flow normally.

If a non-802.1X capable client is connected to an unauthorized 802.1X port, the switch requests the client’s identity. In this situation, the client does not respond to the request, the port remains in the unauthorized state, and the client is not granted access to the network. If a guest VLAN is configured on a port that connects to a client that does not support 802.1X, the port is placed in the configured guest VLAN and in the authorized state. For more information, see the “Using 802.1X for Guest VLANs” section.

In contrast, when an 802.1X-enabled client connects to a port that is not running the 802.1X protocol, the client initiates the authentication process by sending the EAPOL-start frame. When no response is received, the client sends the request a fixed number of times. Because no response is received, the client begins sending frames as if the port is in the authorized state.

You can control the port authorization state by using the authentication port-control interface configuration command and these keywords:

• force-authorized —Disables 802.1X authentication and causes the port to transition to the authorized state without requiring authentication exchange. The port transmits and receives normal traffic without 802.1X-based authentication of the client. This setting is the default.
• force-unauthorized —Causes the port to remain in the unauthorized state, ignoring all attempts by the client to authenticate. The switch cannot provide authentication services to the client using the interface.
• auto —Allows 802.1X authentication and causes the port to begin in the unauthorized state, allowing only EAPOL frames to be sent and received using the port. The authentication process begins when the link state of the port transitions from down to up or when an EAPOL-start frame is received. The switch requests the identity of the client and begins relaying authentication messages between the client and the authentication server. The switch can uniquely identify each client attempting to access the network by the client’s MAC address.

If the client is successfully authenticated (receives an Accept frame from the authentication server), the port state changes to authorized, and all frames from the authenticated client are allowed using the port. If authentication fails, the port remains in the unauthorized state, but authentication can be retried. If the authentication server cannot be reached, the switch can retransmit the request. If no response is received from the server after the specified number of attempts, authentication fails and network access is not granted.

If the link state of a port transitions from up to down, or if an EAPOL-logoff frame is received by the port, the port returns to the unauthorized state.

If Multidomain Authentication (MDA) is enabled on a port, this flow can be used with some exceptions that are applicable to voice authorization. For more information on MDA, see the
“Using Multiple Domain Authentication and Multiple Authentication” section.

Figure 41-3 shows the authentication process.

Figure 41-3 Authentication Flowchart

802.1X Host Mode

The 802.1X port’s host mode determines whether more than one client can be authenticated on the port and how authentication is enforced. You can configure an 802.1X port to use any of the five host modes described in the following sections. In addition, each mode can be modified to allow preauthentication open access:

• Single-Host Mode
• Multiple-Hosts Mode
• Multidomain Authentication Mode
• Multiauthentication Mode
• Pre-authentication Open Access

Single-Host Mode

You can configure an 802.1X port for single-host or multiple-hosts mode. In single-host mode (see Figure 41-1), only one client can be connected to the 802.1X-enabled switch port. The switch detects the client by sending an EAPOL frame when the port link state changes to the up state. If a client leaves or is replaced with another client, the switch changes the port link state to down, and the port returns to the unauthorized state.

Multiple-Hosts Mode

In multiple-hosts mode, you can attach multiple hosts to a single 802.1X-enabled port. Figure 41-4 shows 802.1X port-based authentication in a wireless LAN. In this mode, only one of the attached clients must be authorized for all clients to be granted network access. If the port becomes unauthorized (reauthentication fails or an EAPOL-logoff message is received), the switch denies network access to all of the attached clients. In this topology, the wireless access point is responsible for authenticating the clients attached to it, and it also acts as a client to the switch.

With multiple-hosts mode enabled, you can use 802.1X authentication to authenticate the port and port security to manage network access for all MAC addresses, including that of the client.

Figure 41-4 Multiple Host Mode Example

Multidomain Authentication Mode

Catalyst 4500 series switches support Multidomain Authentication (MDA), which allows an IP phone (Cisco or third-party) and a single host behind the IP phone to authenticate independently, using 802.1X, MAC authentication bypass (MAB) or (for the host only) web-based authentication. In this application, multidomain refers to two domains — data and voice — and only two MAC addresses are allowed per-port. A switch can place the host in the data VLAN and the IP phone in the voice VLAN, even though they appear on the same switch port. The data VLAN and the voice VLAN can be specified in the CLI configuration. The devices are identified as either data or voice depending on the vendor-specific-attributes (VSAs) received from the authentication, authorization, and accounting (AAA) server. The data and voice VLANs can also be obtained from the VSAs received from the (AAA) server during authentication.

Figure 41-5 Multidomain Authentication Mode Example

Figure 41-5Figure 41-5 shows a typical MDA application with a single host behind an IP phone connected to the 802.1X-enabled port. Because the client is not directly connected to the switch, the switch cannot detect a loss of port link if the client is disconnected. To prevent another device from using the established authentication of the disconnected client later, Cisco IP phones send a Cisco Discovery Protocol (CDP) host presence type length value (TLV) to notify the switch of changes in the attached client’s port link state.

For details on how to configure MDA, see the “Using Multiple Domain Authentication and Multiple Authentication” section.

Multiauthentication Mode

Multiauthentication mode allows one client on the voice VLAN and multiple authenticated clients on the data VLAN. When a hub or access point is connected to an 802.1X port, multiauthentication mode provides enhanced security over multiple-hosts mode by requiring authentication of each connected client. For non-802.1X devices, you can use MAB or web-based authentication as the fallback method for individual host authentications, allowing you to authenticate different hosts through different methods on a single port.

Multiauthentication also supports MDA functionality on the voice VLAN by assigning authenticated devices to either a data or voice VLAN depending on the VSAs received from the authentication server.

Note When a port is in multiauthentication mode, Guest VLAN and Authentication Failed VLAN will not activate for data devices.

Pre-authentication Open Access

Any of the four host modes can be additionally configured to allow a device to gain network access before authentication. This preauthentication open access is useful in an application such as the Pre-boot eXecution Environment (PXE), where a device must access the network to download a bootable image containing an authentication client.

Enable preauthentication open access by entering the authentication open command after host mode configuration. It acts as an extension to the configured host mode. For example, if preauthentication open access is enabled with single-host mode, then the port allows only one MAC address. When preauthentication open access is enabled, initial traffic on the port is restricted only by whatever other access restriction, independent of 802.1X, is configured on the port. If no access restriction other than 802.1X is configured on the port, then a client device has full access on the configured VLAN.

802.1X Violation Mode

You can use the authentication violation interface configuration command to configure the violation mode: restrict, shutdown, and replace.

In single-host mode, a security violation is triggered when more than one device are detected on the data vlan. In multidomain authentication mode, a security violation is triggered when more than one device are detected on the data or voice VLAN.

Security violation cannot be triggered in multiple-host mode or multiauthentication mode.

When security violation occurs, the port is protected depending on the configured violation action:

Shutdown—Errdisables the port; the default behavior on a port.

Restrict—The port state is unaffected. However the platform is notified to restrict the traffic from offending MAC-address.

Replace—Replaces existing host with the new host, instead of error-disabling or restricting the port.

For more information see “Configuring Violation Action” section.

Using MAC Move

Hosts should be able to move across ports within a switch on the same or different VLAN without restriction, as if they had moved to a port on another switch.

You can move a MAC address to another port on the same switch. it is not pertinent for directly connected hosts or for hosts behind Cisco phones, where a port-down event or proxy EAPoL-Logoff/CDP TLV is received when the initial host disconnects. It is pertinent for hosts that disconnect from behind a hub, third party phone, or legacy Cisco phone, causing the session to remain up. With MAC move you can disconnect the host from such a device and connect it directly to another port on the same switch.

You can globally enable MAC move so that the device is reauthenticated on the new port. When a host moves to a second port, the session on the first port is deleted, and the host is reauthenticated on the new port.

MAC move is supported on all host modes. (The authenticated host can move to any port on the switch, for any host mode enabled on that port.)

For more information see “Configuring MAC Move” section.

Using MAC Replace

You can allow new hosts to connect to abandoned ports. If the configured violation action is replace, the existing host is replaced by the new host, instead of err-disabling or restricting the port (as happens for single-host and MDA modes).

it is not an issue for directly connected hosts or for hosts behind Cisco phones, where a port-down event or proxy EAPoL-Logoff/CDP TLV is received when the initial host disconnects. It is an issue where a host disconnects from behind a hub, third party phone, or legacy Cisco phone, causing the session to remain up. New hosts connecting to this port violate the host-mode, triggering a violation. When the violation action is replace, the NAD (switch) terminates the initial session and resets the authentication sequence based on the new MAC. This applies to single-host and MDA host modes. In multiple- auth mode, no attempt is made to remove an existing session on the same port.

