DHCP port explained: which ports it uses and what they do
DHCPv4 uses UDP port 67 on the server side and UDP port 68 on the client side. If you are opening a firewall rule or figuring out why a lease is being blocked, those are the two ports to check first. DHCPv6 uses different ports, so IPv4 firewall rules do not carry over to IPv6 networks.
What port does DHCP use?
For DHCPv4, the standard ports are UDP 67 for the server and UDP 68 for the client. That split is fixed: the DHCP server listens on port 67, and the DHCP client listens on port 68.
That is why people often say “DHCP uses ports 67 and 68” instead of one port number. DHCP is a client/server protocol, and each side has its own UDP port so the exchange can be handled correctly.
For example, if a laptop is asking for an address on an IPv4 network, its DHCP client sends from UDP 68 and targets the server side on UDP 67. The server replies from UDP 67 back to UDP 68. That is the normal IPv4 pattern.
Why DHCP needs two UDP ports
DHCP uses two ports because the client and the server do different jobs. The server has to hear requests from many clients, and the client has to receive the reply that belongs to its own request. UDP keeps the exchange lightweight, and the port split keeps the direction clear.
Say a laptop joins a wired network and has no address yet. It starts by sending a DHCP discovery message. At that point, it may not know the server’s address, and in many cases it cannot even talk unicast to a specific host yet. So it sends a broadcast on the local network, using the DHCP client port. Any DHCP server that hears it can answer.
That broadcast behavior matters for firewalls and switches. A rule that only allows unicast traffic to a known server can still fail if it blocks broadcasts on the local segment, or if a relay agent is not in place to forward the request between subnets. If the client never gets past the first broadcast exchange, it never reaches the lease step.
When this matters in real networks
This comes up any time DHCP traffic crosses a boundary you control. The obvious cases are firewall rules and router ACLs, but it also matters in VLAN setups, DHCP relay configurations, and troubleshooting a client that keeps failing to lease an address.
Common places to check:
- Host firewalls on DHCP servers or relay systems
- Network firewalls between VLANs or subnets
- Router or Layer 3 switch DHCP relay settings
- Access switch features that block broadcasts or helper traffic
- Security rules on virtual networks and cloud subnets
If the client and server are on the same subnet, the traffic is usually simple: broadcast discovery, server response, lease offered. If they are on different subnets, a relay agent typically forwards the client’s broadcast as a unicast request to the DHCP server. In that case, the server still speaks on UDP 67, but the path may pass through relay infrastructure instead of staying local.
So if DHCP fails, do not just check whether “port 67 is open” on the server. Make sure the network path allows the full exchange, including broadcast handling where needed and relay forwarding where subnets are separated.
DHCPv4 vs DHCPv6 ports
DHCPv6 does not use the same ports as DHCPv4. That is a common source of mistakes in mixed networks, especially when someone copies an IPv4 firewall rule and expects it to work for IPv6 too.
For DHCPv6, the standard UDP ports are 546 for the client and 547 for the server. So the direction flips in the same general way: the client listens on 546 and the server on 547.
That means you should treat IPv4 and IPv6 separately when you write firewall rules or check packet captures. If an IPv4 lease works but IPv6 addressing does not, the problem may be that UDP 546/547 is blocked even though UDP 67/68 is allowed. The reverse can happen too.
In a dual-stack network, verify both protocol families. DHCPv4 and DHCPv6 solve related problems, but they are not the same service and they do not use the same port numbers.
Common mistakes people make
A few mix-ups come up over and over.
First, DHCP is UDP, not TCP. If you only open TCP ports, DHCP will still fail. That is a common firewall mistake because some admins remember the service name but not the transport protocol.
Second, people sometimes open only one side of the conversation. For IPv4, you need UDP 67 and UDP 68 to work in the right directions. If a rule allows replies but blocks client broadcasts, the lease still fails.
Third, DHCP is often confused with DNS because both are common infrastructure services and both can fail during network setup. DNS usually uses port 53, not 67 or 68, and it serves a different job: name lookup, not address assignment.
Fourth, some people think BOOTP is a separate port problem. DHCP evolved from BOOTP and still uses the same basic UDP port pair for IPv4, so BOOTP-style thinking can help explain the history, but the ports you configure for DHCPv4 are still 67 and 68.
If a packet capture shows traffic on UDP 67/68 and the client still does not get an address, the problem is probably not the port number itself. It is more likely a blocked broadcast, a missing relay, or a policy somewhere else in the path.
Quick check if DHCP still is not working
Start with the simplest path first. If the client is on the same subnet as the server, check that UDP 67 and UDP 68 are allowed in the right direction and that nothing on the local network is dropping broadcasts.
Then check whether a relay agent is supposed to forward requests between subnets. If the client lives on one VLAN and the server on another, the relay or helper configuration has to be correct or the server may never see the request.
A practical checklist:
- Confirm DHCPv4 allows UDP 67 and UDP 68
- Confirm DHCPv6 allows UDP 546 and UDP 547 if IPv6 is involved
- Check the relay/helper address on the router or Layer 3 switch
- Make sure the client is actually on the expected VLAN or subnet
- Verify the server is listening and has available leases
- Test from a client on the same subnet as the server before blaming routing
If the local test works but remote clients fail, the problem is usually in the relay path or a firewall between subnets. If even local clients fail, focus on the server host firewall, the service status, and whether broadcasts are being filtered before they reach the server.
Conclusion
If you just need the firewall answer, remember this: DHCPv4 uses UDP 67 on the server and UDP 68 on the client, and DHCPv6 uses UDP 547 on the server and UDP 546 on the client. From there, check whether the traffic is local broadcast or being relayed across subnets, because that usually explains why a rule that looks right still does not work. If you are troubleshooting, capture the traffic and verify that the request leaves the client, reaches a relay or server, and gets a reply back.
Frequently Asked Questions
Does DHCP use TCP or UDP?
DHCP uses UDP, not TCP. If you open only TCP ports, DHCP traffic will still be blocked.
Why do people say DHCP uses ports 67 and 68?
Because DHCPv4 uses UDP 67 for the server side and UDP 68 for the client side. Both ports are part of the normal exchange.
What ports does DHCPv6 use?
DHCPv6 uses UDP 546 on the client and UDP 547 on the server.
Is DHCP the same as DNS?
No. DNS usually uses port 53 and maps names to addresses. DHCP assigns network configuration such as IP addresses.