Updated: 4 November 2026 · Applies to: iproute2 6.15 to 6.19 and tcpdump 4.99 on Debian 13, Ubuntu 26.04 and AlmaLinux 10.2 (tested); the same commands on Rocky Linux and RHEL; VirtualBox 7.2 (according to Oracle)

Promiscuous mode makes a network interface pass every frame it receives to the operating system, not only the frames addressed to its own MAC address. Packet capture tools, intrusion detection systems and network bridges use it. On Linux, switch it on with sudo ip link set dev eth0 promisc on and off with sudo ip link set dev eth0 promisc off. In VirtualBox, set the adapter's Promiscuous Mode policy to allow-all (or allow-vms). To keep it on after a reboot, see how to set promiscuous mode permanently.

What promiscuous mode does

In normal mode, a network card keeps the frames sent to its own MAC address, broadcasts and the multicast groups it has joined, and drops the rest before the system sees them. In promiscuous mode it keeps everything that arrives on the wire.

That does not mean you see the whole network. A switch, including the virtual switch or bridge on a VPS host, only sends a port the frames meant for it, plus broadcasts and frames for addresses it has not learned yet. Our bridge test further down shows this. To watch other machines' traffic on a physical network, you need port mirroring on the switch.

1. Switch promiscuous mode on and off in Linux

ip -br link                          # find the interface name
sudo ip link set dev eth0 promisc on
sudo ip link set dev eth0 promisc off

Lab output from a Debian 13 test container after switching it on:

$ ip link show eth0
2: eth0@if1275: <BROADCAST,MULTICAST,PROMISC,UP,LOWER_UP> mtu 1500 qdisc noqueue state UP mode DEFAULT group default qlen 1000
$ ip -d link show eth0 | grep -o "promiscuity [0-9]*"
promiscuity 1

Ubuntu 26.04 and AlmaLinux 10 gave the same result. The old ifconfig eth0 promisc from CentOS guides still works where the net-tools package is installed (ifconfig then lists PROMISC in its flags), but Debian and Ubuntu no longer install net-tools by default, and ip does the same job everywhere. The setting is lost at reboot.

2. Check whether an interface is in promiscuous mode

There are two signs, and you should check both:

  • The PROMISC flag in ip link show: it appears only when promiscuous mode was switched on with ip link (or ifconfig).
  • The promiscuity counter in ip -d link show: it counts everything that currently asks for promiscuous mode, including capture tools. Any value above 0 means the card is promiscuous.

In our test, while tcpdump -i eth0 was running, ip link show eth0 showed no PROMISC flag, yet the counter was 1 and the kernel's flag value in /sys/class/net/eth0/flags changed from 0x1003 to 0x1103 (0x100 is the promiscuous bit). With ip link set ... promisc on and tcpdump at the same time, the counter was 2. So ip link | grep PROMISC misses capture tools. List every promiscuous interface with:

ip -d link show | grep -B1 -E "promiscuity [1-9]"

To see which program is capturing, run sudo ss -0 -p; it lists packet sockets and their owners:

Netid Recv-Q Send-Q Local Address:Port Peer Address:PortProcess
p_raw 0      0                  *:eth0             *     users:(("tcpdump",pid=395,fd=3))

3. Capture tools switch it on for you

sudo tcpdump -i eth0 -n       # promiscuous while it runs
sudo tcpdump -i eth0 -n -p    # -p: capture without promiscuous mode

When tcpdump stopped, the counter went back to 0, and with -p it stayed at 0. In Wireshark, the Promiscuous option in the capture options does the same (according to the Wireshark User's Guide). Each time the interface goes into or out of promiscuous mode, the kernel logs a line saying it entered or left promiscuous mode. These lines are normal; to reduce them, see how to stop repeated promiscuous mode messages.

4. Promiscuous mode in VirtualBox

According to the VirtualBox 7.2 manual, each virtual network adapter has a promiscuous mode policy:

  • deny: hides all traffic not meant for the VM's adapter. This is the default.
  • allow-vms: hides the host's traffic but lets the VM see traffic from and to other VMs.
  • allow-all: no restrictions; the adapter sees all traffic.

The policy applies to the NAT Network, Bridged Adapter, Internal Network and Host-only Adapter modes, not to plain NAT. Set it in the VM's Settings > Network, on the adapter's Promiscuous Mode list, or from the command line (the adapter number goes after promisc):

VBoxManage modifyvm "VM name" --nic-promisc1=allow-all     # VM powered off
VBoxManage controlvm "VM name" nicpromisc1 allow-all        # VM running

The policy only allows it: the guest must still put its own interface into promiscuous mode, for example by running tcpdump or Wireshark inside the VM. The manual also notes that most wireless adapters do not support promiscuous mode, so a Bridged Adapter on the host's Wi-Fi card cannot show other devices' traffic.

5. KVM and Linux bridges

On a KVM host, each VM's network port (for example vnet0) is attached to a Linux bridge, and the bridge switches every port it owns to promiscuous mode by itself. In our test, a port added to a bridge showed promiscuity 1 and flags 0x1303 without any command from us. So "entered promiscuous mode" log lines for vnet or tap interfaces on a KVM host are expected.

Inside a guest, promiscuous mode does not make the bridge send it more. We built a bridge with three ports (vm-a, vm-b, vm-c) in a disconnected test container and sent Ethernet frames from vm-a. Lab results:

  1. To a MAC address the bridge had not learned: seen by vm-b and vm-c (flooded to every port).
  2. To vm-c's MAC, after vm-c had sent a frame: seen only by vm-c.
  3. The same, with vm-b in promiscuous mode: still seen only by vm-c.
  4. The same, after ip link set dev br0 type bridge ageing_time 0: seen by vm-b and vm-c.

With an ageing time of 0, the bridge forgets every address at once and floods all frames to all ports, like a hub. That is a quick way to feed a monitoring VM on a test bridge, but every VM on that bridge then receives everyone's traffic. A new bridge in our test showed ageing_time 30000 in ip -d link show br0; note your value before you change it and set it back when you are done. To build the bridge itself, see set up a network bridge for KVM or add a bridge with nmcli.

Security note

A server interface in promiscuous mode that nobody expects can be a sign that someone is capturing traffic. Check the promiscuity counter and ss -0 -p as shown above, and search the kernel log for when it started: journalctl -k | grep -i promiscuous.

Official documentation: ip-link manual page, tcpdump manual and VirtualBox virtual networking.

Related services

KVM VPS hosting

NVMe SSD and full root access, with Free Basic Managed Support.

Unmanaged dedicated servers

full root access and free IPMI KVM, with Free Basic Managed Support.

Managed dedicated servers

our engineers look after the server for you, from $135 a month.

Was this answer helpful?

Still stuck? Open a ticket