For more information see the “Configuring MAC Replace” section.

Using 802.1X with VLAN Assignment

You can use the VLAN assignment to limit network access for certain users. With the VLAN assignment, 802.1X-authenticated ports are assigned to a VLAN based on the username of the client connected to that port. The RADIUS server database maintains the username-to-VLAN mappings. After successful 802.1X authentication of the port, the RADIUS server sends the VLAN assignment to the switch. The VLAN can be a standard VLAN or a PVLAN.

On platforms that support PVLANs, you can isolate hosts by assigning ports into PVLANs.

When configured on the switch and the RADIUS server, 802.1X with VLAN assignment has these characteristics:

• 802.1X with VLAN assignment is unsupported on trunk ports, dynamic ports, or with dynamic-access port assignment through a VLAN Membership Policy Server (VMPS).
• If no VLAN is supplied by the RADIUS server, the port is configured in its access VLAN or isolated PVLAN when authentication succeeds.
• If the authentication server provides invalid VLAN information, the port remains unauthorized. This situation prevents ports from appearing unexpectedly in an inappropriate VLAN due to a configuration error.
• If multi-authentication mode is enabled on an 802.1X port, VLAN Assignment occurs successfully for the first authenticated host. Subsequent authorized (based on user credentials) data hosts, are considered successfully authenticated, provided either they have no VLAN assignment or have a VLAN assignment matching the first successfully authenticated host on the port. This ensures that all successfully authenticated hosts on a port are members of the same VLAN. Flexibility of VLAN assignment is only provided to the first authenticated host.
• If the authentication server provides valid VLAN information, the port is authorized and placed in the specified VLAN when authentication succeeds.
• If the multiple-hosts mode is enabled, all hosts are in the same VLAN as the first authenticated user.
• If 802.1X is disabled on the port, the port is returned to the configured access VLAN.
• A port must be configured as an access port (which can be assigned only into “regular” VLANs), or as a PVLAN host port (which can be assigned only into PVLANs). Configuring a port as a PVLAN host port implies that all hosts on the port are assigned into PVLANs, whether their posture is compliant or non-compliant. If the type of the VLAN named in the Access-Accept does not match the type of VLAN expected to be assigned to the port (regular VLAN to access port, secondary PVLAN to PVLAN host port), the VLAN assignment fails.
• If a guest VLAN is configured to handle non-responsive hosts, the type of VLAN configured as the guest VLAN must match the port type (that is, guest VLANs configured on access ports must be standard VLANs, and guest VLANs configured on PVLAN host ports must be PVLANs). If the guest VLAN’s type does not match the port type, non-responsive hosts are treated as if no guest VLAN is configured (that is, they are denied network access).
• To assign a port into a PVLAN, the named VLAN must be a secondary PVLAN. The switch determines the implied primary VLAN from the locally configured secondary-primary association.

Note If you change the access VLAN or PVLAN host VLAN mapping on a port that is already authorized in a RADIUS assigned VLAN, the port remains in the RADIUS assigned VLAN.

To configure VLAN assignment you need to perform these tasks:

• Enable AAA authorization by using the network keyword to allow interface configuration from the RADIUS server. For an illustration of how to apply the aaa authorization network group radius command, refer to the section “Enabling 802.1X Authentication” section.
• Enable 802.1X. (The VLAN assignment feature is automatically enabled when you configure 802.1X on an access port.)
• Assign vendor-specific tunnel attributes in the RADIUS server. To ensure proper VLAN assignment, the RADIUS server must return these attributes to the switch:

– Tunnel-Type = VLAN

– Tunnel-Medium-Type = 802

– Tunnel-Private-Group-ID = VLAN NAME

Using 802.1X for Guest VLANs

You can use guest VLANs to enable non-802.1X-capable hosts to access networks that use 802.1X authentication. For example, you can use guest VLANs while you are upgrading your system to support 802.1X authentication.

Guest VLANs are supported on a per-port basis, and you can use any VLAN as a guest VLAN as long as its type matches the type of the port. If a port is already forwarding on the guest VLAN and you enable 802.1X support on the network interface of the host, the port is immediately moved out of the guest VLAN and the authenticator waits for authentication to occur.

Enabling 802.1X authentication on a port starts the 802.1X protocol. If the host fails to respond to packets from the authenticator within a certain amount of time, the authenticator brings the port up in the configured guest VLAN.

If the port is configured as a PVLAN host port, the guest VLAN must be a secondary PVLAN. If the port is configured as an access port, the guest VLAN must be a regular VLAN. If the guest VLAN configured on a port is not appropriate for the type of the port, the switch behaves as if no guest VLAN is configured (that is, non-responsive hosts are denied network access).

For details on how to configure guest VLANs, see the “Configuring 802.1X with Guest VLANs” section.

Usage Guidelines for Using 802.1X Authentication with Guest VLANs

When using 802.1X authentication with guest VLANs, consider these guidelines:

• When you reconfigure a guest VLAN to a different VLAN, any authentication failed ports are also moved and the ports stay in their current authorized state.
• When you shut down or remove a guest VLAN from the VLAN database, any authentication failed ports are immediately moved to an unauthorized state and the authentication process is restarted.

Note No periodic reauthentication is allowed with guest VLANs.

Usage Guidelines for Using 802.1X Authentication with Guest VLANs on Windows-XP Hosts

When using 802.1X authentication with guest VLANs on Windows-XP hosts, consider these guidelines:

• If the host fails to respond to the authenticator, the port attempts to connect three times (with a 30 second timeout between each attempt). After this time, the login/password window does not appear on the host, so you must unplug and reconnect the network interface cable.
• Hosts responding with an incorrect login/password fail authentication. Hosts failing authentication are not put in the guest VLAN. The first time that a host fails authentication, the quiet-period timer starts, and no activity occurs for the duration of the quiet-period timer. When the quiet-period timer expires, the host is presented with the login and password window. If the host fails authentication for the second time, the quiet-period timer starts again, and no activity occurs for the duration of the quiet-period timer. The host is presented with the login and password window a third time. If the host fails authentication the third time, the port is placed in the unauthorized state, and you must disconnect and reconnect the network interface cable.

Using 802.1X with MAC Authentication Bypass

The 802.1X protocol has 3 entities: client (supplicant), authenticator, and authentication server. Typically, the host PC runs the supplicant software and tries to authenticate itself by sending its credentials to the authenticator which in turn relays that info to the authentication server for authentication.

However, not all hosts may have supplicant functionality. Devices that cannot authenticate themselves using 802.1X but still need network access can use MAC Authentication Bypass (MAB), which uses the connecting device's MAC address to grant or deny network access.

Typically, you use this feature on ports where devices such as printers are connected. Such devices do not have 802.1X supplicant functionality.

In a typical deployment, the RADIUS server maintains a database of MAC addresses that require access. When this feature detects a new MAC address on a port, it generates a RADIUS request with both username and password as the device's MAC address. After authorization succeeds, the port is accessible to the particular device using the same code path that 802.1X authentication would take when processing an 802.1X supplicant. If authentication fails, the port moves to the guest VLAN if configured, or it remains unauthorized.

The Catalyst 4500 series switch also supports reauthentication of MACs on a per-port level. Be aware that the reauthentication functionality is provided by 802.1X and is not MAB specific. In the reauthentication mode, a port stays in the previous RADIUS-sent VLAN and tries to re-authenticate itself. If the reauthentication succeeds, the port stays in the RADIUS-sent VLAN. Otherwise, the port becomes unauthorized and moves to the guest VLAN if one is configured.

For details on how to configure MAB, see the “Configuring 802.1X with MAC Authentication Bypass” section.

Feature Interaction

This section lists feature interactions and restrictions when MAB is enabled. If a feature is not listed, assume that it interacts seamlessly with MAB (such as Unidirectional Controlled Port).

• MAB can only be enabled if 802.1X is configured on a port. MAB functions as a fall back mechanism for authorizing MACs. If you configure both MAB and 802.1X on a port, the port attempts to authenticate using 802.1X. If the host fails to respond to EAPOL requests and MAB is configured, the 802.1X port is opened up to listen to packets and to grab a MAC address, rather than attempt to authenticate endlessly.

Based on the default 802.1X timer values, the transition between mechanisms takes approximately 90 seconds. You can shorten the time by reducing the value of the transmission period time, which affects the frequency of EAPOL transmission. A smaller timer value results in sending EAPOLs during a shorter time interval. With MAB enabled, after 802.1X performs one full set of EAPOLs, the learned MAC address is forwarded to the authentication server for processing.

