What is an ACL
An ACL (Access Control List) is a packet-processing policy made up of an ordered set of rules that a switch evaluates against traffic to decide what happens to each packet. Each rule is essentially a matching statement — source or destination MAC, source or destination IP, protocol, port number, VLAN, and similar packet fields — paired with an action to take when a packet matches. Once bound to an interface, an ACL lets a switch permit, deny, redirect, or otherwise handle traffic based purely on those defined conditions, giving network operators direct control over what's allowed to cross a given link.
ACLs exist because enterprise and data center networks need to enforce security boundaries between departments, tenants, or services that share the same physical infrastructure. Without packet-level filtering, any device connected to the network can reach any other device it has a route to — which is rarely the intended design. ACLs give operators a way to express "who can talk to what" directly in the forwarding path, without requiring a separate firewall for every segmentation boundary.
How ACLs Work
An ACL implementation is built from two related objects — the table and the rule — bound to the ports where filtering should apply:
An ACL table defines the scope of filtering: its type determines which packet fields it can match (L2 for MAC-based fields, L3/L3v6 for IPv4/IPv6 fields, MIRROR/MIRRORv6 for traffic-mirroring criteria, FLOW_CONTROL for rate-limiting), and its direction (ingress or egress) determines where in the forwarding pipeline it applies — with ingress typically supporting more match fields than egress. A table can be bound to one or many interfaces, and a single interface can have multiple tables applied, since the table-to-port relationship is many-to-many.
Each ACL rule lives inside exactly one table and inherits that table's match-field scope — a rule can't match on fields outside what its table type supports. A rule combines a priority (higher values win when multiple rules could match the same packet), one or more match conditions (source/destination IP or MAC, protocol, port, DSCP, VLAN, and so on, depending on table type), and an action.
On ingress, actions include permit (forward normally), deny (drop, but the packet can still be trapped), drop (dropped outright, never trapped), trap-to-cpu (sent only to the control plane), and copy-to-cpu (forwarded and copied to the control plane), plus the ability to rewrite DSCP, PCP, or traffic-class values, redirect matching traffic to a different interface, next hop, or next-hop group, or hand it off to a traffic-behavior policy for rate limiting. Egress actions are a simpler permit/deny pair, since the forwarding decision has already been made by that point.
Beyond the built-in table types, operators can define user-defined ACL types with a custom combination of match fields and actions, and apply ACLs directly to the control plane to protect management protocols like SSH, SNMP, Telnet, NTP, or BGP from unwanted traffic — independent of the data-plane tables used for regular forwarding.
Why ACLs are Beneficial
ACLs give operators fine-grained, line-rate control over exactly what traffic is allowed to move through the network:
· Enforces segmentation without extra hardware: Because filtering happens directly on the switch, department- or tenant-level access boundaries can be enforced in the forwarding path itself, without routing traffic through a dedicated firewall appliance for every internal boundary.
· Matches control to the traffic type: L2, L3, and L3v6 tables cover MAC-, IPv4-, and IPv6-based policies respectively, while dedicated MIRROR and FLOW_CONTROL table types extend the same rule model to traffic capture and rate limiting, so one policy engine covers several common network functions.
· More than pass/drop: Redirect actions can steer matching traffic to a specific interface, next hop, or load-shared next-hop group — useful for service chaining, traffic steering to inspection tools, or policy-based forwarding — without needing a separate routing mechanism.
· Protects the control plane directly: Control-plane ACLs let operators restrict which sources can reach management protocols like SSH or SNMP, reducing exposure to unauthorized management access independent of data-plane security.
· Extensible beyond built-in types: User-defined ACL types let operators build custom match/action combinations for scenarios the standard L2/L3/L3v6 tables don't directly cover, and packet remarking lets DSCP/PCP be rewritten as part of the same rule set used for filtering.
The trade-off is planning discipline: because rule priority determines which of several matching rules wins, and match fields are limited by table type and traffic direction, ACL design benefits from a clear, deliberate structure — a loosely organized rule set is harder to audit and more prone to unintended matches as it grows.