Catalyst 4500 Series Switch Software Configuration Guide, 15.0(2)SG Configuration Guide

Overview

An ACL is a collection of sequential permit and deny conditions that applies to packets. When a packet is received on an interface, the switch compares the fields in the packet against any applied ACLs to verify that the packet has the permissions required to be forwarded, based on the conditions specified in the access lists. It tests the packets against the conditions in an access list one-by-one. The first match determines whether the switch accepts or rejects the packets. Because the switch stops testing conditions after the first match, the order of conditions in the list is critical. If no conditions match, the switch drops the packet. If no restrictions exist, the switch forwards the packet; otherwise, the switch drops the packet.

Switches traditionally operate at Layer 2, switching traffic within a VLAN. Routers route traffic between VLANs at Layer 3. The Catalyst 4500 series switch can accelerate packet routing between VLANs by using Layer 3 switching. The Layer 3 switch bridges the packet, and then routes the packet internally without going to an external router. The packet is then bridged again and sent to its destination. During this process, the switch can control all packets, including packets bridged within a VLAN.

You configure access lists on a router or switch to filter traffic and provide basic security for your network. If you do not configure ACLs, all packets passing using the switch could be allowed on all parts of the network. You can use ACLs to control which hosts can access different parts of a network or to decide which types of traffic are forwarded or blocked at router interfaces. For example, you can allow e-mail traffic to be forwarded but not Telnet traffic. ACLs can be configured to block inbound traffic, outbound traffic, or both. However, on Layer 2 interfaces, you can apply ACLs only in the inbound direction.

An ACL contains an ordered list of access control entries (ACEs). Each ACE specifies permit or deny and a set of conditions the packet must satisfy in order to match the ACE. The meaning of permit or deny depends on the context in which the ACL is used.

The Catalyst 4500 series switch supports three types of ACLs:

• IP ACLs, which filter IP traffic, including TCP, the User Datagram Protocol (UDP), Internet Group Management Protocol (IGMP), and Internet Control Message Protocol (ICMP).
• IPv6 ACLs (applies only to Supervisor Engine 6E).
• MAC ACLs which match based on Ethernet addresses and Ether Type.

Supported Features That Use ACLs

The switch supports two applications of ACLs to filter traffic:

• Router ACLs are applied to Layer 3 interfaces. They control the access of routed traffic between VLANs. All Catalyst 4500 series switches can create router ACLs, but you must have a Cisco IOS software image on your switch to apply an ACL to a Layer 3 interface and filter packets routed between VLANs.
• Port ACLs perform access control on traffic entering a Layer 2 interface. If insufficient hardware CAM entries exist, the output port ACL is not applied to the port and a warning message is given to user. (This restriction applies to all access group modes for output port ACLs.) When sufficient CAM entries exist, the output port ACL may be reapplied.

If there is any output port ACL configured on a Layer 2 port, then no VACL or router ACL can be configured on the VLANs that the Layer 2 port belongs to. Also, the reverse is true: port ACLs and VLAN-based ACLs (VACLs and router ACLs) are mutually exclusive on a Layer 2 port. This restriction applies to all access group modes. On the input direction, port ACLs, VLAN-based ACLs, and router ACLs can co-exist.

You can apply one IPv4 access list, one IPv6 access list and one MAC access list for a Layer 2 interface.

• VLAN ACLs or VLAN maps control the access of all packets (bridged and routed). You can use VLAN maps to filter traffic between devices in the same VLAN. You do not need the enhanced image to create or apply VLAN maps. VLAN maps are configured to control access based on Layer 3 addresses for IP. MAC addresses using Ethernet ACEs control the access of unsupported protocols. After you apply a VLAN map to a VLAN, all packets (routed or bridged) entering the VLAN are checked against that map. Packets can either enter the VLAN through a switch port or through a routed port after being routed.

You can use both router ACLs and VLAN maps on the same switch.

Router ACLs

You can apply one access list of each supported type to an interface.

Note Catalyst 4500 series switches running Cisco IOS Release 12.2(40)SG do not support IPv6 port ACLs (PACLs).

Multiple features can use one ACL for a given interface, and one feature can use multiple ACLs. When a single router ACL is used by multiple features, it is examined multiple times. The access list type determines the input to the matching operation:

• Standard IP access lists use source addresses for matching operations.
• Extended IP access lists use source and destination addresses and optional protocol type information for matching operations.

The switch examines ACLs associated with features configured on a given interface and a direction. As packets enter the switch on an interface, ACLs associated with all inbound features configured on that interface are examined. After packets are routed and before they are forwarded to the next hop, all ACLs associated with outbound features configured on the egress interface are examined.

ACLs permit or deny packet forwarding based on how the packet matches the entries in the ACL. For example, you can use access lists to allow one host to access a part of a network, but prevent another host from accessing the same part. In Figure 1-1, ACLs applied at the router input allow Host A to access the Human Resources network, but prevent Host B from accessing the same network.

Figure 1-1 Using ACLs to Control Traffic to a Network

Port ACLs

You can also apply ACLs to Layer 2 interfaces on a switch. Port ACLs are supported on physical interfaces and EtherChannel interfaces.

The following access lists are supported on Layer 2 interfaces:

• Standard IP access lists using source addresses
• Extended IP access lists using source and destination addresses and optional protocol type information
• IPv6 access lists using source and destination addresses and optional protocol type information
• MAC extended access lists using source and destination MAC addresses and optional protocol type information

As with router ACLs, the switch examines ACLs associated with features configured on a given interface and permits or denies packet forwarding based on how the packet matches the entries in the ACL. In the example in Figure 1-1, if all workstations were in the same VLAN, ACLs applied at the Layer 2 input would allow Host A to access the Human Resources network, but prevent Host B from accessing the same network.