The MAB module performs authorization for the first MAC address detected on the wire. The port is considered authorized once a valid MAC address is received that RADIUS approves of.

802.1X authentication can re-start if an EAPOL packet is received on a port that was initially authorized as a result of MAB.

Figure 41-6 shows the message exchange during MAB.

Figure 41-6 Message Exchange during MAC Authentication Bypass

• The authentication-failed VLAN is used only with dot1x-authentication-failed users. MAB is not attempted with dot1x-authentication-failed users. If 802.1X authentication fails, a port moves to the authentication-failed VLAN (if configured) whether MAB is configured or not.
• When both MAB and guest VLAN are configured and no EAPOL packets are received on a port, the 802.1X state-machine is moved to a MAB state where it opens the port to listen to traffic and grab MAC addresses. The port remains in this state forever waiting to see a MAC on the port. A detected MAC address that fails authorization causes the port to be moved to the guest VLAN if configured.

While in a guest VLAN, a port is open to all traffic on the specified guest VLAN. Non-802.1X supplicants that normally would be authorized but are in guest VLAN due to the earlier detection of a device that failed authorization, would remain in the guest VLAN indefinitely. However, loss of link or the detection of an EAPOL on the wire causes a transition out of the guest VLAN and back to the default 802.1X mode.

• Once a new MAC is authenticated by MAB, the responsibility to limit access belongs to the 802.1X authenticator (or port security) to secure the port. The 802.1X default host parameter is defined only for a single host. If the port is changed to multiple- user host, port security must be used to enforce the number of MAC addresses allowed through this port.
• Catalyst 4500 series switch supports MAB with VVID, with the restriction that the MAC address appears on a port data VLAN only. All IP phone MACs learned using CDP are allowed on voice VLANs.
• MAB and VMPS are mutually exclusive because their functionality overlaps.

Using 802.1X with Web-Based Authentication

The web-based authentication feature, known as Web Authentication Proxy, allows you to authenticate end users on host systems that do not run the IEEE 802.1X supplicant.

When configuring web-based authentication, consider these guidelines:

• Fallback to web-based authentication is configured on switch ports in access mode. Ports in trunk mode are not supported.
• Fallback to web-based authentication is not supported on EtherChannels or EtherChannel members.
• Although fallback to web-based authentication is an interface-specific configuration, the web-based authentication fallback behavior is defined in a global fallback profile. If the global fallback configuration changes, the new profile is not used until the next instance of authentication fallback.

For detailed information on configuring web-based authentication, see Chapter43, “Configuring Web-Based Authentication”

Using 802.1X with Inaccessible Authentication Bypass

When a switch cannot reach the configured RADIUS servers and clients (supplicants) cannot be authenticated, you can configure a switch to allow network access to hosts connected to critical ports that are enabled for Inaccessible Authentication Bypass.

When Inaccessible Authentication Bypass is enabled, a switch monitors the status of the configured RADIUS servers. If no RADIUS servers are available, clients that fail authentication due to server unavailability are authorized. Inaccessible Authentication Bypass can be enabled for data clients and voice clients. For data clients, you can specify an Inaccessible Authentication Bypass VLAN on a per-port basis. For voice clients they are authorized in the configured voice vlan. Inaccessible Authentication Bypass for voice clients can activate in Multiple Domain Authentication and Multiple Authentication modes, in which authentication is enforced for voice devices.

Note Inaccessible Authentication Bypass allows a voice client to access configured voice VLAN when RADIUS becomes unavailable. For the voice device to operate properly, it must learn the voice VLAN ID through other protocols such as CDP, LLDP, or DHCP, wherever appropriate. When a RADIUS server is unavailable, it may not be possible for a switch to recognize a MAC address as that of a voice device. Therefore, when Inaccessible Authentication Bypass is configured for voice devices, it should also be configured for data. Voice devices may be authorized on both critical data and voice VLANs. If port security is enabled, this may affect the maximum port security entries enforced on the port.

By default, data clients that were already authorized when RADIUS becomes unavailable are unaffected by Inaccessible Authentication Bypass. To reauthenticate all authorized data clients on the port when RADIUS becomes unavailable, use the authentication server dead action reinitialize vlan interface configuration command. This command is intended for multiauthentication mode and is mutually exclusive with the authentication server dead action authorize vlan command.

Note In multiauthentication mode, you cannot use the authentication server dead action authorize vlan command to enable Inaccessible Authentication Bypass for data clients; it has no effect. Instead, use the authentication server dead action reinitialize vlan vlan-id command.

When RADIUS becomes available, critically authorized ports can be configured to automatically reauthenticate themselves.

Note To properly detect RADIUS server availability, the test username name option should be enabled in the radius-server host command. For details on how to configure RADIUS server, see the “Configuring Switch-to-RADIUS-Server Communication” section.

Inaccessible Authentication Bypass cannot activate after a port falls back to Web-based authentication. For details on how to configure Web-based authentication, see Chapter43, “Configuring Web-Based Authentication”

For details on how to configure Inaccessible Authentication Bypass, see Chapter 43, “Configuring Web-Based Authentication”.

Using 802.1X with Unidirectional Controlled Port

Unidirectional Controlled Port is a combined hardware and software feature that allows dormant PCs to be powered on based on the receipt of a specific Ethernet frame, known as the magic packet. Generally, Unidirectional Controlled Port is used in environments where administrators plan to manage remote systems during off-hours, when the systems usually have been powered down.

Use of Unidirectional Controlled Port with hosts attached through 802.1X ports presents a unique problem: when the host powers down, a 802.1X port becomes unauthorized. In this state, the port allows the receipt and transmission of EAPoL packets only. The Unidirectional Controlled Port magic packet cannot reach the host; without powering up, the PC cannot authenticate and open the port.

Unidirectional Controlled Port solves this problem by allowing packets to be transmitted on unauthorized 802.1X ports.

Note Unidirectional Controlled Port only works when Spanning Tree PortFast is enabled on the port.

For details on how to configure 802.1X with Unidirectional Controlled Port, see the “Configuring 802.1X with Unidirectional Controlled Port” section.

Unidirectional State

A unidirectional controlled port is typically configured when a connected host might enter a sleeping mode or power-down state. When either occurs, the host does not exchange traffic with other devices in the network. A host connected to the unidirectional port cannot send traffic to the network; it can only receive traffic from other devices in the network.

When you configure a port as unidirectional (with the authentication control-direction in interface configuration command), the port will receive traffic in VLANs on that port, but it is not put into a spanning-tree forwarding state. If a VLAN contains only unauthenticated ports, any SVI on that VLAN will be in a down state, during which packets will not be routed into the VLAN. For the SVI to be up, and so enable packets to be routed into the VLAN, at least one port in the VLAN must either be authenticated or in the spanning-tree forwarding state.

Bidirectional State

When you configure a port as bidirectional by using the authentication control-direction both interface configuration command, the port is access-controlled in both directions. In this state, except for EAPOL packets, a switch port does not receive or send packets.

Using 802.1X with VLAN User Distribution

An alternative to dynamically assigning a VLAN ID or a VLAN name is to assign a VLAN group name. The 802.1X VLAN User Distribution feature allows you to distribute users belonging to the same group (and characterized by a common VLAN group name) across multiple VLANs. You usually do this to avoid creating an overly large broadcast domain.

For example, with this feature, you can download a common VLAN group name (similar to ENG-Group, for all the users belonging to the engineering organization) from the authentication server to all the access-layer switches. The VLAN group name is then individually mapped to a different VLAN on each access-layer switch. The same VLAN number need not be spanned across separate switches. Similarly, the VLANs does not need to be renamed at the edge devices.

When the authentication server returns more than one VLAN group name or VLANs, this feature attempts to distribute users evenly across those groups. It internally maintains the count of users assigned to each VLAN on that switch by authentication or port security. Based on this information, this feature assigns a newly authenticated user to the least loaded VLAN on that switch among all the VLANs or VLAN group names obtained from the RADIUS server.

This VLAN distribution considers the load of all the valid VLANs only during initial user authentication, and not during reassignment. When some of the existing authenticated users are removed, the feature does not attempt to redistribute the remaining authenticated users. Group distribution does not guarantee perfect load distribution all the time.

Deployment Example

In a large campus LAN design, you might want to design the VLAN infrastructure without large Layer 2 domain. For the same employee VLAN, customers might have different VLANs at different campus access switches. When you deploy 802.1X with VLAN assignment, it does not assign one employee VLAN to all employees. You have to know the real VLANs configured on the switch. User distribution allows you to send a list of VLAN or VLAN group name(s) to the switch. Your switch can then do a local mapping to the corresponding VLAN. (Figure 41-7).

