MikroTik RouterOS Firewall Mangle: Packet Marking and Traffic Classification

MikroTik RouterOS Firewall Mangle: Packet Marking and Traffic Classification

RouterOS Version

Primary Version: 7.16 | Minimum: 6.0 | Difficulty: Intermediate | Time: 45 minutes

TL;DR (Quick Start)

For the impatient: here’s the 30-second version.

# Mark web traffic connections and packets for QoS
/ip firewall mangle add chain=forward connection-state=new protocol=tcp dst-port=80,443 action=mark-connection new-connection-mark=web_conn
/ip firewall mangle add chain=forward connection-mark=web_conn action=mark-packet new-packet-mark=web_pkt passthrough=no

Overview

The firewall mangle is RouterOS’s traffic classification engine - it marks packets, connections, and routes for processing by other facilities like queues, NAT, and routing tables. Unlike the firewall filter (which allows or blocks traffic), mangle identifies and categorizes traffic without affecting whether it passes or not.

Understanding mangle is essential because:

• QoS depends on it: Queue trees use packet marks to apply bandwidth policies
• Policy routing needs it: Route marks direct traffic through specific gateways
• Traffic shaping requires it: Connection marks enable per-client bandwidth management
• Advanced NAT uses it: Marked connections can be treated differently by NAT rules

This guide explains how mangle fits into the packet processing pipeline, the three types of marks, and practical patterns for traffic classification.

The Mangle Processing Pipeline

Mangle operates at specific points in the packet flow. Understanding where each chain processes packets is critical for effective traffic marking:

                              ┌─────────────┐
                              │   INPUT     │
                              │   chain     │──→ Local Process
                              └─────────────┘
                                    ↑
Internet ──→ [PREROUTING] ──→ [Routing] ──┤
              chain          Decision     │
                                          ↓
                              ┌─────────────┐
                              │  FORWARD    │
                              │   chain     │
                              └─────────────┘
                                    │
                                    ↓
Local Process ──→ [OUTPUT] ──→ [Routing] ──→ [POSTROUTING] ──→ Internet
                  chain       Decision         chain

Key insight: PREROUTING runs before the routing decision, making it ideal for policy routing marks. FORWARD only processes transit traffic, making it efficient for QoS marking of client traffic.

The Five Mangle Chains

RouterOS has five predefined mangle chains that cannot be deleted. Each processes packets at a specific point:

PREROUTING Chain: First Contact

When: Immediately after packet arrives on interface, before routing decision.

Use cases:

• Policy routing (mark-routing)
• Early traffic classification
• Marking traffic before it splits into INPUT or FORWARD

Example: Mark all traffic from a specific VLAN for a separate routing table:

/ip firewall mangle add chain=prerouting in-interface=vlan100 action=mark-routing new-routing-mark=isp2

INPUT Chain: Traffic to the Router

When: After routing decision, for packets destined to the router itself.

Use cases:

• Mark management traffic for QoS
• Classify services running on the router
• Monitor traffic to router services

Example: Mark DNS queries to the router:

/ip firewall mangle add chain=input protocol=udp dst-port=53 action=mark-packet new-packet-mark=dns_queries

FORWARD Chain: Transit Traffic

When: After routing decision, for packets passing through the router.

Use cases:

• QoS packet marking for client traffic
• Connection marking for bandwidth management
• Traffic classification between networks

Example: Mark HTTP traffic for queue tree:

/ip firewall mangle add chain=forward protocol=tcp dst-port=80 action=mark-packet new-packet-mark=http_traffic

OUTPUT Chain: Router-Generated Traffic

When: After local process generates packet, before routing decision.

Use cases:

• Mark router-generated traffic for policy routing
• Classify traffic from router services
• QoS for router’s own traffic

Example: Mark router’s backup traffic:

/ip firewall mangle add chain=output protocol=tcp dst-port=21 action=mark-routing new-routing-mark=backup_link

POSTROUTING Chain: Final Processing

When: Just before packet leaves the router, after all routing decisions.