When you apply a port ACL to a trunk port, the ACL filters traffic on all VLANs present on the trunk port. When you apply a port ACL to a port with voice VLAN, the ACL filters traffic on both data and voice VLANs.

With port ACLs, you can filter IP traffic by using IP access lists and non-IP traffic by using MAC addresses. You can filter both IP and non-IP traffic on the same Layer 2 interface by applying both an IP access list and a MAC access list to the interface.

With port ACLs, you can filter IPv4 traffic with IPv4 access lists, IPv6 traffic with IPv6 access lists, and non-IP traffic with MAC access lists. You can filter multiple types of traffic simultaneously by applying ACLs of the appropriate type to the Layer 2 interface simultaneously.

Note You cannot simultaneously apply more than one access list of a given type to a Layer 2 interface. If an IPv4, IPv6, or MAC access list is already configured on a Layer 2 interface, and you apply a new IPv4, IPv6 or MAC access list to the interface, the new ACL replaces the previously configured ACL of the same type.

Dynamic ACLs

Various security features, such as 802.1X, NAC and Web Authentication, are capable of downloading ACLs from a central server and applying them to interfaces. Prior to Cisco IOS Release 12.2(54)SG, these features required the explicit configuration of a standard port ACL

Starting with Cisco IOS Release 12.2(54)SG, a port ACL does not require configuration. For more details refer to the “Removing the Requirement for a Port ACL”.

VLAN Maps

VLAN maps can control the access of all traffic in a VLAN. You can apply VLAN maps on the switch to all packet s that are routed into or out of a VLAN or are bridged within a VLAN. VLAN maps are not defined by direction (input or output).

You can configure VLAN maps to match Layer 3 addresses for IP traffic. Access of all non-IP protocols is controlled with a MAC address and an Ethertype using MAC ACLs in VLAN maps. (IP traffic is not controlled by MAC ACLs in VLAN maps.) You can enforce VLAN maps only on packets heading to the switch; you cannot enforce VLAN maps on traffic between hosts on a hub or on another switch connected to this switch.

With VLAN maps, forwarding packets is permitted or denied, based on the action specified in the map. Figure 1-2 illustrates how a VLAN map is applied to deny a specific type of traffic from Host A in VLAN 10 from being forwarded.

Figure 1-2 Using VLAN Maps to Control Traffic

TCAM Programming Algorithms

Note The TCAM programming algorithm is neither available on Supervisor Engine 6-E nor 6L-E.

Beginning with Cisco IOS Release 12.2(25)EWA, two TCAM programming algorithms are supported on Catalyst 4500 and 4900 series switches: packed and scattered. The packed mode algorithm programs the entries in the same 8-entry TCAM block if the entries’ masks match. If the current entry’ mask differs from previous entries, the switch software programs the entry in a new 8-entry block. If the mask does not change, or if the mask changes every eight entries across ACLs from the beginning to the end of the configuration, the TCAM may be fully utilized in packed mode for Supervisor Engines II-Plus-TS, IV, V, and the Catalyst 4948 series switch.

In scattered mode, the entries from a single ACL are distributed across different 8-entry blocks until the ACL is fully programmed. If successive ACLs have the same mask pattern as the first ACL, the TCAM on Supervisor Engines II-Plus-TS, IV, V, and the Catalyst 4948 series switch may be fully utilized.

Scattered mode is recommended for IP Source Guard configurations on Supervisor Engines II-Plus-TS, IV, V, and the Catalyst 4948 switch. it is because the mask pattern for per-VLAN ACLs is the same for all ports configured for IP Source Guard: permit ARP packets, permit Layer 2 traffic if port security is not configured, permit IP traffic from a particular source IP address with a 32-bit mask, deny unknown, and permit all.

Note The TCAM programming algorithm can be configured on Supervisor Engine V-10GE and the Catalyst 4948-10GE switch running Cisco IOS Release 12.2(25)EWA or its subsequent maintenance releases. On Supervisor Engine V-10GE and the Catalyst 4948-10GE switch, however, because ACL masks are not shared among ACEs, TCAM utilization is the same for all configured programming algorithms.

Note The TCAM programming algorithm cannot be configured on Supervisor Engines II-Plus-10GE or V-10GE or the Catalyst 4948-10GE switch running Cisco IOS Release 12.2(25)SG and later.

Note The TCAM utilization should not change after you successively configure the same TCAM programming algorithm. For example, configuring access list hardware entries packed twice should not affect TCAM utilization. However, TCAM utilization may change if one or more commands are entered between successive configurations of the same TCAM programming algorithm.

This is a summary of scenarios that cause TCAM utilization to change:

• Adding or deleting ACLs or ACEs in the running-config
• Copying or recopying the ACL configuration from bootflash, a TFTP server, or CompactFlash memory to the running-config
• Changing the TCAM programming algorithm
• Saving the running-config to NVRAM and reloading the switch
• Resizing the feature ACL or QoS regions of the TCAM by using the
  access list hardware region [feature | qos] [input | output] balance percent command on
  Cisco IOS Release 12.2(31)SGA and later.
• Upgrading from images based on Cisco IOS Release 12.2(25)EWA to images based on Cisco IOS Release 12.2(31)SGA

Two types of hardware resources are consumed when you program ACLs: entries and masks. If either one of these resources is exhausted, no additional ACLs can be programmed into hardware.

If you run out of resources, refer to the following sections:

• Changing the Programming Algorithm
• Resizing the TCAM Regions
• Selecting Mode of Capturing Control Packets

Changing the Programming Algorithm

If the masks on a system are exhausted, but entries are available, changing the programming scheme from packed to scattered might free up masks, allowing additional ACLs to be programmed into hardware.

Note Changing the ACL programming algorithm or resizing the TCAM regions causes all ACLs to be temporarily unloaded from the hardware and then reloaded in accordance with the new TCAM parameters. ACLs are inoperative until the reloading process is complete.