Figure 41-7 802.1X with VLAN User Distribution

For details on how to configure VLAN User Distribution, see the “Configuring 802.1X with VLAN User Distribution” section.

Using 802.1X with Authentication Failed VLAN Assignment

You can use authentication-failed VLAN assignment on a per-port basis to provide access for authentication failed users. Authentication failed users are end hosts that are 802.1X- capable but do not have valid credentials in an authentication server or end hosts that do not give any username and password combination in the authentication pop-up window on the user side.

If a user fails the authentication process, that port is placed in the authentication-failed VLAN. The port remains in the authentication-failed VLAN until the reauthentication timer expires. When the reauthentication timer expires the switch starts sending the port reauthentication requests. If the port fails reauthentication it remains in the authentication-failed VLAN. If the port is successfully reauthenticated, the port is moved either to the VLAN sent by RADIUS server or to the newly authenticated ports configured VLAN; the location depends on whether RADIUS is configured to send VLAN information.

Note When enabling periodic reauthentication (see the “Enabling Periodic Reauthentication” section), only local reauthentication timer values are allowed. You cannot use a RADIUS server to assign the reauthentication timer value.

You can set the maximum number of authentication attempts that the authenticator sends before moving a port into the authentication-failed VLAN. The authenticator keeps a count of the failed authentication attempts for each port. A failed authentication attempt is either an empty response or an EAP failure. The authenticator tracks any mix of failed authentication attempts towards the authentication attempt count. After the maximum number of attempts is reached the port is placed in the authentication-failed VLAN until the reauthentication timer expires again.

Note RADIUS can send a response without an EAP packet in it when it does not support EAP, and sometimes third-party RADIUS servers also send empty responses. When this behavior occurs, the authentication attempt counter is incremented.

For details on how to configure Authentication Failed VLAN Assignment, see the “Configuring 802.1X with Authentication Failed” section.

Usage Guidelines for Using Authentication Failed VLAN Assignment

Usage guidelines include the following:

• You should enable reauthentication. The ports in authentication-failed VLANs do not receive reauthentication attempts if reauthentication is disabled. To start the reauthentication process the authentication-failed VLAN must receive a link-down event or an EAP logoff event from the port. If the host is behind a hub, you may never get a link-down event and may not detect the new host until the next reauthentication occurs.
• EAP failure messages are not sent to the user. If the user failures authentication the port is moved to an authentication-failed VLAN and a EAP success message is sent to the user. Because the user is not notified of the authentication failure there may be confusion as to why there is restricted access to the network. A EAP Success message is sent for the following reasons:

– If the EAP Success message is not sent, the user tries to authenticate every 60 seconds (by default) by sending an EAP-start message.

– In some cases, users have configured DHCP to EAP-Success and unless the user sees a success, DHCP does not work on the port.

• Sometimes a user caches an incorrect username and password combination after receiving a EAP success message from the authenticator and reuses that information in every reauthentication. Until the user passes the correct username and password combination the port remains in the authentication-failed VLAN.
• When an authentication failed port is moved to an unauthorized state the authentication process is restarted. If you should fail the authentication process again the authenticator waits in the held state. After you have correctly reauthenticated all 802.1X ports are reinitialized and treated as normal 802.1X ports.
• When you reconfigure an authentication-failed VLAN to a different VLAN, any authentication failed ports are also moved and the ports stay in their current authorized state.
• When you shut down or remove an authentication-failed VLAN from the VLAN database, any authentication failed ports are immediately moved to an unauthorized state and the authentication process is restarted. The authenticator does not wait in a held state because the authentication-failed VLAN configuration still exists. While the authentication-failed VLAN is inactive, all authentication attempts are counted, and as soon as the VLAN becomes active the port is placed in the authentication-failed VLAN.
• If you reconfigure the maximum number of authentication failures allowed by the VLAN, the change takes affect after the reauthentication timer expires.
• Internal VLANs that are used for Layer 3 ports cannot be configured as authentication-failed VLANs.
• The authentication-failed VLAN is supported only in single-host mode (the default port mode).
• When a port is placed in an authentication-failed VLAN the user’s MAC address is added to the mac-address-table. If a new MAC address appears on the port, it is treated as a security violation.
• When an authentication failed port is moved to an authentication-failed VLAN, the Catalyst 4500 series switch does not transmit a RADIUS-Account Start Message as it does for standard 802.1X authentication.

Using 802.1X with Port Security

You can enable port security on an 802.1X port in either single- or multiple-host mode. (To do so, you must configure port security by using the switchport port-security interface configuration command.) When you enable port security and 802.1X on a port, 802.1X authenticates the port, and port security manages the number of MAC addresses allowed on that port, including that of the client. You can use an 802.1X port with port security enabled to limit the number or group of clients that can access the network.

For information on selecting multiple host mode, see the “Resetting the 802.1X Configuration to the Default Values” section.

These examples describe the interaction between 802.1X and port security on a switch:

• When a client is authenticated, and the port security table is not full, the client’s MAC address is added to the port security list of secure hosts. The port then proceeds to come up normally.

When a client is authenticated and manually configured for port security, it is guaranteed an entry in the secure host table (unless port security static aging was enabled).

A security violation occurs if an additional host is learned on the port. The action taken depends on which feature (802.1X or port security) detects the security violation:

– If 802.1X detects the violation, the action is to error-disable the port.

– If port security detects the violation, the action is to shut down or restrict the port (the action is configurable).

The following describes when port security and 802.1X security violations occur:

– In single-host mode, after the port is authorized, any MAC address received other than the client’s causes a 802.1X security violation.

– In single-host mode, if installation of an 802.1X client’s MAC address fails because port security has already reached its limit (due to a configured secure MAC addresses), a port security violation is triggered.

– In multiple-host mode, once the port is authorized, any additional MAC addresses that cannot be installed because the port security has reached its limit triggers a port security violation.

• When an 802.1X client logs off, the port transitions back to an unauthenticated state, and all dynamic entries in the secure host table are cleared, including the entry for the client. Normal authentication then ensues.
• If you administratively shut down the port, the port becomes unauthenticated, and all dynamic entries are removed from the secure host table.
• Only 802.1X can remove the client’s MAC address from the port security table. Note that in multiple-host mode, with the exception of the client’s MAC address, all MAC addresses that are learned by port security can be deleted using port security CLIs.
• Whenever port security ages out a 802.1X client’s MAC address, 802.1X attempts to reauthenticate the client. Only if the reauthentication succeeds is the client’s MAC address be retained in the port security table.
• All of the 802.1X client’s MAC addresses are tagged with (dot1x) when you display the port security table by using CLI.

Using 802.1X Authentication with ACL Assignments and Redirect URLs

You can download per-host policies such as ACLs and redirect URLs to the switch from the RADIUS server during 802.1X or MAB authentication of the host. ACL download is also supported with web authentication after a fallback from 802.1X or MAB.

When the 802.1X host mode of the port is either single-host, MDA, or multiple authentication, the downloaded ACLs (DACLs) are modified to use the authenticated hosts’ IP address as the source address. When the host mode is multiple-hosts, the source address is configured as ANY, and the downloaded ACLs or redirects apply to all devices on the port.

If no ACLs are provided during the authentication of a host, the static default ACL configured on the port is applied to the host. On a voice VLAN port, only the static default ACL of the port is applied to the phone.

This section includes these topics:

• Cisco Secure ACS and AV Pairs for URL-Redirect
• ACLs

For details on how to configure downloadable ACL and URL redirect, refer to the “Configuring 802.1X Authentication with ACL Assignments, and Redirect URLs” section.

Cisco Secure ACS and AV Pairs for URL-Redirect

When downloadable ACL is enabled, Cisco Secure ACS provides AAA services through RADIUS.

You can set these Attribute-Value (AV) pairs on the Cisco Secure ACS with RADIUS cisco-av-pair vendor-specific attributes (VSAs):

• CiscoSecure-Defined-ACL specifies the names of the DACLs on the Cisco Secure ACS. The switch receives the ACL name using the CiscoSecure-Defined-ACL AV pair in the format:

#ACL#-IP-name-number

name is the ACL name and number is the version number (similar to 3f783768).

