Release 15.1SY Supervisor Engine 720 Software Configuration Guide

Prerequisites for Global Protocol Packet Policing

None.

Restrictions for Global Protocol Packet Policing

• The minimum values supported by the mls qos protocol arp police command are too small for use in production networks.
• ARP packets are approximately 40 bytes long and ARP reply packets are approximately 60 bytes long. The policer rate value is in bits per second. The burst value is in bytes per second. Together, an ARP request and reply are approximately 800 bits.
• The configured rate limits are applied separately to the PFC and each DFC. The RP CPU will receive the configured value times the number of forwarding engines.
• Policy-based protocol packet policing is applied per-forwarding engine (PFC and any DFCs).
• The protocol packet policing mechanism effectively protects the RP CPU against attacks such as line-rate ARP attacks, but it polices both routing protocols and ARP packets to the switch and also polices traffic through the switch with less granularity than CoPP.
• The policing mechanism shares the root configuration with a policing-avoidance mechanism. The policing-avoidance mechanism lets the routing protocol and ARP packets flow through the network when they reach a QoS policer. This mechanism can be configured using the mls qos protocol protocol_name pass-through command.
• Policy-based protocol packet policing does not support microflow policers.
• Only ingress Policy-based protocol packet policing is supported.
• Policy-based protocol packet policing does not support Layer 4 ACL operators (see the “Restrictions for Layer 4 Operators in ACLs” section), which imposes these subsequent restrictions:

– For IPv4 or IPv6 traffic, no support for UDP or TCP port range matching

– For IPv6 traffic, no support for precedence or DSCP matching

• Protocol packet policing policies can share an aggregate policer with QoS policies.
• An aggregate policer cannot be applied to both ingress and egress traffic.
• Policy-based protocol packet policing supports the class default and permit protocol_name any any commands, but traffic flow might be affected significantly because the protocol packet policing policy processes all matched traffic.
• With Supervisor Engine 720, policy-based protocol packet policing is applied only to untrusted ports.
• You can configure both single-command protocol packet policing and policy-based protocol packet policing. Single-command protocol packet policing is applied first and then policy-based protocol packet policing.

Note The software does not detect or attempt to resolve any configuration conflicts between single-command protocol packet policing and policy-based protocol packet policing.

• You can configure both policy-based protocol packet policing and control plane policing (see Chapter 1, “Control Plane Policing (CoPP)”). Policy-based protocol packet policing is applied first and then CoPP.
• Single-command protocol packet policing programs the configured protocol-specific action for ingress traffic and automatically programs a corresponding egress-traffic pass-through action to preserve the ingress result egress traffic.
• Policy-based protocol packet policing does not automatically preserve the ingress policing result in egress traffic.

– To preserve the ingress policing result in egress traffic with policy-based protocol packet policing, configure an appropriate output policy. To pass egress traffic through unchanged, duplicate each ingess class in the output policy and configure trust dscp as the class-map action.

– Without an output policy-map, egress traffic is processed by any configured interface-based policy-map and ingress global policy result will be overwritten.

• The PFC and any DFCs supports a single match command in class-map match-all class maps, except that the match protocol command can be configured in a class map with the match dscp or match precedence command.
• The PFC and any DFCs supports multiple match commands in class-map match-any class maps.
• Class maps can use the match commands listed in Table 1-1 to configure a traffic class that is based on the match criteria.

The PFC and any DFCs supports these ACL types for use with the match access group command:

Information About Global Protocol Packet Policing

Attackers may try to overwhelm the RP CPU with routing protocol control packets (for example, ARP packets). Protocol packet policing rate limits this traffic in hardware. Release 15.1(1)SY1 and later releases support policy-based global protocol packet policing, shown in Cisco Feature Navigator as the Global QoS Policy feature.

How to Configure Single-Command Global Protocol Packet Policing

Enter the mls qos protocol ? to display the supported routing protocols.