Use cases:

• Final packet modifications (DSCP, TTL)
• Marking after NAT has been applied
• Last chance to classify outbound traffic

Example: Set DSCP for all outbound VoIP traffic:

/ip firewall mangle add chain=postrouting protocol=udp dst-port=5060 action=change-dscp new-dscp=46

Understanding Mark Types

Mangle creates three distinct types of marks, each serving different purposes:

Packet Marks: Per-Packet Classification

Scope: Individual packet only Persistence: Transient - exists only while packet is processed Primary use: Queue trees for bandwidth management

/ip firewall mangle add chain=forward protocol=tcp dst-port=80 action=mark-packet new-packet-mark=web_traffic

How queues use packet marks:

/queue tree add name=web_queue parent=global packet-mark=web_traffic max-limit=10M

Important: Packet marks are router-internal only. They are not transmitted across the network and cannot be seen by other devices.

Connection Marks: Efficient Classification

Scope: All packets in a connection Persistence: Stored in connection tracking table Primary use: Reduce CPU by marking once, matching many

The connection marking pattern is essential for performance:

# Rule 1: Mark new connections (inspects headers once)
/ip firewall mangle add chain=forward connection-state=new src-address=192.168.88.0/24 \
    action=mark-connection new-connection-mark=lan_conn

# Rule 2: Mark packets based on connection mark (fast lookup)
/ip firewall mangle add chain=forward connection-mark=lan_conn \
    action=mark-packet new-packet-mark=lan_traffic

Why this matters: Without connection marks, every packet requires header inspection. With connection marks, only the first packet of each connection is fully analyzed - subsequent packets use a fast connection table lookup.

Routing Marks: Policy Routing

Scope: Packet’s routing decision Persistence: Until routing decision is made Primary use: Direct traffic through specific routing tables

# RouterOS v7: Create routing table first (required)
/routing table add name=via_isp2 fib

# Mark traffic for secondary ISP (exclude local traffic!)
/ip firewall mangle add chain=prerouting src-address=192.168.100.0/24 \
    dst-address-type=!local action=mark-routing new-routing-mark=via_isp2

# Route marked traffic through ISP2
/ip route add dst-address=0.0.0.0/0 gateway=10.0.0.1 routing-table=via_isp2

Critical notes:

• RouterOS v7 requires pre-created routing tables: Create with /routing table add name=myTable fib before using mark-routing
• Always exclude local traffic: Use dst-address-type=!local to prevent breaking router’s own connectivity
• FastTrack incompatible: Traffic with routing marks cannot use FastTrack

The Passthrough Parameter

The passthrough parameter controls whether rule processing continues after a match:

passthrough=yes (default): Continue to next rule after action passthrough=no: Stop processing after action

When to Use passthrough=yes

Use when you need multiple marks on the same packet:

# Mark connection first (passthrough=yes by default)
/ip firewall mangle add chain=forward connection-state=new src-address=192.168.88.0/24 \
    action=mark-connection new-connection-mark=internal_conn

# Then mark packet (same packet, second mark)
/ip firewall mangle add chain=forward connection-mark=internal_conn \
    action=mark-packet new-packet-mark=internal_pkt passthrough=no

When to Use passthrough=no

Use when the mark is final and no further processing is needed:

# Final classification - stop processing
/ip firewall mangle add chain=forward protocol=tcp dst-port=22 \
    action=mark-packet new-packet-mark=ssh_traffic passthrough=no

Performance tip: Use passthrough=no on your final rules to avoid unnecessary rule traversal.

Mangle Actions Reference

Action	Description	Use Case
mark-packet	Add packet mark	Queue tree classification
mark-connection	Add connection mark	Efficient per-connection classification
mark-routing	Add routing mark	Policy-based routing
change-dscp	Modify DSCP field (0-63)	QoS marking for upstream devices
change-ttl	Modify TTL value	Prevent traceroute, hide hops
change-mss	Modify Maximum Segment Size	Fix path MTU discovery issues
set-priority	Set VLAN/WMM priority	Layer 2 QoS
clear-df	Remove Don’t Fragment flag	Allow fragmentation
sniff-tzsp	Export to TZSP receiver	Packet capture to Wireshark

Configuration Steps

This section provides a practical configuration that demonstrates the core mangle concepts.