The Auth-Manager code verifies whether the access control entries (ACEs) of the specified downloadable ACL were previously downloaded. If not, the Auth-Manager code sends an AAA request with the downloadable ACL name as the username so that the ACEs are downloaded. The downloadable ACL is then created as a named ACL on the switch. This ACL has ACEs with a source address of any and does not have an implicit deny statement at the end. When the downloadable ACL is applied to an interface after authentication completes, the source address changes from any to the host source IP address depending on the host mode of the interface. The ACEs are prepended to the downloadable ACL applied to the switch interface to which the endpoint device is connected. If traffic matches the CiscoSecure-Defined-ACL ACEs, the appropriate actions are taken.

• url-redirect and url-redirect-acl specify the local URL policy on the switch. The switches use these cisco-av-pair VSAs as follows:

– url-redirect = <HTTP or HTTPS URL>

– url-redirect-acl = switch ACL name or number

These AV pairs enable the switch to intercept an HTTP or HTTPS request from the endpoint device and forward the client web browser to the specified redirect address from which the latest antivirus files can be downloaded. The url-redirect AV pair on the Cisco Secure ACS contains the URL to which the web browser is redirected. The url-redirect-acl AV pair contains the name or number of an ACL that specifies the HTTP or HTTPS traffic to be redirected. Traffic that matches a permit entry in the redirect ACL is redirected.

Note The redirect or default ACL must be defined on the switch.

ACLs

If downloadable ACL is configured for a particular client on the authentication server, you must configure a default port ACL on a client-facing switch port.

If the default ACL is configured on the switch and the Cisco Secure ACS sends a host access policy to the switch, it applies the policy to traffic from the host connected to a switch port. If the policy does not apply, the switch applies the default ACL. If the Cisco Secure ACS sends the switch a downloadable ACL, this ACL takes precedence over the default ACL already configured on the switch port. However, if the switch receives a host access policy from the Cisco Secure ACS, but the default ACL is not configured, the authorization failure is declared.

For details on how to configure a downloadable policy, refer to the “Configuring a Downloadable Policy” section.

Using 802.1X with RADIUS-Provided Session Timeouts

You can specify whether a switch uses a locally configured or a RADIUS-provided reauthentication timeout. If the switch is configured to use the local timeout, it reauthenticates the host when the timer expires.

If the switch is configured to use the RADIUS-provided timeout, it scans the RADIUS Access-Accept message for the Session-Timeout and optional Termination-Action attributes. The switch uses the value of the Session-Timeout attribute to determine the duration of the session, and it uses the value of the Termination-Action attribute to determine the switch action when the session's timer expires.

If the Termination-Action attribute is present and its value is RADIUS-Request, the switch reauthenticates the host. If the Termination-Action attribute is not present, or its value is Default, the switch terminates the session.

Note The supplicant on the port detects that its session was terminated and attempts to initiate a new session. Unless the authentication server treats this new session differently, the client may see only a brief interruption in network connectivity as the switch sets up a new session.

If the switch is configured to use the RADIUS-supplied timeout, but the Access-Accept message does not include a Session-Timeout attribute, the switch never reauthenticates the supplicant. This behavior is consistent with Cisco's wireless access points.

For details on how to configure RADIUS-provided session timeouts, see the “Configuring RADIUS-Provided Session Timeouts” section.

Using 802.1X with Voice VLAN Ports

A voice VLAN port is a special access port associated with two VLAN identifiers:

• Voice VLAN ID (VVID) to carry voice traffic to and from the IP phone. The VVID is used to configure the IP phone connected to the port.
• Port VLAN ID (PVID) to carry the data traffic to and from the workstation connected to the switch using the IP phone. The PVID is the native VLAN of the port.

Each port that you configure for a voice VLAN is associated with a VVID and a PVID. This configuration allows voice traffic and data traffic to be separated onto different VLANs.

A voice VLAN port becomes active when a link exists whether the port is AUTHORIZED or UNAUTHORIZED. All traffic exiting the voice VLAN is obtained correctly and appears in the MAC address table. Cisco IP phones do not relay CDP messages from other devices. If several Cisco IP phones are connected in a series, the switch recognizes only the one directly connected to it. When 802.1X is enabled on a voice VLAN port, the switch drops packets from unrecognized Cisco IP phones more than one hop away.

When 802.1X is enabled on a port, you cannot configure a PVID that is equal to a VVID. For more information about voice VLANs, see Chapter38, “Configuring Voice Interfaces”

Observe the following feature interactions:

• 802.1X VLAN assignment cannot assign to the port the same VLAN as the voice VLAN; otherwise, the 802.1X authentication fails. The same holds true for dynamic VLAN assignment.
• 802.1X guest VLAN works with the 802.1X voice VLAN port feature. However, the guest VLAN cannot be the same as the voice VLAN.
• 802.1X port security works with the 802.1X voice VLAN port feature and is configured per-port. Two MAC addresses must be configured: one for the Cisco IP phone MAC address on the VVID and one for the PC MAC address on PVID.

However, you cannot use the 802.1X voice VLAN port feature with 802.1X port security’s sticky MAC address configuration and statically configured MAC address configuration.

• 802.1X accounting is unaffected by the 802.1X voice VLAN port feature.
• When 802.1X is configured on a port, you cannot connect multiple IP phones to a
  Catalyst 4500 series switch through a hub.
• Because voice VLANs cannot be configured as PVLAN host ports, and because only PVLANs can be assigned to PVLAN host ports, VLAN assignment cannot assign a PVLAN to a port with a voice VLAN configured.

For details on how to configure 802.1X with voice VLANs, see the “Configuring 802.1X with Voice VLAN” section.

Using Multiple Domain Authentication and Multiple Authentication

Multiple Domain Authentication (MDA) allows both a data device and a voice device, such as an IP phone (Cisco or third party non-Cisco), to authenticate on the same switch port, which is divided into a data domain and a voice domain.

Multi Auth allows multiple data devices and a voice device. When a voice VLAN is configured on a multiple- authentication port, the port can perform authentication in the voice domain as on an MDA port.

MDA does not enforce the order of device authentication. For best results, however, you should authenticate a voice device before you authenticate a data device on an MDA-enabled port.

When configuring MDA, consider the following guidelines.

Note The same guidelines also apply for Multiple Authentication when voice VLAN is configured.

• We recommend that you enable CoPP on an MDA-enabled port to protect against a DoS attack. Refer to Chapter45, “Configuring Control Plane Policing and Layer 2 Control Packet QoS”
• To configure a switch port for MDA or Multiple Authentication, see the “Configuring Multiple Domain Authentication and Multiple Authorization” section.
• You must configure the voice VLAN for the IP phone when the host mode is set to multidomain. For more information, see Chapter38, “Configuring Voice Interfaces”
• To authorize a voice device, the AAA server must be configured to send a Cisco Attribute-Value (AV) pair attribute with a value of device-traffic-class=voice. Without this value, the switch treats the voice device as a data device.
• The guest VLAN and restricted VLAN features only apply to the data devices on an MDA-enabled port. The switch treats a voice device that fails authorization as a data device.
• If more than one device attempts authorization on either the voice or the data domain of a port, it is error-disabled.
• Until a device is authorized, the port drops its traffic. Non-Cisco IP phones or voice devices are allowed into both the data and voice VLANs. The data VLAN allows the voice device to contact a DHCP server to obtain an IP address and acquire the voice VLAN information. After the voice device starts sending on the voice VLAN, its access to the data VLAN is blocked. A security violation may occur in MDA if the voice device continues to send traffic on the data VLAN.
• MDA can use MAC authentication bypass as a fallback mechanism to allow the switch port to connect to devices that do not support 802.1X authentication. it is especially useful for third party phones without 802.1X supplicant. For more information, see the “Using 802.1X with MAC Authentication Bypass” section.
• When a data or a voice device is detected on a port, its MAC address is blocked until authorization succeeds. If the authorization fails, the MAC address remains blocked for 5 minutes.
• If more than one device is detected on the data VLAN or more than one voice device is detected on the voice VLAN while a port is unauthorized, the port is error-disabled.
• When a port host mode is changed from single- or multihost to multidomain mode, an authorized data device remains authorized on the port. However, a Cisco IP phone that was allowed on the port in the voice VLAN is automatically removed and must be reauthenticated on that port.
• Active fallback mechanisms such as guest VLAN and restricted VLAN remain configured after a port changes from single- or multihost mode to multidomain mode.
• Switching a port host mode from multidomain to single- or multihost mode removes all authorized devices from the port.
• If a data domain is authorized first and placed in the guest VLAN, non-802.1X-capable voice devices need to tag their packets on the voice VLAN to trigger authentication.
• We do not recommend per-user ACLs with an MDA-enabled port. An authorized device with a per-user ACL policy might impact traffic on both the voice and data VLANs of the port. If used, only one device on the port should enforce per-user ACLs.