The mls qos protocol arp police command rate limits ARP packets. This example shows how to allow 200 ARP requests and replies per second:

This example shows how to display the available protocols to use with protocol packet policing:

This example shows how to display the available keywords to use with the mls qos protocol command:

How to Configure Policy-Based Global Protocol Packet Policing

Use these QoS sections and the global protocol packet policing policy map configuration section:

• Configuring a Class Map
• Configuring a Policy Map
• Configuring a Global Protocol Packet Policing Policy Map

Configuring a Global Protocol Packet Policing Policy Map

To configure a global protocol packet policing policy map, perform this task:

Prerequisites for uRPF Check

None.

Restrictions for uRPF Check

• Unicast RPF does not provide complete protection against spoofing. Spoofed packets can enter a network through unicast RPF-enabled interfaces if an appropriate return route to the source IP address exists.
• Avoid configurations that overload the route processor with unicast RPF checks.

– Do not configure unicast RPF to filter with an ACL.

– Do not configure the global unicast RPF “punt” check mode.

• The PFC does not provide hardware support for the unicast RPF check for policy-based routing (PBR) traffic. ( CSCea53554)
• The switch applies the same unicast RPF mode to all interfaces where unicast RPF check is configured. Any change that you make in the unicast RPF mode on any interfaces is applied to all interfaces where the unicast RPF check is configured.
• The “allow default” options of the unicast RPF modes do not offer significant protection against spoofing.

– Strict unicast RPF Check with Allow Default—Received IP traffic that is sourced from a prefix that exists in the routing table passes the unicast RPF check if the prefix is reachable through the input interface. If a default route is configured, any IP packet with a source prefix that is not in the routing table passes the unicast RPF check if the ingress interface is a reverse path for the default route.

– Loose unicast RPF Check with Allow Default—If a default route is configured, any IP packet passes the unicast RPF check.

– If, on a maximum of 24 interfaces, divided into four groups of six interfaces each, the switch receives valid IP packets that have up to six reverse-path interfaces per source prefix, you can configure the unicast RPF strict mode with the mls ip cef rpf multipath interface-group command.

This option requires you to identify the source prefixes and the interfaces that serve as reverse paths for the source prefixes and to configure interface groups for those reverse path interfaces. All of the reverse-path interfaces for each source prefix must be in the same interface group. You can configure four interface groups, with each group containing up to six reverse-path interfaces. There is no limit on the number of source prefixes that an interface group can support.

To ensure that no more than six reverse-path interfaces exist in the routing table for each prefix, enter the maximum-paths 6 command in config-router mode when configuring OSPF, EIGRP, or BGP.

IP traffic with one or two reverse-path interfaces that is received on uPPF-check enabled interfaces outside the interface groups passes the unicast RPF check if the ingress interface and at most one other interface are reverse paths.

With maximum paths set to six, IP traffic with more than two reverse-path interfaces that is received on uPPF-check enabled interfaces outside the interface groups always pass the unicast RPF check.

– If, on any number of interfaces, the switch receives valid IP packets that have one or two reverse path interfaces per source prefix, you can configure the unicast RPF strict mode with the mls ip cef rpf multipath pass command.

To ensure that no more than two reverse-path interfaces exist in the routing table for each prefix, enter the maximum-paths 2 command in config-router mode when configuring OSPF, EIGRP, or BGP.

• Unicast RPF Loose Mode with the Pass Global Mode—The unicast RPF loose mode provides less protection than strict mode, but it is an option on switches that receive valid IP traffic on interfaces that are not reverse paths for the traffic. The unicast RPF loose mode verifies that received traffic is sourced from a prefix that exists in the routing table, regardless of the interface on which the traffic arrives.

Information about uRPF Check

The unicast RPF check verifies that the source address of received IP packets is reachable. The unicast RPF check discards IP packets that lack a verifiable IP source prefix (route), which helps mitigate problems that are caused by traffic with malformed or forged (spoofed) IP source addresses.

The unicast RPF check is performed in software on the RP and supports up to 16 reverse-path interfaces.

To ensure that no more than 16 reverse-path interfaces exist in the routing table for each prefix, enter the maximum-paths 16 command in config-router mode when configuring OSPF, EIGRP, or BGP.

For unicast RPF check without ACL filtering, the PFC3 provides hardware support for the RPF check of traffic from multiple interfaces.

For unicast RPF check with ACL filtering, the PFC determines whether or not traffic matches the ACL. The PFC sends the traffic denied by the RPF ACL to the route processor (RP) for the unicast RPF check. Packets permitted by the ACL are forwarded in hardware without a unicast RPF check.

How to Configure Unicast RPF Check

• Configuring the Unicast RPF Check Mode
• Configuring the Multiple-Path Unicast RPF Check Mode
• Configuring Multiple-Path Interface Groups
• Enabling Self-Pinging

Configuring the Unicast RPF Check Mode

To configure unicast RPF check mode, perform this task:

Note The most recently configured mode is automatically applied to all ports configured for unicast RPF check.

• Use the rx keyword to enable strict check mode.
• Use the any keyword to enable exist-only check mode.
• Use the allow-default keyword to allow use of the default route for RPF verification.
• Use the list option to identify an access list.

– If the access list denies network access, denied packets are dropped at the port.

– If the access list permits network access, packets are forwarded to the destination address. Forwarded packets are counted in the interface statistics.

– If the access list includes the logging action, information about the packets is sent to the log server.

This example shows how to enable unicast RPF exist-only check mode on Gigabit Ethernet port 4/1:

This example shows how to enable unicast RPF strict check mode on Gigabit Ethernet port 4/2:

Note ACL-based unicast RPF check is not supported in the hardware on Supervisor Engine 720. If you configure ACL-based uRPF check on an interface, all packets denied in the ACL are redirected to the MSFC3 CPU for the uRPF check, while packets permitted by the ACL are forwarded in hardware without a uRPF check. Redirected packets are subject to the global ACL bridged input hardware-to-CPU rate limiter (that limits the amount of such traffic that reaches the MSFC3 CPU) and may be dropped in hardware. Configure this rate limiter with the mls rate-limit unicast acl input global configuration command.

Beginning in Cisco IOS release 12.2(18)SXF6, you can change this behavior using the mls ip cef rpf hw-enable-rpf-acl global configuration command. When you add this command, packets permitted by the exception ACL are redirected to the MSFC3 CPU for the uRPF check, while packets denied by the ACL are forwarded in hardware with uRPF check. If the packets fail the hardware RPF check, they are punted to CPU. RPF-fail packets are subject to the global RPF fail hardware-to-CPU rate limiter. Configure this rate limiter with the mls rate-limit unicast ip rpf-failure global configuration command.

Configuring the Multiple-Path Unicast RPF Check Mode

To configure the multiple-path unicast RPF check mode, perform this task:

When configuring multiple path RPF check, note the following information:

• punt mode (default)—The PFC3 performs the unicast RPF check in hardware for up to two interfaces per prefix. Packets arriving on any additional interfaces are redirected (punted) to the RP for unicast RPF check in software.
• pass mode—The PFC3 performs the unicast RPF check in hardware for single-path and two-path prefixes. Unicast RPF check is disabled for packets coming from multipath prefixes with three or more reverse-path interfaces (these packets always pass the unicast RPF check).
• interface-group mode—The PFC3 performs the unicast RPF check in hardware for single-path and two-path prefixes. The PFC3 also performs the unicast RPF check for up to four additional interfaces per prefix through user-configured multipath unicast RPF check interface groups. Unicast RPF check is disabled for packets coming from other multipath prefixes that have three or more reverse-path interfaces (these packets always pass the unicast RPF check).

This example shows how to configure punt as the multiple path RPF check mode:

Configuring Multiple-Path Interface Groups

To configure multiple-path unicast RPF interface groups, perform this task:

This example shows how to configure interface group 2:

Enabling Self-Pinging

With unicast RPF check enabled, by default the switch cannot ping itself. To enable self-pinging, perform this task:

This example shows how to enable self-pinging:

Prerequisites for Rate-Limiters

None.

Restrictions for Rate-Limiters

• These are Layer 2 rate limiters:

– Layer 2 PDUs

– Layer 2 protocol tunneling

– Layer 2 Multicast IGMP

• Do not use the CEF receive limiter if CoPP is being used. The CEF receive limiter will override the CoPP traffic.
• Configured rate limits is applied to each forwarding engine (except for the Layer 2 hardware rate limiter which is applied globally).
• Layer 2 rate limiters are not supported in truncated mode.
• The following restrictions apply when using the ingress and egress ACL-bridged packet rate limiters:

– The ingress and egress ACL-bridged packet rate limiter is available for unicast traffic only.

– The ingress and egress ACL-bridged packet rate limiters share a single rate-limiter register. If you enable the ACL-bridge ingress and egress rate limiters, both the ingress and the egress ACLs must share the same rate-limiter value.

Recommended Rate-Limiter Configuration

The recommended rate-limiter configuration is as follows:

• Enable the rate limiters for the traffic types most likely to be used in a DoS attack.
• Do not use a rate limiter on VACL logging unless you configure VACL logging.
• Disable redirects.
• Disable unreachables.
• Do not enable the MTU rate limiter if all interfaces have the same MTU.
• When configuring the Layer 2 PDU rate limiter, note the following information:

– Calculate the expected or possible number of valid PDUs and double or triple the number.

– PDUs include BPDUs, DTP, VTP, PAgP, LACP, UDLD, etc.

– Rate limiters do not discriminate between good frames or bad frames.

Ingress-Egress ACL Bridged Packets (Unicast Only)

Commands:

mls rate-limit unicast acl input

mls rate-limit unicast acl output

The PFC and DFC provide separate ACL-bridge rate-limiters.

This rate limiter rate limits packets sent to the RP because of an ingress or egress ACL bridge results. The switch accomplishes this by altering existing and new ACL TCAM entries with a TCAM bridge result to a Layer 3 redirect result pointing to the RP. Packets hitting the TCAM entries with the altered Layer 3 redirect rate limit result will be rate limited according to the instructions set in CLI by the network administrator. If the ACL bridge rate limiting is disabled, the Layer 3 redirect rate limit results are converted to the bridge result.

Burst values regulate how many packets can be allowed in a burst. Each allowed packet consumes a token and a token must be available for a packet to be allowed. One token is generated per millisecond. When packets are not coming in, tokens can be accumulated up to the burst value. For example, if the burst value is set to 50, the switch can accumulate up to 50 tokens and absorb a burst of 50 packets.

uRPF Check Failure

Command: mls rate-limit unicast ip rpf-failure

The uRPF check failure rate limiter allows you to configure a rate for the packets that need to be sent to the RP because they failed the uRPF check. The uRPF checks validate that incoming packets on an interface are from a valid source, which minimizes the potential threat of DoS attacks from users using spoofed addresses. When spoofed packets fail the uRPF check, those failures can be sent to the RP. The uRPF check rate limiters allow you to rate limit the packets per second that are bridged to the RP CPU when a uRPF check failure occurs.

TTL Failure

Command: mls rate-limit all ttl-failure

This rate limiter rate limits packets sent to the RP because of a time-to-live (TTL) check failure. As indicated by the all keyword in the following example, this rate limiter applies to both multicast and unicast traffic.

Note The TTL failure rate limiter is not supported for IPv6 multicast.

ICMP Unreachable (Unicast Only)

Commands:

mls rate-limit unicast ip icmp unreachable acl-drop

mls rate-limit unicast ip icmp unreachable no-route

In an ICMP unreachable attack, a device is flooded with a large number of packets that contain a destination address that is unreachable from the flooded device (in this case, the RP). The ICMP unreachable rate limiter allows you to rate limit the packets that are sent to the RP containing unreachable addresses.

FIB (CEF) Receive Cases (Unicast Only)

Command: mls rate-limit unicast cef receive

The FIB receive rate limiter provides the capability to rate limit all packets that contain the RP IP address as the destination address. The rate limiters do not discriminate between good frames and bad frames.

Note Do not enable the FIB receive rate limiter if you are using CoPP. The FIB receive rate limiter overrides the CoPP policies.

FIB Glean (Unicast Only)

Command: mls rate-limit unicast cef glean

The FIB glean rate limiter does not limit ARP traffic, but provides the capability to rate limit traffic that requires address resolution (ARP) and requires that it be sent to the RP. This situation occurs when traffic enters a port and contains the destination of a host on a subnet that is locally connected to the RP, but no ARP entry exists for that destination host. In this case, because the MAC address of the destination host will not be answered by any host on the directly connected subnet that is unknown, the “glean” adjacency is hit and the traffic is sent directly to the RP for ARP resolution. This rate limiter limits the possibility of an attacker overloading the CPU with such ARP requests.

Layer 3 Security Features (Unicast Only)

Command: mls rate-limit unicast ip features

Some security features are processed by first being sent to the RP. For these security features, you need to rate limit the number of these packets being sent to the RP to reduce any potential overloading. The security features include authentication proxy (auth-proxy), IPSEC, and inspection.

Authentication proxy is used to authenticate inbound or outbound users or both. These users are normally blocked by an access list, but with auth-proxy, the users can bring up a browser to go through the firewall and authenticate on a terminal access controller access control system plus (TACACS+) or RADIUS server (based on the IP address). The server passes additional access list entries down to the switch to allow the users through after authentication. These ACLs are stored and processed in software, and if there are many users utilizing auth-proxy, the RP may be overwhelmed. Rate limiting would be advantageous in this situation.

IPsec and inspection are also done by the RP and may require rate limiting. When the Layer 3 security feature rate limiter is enabled, all Layer 3 rate limiters for auth-proxy, IPsec and inspection are enabled at the same rate.

ICMP Redirect (Unicast Only)

Command: mls rate-limit unicast ip icmp redirect

The ICMP-redirect rate limiter allows you to rate limit ICMP traffic. For example, when a host sends packets through a nonoptimal switch, the RP sends ICMP-redirect messages to the host to correct its sending path. If this traffic occurs continuously, and is not rate limited, the RP will continuously generate ICMP-redirect messages.

VACL Log (Unicast Only)

Command: mls rate-limit unicast acl vacl_log

Packets that are sent to the RP because of VLAN-ACL logging can be rate limited to ensure that the CPU is not overwhelmed with logging tasks. VACLs are processed in hardware, but the RP does the logging. When VACL logging is configured on the switch, IP packets that are denied in the VACL generate log messages.

MTU Failure

Command: mls rate-limit all mtu

Like the TTL failure rate limiter, the rate limiter for MTU failures is supported for both unicast and multicast traffic. Packets that fail an MTU check are sent to the RP CPU. This might cause the RP to be overwhelmed.

Layer 2 Multicast IGMP Snooping

Command: mls rate-limit multicast ipv4 igmp

The IGMP snooping rate limiter limits the number of Layer 2 IGMP packets destined for the RP. IGMP snooping listens to IGMP messages between the hosts and the switch. You cannot enable the Layer 2 PDU rate limiter if the switch is operating in truncated mode. The switch uses truncated mode for traffic between fabric-enabled modules when there are both fabric-enabled and nonfabric-enabled modules installed. In this mode, the switch sends a truncated version of the traffic (the first 64 bytes of the frame) over the switch fabric channel. The IGMP snooping rate limiter does not rate limit PIM messages.

Layer 2 PDU

Command: mls rate-limit layer2 pdu

The Layer 2 protocol data unit (PDU) rate limiter allows you to limit the number of hardware-switched Layer 2 PDU protocol packets (including BPDUs, DTP, PAgP, CDP, STP, and VTP packets). You cannot enable the Layer 2 PDU rate limiter if the switch is operating in truncated mode. The switch uses truncated mode for traffic between fabric-enabled modules when there are both fabric-enabled and nonfabric-enabled modules installed. In this mode, the switch sends a truncated version of the traffic (the first 64 bytes of the frame) over the switch fabric channel.

Layer 2 Protocol Tunneling

Command: mls rate-limit layer2 l2pt

This rate limiter limits the hardware-switched Layer 2 protocol tunneling packets, which include control PDUs, CDP, STP, and VTP packets. These packets are encapsulated in software (rewriting the destination MAC address in the PDU), and then forwarded to a proprietary multicast address (01-00-0c-cd-cd-d0). You cannot enable the Layer 2 PDU rate limiter if the switch is operating in truncated mode. The switch uses truncated mode for traffic between fabric-enabled modules when there are both fabric-enabled and nonfabric-enabled modules installed. In this mode, the switch sends a truncated version of the traffic (the first 64 bytes of the frame) over the switch fabric channel.

IP Errors

Command: mls rate-limit unicast ip errors

This rate limiter limits the packets with IP checksum and length errors. When a packet reaches the PFC or DFC with an IP checksum error or a length inconsistency error, it must be sent to the RP for further processing. An attacker might use the malformed packets to carry out a DoS attack, but the network administrator can configure a rate for these types of packets to protect the control path.

IPv4 Multicast

Commands:

mls rate-limit multicast ipv4 connected

mls rate-limit multicast ipv4 fib-miss

mls rate-limit multicast ipv4 igmp

mls rate-limit multicast ipv4 ip-options

mls rate-limit multicast ipv4 pim

These rate limiters limit IPv4 multicast packets. The rate limiters can rate limit the packets that are sent from the data path in the hardware up to the data path in the software. The rate limiters protect the control path in the software from congestion and drop the traffic that exceeds the configured rate.

IPv6 Multicast

Commands:

mls rate-limit multicast ipv6 connected

mls rate-limit multicast ipv6 control-packet

mls rate-limit multicast ipv6 mld

These rate limiters limit IPv6 multicast packets. The rate limiters can rate limit the packets that are sent from the data path in the hardware up to the data path in the software. The rate limiters protect the control path in the software from congestion and drop the traffic that exceeds the configured rate.

Hardware-Based Rate Limiters Default Configuration

Table 1-1 shows the DoS protection default configuration for the hardware-based rate limiters.

Displaying Rate-Limiter Information

The show mls rate-limit command displays information about the configured rate limiters.

The show mls rate-limit usage command displays the hardware register that is used by a rate-limiter type. If the register is not used by any rate-limiter type, Free is displayed in the output. If the register is used by a rate-limiter type, Used and the rate-limiter type are displayed.

In the command output, the rate-limit status could be one of the following:

• On indicates that a rate for that particular case has been set.
• Off indicates that the rate-limiter type has not been configured, and the packets for that case are not rate limited.
• On/Sharing indicates that a particular case (not manually configured) is affected by the configuration of another rate limiter belonging to the same sharing group.
• A hyphen indicates that the multicast partial-SC rate limiter is disabled.

In the command output, the rate-limit sharing indicates the following information:

• Whether sharing is static or dynamic
• Group dynamic sharing codes

To display the configured rate limiters, use the show mls rate-limit command:

To display the usage of the hardware rate limiters, use the show mls rate-limit usage command:

Prerequisites for Packet Drop Statistics

None.

Restrictions for Packet Drop Statistics

• The incoming captured traffic is not filtered.
• The incoming captured traffic is not rate limited to the capture destination.

Information About Packet Drop Statistics

You can use show commands to display packet drop statistics. You can capture the traffic on an interface and send a copy of this traffic to a traffic analyzer connected to a port, which can aggregate packet drop statistics.

How to Monitor Dropped Packets

• Using show Commands
• Using SPAN
• Using VACL Capture

Using show Commands

The PFC and DFCs support ACL hit counters in hardware. You can use the show tcam interface command to display each entry in the ACL TCAM.

This example shows how to use the show tcam interface command to display the number of times the entry was hit:

You can also use the TTL and IP options counters to monitor the performance of the Layer 3 forwarding engine.

This example shows how to use the show mls statistics command to display packet statistics and errors associated with the Layer 3 forwarding engine:

Using SPAN

This example shows how to use the monitor session command to capture and forward traffic to an external interface:

This example shows how to use the show monitor session command to display the destination port:

For more information, see Chapter1, “Local SPAN, RSPAN, and ERSPAN”

Using VACL Capture

The VACL capture feature allows you to direct traffic to ports configured to forward captured traffic. The capture action sets the capture bit for the forwarded packets so that ports with the capture function enabled can receive the packets. Only forwarded packets can be captured.

You can use VACL capture to assign traffic from each VLAN to a different interface.

VACL capture does not allow you to send one type of traffic, such as HTTP, to one interface and another type of traffic, such as DNS, to another interface. Also, VACL capture granularity is only applicable to traffic switched locally; you cannot preserve the granularity if you direct traffic to a remote switch.

For more information, see Chapter1, “VLAN ACLs (VACLs)”

http://www.cisco.com/en/US/products/hw/switches/ps708/tsd_products_support_series_home.html

Participate in the Technical Documentation Ideas forum