Step 1: Create Connection Marks for Traffic Classification

Mark connections by traffic type for efficient processing:

/ip firewall mangle
add chain=forward connection-state=new protocol=tcp dst-port=80,443 \
    action=mark-connection new-connection-mark=web_conn comment="Mark web connections"
add chain=forward connection-state=new protocol=udp dst-port=53 \
    action=mark-connection new-connection-mark=dns_conn comment="Mark DNS connections"
add chain=forward connection-state=new protocol=tcp dst-port=22 \
    action=mark-connection new-connection-mark=ssh_conn comment="Mark SSH connections"

Step 2: Create Packet Marks from Connection Marks

Convert connection marks to packet marks for queue trees:

/ip firewall mangle
add chain=forward connection-mark=web_conn action=mark-packet \
    new-packet-mark=web_pkt passthrough=no comment="Web packet mark"
add chain=forward connection-mark=dns_conn action=mark-packet \
    new-packet-mark=dns_pkt passthrough=no comment="DNS packet mark"
add chain=forward connection-mark=ssh_conn action=mark-packet \
    new-packet-mark=ssh_pkt passthrough=no comment="SSH packet mark"

Step 3: Use Packet Marks in Queue Tree (Optional)

Apply bandwidth policies based on marks:

/queue tree
add name=download parent=global max-limit=100M
add name=web_download parent=download packet-mark=web_pkt max-limit=50M
add name=dns_download parent=download packet-mark=dns_pkt max-limit=5M priority=1
add name=ssh_download parent=download packet-mark=ssh_pkt max-limit=10M priority=2

Verification

Confirm your mangle rules are active and processing traffic.

Check 1: Verify Rules Are Created

/ip firewall mangle print

Expected Output:

Flags: X - disabled, I - invalid, D - dynamic
 #   CHAIN    ACTION          CONNECTION-MARK  NEW-CONN-MARK  NEW-PKT-MARK
 0   forward  mark-connection                  web_conn
 1   forward  mark-connection                  dns_conn
 2   forward  mark-connection                  ssh_conn
 3   forward  mark-packet     web_conn                        web_pkt
 4   forward  mark-packet     dns_conn                        dns_pkt
 5   forward  mark-packet     ssh_conn                        ssh_pkt

Check 2: Monitor Rule Statistics

/ip firewall mangle print stats

Expected Output:

 #   CHAIN    ACTION           BYTES        PACKETS
 0   forward  mark-connection  0            0
 1   forward  mark-connection  0            0
 2   forward  mark-connection  0            0
 3   forward  mark-packet      1,234,567    8,901
 4   forward  mark-packet      45,678       234
 5   forward  mark-packet      98,765       543

Check 3: Verify Connection Marks in Tracking Table

/ip firewall connection print where connection-mark!=""

Expected Output:

 #   PROTOCOL  SRC-ADDRESS        DST-ADDRESS        CONNECTION-MARK
 0   tcp       192.168.88.10:4521 93.184.216.34:443  web_conn
 1   udp       192.168.88.10:5353 8.8.8.8:53         dns_conn

Common Patterns

Pattern 1: Per-Client Bandwidth Management

Mark each client’s traffic separately for individual queue limits:

# Mark connections per client
/ip firewall mangle
add chain=forward connection-state=new src-address=192.168.88.10 \
    action=mark-connection new-connection-mark=client10_conn
add chain=forward connection-state=new src-address=192.168.88.11 \
    action=mark-connection new-connection-mark=client11_conn

# Mark packets for queuing
add chain=forward connection-mark=client10_conn action=mark-packet \
    new-packet-mark=client10_pkt passthrough=no
add chain=forward connection-mark=client11_conn action=mark-packet \
    new-packet-mark=client11_pkt passthrough=no

Pattern 2: Dual-WAN Load Balancing

Mark traffic for different ISPs based on source:

# Mark traffic from different subnets for different ISPs
/ip firewall mangle
add chain=prerouting src-address=192.168.10.0/24 action=mark-routing \
    new-routing-mark=to_isp1
add chain=prerouting src-address=192.168.20.0/24 action=mark-routing \
    new-routing-mark=to_isp2

# Create routing table entries
/ip route
add dst-address=0.0.0.0/0 gateway=10.0.1.1 routing-mark=to_isp1
add dst-address=0.0.0.0/0 gateway=10.0.2.1 routing-mark=to_isp2

Pattern 3: MSS Clamping for PPPoE/VPN

Fix path MTU discovery issues by clamping TCP MSS for tunneled connections:

# Clamp MSS for outbound traffic through PPPoE
/ip firewall mangle add chain=forward protocol=tcp tcp-flags=syn \
    out-interface=pppoe-out tcp-mss=1301-65535 action=change-mss new-mss=1300 \
    passthrough=yes comment="Clamp MSS for PPPoE outbound"

# Clamp MSS for inbound traffic through PPPoE
/ip firewall mangle add chain=forward protocol=tcp tcp-flags=syn \
    in-interface=pppoe-out tcp-mss=1301-65535 action=change-mss new-mss=1300 \
    passthrough=yes comment="Clamp MSS for PPPoE inbound"

Why this works: PPPoE adds 8 bytes of overhead, reducing effective MTU. When PMTUD fails (ICMP blocked), TCP connections stall with large packets. MSS clamping forces smaller TCP segments, avoiding fragmentation issues.

Pattern 4: DSCP Marking for Upstream QoS

Set DSCP values for traffic that will be prioritized by upstream devices:

/ip firewall mangle
add chain=postrouting protocol=udp dst-port=5060-5061 action=change-dscp \
    new-dscp=46 comment="VoIP signaling - EF"
add chain=postrouting protocol=udp dst-port=10000-20000 action=change-dscp \
    new-dscp=46 comment="VoIP RTP - EF"
add chain=postrouting protocol=tcp dst-port=22 action=change-dscp \
    new-dscp=26 comment="SSH - AF31"

Pattern 5: Bypass FastTrack for Marked Traffic

When using policy routing, ensure marked traffic bypasses FastTrack:

# FastTrack unmarked traffic only
/ip firewall filter
add chain=forward connection-state=established,related routing-mark="" \
    action=fasttrack-connection
add chain=forward connection-state=established,related action=accept

Troubleshooting

Problem: “Mangle rules show zero packet counts”

Cause: Rules may be unreachable, or traffic doesn’t match criteria.

Solution:

1. Check rule order - earlier rules may match first with passthrough=no
2. Verify traffic exists: /ip firewall connection print
3. Check chain selection - use forward for transit traffic, not input
4. Ensure connection tracking is enabled: /ip firewall connection tracking print

Problem: “Connection marks not appearing in connection table”

Cause: Rule may be in wrong chain or connection-state not specified.

Solution:

1. For new connection marks, use connection-state=new
2. Use prerouting or forward chain, not postrouting
3. Verify rule matches: add log=yes temporarily to debug

Problem: “Policy routing marks don’t affect traffic”

Cause: Routing table or routes may be misconfigured.

Solution:

1. Verify routing mark is set in prerouting chain (before routing decision)
2. Check route exists with matching routing-mark: /ip route print where routing-mark=your_mark
3. Ensure FastTrack isn’t processing marked connections
4. Verify with: /ip firewall mangle print stats

Problem: “Packet mark limit exceeded” error

Cause: RouterOS has a limit of 4096 unique packet marks.

Solution:

1. Use connection marks instead of packet marks where possible
2. Consolidate similar traffic under fewer marks
3. Review if all marks are necessary
4. Check for dynamic mark generation that may be creating excessive marks

Problem: “Queue tree doesn’t apply to marked traffic”

Cause: Packet marks may not match queue configuration.

Solution:

1. Verify exact packet mark name in both mangle and queue
2. Check queue tree parent interface matches traffic path
3. Ensure passthrough=no on final packet mark rule
4. Verify marks are created: /ip firewall mangle print stats

Performance Considerations

Use Connection Marks for Efficiency

Wrong approach (CPU intensive):

# Every packet inspected
/ip firewall mangle add chain=forward src-address=192.168.88.0/24 \
    action=mark-packet new-packet-mark=lan_pkt

Right approach (efficient):

# First packet inspected, connection marked
/ip firewall mangle add chain=forward connection-state=new src-address=192.168.88.0/24 \
    action=mark-connection new-connection-mark=lan_conn
# Subsequent packets use fast connection lookup
/ip firewall mangle add chain=forward connection-mark=lan_conn \
    action=mark-packet new-packet-mark=lan_pkt

Rule Order Matters

1. Place most frequently matched rules at the top
2. Use passthrough=no when no further matching is needed
3. Mark connections first, then derive packet marks
4. Group related rules together to improve cache efficiency

FastTrack Interaction

FastTrack bypasses firewall processing for marked connections. This is incompatible with:

• Policy routing (routing marks)
• Queue trees requiring packet marks
• Any mangle processing of established connections

To use mangle with FastTrack, exclude marked traffic:

/ip firewall filter add chain=forward connection-state=established,related \
    connection-mark="" action=fasttrack-connection

Common Mistakes

Mistake 1: Using Wrong Chain

Wrong: Marking transit traffic in INPUT chain

# INPUT is for traffic TO the router, not through it
/ip firewall mangle add chain=input src-address=192.168.88.0/24 action=mark-packet new-packet-mark=client_pkt

Right: Use FORWARD for transit traffic

/ip firewall mangle add chain=forward src-address=192.168.88.0/24 action=mark-packet new-packet-mark=client_pkt

Mistake 2: Routing Marks in Wrong Chain

Wrong: Setting routing mark after routing decision

# Routing decision already made - mark has no effect
/ip firewall mangle add chain=forward action=mark-routing new-routing-mark=isp2

Right: Set routing marks in PREROUTING

# Before routing decision - mark affects route selection
/ip firewall mangle add chain=prerouting action=mark-routing new-routing-mark=isp2

Mistake 3: Forgetting passthrough=no

Problem: Packet gets marked multiple times, wasting CPU

# All rules process because passthrough=yes (default)
/ip firewall mangle add chain=forward protocol=tcp dst-port=80 action=mark-packet new-packet-mark=web
/ip firewall mangle add chain=forward protocol=tcp dst-port=443 action=mark-packet new-packet-mark=https
/ip firewall mangle add chain=forward action=mark-packet new-packet-mark=other

Solution: Add passthrough=no to terminal rules

/ip firewall mangle add chain=forward protocol=tcp dst-port=80 action=mark-packet new-packet-mark=web passthrough=no
/ip firewall mangle add chain=forward protocol=tcp dst-port=443 action=mark-packet new-packet-mark=https passthrough=no
/ip firewall mangle add chain=forward action=mark-packet new-packet-mark=other passthrough=no

Mistake 4: Not Using Connection Marks

Inefficient: Inspecting every packet

/ip firewall mangle add chain=forward src-address=192.168.88.100 action=mark-packet new-packet-mark=client100

Efficient: Mark connection once, then match connection mark

/ip firewall mangle add chain=forward connection-state=new src-address=192.168.88.100 \
    action=mark-connection new-connection-mark=client100_conn
/ip firewall mangle add chain=forward connection-mark=client100_conn \
    action=mark-packet new-packet-mark=client100

Related Topics

Prerequisites

• IP Address Configuration - interface addressing fundamentals
• Static Routes - routing concepts for policy routing

Related Firewall Topics

• Firewall Filter Basics - packet filtering (complementary to mangle)
• NAT Masquerade - NAT works with mangle marks
• Address Lists - address groups for mangle matching

Uses Mangle Marks

• Simple Queues - bandwidth limiting
• Queue Tree - advanced QoS using packet marks

Multi-WAN

• DHCP Client - multi-WAN with mangle-based load balancing

References

• MikroTik Mangle Documentation
• Packet Flow in RouterOS
• Firewall Overview
• Queue Trees Documentation