802.1X Supplicant and Authenticator Switches with Network Edge Access Topology

The Network Edge Access Topology (NEAT) feature extends identity to areas outside the wiring closet (such as conference rooms).

You can enable any authentication host mode on the authenticator switch interface that connects to a supplicant switch. Once the supplicant switch authenticates successfully, the port mode changes from access to trunk. To ensure that NEAT works on all host modes, use the dot1x supplicant force-multicast global configuration command on the supplicant switch. If the access VLAN is configured on the authenticator switch, it becomes the native VLAN for the trunk port after successful authentication.

Note MAB is not supported or recommended for use with NEAT. Only use 802.1X to authenticate the supplicant switch.

Note The Catalyst 4500 series switch only supports authenticator ports.

Deployment

NEAT is intended for deployment scenarios where a switch acting as 802.1X authenticator to end-hosts (PC or Cisco IP-phones) is placed in an unsecured location (outside wiring closet).

Because of this topology, the authenticator switch cannot always be trusted. For example, compact switches (8-port Catalyst 3560 and Catalyst 2960) are generally deployed outside the wiring closet. This enables hacker devices to swamp them to gain access to the network, compromising security. An edge switch must be able to authenticate itself against another switch, referred to as Network Edge Authentication Topology (NEAT).

Figure 41-8 illustrates a typical NEAT topology.

Figure 41-8 Typical NEAT Topology

NEAT facilitates the following functionality in such scenarios:

Host Authorization— Ensures that only traffic from authorized hosts (connecting to the switch with a supplicant) is allowed on the network. The switches use Client Information Signalling Protocol (CISP) to send the MAC addresses connecting the supplicant switch to the authenticator switch.

Auto enablement—Automatically enables trunk configuration on the authenticator switch, allowing user traffic from multiple VLANs arising from supplicant switches. At the ACS, you must configure the Cisco AV pair as device-traffic-class=switch. For details on how to do this, see the “Configuring an Authenticator and a Supplicant Switch with NEAT” section.

How 802.1X Fails on a Port

802.1X may fail on a port in three ways: timeout, explicit failure, and protocol timeout.

Timeout—A switch attempts 802.1X at link up but the attached endpoint is not 802.1X-capable. After the configured number of retries and timeouts, the switch attempts the next authentication method if one is configured (like MAB). If MAB fails, the switch deploys the Guest VLAN (also called the no-response VLAN), if configured. The Guest VLAN is configured with the authentication event no-response interface command.

Explicit Failure—A switch and the endpoint perform the entire 802.1X authentication sequence and the result is an explicit failure (usually indicated by an Access-Reject from the RADIUS server to the switch and an EAP-Failure sent from the switch to the endpoint). In this case, the switch attempts MAB (if "authentication event failure action next-method" is configured) or deploy the AuthFail VLAN (if "authentication event failure action authorize vlan" is configured).

Protocol Timeout—A switch and the endpoint start the 802.1X authentication process but do not complete it. For example, the endpoint may send an 802.1X EAPoL-Start message and then stop responding to the switch (perhaps, because the endpoint lacks a credential or because it is waiting for end user to enter some information). In this case, the switch knows that the connected device is EAPoL-capable, so it will not deploy the Guest VLAN after timing out. Instead, it restarts authentication after a timeout. The switch continues to label the port as EAPoL-capable until a physical link down event is detected. To force the switch to deploy the Guest VLAN in the case of a protocol timeout, configure dot1x guest-vlan supplicant globally. If the port is configured for hostmode multi-domain authentication, the switch behaves as if dot1x guest-vlan supplicant is configured.

Supported Topologies

The 802.1X port-based authentication supports two topologies:

• Point-to-point
• Wireless LAN

In a point-to-point configuration (see Figure 41-1), only one client can be connected to the 802.1X-enabled switch port when the multiple- host mode is not enabled (the default). The switch detects the client when the port link state changes to the up state. If a client leaves or is replaced with another client, the switch changes the port link state to down, and the port returns to the unauthorized state.

For 802.1X port-based authentication in a wireless LAN (Figure 41-9), you must configure the 802.1X port as a multiple-host port that is authorized as a wireless access point once the client is authenticated. (See the “Resetting the 802.1X Configuration to the Default Values” section.) When the port is authorized, all other hosts that are indirectly attached to the port are granted access to the network. If the port becomes unauthorized (reauthentication fails or an EAPOL-logoff message is received), the switch denies access to the network for all wireless access point-attached clients. In this topology, the wireless access point is responsible for authenticating clients attached to it, and the wireless access point acts as a client to the switch.

Figure 41-9 Wireless LAN Example

Default 802.1X Configuration

Table 41-1 shows the default 802.1X configuration.

802.1X Configuration Guidelines

Guidelines for configuring 802.1X authentication include the following:

• The 802.1X protocol is supported only on Layer 2 static access, PVLAN host ports, and Layer 3 routed ports. You cannot configure 802.1X for any other port modes.
• If you are planning to use VLAN assignment, be aware that the features use general AAA commands. For information on how to configure AAA, refer to the “Enabling 802.1X Authentication” section. Alternatively, you can refer to the Cisco IOS security documentation at this location:

http://www.cisco.com/en/US/products/ps6586/products_ios_technology_home.html

Enabling 802.1X Authentication

To enable 802.1X port-based authentication, you first must enable 802.1X globally on your switch, then enable AAA and specify the authentication method list. A method list describes the sequence and authentication methods that must be queried to authenticate a user.

The software uses the first method listed in the method list to authenticate users; if that method fails to respond, the software selects the next authentication method in the list. This process continues until there is successful communication with a listed authentication method or until all defined methods are exhausted. If authentication fails at any point in this cycle, the authentication process stops, and no other authentication methods are attempted.

Note To allow VLAN assignment, you must enable AAA authorization to configure the switch for all network-related service requests.

To configure 802.1X port-based authentication, perform this task:

Note Enabling Spanning Tree PortFast ensures that a port comes up immediately after authorization.

Note Whenever you configure any 802.1X parameter on a port, a dot1x authenticator is automatically created on the port. As a result, dot1x pae authenticator appears in the configuration, ensuring that dot1x authentication still works on legacy configurations without manual intervention.

This example shows how to enable 802.1X and AAA on Fast Ethernet port 2/1, and how to verify the configuration:

The following example illustrates when a port is authorized:

Configuring Switch-to-RADIUS-Server Communication

A RADIUS security server is identified by its host name or IP address, host name and specific UDP port number, or IP address and specific UDP port numbers. The combination of the IP address and UDP port number creates a unique identifier, which enables RADIUS requests to be sent to multiple UDP ports on a server at the same IP address. If two different host entries on the same RADIUS server are configured for the same service (for example, authentication), the second host entry configured acts as the failover backup to the first one. The RADIUS host entries are tried in the order they were configured.

To configure the RADIUS server parameters on the switch, perform this task:

This example shows how to specify the server with IP address 172.120.39.46 as the RADIUS server. The first command specifies port 1612 as the authorization port, sets the encryption key to rad123.

The second command dictates that key matches are performed on the RADIUS server:

You can globally configure the timeout, retransmission, and encryption key values for all RADIUS servers by using the radius-server host global configuration command. If you want to configure these options on a per-server basis, use the radius-server timeout , radius-server retransmit , and the radius-server key global configuration commands.

You also need to create a AAA client setting on the RADIUS server. These settings include the IP address of the switch and the key string to be shared by both the server and the switch.

Configuring Multiple Domain Authentication and Multiple Authorization

To configure Multiple Domain Authentication (MDA) and Multiple Authorization, perform this task.

This example shows how to enable 802.1X authentication and to allow multiple hosts:

This example shows how to enable MDA and to allow both a host and a 802.1X voice device (a Cisco or third-party phone with 802.1X supplicant) on the port:

This example shows how to enable MDA and to allow both a host and a non-802.1X voice device on the port:

This example shows how to verify the dot1x MDA settings on interface FastEthernet3/1:

This example shows how to enable MDA and to authentication of multiple hosts and a voice device on an IEEE 802.1x-authorized port:

Configuring 802.1X Authentication with ACL Assignments, and Redirect URLs

This section includes these topics:

• Downloadable ACL
• URL-Redirect
• Configuring a Downloadable Policy

Downloadable ACL

The downloadable ACL (DACL) feature allows you to download device specific authorization policies from the authentication server. These policies activate after authentication succeeds for the respective client and the client’s IP address was populated in the IP device tracking table. (Downloadable ACL is applied on the port, once the port is authenticated and the IP device tracking table has the host IP address entry).