The goal is to use TCAM resources more efficiently by minimizing the number of masks per ACL entry.

The following table shows how to change the algorithm that controls TCAM utilization:

Note access-list hardware entries packed is the default and does not appear in the output of the show running-config command. If you configure scattered mode, access-list hardware entries scattered appears in the output. If you configure packed mode, nothing appears in the output.

Note The default TCAM programming algorithm is packed.

The following output was collected from a switch running in packed mode. Observe that 89 percent of the masks are required to program only 49 percent of the ACL entries.

The following output was collected after the algorithm was switched to scattered. Observe that the number of masks required to program 49 percent of the entries has decreased to 49 percent.

Note When you enable DHCP snooping and IP Source Guard on all ports on a chassis, you must use the scattered keyword.

Resizing the TCAM Regions

The TCAM is divided into regions, each of which holds different kinds of entries. Four main TCAM types exist: Input ACL, Output ACL, Input QoS, and Output QoS. Each is divided into a PortAndVlan region and a PortOrVlan region. By default the PortAndVlan region and PortOrVlan region are of equal size.

The following table lists the entries and mask counts for each of the supported supervisor engines. The entry and mask counts for the various supervisor engines are for each TCAM type; for example, the input feature TCAM has 16,000 entries, and the output feature TCAM has 16,000 entries.

Note Because of the ratio of entries to masks on Supervisor Engines II-Plus-TS, IV, V, and the
Catalyst 4948 switch is 8:1, TCAM space for masks may be exhausted before space for entries is exhausted.

Note One region in a TCAM type might be filled while the other region still has free space. When this occurs, the regions can be resized to move the free entries from the region where they are not needed to the region where they are needed. Regions are resized using the
access-list hardware region {feature | qos} {input | output} balance command. Each TCAM type has its own independent region balance.

Note Higher balance values indicate more entries to the PortAndVlan region and less entries to the PortOrVlan region. Lower balance values indicate less entries to the PortAndVlan region and more entries to the PortOrVlan region. A balance of 50 indicates equal allocation to both PortAndVlan and PortOrVlan regions.

Note You can only shift entries between the PortAndVlan region and the PortOrVlan region for a specific TCAM type (for example, from the Input ACL TCAM PortOrVlan region to the Input ACL TCAM PortAndVlan region). You cannot shift entries between TCAM types.

To determine whether region resizing would be beneficial, enter the following command:

The above output indicates that the Input ACL PortAndVlan region is out of masks, but there is free space in the Input ACL PortOrVlan region that could be repurposed.

The following example shows how to change the region balance of the Input ACL TCAM so that 75 percent of the entries are allocated to the PortAndVlan region and only 25 percent to the PortOrVlan region.

After adjusting the region balance, the PortAndVlan region has more resources allocated to it, and the PortOrVlan region has fewer resources.

Note Both the no access-list hardware region {feature | qos} {input | output} balance command and a balance of 50 force the configuration to the default values. A similar configuration can also be performed for QoS.

Troubleshooting High CPU Due to ACLs

Packets that match entries in fully programmed ACLs are processed in hardware. However, large ACL and IPSG configurations may exhaust TCAM masks on Supervisor Engines II-Plus-TS, IV, V, and the Catalyst 4948 switch before the ACLs are fully programmed.

Packets that match entries in partially programmed ACLs are processed in software using the CPU. This may cause high CPU utilization and packets to be dropped. To determine whether packets are being dropped due to high CPU utilization, reference the following:

http://www.cisco.com/en/US/products/hw/switches/ps663/products_tech_note09186a00804cef15.shtml

If the ACL and/or IPSG configuration is partially programmed in hardware, upgrading to
Cisco IOS Release 12.2(31)SGA or later and resizing the TCAM regions may enable the ACLs to be fully programmed.

Note Removal of obsolete TCAM entries can take several CPU process review cycles to complete. This process may cause some packets to be switched in software if the TCAM entry or mask utilization is at or near 100 percent.

Selecting Mode of Capturing Control Packets

In some deployments, you might want to bridge control packets in hardware rather than globally capture and forward them in software (at the expense of the CPU). The per-VLAN capture mode feature allows a Catalyst 4500 series switch to capture control packets only on selected VLANs and bridge traffic in hardware on all other VLANs.

With Supervisor Engine 6-E and 6L-E, when you use per-VLAN capture mode on your switch, it partially disables the global TCAM capture entries internally and attaches feature-specific capture ACLs on those VLANs that are enabled for snooping features. (All IP capture entries, and other non-IP entries are still captured through global TCAM.)

For all other supervisor engines, when you use per-VLAN capture mode on your switch, it partially disables the global TCAM capture entries internally and attaches feature-specific capture ACLs on those VLANs that are enabled for snooping or routing features. (All IP capture entries, CGMP, and other non-IP entries are still captured through global TCAM.)

Because this feature controls specific control packets, they are captured only on the VLANs on which the internal ACLs are installed. On all other VLANs, the control traffic is bridged in hardware rather than forwarded to CPU.

The per-VLAN capture mode allows you to apply user-defined ACLs and QoS policers (in hardware) on control packets. You can also subject the aggregate control traffic ingressing the CPU to control plane policing.

When you use per-VLAN capture mode, the following four protocol groups are selectable per-VLAN. The breakdown of protocols intercepted by each group is as follows:

• IGMP Snooping—Cgmp, Ospf, Igmp, RipV2, Pim, 224.0.0.1, 224.0.0.2, 224.0.0.*
• DHCP Snooping—Client to Server, Server to Client, Server to Server

The following bullets only apply to supervisor engines other than 6-E and 6L-E.

• Unicast Routing—Ospf, Rip v2, 224.0.0.1, 224.0.0.2, 224.0.0.*
• Multicast Routing—Ospf, Rip v2, Igmp, Pim, 224.0.0.1, 224.0.0.2, 224.0.0.*

Because some of the groups have multiple overlapping ACEs (for example, 224.0.0.* is present in all the groups except for DHCP Snooping), turning on a certain group will also trigger the interception of some protocols from other groups.

Following are the programming triggers for the four protocol groups per-VLAN:

• IGMP Snooping should be enabled globally on a given VLAN.
• DHCP Snooping should be enabled globally on a given VLAN.

The following bullets only apply to supervisor engines other than 6-E and 6L-E.

• Unicast Routing should be enabled and SVI (or a Layer 3 physical) interface should be up and configured with an IP protocol address. it is because interfaces immediately become part of the routing process once the SVI interface comes up and the protocol family address is configured.
• Multicast Routing should be enabled and one of the multicast routing protocols should be configured on the interface (IGMP, PIMv1, PIMv2, MBGP, MOSPF, DVMRP, and IGMP snooping).

Guidelines and Restrictions

Note Before configuring per-VLAN capture mode, you should examine your configuration to ensure that only the necessary features are enabled on the desired VLANs.

The following guidelines and restrictions apply to per-VLAN capture mode:

• Starting with Cisco IOS Release 15.0(2)SG, for Supervisor Engine 6-E and Supervisor Engine 6L-E, globally reserved static ACL entries in the TCAM region for Layer 3 control packets are removed. The per-VLAN CTI command is not needed and does not apply for Layer 3 control packets because these packets are captured in per-VLAN fashion by default.

The following still function:

– Global static capture and CTI commands for IGMP or PIM packets (both use MAC addresses 224.0.0.1 and 224.0.0.2)

– Global and per-VLAN CTI for DHCP packets

With Cisco IOS Release 15.0(2)SG, per-VLAN capture of Layer 3 control packets is driven by SVI configuration. Except for IGMP, PIM, or DHCP, no special configuration is required.

• Enabling per-VLAN capture mode consumes additional entries in the ACL/feature TCAM.

The number of available TCAM entries depends on the type of supervisor engine. The entry and mask count further limits the utilization of the ACL/feature TCAM.

• Certain configurations can exhaust TCAM resource earlier in per-VLAN capture mode than in global capture mode (such as, when IP Source Guard is enabled on several interfaces or on a user-configured PACL).

You can resize TCAM regions to make more entries available to the PortAndVlan or PortOrVlan region based on the configuration. This allows more entries to be programmed in hardware before reaching the limit. When TCAM resources are exhausted, the packets are forwarded in software.

• In per-VLAN capture mode, you can configure ACLs to permit or deny control traffic on a VLAN or port.

Because security ACLs are terminated by an implicit deny, you must ensure that the ACLs are configured to permit the control packets necessary for the feature (protocol) to operate. However, this rule does not differ from the default behavior.

Selecting Control Packet Capture

To select the mode of capturing control packets, perform this task:

This example shows how to configure a Catalyst 4500 series switch to capture control packets only on VLANs where features are enabled:

This example shows how to configure a Catalyst 4500 series switch to capture control packets globally across all VLANs (using static ACL, the default mode):

When the capture mode changes from global to VLAN, the static CAM entries are invalidated. This creates a window during which control packets may pass through a Catalyst 4500 series switch without being intercepted to the CPU. This temporary situation is restored when the new per-VLAN capture entries are programmed in the hardware.

When you configure per-VLAN capture mode, you should examine the show commands for individual features to verify the appropriate behavior. In per-VLAN capture mode, the invalidated static CAM entries will appear as inactive in the output of the show platform hardware acl input entries static command. For example, the hit count for inactive entries will remain frozen because those entries are invalidated and applied per-VLAN where the feature is enabled. The following table lists the CamIndex entry types and the Cam regions.

Restrictions for Layer 4 Operations

You can specify these operator types, each of which uses one Layer 4 operation in the hardware:

• gt (greater than)
• lt (less than)
• neq (not equal)
• range (inclusive range)

For Supervisor Engine II+ through V-10GE, you should not specify more than six different operations on the same ACL. If you exceed this number, each new operation might cause the affected ACE (access control entry) to be translated into multiple ACEs in hardware. If you exceed this number, the affected ACE might be processed in software.

For Supervisor Engine 6-E and 6L-E, the limits on the number of Layer 4 operations differ for each type of ACL, and can also vary based on other factors: whether an ACL is applied to incoming or outgoing traffic, whether the ACL is a security ACL or is used as a match condition for a QoS policy, and whether IPv6 ACLs are being programmed using the compressed flow label format.

Note The IPv6 compressed flow label format uses the Layer 2 Address Table to compress a portion of the IPv6 source address of each ACE in the ACL. The extra space freed in the flow label can then be used to support more Layer 4 operations. For this compression to be used, the IPv6 ACL cannot contain any ACEs that mask in only a portion of the bottom 48 bits of the source IPv6 address.

Generally, you will receive at most the following number of Layer 4 operations on the same ACL:

Note Where up to 16 operations are supported, the seventeenth will trigger an expansion.

If you exceed the number of available Layer 4 operations, each new operation might cause the affected ACE to be translated into multiple ACEs in the hardware. If this translation fails, packets are sent to the CPU for software processing.

Configuration Guidelines for Layer 4 Operations

When using Layer 4 operators, consider these guidelines:

• Layer 4 operations are considered different if the operator or operand differ. For example, the following ACL contains three different Layer 4 operations because gt 10 and gt 11 are considered two different Layer 4 operations:

Note The eq operator can be used an unlimited number of times because eq does not use a Layer 4 operation in hardware.

• Layer 4 operations are considered different if the same operator/operand couple applies once to a source port and once to a destination port, as in the following example:

A more detailed example follows:

Access lists 101 and 102 use the following Layer 4 operations:

• Access list 101 Layer 4 operations: 5

– gt 10 permit and gt 10 deny both use the same operation because they are identical and both operate on the destination port.

• Access list 102 Layer 4 operations: 4
• Total Layer 4 operations: 8 (due to sharing between the two access lists)

– neq6 permit is shared between the two ACLs because they are identical and both operate on the same destination port.

• A description of the Layer 4 operations usage is as follows:

– Layer 4 operation 1 stores gt 10 permit and gt 10 deny from ACL 101

– Layer 4 operation 2 stores lt 9 deny from ACL 101

– Layer 4 operation 3 stores gt 11 deny from ACL 101

– Layer 4 operation 4 stores neg 6 permit from ACL 101 and 102

– Layer 4 operation 5 stores neg 6 deny from ACL 101

– Layer 4 operation 6 stores gt 20 deny from ACL 102

– Layer 4 operation 7 stores lt 9 deny from ACL 102

– Layer 4 operation 8 stores range 11 13 deny from ACL 102

How ACL Processing Impacts CPU

ACL processing can impact the CPU in two ways:

• For some packets, when the hardware runs out of resources, the software must perform the ACL matches:

– For Supervisor Engine 2-Plus to V-10GE, the TCP flag combinations rst ack and syn fin rst are processed in hardware. rst ack is equivalent to the keyword established. Other TCP flag combinations are supported in software.

– For Supervisor Engine 6-E and 6L-E, the TCP flag combinations rst ack, syn fin rst, urg and psh are processed in hardware. rst ack is equivalent to the keyword established. Other TCP flag combinations are supported in software.

Note Match-all is not supported. Match-any is supported only when used in the following combinations of positive flags: "rst and ack" (must be combined), "sync and fin and rst" (must be combined), "psh" and "urg".

– For Supervisor Engine 2-Plus to V-10GE, you can specify up to six Layer 4 operations (lt, gt, neq, and range) in an ACL to ensure that all operations are processed in hardware. More than six Layer 4 operations trigger an attempt to translate the excess operations into multiple ACEs in hardware. If this attempt fails, packets are processed in software. The translation process is less likely to succeed on large ACLs with a great number of Layer 4 operations, and on switches with large numbers of ACLs configured. The precise limit depends on how many other ACLs are configured and which specific Layer 4 operations are used by the ACLs being translated. The eq operator does not require any Layer 4 operations and can be used any number of times.

– For Supervisor Engine and 6L-E, refer to the “Restrictions for Layer 4 Operations” section.

– If the total number of Layer 4 operations in an ACL is less than six, you can distribute the operations in any way you choose.

Examples

The following access lists are processed completely in hardware:

Access lists 104 and 105 are identical; established is shorthand for rst and ack.

Access list 101, is processed completely in software:

Because four source and two destination operations exist, access list 106 is processed in hardware:

In the following code, the Layer 4 operations for the third ACE trigger an attempt to translate dst lt 1023 into multiple ACEs in hardware, because three source and three destination operations exist. If the translation attempt fails, the third ACE is processed in software.

Similarly, for access list 103, the third ACE triggers an attempt to translate dst gt 1023 into multiple ACEs in hardware. If the attempt fails, the third ACE is processed in software. Although the operations for source and destination ports look similar, they are considered different Layer 4 operations.

Note Remember that source port lt 80 and destination port lt 80 are considered different operations.

• Some packets must be sent to the CPU for accounting purposes, but the action is still performed by the hardware. For example, if a packet must be logged, a copy is sent to the CPU for logging, but the forwarding (or dropping) is performed in the hardware. Although logging slows the CPU, it does not affect the forwarding rate. This sequence of events would happen under the following conditions:

– When a log keyword is used

– When an output ACL denies a packet

– When an input ACL denies a packet, and on the interface where the ACL is applied,
ip unreachable is enabled (ip unreachable is enabled by default on all the interfaces)

VLAN Map Configuration Guidelines

When configuring VLAN maps, consider these guidelines:

• VLAN maps do not filter IPv4 ARP packets.
• If there is no router ACL configured to deny traffic on a routed VLAN interface (input or output), and no VLAN map configured, all traffic is permitted.
• Each VLAN map consists of a series of entries. The order of entries in a VLAN map is important. A packet that comes into the switch is tested against the first entry in the VLAN map. If it matches, the action specified for that part of the VLAN map is taken. If there is no match, the packet is tested against the next entry in the map.
• If the VLAN map has at least one match clause for the type of packet (IP or MAC) and the packet does not match any of these match clauses, the default is to drop the packet. If there is no match clause for that type of packet in the VLAN map, the default is to forward the packet.
• The system might take longer to boot if you have configured a very large number of ACLs.

Creating and Deleting VLAN Maps

Each VLAN map consists of an ordered series of entries. To create, add to, or delete a VLAN map entry, perform this task:

You can use the no vlan access-map name global configuration command to delete a map. You can use the no vlan access-map name number global configuration command to delete a single sequence entry from within the map. You can use the no action access-map configuration command to enforce the default action, which is to forward.

VLAN maps do not use the specific permit or deny keywords. To deny a packet by using VLAN maps, create an ACL that would match the packet, and then set the action to drop. A permit in the ACL is the same as a match. A deny in the ACL means no match.

Examples of ACLs and VLAN Maps

These examples show how to create ACLs and VLAN maps for specific purposes.

Example 1

This example shows how to create an ACL and a VLAN map to deny a packet. In the first map, any packets that match the ip1 ACL (TCP packets) would be dropped. You first create the ip1 ACL to permit any TCP packet and no other packets. Because there is a match clause for IP packets in the VLAN map, the default action is to drop any IP packet that does not match any of the match clauses.

This example shows how to create a VLAN map to permit a packet. ACL ip2 permits UDP packets; and any packets that match the ip2 ACL are forwarded.

In this map, any IP packets that did not match any of the previous ACLs (that is, packets that are not TCP packets or UDP packets) would get dropped.

Example 2

In this example, the VLAN map is configured to drop IP packets and to forward MAC packets by default. By applying standard ACL 101 and the extended named access lists igmp-match and tcp-match, the VLAN map is configured to do the following:

• Forward all UDP packets
• Drop all IGMP packets
• Forward all TCP packets
• Drop all other IP packets
• Forward all non-IP packets

Example 3

In this example, the VLAN map is configured to drop MAC packets and forward IP packets by default. By applying MAC extended access lists, good-hosts and good-protocols, the VLAN map is configured to do the following:

• Forward MAC packets from hosts 0000.0c00.0111 and 0000.0c00.0211
• Forward MAC packets of DECnet or VINES (Virtual Integrated Network Service) protocol-family
• Drop all other non-IP packets
• Forward all IP packets

Example 4

In this example, the VLAN map is configured to drop all packets (IP and non-IP). By applying access lists tcp-match and good-hosts, the VLAN map is configured to do the following:

• Forward all TCP packets
• Forward MAC packets from hosts 0000.0c00.0111 and 0000.0c00.0211
• Drop all other IP packets
• Drop all other MAC packets

Applying a VLAN Map to a VLAN

To apply a VLAN map to one or more VLANs, perform this task:

Note You cannot apply a VLAN map to a VLAN on a switch that has ACLs applied to Layer 2 interfaces (port ACLs).

This example shows how to apply VLAN map 1 to VLANs 20 through 22:

Using VLAN Maps in Your Network

Figure 1-3 shows a typical wiring closet configuration. Host X and Host Y are in different VLANs, connected to wiring closet switches A and C. Traffic moving from Host X to Host Y is routed by Switch B. Access to traffic moving from Host X to Host Y can be controlled at the entry point of Switch A. In the following configuration, the switch can support a VLAN map and a QoS classification ACL.

Figure 1-3 Wiring Closet Configuration

For example, if you do not want HTTP traffic to be switched from Host X to Host Y, you could apply a VLAN map on Switch A to drop all HTTP traffic moving from Host X (IP address 10.1.1.32) to Host Y (IP address 10.1.1.34) at Switch A and not bridge the traffic to Switch B. To configure this scenario, you would do the following.

First, define an IP access list HTTP to permit (match) any TCP traffic on the HTTP port, as follows:

Next, create a VLAN access map named map2 so that traffic that matches the HTTP access list is dropped and all other IP traffic is forwarded, as follows:

You then apply the VLAN access map named map2 to VLAN 1, as follows:

Denying Access to a Server on Another VLAN

Figure 1-4 shows how to restrict access to a server on another VLAN. In this example, server 10.1.1.100 in VLAN 10 has the following access restrictions:

• Hosts in subnet 10.1.2.0/8 in VLAN 20 should not have access.
• Hosts 10.1.1.4 and 10.1.1.8 in VLAN 10 should not have access.

Figure 1-4 Deny Access to a Server on Another VLAN

This procedure configures ACLs with VLAN maps to deny access to a server on another VLAN. The VLAN map SERVER 1_ACL denies access to hosts in subnet 10.1.2.0/8, host 10.1.1.4, and host 10.1.1.8. Then it permits all other IP traffic. In Step 3, VLAN map SERVER1 is applied to VLAN 10.

To configure this scenario, follow these steps:

Step 1 Define the IP ACL to match and permit the correct packets.

Step 2 Define a VLAN map using the ACL to drop IP packets that match SERVER1_ACL and forward IP packets that do not match the ACL.

Step 3 Apply the VLAN map to VLAN 10.

Guidelines for Using Router ACLs and VLAN Maps on the Same VLAN

Because the switch hardware performs one lookup for each direction (input and output), you must merge a router ACL and a VLAN map when they are configured on the same VLAN. Merging the router ACL with the VLAN map can significantly increase the number of ACEs.

When possible, try to write the ACL so that all entries have a single action except for the final, default action. You should write the ACL using one of these two forms:

permit...
permit...
permit...
deny ip any any

or

deny...
deny...
deny...
permit ip any any

To define multiple permit or deny actions in an ACL, group each action type together to reduce the number of entries.

If you need to specify the full-flow mode and the ACL contains both IP ACEs and TCP/UDP/ICMP ACEs with Layer 4 information, put the Layer 4 ACEs at the end of the list. Doing this gives priority to the filtering of traffic based on IP addresses.

Examples of Router ACLs and VLAN Maps Applied to VLANs

These examples show how router ACLs and VLAN maps are applied on a VLAN to control the access of switched, bridged, routed, and multicast packets. Although the following illustrations show packets being forwarded to their destination, each time a packet crosses a line indicating a VLAN map or an ACL, the packet could be dropped rather than forwarded.

ACLs and Switched Packets

Figure 1-5 shows how an ACL processes packets that are switched within a VLAN. Packets switched within the VLAN are not processed by router ACLs.

Figure 1-5 Applying ACLs on Switched Packets

ACLs and Routed Packets

Figure 1-6 shows how ACLs are applied on routed packets. For routed packets, the ACLs are applied in this order:

1. VLAN map for input VLAN

2. Input router ACL

3. Output router ACL

4. VLAN map for output VLAN

Figure 1-6 Applying ACLs on Routed Packets

Creating a PACL

To create a PACL and apply it to one or more interfaces, follow these steps:

Step 1 Create the standard or extended IPv4 ACLs, IPv6 ACLs, or named MAC extended ACLs that you want to apply to the interface.

Step 2 Use the IP access-group, IPv6 traffic-filter, or mac access-group interface command to apply IPv4, IPv6, or MAC ACLs to one or more Layer 2 interfaces.

PACL Configuration Guidelines

When configuring PACLs, consider these guidelines:

• There can be at most one IPv4, one IPv6, and one MAC access list applied to the same Layer 2 interface per direction.
• The IPv4 access list filters only IPv4 packets, the IPv6 access list filters only IPv6 packets, and the MAC access list filters only non-IP packets.
• The number of ACLs and ACEs that can be configured as part of a PACL are bounded by the hardware resources on the switch. Those hardware resources are shared by various ACL features
  (for example, RACL, VACL) that are configured on the system. If insufficient hardware resources to program PACL exist in hardware, the actions for input and output PACLs differ:

– For input PACLs, some packets are sent to CPU for software forwarding.

– For output PACLs, the PACL is disabled on the port.

• These restrictions pertain to output PACLs only:

– If insufficient hardware resources exist to program the PACL, the output PACL is not applied to the port, and you receive a warning message.

– On Supervisor Engine II+, if an output PACL is configured on a Layer 2 port, then neither a VACL nor a Router ACL can be configured on the VLANs to which the Layer 2 port belongs.

If any VACL or Router ACL is configured on the VLANs to which the Layer 2 port belongs, the output PACL cannot be configured on the Layer 2 port. PACLs and VLAN-based ACLs (VACL and Router ACL) are mutually exclusive on Layer 2 ports.

• The input ACL logging option is supported, although logging is not supported for output ACLs.
• The access group mode can change the way PACLs interact with other ACLs. To maintain consistent behavior across Cisco platforms, use the default access group mode.
• If a PACL is removed when there are active sessions on a port, a hole (permit ip any any) is installed on the port.

Removing the Requirement for a Port ACL

Prior to Cisco IOS Release 12.2(54)SG, a standard port ACL was necessary if you planned to download an ACL from a AAA server. This was because ACL infrastructure was insufficient to provide dynamic creation of access control entries without associating an ACL with the port.

Starting with Cisco IOS Release 12.2(54)SG, configuring a port ACL is not mandatory. If a port ACL is not configured on the port (by entering the ip access-group number in command), a default ACL (AUTH-DEFAULT-ACL) is attached automatically to the port when an ACL is downloaded. It allows only DHCP traffic and consists of the following ACEs:

AUTH-DEFAULT-ACL is automatically created. To modify it, enter the following command:

This ACL is not nvgened. AUTH-DEFAULT-ACL is attached provided there are sessions applying dynamic ACLs (Per-user/Filter-Id/DACL). AUTH-DEFAULT-ACL is removed when the last authenticated session with policies is cleared. It remains attached to the port provided at least one session is applying dynamic policies.

Configuration Restrictions

The following restrictions apply:

• Starting with Cisco IOS Release 12.2(54)SG, the port ACL does not require configuration; the default ACL is created automatically.
• Even if AUTH-DEFAULT-ACL is modified, it is not nvgened.

Debugging Considerations

Syslog messages appear when AUTH-DEFAULT-ACL is attached or detached from an interface provided you enter the epm logging command in configuration mode.

The following syslog displays when the default ACL is attached:

The following syslog displays when the ACL is detached:

Webauth Fallback

Many authentication methods require specific capabilities on the end-point device to respond to the network authenticating device with its identity or credentials. If the end-point lacks the required capability, the authenticator must fallback to alternative methods to gather host or user credentials. If the 802.1X/MAB authentication mechanism fails, a fallback to webauth might occur.

Prior to Cisco IOS Release 12.2(54)SG, webauth fallback implementation required a fallback profile configured on the authenticating device. As part of this profile, an admission rule must be configured along with the access policies (the fallback ACL).

Consider a situation where no port ACL is configured on a port. The first few hosts authenticated through 802.1X/MAB do not download any ACLs. All traffic from these hosts is allowed through. Now, suppose a host connects to the port, and there is a fallback to webauth to authenticate the host. The fallback ACL will be installed on the port, and traffic from previously authenticated hosts will also be restricted by this fallback ACL.

Starting with Cisco IOS Release 12.2(54)SG, Cisco uses a different approach to address this issue. When a host falls back to webauth for authentication, the ACE entries in the fallback ACL are converted into entries with Host IP insertion for a host that has fallen back and will be applied until the host authenticates. Once the host successfully authenticates, the fallback ACL is removed. The resultant host ACLS will be: dynamic ACLs and Port ACL/AUTH-DEFAULT-ACL. Refer to the previous section for an explanation of AUTH-DEFAULT -ACL.

Configuring IPv4, IPv6, and MAC ACLs on a Layer 2 Interface

Note Only IPv4, IPv6 and MAC ACLs can be applied to Layer 2 physical interfaces.

Standard (numbered, named), Extended (numbered, named) IP ACLs, and Extended Named MAC ACLs are also supported.

To apply IPv4 or MAC ACLs on a Layer 2 interface, perform this task:

To apply IPv6 ACLs on a Layer 2 interface, perform this task:

The following example shows how to configure the Extended Named IP ACL simple-ip-acl to permit all TCP traffic and implicitly deny all other IP traffic:

The following example shows how to configure the Extended Named MACL simple-mac-acl to permit source host 000.000.011 to any destination host:

Using PACL with Access-Group Mode

You can use the access group mode to change the way PACLs interact with other ACLs. For example, if a Layer 2 interface belongs to VLAN100, VACL (VLAN filter) V1 is applied on VLAN100, and PACL P1 is applied on the Layer 2 interface. In this situation, you must specify how P1 and V1 impact the traffic with the Layer 2 interface on VLAN100. In a per-interface method, you can use the access-group mode command to specify one of the following desired modes:

• prefer port mode—If PACL is configured on a Layer 2 interface, then PACL takes effect and overwrites the effect of other ACLs (Router ACL and VACL). If no PACL feature is configured on the Layer 2 interface, other features applicable to the interface are merged and applied on the interface. it is the default access group mode.
• prefer VLAN mode—VLAN-based ACL features take effect on the port if they have been applied on the port and no PACLs are in effect. If no VLAN-based ACL features are applicable to the Layer 2 interface, then the PACL feature already on the interface is applied.
• merge mode—Merges applicable ACL features before they are programmed into the hardware.

Note Output PACLs are mutually exclusive with VACL and Router ACLs on Supervisor Engine II+. Access group mode does not change the behavior of output traffic filtering.

Configuring Access-group Mode on Layer 2 Interface

To configure an access mode on a Layer 2 interface, perform this task:

This example shows how to merge and apply features other than PACL on the interface:

This example shows how to merge applicable ACL features before they are programmed into hardware:

Applying ACLs to a Layer 2 Interface

To apply IPv4, IPv6, and MAC ACLs to a Layer 2 interface, perform one of these tasks:

Note Supervisor Engines III and Supervisor Engine IV running on a Catalyst 4500 series switch support both input and output PACLs on an interface.

This example applies the extended named IP ACL simple-ip-acl to interface FastEthernet 6/1 ingress traffic:

This example applies the IPv6 ACL simple-ipv6-acl to interface FastEthernet 6/1 ingress traffic:

This example applies the extended named MAC ACL simple-mac-acl to interface FastEthernet 6/1 egress traffic:

Displaying an ACL Configuration on a Layer 2 Interface

To display information about an ACL configuration on Layer 2 interfaces, perform one of these tasks:

This example shows that the IP access group simple-ip-acl is configured on the inbound direction of interface fa6/1:

This example shows that MAC access group simple-mac-acl is configured on the inbound direction of interface fa6/1:

This example shows that access group merge is configured on interface fa6/1:

Introduction

When deploying IPv6 networks, routers are configured to use IPv6 Router Advertisements to convey configuration information to hosts onlink. Router Advertisement is a critical part of the autoconfiguration process. The conveyed information includes the implied default router address obtained from the observed source address of the Router-Advertisement (RA) message. However, in some networks, invalid RAs are observed. This may happen because of misconfigurations or a malicious attacks on the network.

Devices acting as rogue routers may send illegitimate RAs.When using IPv6 within a single Layer 2 network segment, you can enable Layer 2 devices to drop rogue RAs before they reach end-nodes.

RA Guard examines incoming Router-Advertisement and Router-Redirect packets and decides whether to switch or block them based solely on information found in the message and in the Layer 2 device configuration.

You can configure RA Guard in two modes (host and router) based on the device connected to the port.

• Host mode—All the Router-Advertisement and Router-Redirect messages are disallowed on the port.
• Router mode—All messages (RA/RS/Redirect) are allowed on the port; only host mode is supported.

You can configure Catalyst 4500 host ports to allow or disallow RA messages. Once a port is configured to disallow the Router-Advertisement and Router-Redirect packets, it filters the content of the received frames on that port and blocks Router-Advertisement or Router-Redirect frames.

When RA Guard is configured on a port, the following packets are dropped in hardware:

• Router-Advertisement packets —IPv6 ICMP packets with ICMP type = 134
• Router-Redirect packets—IPv6 ICMP packets with ICMP type = 137

Per port RA Guard ACL statistics are supported and displayed when you enter a show ipv6 snooping counters interface command. The statistics output displays the number of packets that have been dropped per port due to the RA Guard.

Note Beginning with Cisco IOS Release 15.0(2)SG, per port RA Guard ACL statistics are supported and displayed when you enter a show ipv6 snooping counters interface command. (Previous to this release, you enter the show ipv6 first-hop counters interface command.)

Deployment

Figure 1-10 illustrates a deployment scenario for RA Guard. We drop RA packets from ports that are connected to hosts and permit RA packets from ports connected to the Router.

Figure 1-10 Typical RA Guard Deployment

Configuring RA Guard

To configure RA Guard, perform this step:

Examples

This examples shows how to enable RA Guard on the switch:

The following example shows a sample output of the show ipv6 commands:

Note Beginning with Cisco IOS Release 15.0(2)SG, per port RA Guard ACL statistics are supported and displayed when you enter a show ipv6 snooping counters interface command. (Previous to this release, you enter the show ipv6 first-hop counters interface command.)

Note Be aware that only RA (Router Advertisement) and REDIR (Router Redirected packets) counters are supported in 12.2(54)SG.

Usage Guidelines

Observe the following restrictions:

• RA Guard is an ingress feature; only IPv6 Router-Advertisement and Router-Redirect packets entering through the port are filtered.
• RA Guard does not offer protection in environments where IPv6 traffic is tunneled.
• This feature is supported only in hardware; packets are not punted to software except under resource exhaustion (for example, TCAM memory exhaustion).
• RA Guard is purely an Layer 2 port based feature and can be configured only on switchports. It works irrespective of whether IPv6 routing is enabled. It is not supported on router interfaces and VLANs.
• RA Guard is supported on trunk ports; filtering is performed on packets arriving from all the allowed VLANs.
• RA Guard is supported on EtherChannel; the RA Guard configuration (whether present or not) on the EtherChannel overrides the RA Guard configuration on the member ports.
• RA Guard is supported on ports that belong to PVLANs (for example, isolated secondary host ports, community secondary host ports, promiscuous primary host ports, (primary/secondary) trunk ports. Primary VLAN features are inherited and merged with port features.
• Because of hardware limitations, it may not be possible for Supervisor Engine 6-E and 6L-E to collect statistics for RA Guard in hardware. If so, an error message is displayed. The show ipv6 snooping counter interface commands display the estimated counters.

Note Beginning with Cisco IOS Release 15.0(2)SG, per port RA Guard ACL statistics are supported and displayed when you enter a show ipv6 snooping counters interface command. (Previous to this release, you enter the show ipv6 first-hop counters interface command.)