The following sections describe the configuration that is necessary to complement the related authentication (802.1X or MAB) configuration. (No unique configuration is required on the switch. All of the configuration is on the ACS.) After authentication succeeds, enter the show ip access-list command to display the downloadable ACLs.

Configuring the Switch for Downloadable ACL

To configure the switch for downloadable ACL, follow these steps:

Step 1 Configure the IP device tracking table.

Step 2 Configure RADIUS VSA to forward authentication.

Step 3 Configure static ACL for the interface.

Interface Configuration Example

Debug Commands for DACL

The IP device tracking table contains the host IP address learned through ARP or DHCP.

The following command displays the constraints on the IP device tracking table:

The following show authentication sessions command displays the authentication sessions that contains the downloadable ACL obtained from ACS:

Note The show epm command will be deprecated, displaying a warning message when used. Use the show authentication sessions command instead.

The show authentication sessions interface interface-name policy displays session information in the form of Local Policies(features defined locally on the box), Server policies(features downloaded from radius) and Resultant Policies(the one with higher precedence when both local and server policies are present). By default, server policies have higher precedence than those defined locally.

The following command displays the contents of the downloadable ACL:

Cisco ACS Configuration for DACL

Note Only Cisco ACS supports DACL.

To ensure correct functioning of the ACS configuration required for DACL, follow these steps:

Step 1 Configure a downloadable IP ACL on the window that appears when you select
Radius Shared Profile > Downloadable IP ACL Content (Shared Profile Components).

Figure 41-10 Shared Profile Components

Step 2 Attach this downloadable ACL with the USER on the window that appears when you select User > DACLs (Downloadable ACLs).

Figure 41-11 Downloadable ACLs

URL-Redirect

To configure URL-direct, you need to configure it on the ACS, and on the switch.

Configuring ACS

To configure two Cisco-AV pairs, add the following statements under the user or group Cisco IOS/PIX 6x RADIUS attributes:

Note A default port ACL must be configured on the interface.

Configuring the Switch

To configure the switch for URL redirect, follow these steps:

Step 1 Configure the IP device tracking table.

Step 2 Configure RADIUS by using the send authentication command.

Step 3 Configure the URL redirect ACL (URLACL).

Step 4 Configure static ACL (PACL) for the interface.

Interface Configuration Example

Verify URL-redirect by using the following commands.

The show ip device tracking command displays the constraints on the IP device tracking table:

The show authentication sessions interface details command displays the URL-redirect-acl and URL-redirect URL information that downloads from the ACS:

For more information about AV pairs that are supported by Cisco IOS software, see the
ACS configuration and command reference documentation about the software releases running on the AAA clients.

Guideline for DACL and URL Redirect

For downloadable ACL or URL redirect, the ACL source must be ANY
(permit TCP ANY host 1.1.1.1 eq 80 or permit TCP ANY host 1.1.1.1 eq 443).

Configuring a Downloadable Policy

To configure downloadable policies, perform this task:

The following example illustrates how to configure a switch for downloadable policy:

Configuring 802.1X Authentication with Per-User ACL and Filter-ID ACL

This section includes the following topics:

• Per-User ACL and Filter-ID ACL
• Configuring a Per-User ACL and Filter-ID ACL

Per-User ACL and Filter-ID ACL

The Catalyst 4500 switch offers the Filter-ID/Per-user-acl enhancement, which allows ACL policy enforcement using a third-party AAA server.

The Filter-ID feature provides the following capabilities:

Filter-ID option allows an administrator to define the ACL name on the AAA server using IETF standard RADIUS attribute. The ACL itself must be preconfigured locally on the switch.

The Per-user-acl feature provides the following capabilities:

Per-user ACL allows an administrator to define the per-user ACL on the AAA server using Cisco RADIUS AV pairs. This action allows a third-party AAA server to interoperate by loading the Cisco RADIUS dictionary, which has Cisco Radius AV pairs configured as a VSA.

Note The RADIUS vendor-specific attributes (VSAs) allow vendors to support their own proprietary RADIUS attributes that are not included in standard RADIUS attributes.

Configuring the Switch

To configure the switch for per-user ACL and filter-ID ACL:

Step 1 Configure the IP device tracking table.

Step 2 Configure static ACL for the interface.

Interface Configuration Example

Per-User ACL Configuration in ACS

In the Group/User Setting page, scroll down to the Cisco IOS/PIX 6.x RADIUS Attributes section. Select the box next to [009\001 cisco-av-pair] and enter the elements of the per-user ACL. Per-user ACLS take this format:

Figure 41-12 shows how members of the group you are configuring are denied all access to the 10.100.60.0 subnet, are denied HTTP access to the server at 10.100.10.116, and are permitted everywhere else.

Figure 41-12 Define the ACEs for the Per-User ACL

Note Outbound ACLs (OUTACL) are not supported.

Filter-Id Configuration in ACS

In the Group/User Setting page, scroll down to the IETF RADIUS Attributes section. Select the box next to Filter-Id and enter the ACL to apply for members of this group (Configuring the Filter-ID Attribute).

The Filter-Id ia in this format:

Figure 41-13 Configuring the Filter-ID Attribute

Note Outbound ACLs (for example, 100.out) are not supported.

Debug Commands for Per-User ACL and Filter-ID ACL

The IP device tracking table contains the host IP address learned through ARP or DHCP. The following command displays the constraints on the IP device tracking table:

The following command shows authentication sessions that contains the Filter-Id 100:

The following command displays the contents of the per-user-acl (note that per-user-acl are shown above as the default port ACL configured on the interface, 151 is the default port ACL in the following example):

The following command displays the number of sessions:

The following command displays authentication sessions that contains the per-user-acl:

The following command displays the contents of the per-user-acl (note that per-user-acl are shown above as the default port ACL configured on the interface, 151 is the default port ACL in the preceding example below):

The following command shows that the Policy Enforced Module (EPM) session contains the Filter-Id 155 from ACS:

Note The 156 IP extended ACL is to be preconfigured on the switch, so that the policy enforcement can happen.

The following command shows authentication sessions that contains the Filter-Id TEST-ACL. TEST-ACL has been defined locally:

The following command displays the contents of the Filter-Id applied on the interface:

Guidelines for Per-User ACL and Filter-ID ACL

For per user ACL and Filter-ID ACL, the ACL source must be ANY
(permit TCP ANY host 1.1.1.1 eq 80 or permit TCP ANY host 1.1.1.1 eq 443).

Configuring a Per-User ACL and Filter-ID ACL

To configure per-user ACL and Filter-ID ACL, perform this task:

The following example illustrates how to configure a switch for downloadable policy:

Configuring RADIUS-Provided Session Timeouts

You can configure the Catalyst 4500 series switch to use a RADIUS-provided reauthentication timeout.

To configure RADIUS-provided timeouts, perform this task:

This example shows how to configure a switch to derive the reauthentication period from the server and to verify the configuration:

Configuring MAC Move

MAC move allows an authenticated host to move from one switch port to another.

Note You should remove port security before configuring MAC move.

To globally enable MAC move on the switch, perform this task:

This example shows how to globally enable MAC move on a switch:

The following syslog messages displays when MAC-move happens:

Configuring MAC Replace

MAC replace allows new users to connect to abandoned ports.

If a user disconnects but the switch has not received the EAPoL-Logoff, the session will remain up. For single or multiple- domain modes, no new hosts can connect to that port. If a new host tries to connect, a violation is triggered on the port. Where the violation action is configured as replace, the desired behavior is for the NAD (switch) to terminate the initial session and reset the authentication sequence based on the new MAC.

To enable MAC replace on a switch, perform this task:

This example shows how to globally enable MAC replace on a switch:

The following syslog messages displays when MAC-replace occurs:

Configuring Violation Action

You can configure 802.1X security violation behavior as either shutdown, restrict, or replace mode, based on the response to the violation.

To configure the violation action, performing the following task:

This example shows how to configure the violation mode shutdown on a switch:

A port is error-disabled when a security violation triggers on shutdown mode. The following syslog messages displays:

Configuring 802.1X with Guest VLANs

You can configure a guest VLAN for each 802.1X port on the Catalyst 4500 series switch to provide limited services to clients, such as downloading the 802.1X client. These clients might be upgrading their system for 802.1X authentication, and some hosts, such as Windows 98 systems, might not be 802.1X-capable.

When you enable a guest VLAN on an 802.1X port, the Catalyst 4500 series switch assigns clients to a guest VLAN, provided one of the following apply:

• The authentication server does not receive a response to its EAPOL request or identity frame.
• The EAPOL packets are not sent by the client.

The Catalyst 4500 series switch maintains the EAPOL packet history. If another EAPOL packet is detected on the interface during the lifetime of the link, network access is denied. The EAPOL history is reset upon loss of the link.

Any number of 802.1X-incapable clients are allowed access when the switch port is moved to the guest VLAN. If an 802.1X-capable client joins the same port on which the guest VLAN is configured, the port is put into the unauthorized state in the user-configured access VLAN, and authentication is restarted.

Guest VLANs are supported on 802.1X ports in single-host or multiple-hosts mode.

Note When a port is put into a guest VLAN, it is automatically placed into multihost mode, and an unlimited number of hosts can connect using the port. Changing the multihost configuration does not effect a port in a guest VLAN.

Note Except for an RSPAN VLAN or a voice VLAN, you can configure any active VLAN as an 802.1X guest VLAN.

To configure 802.1X with guest VLAN on a port, perform this task:

This example shows how to enable regular VLAN 50 on Fast Ethernet 4/3 as a guest VLAN on a static access port:

This example shows how to enable a secondary PVLAN 100 as a guest VLAN on a PVLAN host port:

To allow supplicants into a guest VLAN on a switch, perform this task:

This example shows how to enable the guest VLAN feature and to specify VLAN 5 as a guest VLAN:

Configuring 802.1X with MAC Authentication Bypass

To enable MAC Authentication Bypass (MAB), perform this task:

Note Removing a 802.1X MAB configuration from a port does not impact the authorized or authenticated state of the port. If the port is in an unauthenticated state, it remains in that state. If the port is in an authenticated state because of MAB, the switch reverts to the 802.1X Authenticator. If the port was already authorized with a MAC address and the MAB configuration was removed, the port remains in an authorized state until reauthentication occurs. At that time, if an 802.1X supplicant is detected on the wire, the MAC address is removed.

This example shows how to enable MAB on Gigabit Ethernet interface 3/3 and to verify the configuration:

Configuring 802.1X with Inaccessible Authentication Bypass

To configure a port as a critical port and to enable the Inaccessible Authentication Bypass feature, perform this task:

The following example shows a full configuration of 802.1X with Inaccessible Authentication Bypass, including required AAA and RADIUS configuration as specified in the “Enabling 802.1X Authentication” section and “Configuring Switch-to-RADIUS-Server Communication” section.

The RADIUS server configured is at IP address 10.1.2.3, using port 1645 for authentication and 1646 for accounting. The RADIUS secret key is mykey. The username used for the test server probes is randomizes. The test probes for both living and dead servers are generated once per minute. The interface FastEthernet 3/1 is configured to critically authenticate into VLAN 17 when AAA becomes unresponsive, and to reinitialize automatically when AAA becomes available again.

Configuring 802.1X with Unidirectional Controlled Port

To configure unidirectional controlled port, perform this task:

Note Unidirectional controlled port only works when Spanning Tree PortFast is enabled on the port. Unidirectional controlled port and Spanning Tree PortFast should be configured on a switch port that connects to a host. If two such ports are connected together with an Ethernet cable, high CPU utilization may result because host learning will be flapping between the two ports.

This example shows how to enable unidirectional port control:

Configuring 802.1X with VLAN User Distribution

You will need to configure the switch and ACS to configure 802.1X with VLAN user distribution.

Configuring the Switch

To configure the switch, follow these steps:

Step 1 Create a VLAN group on the switch.

Enter the following commands to create a VLAN group and assign some VLANs to the VLAN group. The following example creates the VLAN group eng-group and maps VLANs 20 to 24 to that group:

Note Ensure that the VLANs you specify as part of the VLAN group are enabled on the switch. Only specified VLANs are considered for assignment.

Step 2 Configure the individual ports for multidomain, single-host or multiple- host.

For details, refer to the “Enabling 802.1X Authentication” section.

show commands

Use the following show commands to display the member VLANs in a VLAN group:

The following examples show outputs of the show vlan group command:

In this example, VLANs 3,4, 5, 7, and 9 are members of the VLAN group my group.

ACS Configuration

After configuring the switch, you must provide the VLAN group name in the ACS configuration.

By default, ACS sends only one VLAN name or group per user. However, you can configure ACS to send more than one tag per attribute. To do this, you must modify the configuration in ACS for user or group. (See the example shown in Figure 41-14.)

Figure 41-14 VLAN User Distribution on ACS: Interface Configuration to Modify Tags per Attribute

After you add the number of tags required per attribute, the user or group set up presents multiple fields to be filled with values from the RADIUS server (Figure 41-15).

Figure 41-15 VLAN User Distribution on ACS: Multiple VLAN Numbers Configured per User

After you complete these two tasks and receive authorization, ACS sends the configured VLAN group to the switch. The switch is alerted to the list of VLANs configured under the VLAN group, and the least loaded valid VLAN in the group is assigned to the port.

Configuring 802.1X with Authentication Failed

By configuring authentication-failed VLAN alignment on any Layer 2 port on the Catalyst 4500 series switch, you can provide limited network services to clients that fail the authentication process.

Note You can use authentication-failed VLAN assignment with other security features, such as Dynamic ARP Inspection (DAI), Dynamic Host Configuration Protocol (DHCP) snooping, and IP Source Guard. Each of these features can be enabled and disabled independently on the authentication-failed VLAN.

To configure 802.1X with authentication-failed VLAN assignment, perform this task:

This example shows how to enable a regular VLAN 40 on Fast Ethernet 4/3 as a authentication-failed VLAN on a static access port:

Configuring 802.1X with Voice VLAN

Note You must configure 802.1X and voice VLAN simultaneously.

Note You cannot configure an authentication-failed VLAN and a voice VLAN on the same port. When you try to configure these two features on the same port, a syslog message appears.

To enable 802.1X with voice VLAN, perform this task:

This example shows how to enable 802.1X with voice VLAN feature on Fast Ethernet interface 5/9:

Configuring 802.1X with VLAN Assignment

For enabling dynamic VLAN assignment, no additional configuration is required in the switch. For information on configuring Multiple- authentication (MDA), refer to the “Configuring Multiple Domain Authentication and Multiple Authorization” section. To enable VLAN assignment, you must configure the Cisco ACS server.

To enable 802.1X with VLAN assignment, perform this task:

The following example shows how to configure MDA on an interface and 802.1X as the authentication mechanism:

Note You must configure VLAN assignment in the ACS server. No configuration changes are required on the switch.

Cisco ACS Configuration for VLAN Assignment

The procedure for enabling MDA with voice VLAN assignment is the same as that for activating MDA except for one step: Configure a VLAN for dynamic VLAN assignment after selecting
User > IETF RADIUS Attributes (Figure 41-16). This step ensures correct functioning of the ACS configuration required for dynamic VLAN assignment.

Figure 41-16 User Set Up

Note The procedure is the same for voice devices except that the AAA server must be configured to send a Cisco Attribute-Value (AV) pair attribute with a value of device-traffic-class=voice.

Enabling Fallback Authentication

On a port in multiauthentication mode, either or both of MAB and web-based authentication can be configured as fallback authentication methods for non-802.1X hosts (those that do not respond to EAPOL). You can configure the order and priority of the authentication methods.

For detailed configuration information for MAB, see the “Configuring 802.1X with MAC Authentication Bypass” section.

For detailed configuration information for web-based authentication, see Chapter43, “Configuring Web-Based Authentication”

Note When web-based authentication and other authentication methods are configured on an MDA or multiauthentication port, downloadable ACL policies must be configured for all devices attached to that port.

To enable fallback authentication, perform this task:

This example shows how to enable 802.1X fallback to MAB, and then to enable web-based authentication, on an 802.1X-enabled port:

Note If web-based authentication also fails then reauthentication starts. 802.1X, MAB, and web-based authentication are attempted in that order.

To determine if a host was authenticated using 802.1X when fallback authentication is configured on the port, enter the following commands:

To determine if a host was authenticated using MAB when fallback authentication is configured on the port, enter the following commands:

To determine if a host was authenticated using web authentication when fallback authentication is configured on the port, enter the following commands:

Enabling Periodic Reauthentication

You can enable periodic 802.1X client reauthentication and specify how often it occurs. If you do not specify a time value before enabling reauthentication, the interval between reauthentication attempts is 3600 seconds.

Automatic 802.1X client reauthentication is a per-interface setting and can be set for clients connected to individual ports. To manually reauthenticate the client connected to a specific port, see the “Changing the Quiet Period” section.

To enable periodic reauthentication of the client and to configure the number of seconds between reauthentication attempts, perform this task: