SYSTEMD.N...ING-SCHEME(7)

📦
version 262.1 · 16 versions
, in arch rolling

🌸 SYSTEMD.NET-NAMING-SCHEME(7) 🌸

okay bestie, welcome to the manual page for systemd.net-naming-scheme — aka Network device naming schemes. this is the one that explains why your ethernet card is called enp3s0 and not eth0 like it's 2009. strap in 💅

✨ NAME ✨

systemd.net-naming-scheme — Network device naming schemes

🦋 DESCRIPTION 🦋

so, here's the tea: network interface names and MAC addresses can be generated based on certain stable interface attributes. that's only possible when there's enough info about the device to generate those attributes and the use of that info is configured. this page is all about interface naming — like, what possible names may be generated. those names get generated by the systemd-udevd.service(8) builtin net_id and exported as udev(7) properties: ID_NET_NAME_ONBOARD=, ID_NET_LABEL_ONBOARD=, ID_NET_NAME_PATH=, ID_NET_NAME_SLOT=.

Names and MAC addresses are derived from various stable device metadata attributes. Newer versions of systemd-udevd take more of these attributes into account, which improves (and thus possibly changes) the names and addresses used for the same devices. different versions of those generation rules are called "naming schemes". the default naming scheme is chosen at compilation time. usually that's the latest implemented version, but you can also set one of the older versions to preserve compatibility. this is genuinely useful, for example for distributions, which may introduce new versions of systemd in stable releases without changing the naming scheme. the naming scheme may also be overridden using the net.naming_scheme= kernel command line switch — see systemd-udevd.service(8). available naming schemes are described below 👇

once the udev properties have been generated, appropriate udev rules may be used to actually rename devices based on those properties. see the description of NamePolicy= and MACAddressPolicy= in systemd.link(5).

🫶 note that while the concept of network interface naming schemes is primarily relevant in the context of systemd-udevd.service, the systemd-nspawn(1) container manager also takes it into account when naming network interfaces — see below.

👑 POLICIES 👑

all names start with a two-character prefix that signifies the interface type. it's giving very "hello my name is" energy.

⭐ Table 1. Two character prefixes based on the type of interface

  • en — Ethernet
  • ib — InfiniBand
  • sl — Serial line IP (slip)
  • wl — Wireless local area network (WLAN)
  • ww — Wireless wide area network (WWAN)
  • mc — Management Component Transport Protocol (MCTP)

the udev(7) net_id builtin exports the following device properties, and yes we're going through all of them 💖

🎀 ID_NET_NAME_ONBOARD=prefixonumber, ID_NET_NAME_ONBOARD=prefixdnumber

This name is set based on the numeric ordering information given by the firmware for on-board devices. Different schemes are used depending on the firmware type, as described in the table below.

Table 2. On-board naming schemes

  • prefixonumber — PCI on-board index
  • prefixdnumber — DeviceTree alias index

Added in version 243.

🌸 ID_NET_LABEL_ONBOARD=label

This property is set based on textual label given by the firmware for on-board devices. This is only available for PCI devices.

Added in version 243.

💜 ID_NET_NAME_MAC=prefixxAABBCCDDEEFF

This name consists of the prefix, letter x, and 12 hexadecimal digits of the MAC address. It is available if the device has a fixed MAC address. Because this name is based on an attribute of the card itself, it remains "stable" when the device is moved (even between machines), but will change when the hardware is replaced. so like — it'll follow your NIC across machines, but the second you swap the card, new name, no cap.

Added in version 243.

🌈 ID_NET_NAME_SLOT=prefix[Pdomain]sslot[ffunction][nport_name|ddev_port], ID_NET_NAME_SLOT=prefixvslot, ID_NET_NAME_SLOT=prefixxslot, ID_NET_NAME_SLOT=prefix[Pdomain]sslot[ffunction][nport_name|ddev_port]bnumber, ID_NET_NAME_SLOT=prefix[Pdomain]sslot[ffunction][nport_name|ddev_port]uport...[cconfig][iinterface], ID_NET_NAME_SLOT=prefix[Pdomain]sslot[ffunction][nport_name|ddev_port]vslot, ID_NET_NAME_SLOT=prefix[Pdomain]sslot[ffunction][nport_name|ddev_port]rslot, ID_NET_NAME_SLOT=prefix[Pdomain]sslot[ffunction][nport_name|ddev_port]Ssfnum

This property describes the slot position. Different schemes are used depending on the bus type, as described in the table below. In case of USB, BCMA, and SR-VIO devices, the full name consists of the prefix, PCI slot identifier, and USB or BCMA or SR-VIO slot identifier. The first two parts are denoted as ... in the table below.

Table 3. Slot naming schemes

  • prefix [Pdomain] sslot [ffunction] [nport_name | ddev_port] — PCI slot number
  • prefix vslot — VIO slot number (IBM PowerVM)
  • prefix Xnumber — VIF interface number (Xen)
  • ... bnumber — Broadcom bus (BCMA) core number
  • ... uport... [cconfig] [iinterface] — USB port number chain
  • ... vslot — SR-VIO slot number
  • ... rslot — SR-IOV slot number
  • ... Ssfnum — Auxiliary sub-function (SF) number

okay, now the fine print because the details matter:

  • The PCI domain is only prepended when it is not 0.
  • All multi-function PCI devices will carry the ffunction number in the device name, including the function 0 device.
  • For non-multi-function devices, the number is suppressed if 0.
  • The port name port_name is used, or the port number ddev_port if the name is not known.
  • For BCMA devices, the core number is suppressed when 0.
  • For USB devices the full chain of port numbers of hubs is composed. If the name gets longer than the maximum number of 15 characters, the name is not exported. The usual USB configuration number 1 and interface number 0 values are suppressed.
  • SR-IOV virtual devices are named based on the name of the parent interface, with a suffix of v and the virtual device number, with any leading zeros removed. The bus number is ignored.

to be continued bestie, don't go anywhere 💅✨

SR-IOV virtual device representors get their names from the name of the physical device interface, with a suffix of r plus the number of the virtual device linked to that particular representor, and any leading zeros get yeeted. The physical port name and the bus number are just ignored, periodt. 🙈

Auxiliary sub-function (SF) network devices — where the network device's parent is an auxiliary device exposing a sfnum sysfs attribute (currently mlx5_core SFs, iykyk) — are named based on the underlying PCI function (the PF, or for VF-SF the PF behind the VF), with a suffix of S plus the user-defined sub-function number from sfnum. It's literally analogous to how SR-IOV virtual function devices get named with a v suffix. ✨

If the SF's parent PCI function is itself an SR-IOV virtual function (VF-SF), the name is rooted at the PF and both suffixes are chained together, with the v suffix coming before the S suffix (e.g. "enp193s0f0v0S88"). So the PF, the VF, and the SF form a stable, hierarchical sequence no matter what the VF's underlying PCI bus/device/function numbering is doing. 🎀

In some configurations a parent PCI bridge of a given network controller may be associated with a slot. In that case we do not generate this device property, so we can avoid possible naming conflicts. 🫶

Added in version 243.

🌸 ID_NET_NAME_PATH — the whole roster 🌸

ID_NET_NAME_PATH=prefixcbus_id
ID_NET_NAME_PATH=prefixavendormodeliinstance
ID_NET_NAME_PATH=prefixiaddressnport_name
ID_NET_NAME_PATH=prefixuport...
ID_NET_NAME_PATH=prefix[Pdomain]pbussslot[ffunction][nphys_port_name|ddev_port]
ID_NET_NAME_PATH=prefix[Pdomain]pbussslot[ffunction][nphys_port_name|ddev_port]bnumber
ID_NET_NAME_PATH=prefix[Pdomain]pbussslot[ffunction][nphys_port_name|ddev_port]uport...[cconfig][iinterface]
ID_NET_NAME_PATH=prefix[Pdomain]pbussslot[ffunction][nphys_port_name|ddev_port]Ssfnum

This property describes the device installation location — where the thing physically lives, basically. Different schemes get used depending on the bus type, and they're all laid out in the table below. 👇

For BCMA and USB devices, the PCI path information must be known, and the full name is made up of the prefix, the PCI slot identifier, and the USB or BCMA location. Those first two parts show up as ... in the table below, so don't be confused bestie. 💅

📍 Table 4. Path naming schemes

  • prefix cbus_id — CCW or grouped CCW device identifier
  • prefix avendor model iinstance — ACPI path names for ARM64 platform devices
  • prefix iaddress nport_name — Netdevsim (simulated networking device) device number and port name
  • prefix [Pdomain] pbus sslot [ffunction] [nphys_port_name | ddev_port] — PCI geographical location
  • ... bnumber — Broadcom bus (BCMA) core number
  • ... uport... [cconfig] [iinterface] — USB port number chain
  • ... Ssfnum — Auxiliary sub-function (SF) number

CCW and grouped CCW devices are found in IBM System Z mainframes, iconic honestly. Any leading zeros and dots are suppressed. 🦋

For PCI, BCMA, and USB devices, the same rules as described above for slot naming are used.

Added in version 243.

✨ HISTORY ✨

okay bestie, storytime 💖 here's the full lore of every "naming scheme" that's ever been defined — and yes, you can pick one of these at system boot-up time with the net.naming_scheme= kernel command line switch (the one from up above, iykyk):

v238

This is the naming scheme that was implemented in systemd 238.

Added in version 243.

v239

Naming got a glow-up for virtual network interfaces created with SR-IOV and NPAR, and for devices where the PCI network controller device doesn't have a slot number associated. 💅

SR-IOV virtual devices are now named based on the name of the parent interface, with a suffix of vport, where port is the virtual device number. Before this, those virtual devices were named as if they were completely independent. not it pretending it had no parents 😭

The ninth and later NPAR virtual devices are named following the scheme used for the first eight NPAR partitions. Previously those devices were not renamed and the kernel default (ethN) was used.

Names are also generated for PCI devices where the PCI network controller device does not have an associated slot number itself, but one of its parents does. Previously those devices were not renamed and the kernel default was used.

Added in version 243.

v240

The ib prefix and stable names for infiniband devices are introduced ✨ Previously those devices were not renamed.

The ACPI index field (used in ID_NET_NAME_ONBOARD=) is now also used when 0. (yes, even zero gets to be the main character now 👑)

A new naming policy NamePolicy=keep was introduced. With this policy, if the network device name was already set by userspace, the device will not be renamed again. Previously, this naming policy applied implicitly, and now it must be explicitly requested. Effectively, this means that network devices will be renamed according to the configuration, even if they have been renamed already, if keep is not specified as the naming policy in the .link file. See systemd.link(5) for a description of NamePolicy=.

Added in version 243.

v241

MACAddressPolicy=persistent was extended to set MAC addresses based on the device name. Previously addresses were only based on the ID_NET_NAME_* attributes, which meant that interface names would never be generated for virtual devices. Now a persistent address will be generated for most devices, including in particular bridges. 🌈

Note: when userspace does not set a MAC address for a bridge device, the kernel will initially assign a random address, and then change it when the first device is enslaved to the bridge. With this naming policy change, bridges get a persistent MAC address based on the bridge name instead of the first enslaved device. no cap, the bridge finally got its own identity 💖

Added in version 243.

v243

Support for renaming netdevsim (simulated networking) devices was added. Previously those devices were not renamed.

Previously two-letter interface type prefix was prepended to ID_NET_LABEL_ONBOARD=. This is not done anymore. (we said okay, we're over it 🫶)

Added in version 243.

v245

When systemd-nspawn(1) derives the name for the host side of the network interface created with --network-veth from the container name, it previously simply truncated the result at 15 characters if longer (since that's the maximum length for network interface names). From now on, for any interface name that would be longer than 15 characters the last 4 characters are set to a 24bit hash value of the full interface name. This way network interface name collisions between multiple similarly named containers (who only differ in container name suffix) should be less likely (but still possible, since the 24bit hash value is very small). like, it's giving effort but it's still not a guarantee bestie 💀

Added in version 245.

v247

When a PCI slot is associated with a PCI bridge that has multiple child network controllers, the same value of the ID_NET_NAME_SLOT property might be derived for those controllers. This would cause a naming conflict if the property is selected as the device name. Now, we detect this situation and do not produce the ID_NET_NAME_SLOT property. caught it, fixed it, iconic ✨

Added in version 247.

v249

PCI hotplug slot names for the s390 PCI driver are a hexadecimal representation of the function_id device attribute. This attribute is now used to build the ID_NET_NAME_SLOT. Before that, all slot names were parsed as decimal numbers, which could either result in an incorrect value of the ID_NET_NAME_SLOT property or none at all. 💅

Some firmware and hypervisor implementations report unreasonably high numbers for the on-board index. To prevent the generation of bogus on-board interface names, index numbers greater than 16381 (2¹⁴-1) were ignored. For s390 PCI devices index values up to 65535 (2¹⁶-1) are valid. To account for that, the limit was increased to 65535.

The udev(7) rule NAME= replaces ":", "/", and "%" with an underscore ("_"), and refuses strings which contain only digits.

Added in version 249.

v250

Added naming scheme for Xen netfront "vif" interfaces based on the guest side VIF number set from the Xen config (or the interface index in AWS EC2). 🦋

Added in version 250.

v251

Since version v247 we no longer set ID_NET_NAME_SLOT if we detect that a PCI device associated with a slot is a PCI bridge, as that would create naming conflict when there are more child devices on that bridge. Now, this is relaxed and we will use slot information to generate the name based on it, but only if the PCI device has multiple functions. This is safe because distinct function number is a part of the device name for multifunction devices. Note, this is reverted in v255. See below. 🙈

Added in version 251.

v252

Added naming scheme for platform devices with devicetree aliases. 🌸

Added in version 252.

v253

Set ID_NET_NAME_PATH for usb devices not connected via a PCI bus.

Added in version 253.

v254

Naming was changed for SR-IOV virtual device representors, optionally settable at compilation time. The "rslot" suffix was added to differentiate SR-IOV virtual device representors attached to a single physical device interface. Because of a mistake, this scheme was not the default scheme for systemd version 254. 😭 (we fumbled, bestie)

Added in version 255.

v255

Naming was changed for SR-IOV virtual device representors to enable the change introduced in v254 by default.

If we detect that a PCI device associated with a slot is a PCI bridge, we no longer set ID_NET_NAME_SLOT, reverting a change that was introduced in v251.

Added in version 255.

v257

PCI slot number is now read from firmware_node/sun sysfs file.

The naming scheme based on devicetree aliases was extended to support aliases for individual interfaces of controllers with multiple ports. 👑

Added in version 257.

v258

When no renaming is requested, ID_NET_NAME property is now equivalent to INTERFACE property.

Added in version 258.

v259

The naming scheme based on devicetree aliases was extended to support WLAN devices. ✨

Added in version 259.

v260

MCTP interfaces are now assigned persistent names, using a mc prefix.

Added in version 260.

v261

Stable names are now generated for auxiliary sub-function (SF) network devices (such as "mlx5_core" SFs). The name is built from the parent PCI Physical Function's path with an "Ssfnum" suffix, where sfnum is the user-defined SF number (the value passed to devlink port add ... sfnum N, exposed by the kernel as the sfnum sysfs attribute on the auxiliary device). This is analogous to the "vN" suffix used for SR-IOV virtual functions; for SFs hosted on SR-IOV VFs (VF-SF), the two suffixes are chained on top of the PF's base name. it's giving family tree energy 🌈

Added in version 261.

Note that latest may be used to denote the latest scheme known (to this particular version of systemd).

✨ LIMITING THE USE OF SPECIFIC SYSFS ATTRIBUTES ✨

When creating names for network cards, some naming schemes use data from sysfs populated by the kernel. This means that although a specific naming scheme in udev(7) is picked, the network card's name can still change when a new kernel version adds a new sysfs attribute. For example, if kernel starts setting the phys_port_name, udev will append the "nphys_port_name" suffix to the device name. 😭 like okay, plot twist

  • ID_NET_NAME_ALLOW=BOOL — This udev property sets a fallback policy for reading a sysfs attribute. If set to 0 udev will not read any sysfs attribute by default, unless it is explicitly allowlisted, see below. If set to 1 udev can use any sysfs attribute unless it is explicitly forbidden. The default value is 1.

Added in version 256.

  • ID_NET_NAME_ALLOW_sysfsattr=BOOL — This udev property explicitly states if udev shall use the specified sysfsattr, when composing the device name.

Added in version 256.

With these options, users can set an allowlist or denylist for sysfs attributes. To create an allowlist, the user needs to set ID_NET_NAME_ALLOW=0 for the device and then list the allowed attributes with the ID_NET_NAME_ALLOW_sysfsattr=1 options. In case of a denylist, the user needs to provide the list of denied attributes with the ID_NET_NAME_ALLOW_sysfsattr=0 options. pick your fighter: allowlist or denylist 💅

💖 EXAMPLES 💖

ok bestie, this is the part where it all clicks. real devices, real names, real udevadm output. no cap, these are literally the receipts. 🧾✨

🌸 Example 1. Using udevadm test-builtin to display device properties

$ udevadm test-builtin net_id /sys/class/net/enp0s31f6
...
Using default interface naming scheme 'v243'.
ID_NET_NAMING_SCHEME=v243
ID_NET_NAME_MAC=enx54ee75cb1dc0
ID_OUI_FROM_DATABASE=Wistron InfoComm(Kunshan)Co.,Ltd.
ID_NET_NAME_PATH=enp0s31f6
...

see how it just tells you which naming scheme it's on? ID_NET_NAMING_SCHEME=v243 — that's your vibe check right there. and then it hands you the MAC-based name, the OUI lookup, and the path name, all in one go. iconic behaviour fr fr 💅

🌸 Example 2. PCI Ethernet card with firmware index "1"

ID_NET_NAME_ONBOARD=eno1
ID_NET_NAME_ONBOARD_LABEL=Ethernet Port 1

firmware said "this is port 1" and the kernel was like bet, eno1. the label is the human-readable glow-up. 🎀

🌸 Example 3. PCI Ethernet card in slot with firmware index number

# /sys/devices/pci0000:00/0000:00:1c.3/0000:05:00.0/net/ens1
ID_NET_NAME_MAC=enx000000000466
ID_NET_NAME_PATH=enp5s0
ID_NET_NAME_SLOT=ens1

three names, one card. the # line is the sysfs path it lives at, and everything under it is what net_id cooked up. ✨

🌸 Example 4. PCI Ethernet multi-function card with 2 ports

# /sys/devices/pci0000:00/0000:00:1c.0/0000:02:00.0/net/enp2s0f0
ID_NET_NAME_MAC=enx78e7d1ea46da
ID_NET_NAME_PATH=enp2s0f0

# /sys/devices/pci0000:00/0000:00:1c.0/0000:02:00.1/net/enp2s0f1
ID_NET_NAME_MAC=enx78e7d1ea46dc
ID_NET_NAME_PATH=enp2s0f1

two ports, same card, f0 and f1 to tell them apart. see how the MACs are almost the same but the last digit flips? that's the function number doing its thing. 😌

🌸 Example 5. PCI WLAN card

# /sys/devices/pci0000:00/0000:00:1c.1/0000:03:00.0/net/wlp3s0
ID_NET_NAME_MAC=wlx0024d7e31130
ID_NET_NAME_PATH=wlp3s0

note the wl prefix, bestie — wireless gets its own little letter so you never confuse it with a wired en. 🦋

🌸 Example 6. PCI IB host adapter with 2 ports

# /sys/devices/pci0000:00/0000:00:03.0/0000:15:00.0/net/ibp21s0f0
ID_NET_NAME_PATH=ibp21s0f0

# /sys/devices/pci0000:00/0000:00:03.0/0000:15:00.1/net/ibp21s0f1
ID_NET_NAME_PATH=ibp21s0f1

ib for InfiniBand, and notice: no ID_NET_NAME_MAC here, only the path name. sometimes that's just how it is. 💜

🌸 Example 7. USB built-in 3G modem

# /sys/devices/pci0000:00/0000:00:1d.0/usb2/2-1/2-1.4/2-1.4:1.6/net/wwp0s29u1u4i6
ID_NET_NAME_MAC=wwx028037ec0200
ID_NET_NAME_PATH=wwp0s29u1u4i6

ww = WWAN, aka the modem vibe. that path name is a whole saga of USB hubs and it's lowkey beautiful. 😭

🌸 Example 8. USB Android phone

# /sys/devices/pci0000:00/0000:00:1d.0/usb2/2-1/2-1.2/2-1.2:1.0/net/enp0s29u1u2
ID_NET_NAME_MAC=enxd626b3450fb5
ID_NET_NAME_PATH=enp0s29u1u2

your phone tethering shows up as en something because to the kernel it's just... ethernet. main character energy honestly. 📱

🌸 Example 9. s390 grouped CCW interface

# /sys/devices/css0/0.0.0007/0.0.f5f0/group_device/net/encf5f0
ID_NET_NAME_MAC=enx026d3c00000a
ID_NET_NAME_PATH=encf5f0

en again, but the c in there is the s390 channel-subsystem flavour. every architecture gets its moment. 👑

🌸 Example 10. Set an allowlist for reading sysfs attributes for network card naming

/etc/udev/hwdb.d/50-net-naming-allowlist.hwdb
net:naming:drvirtio_net:*
  ID_NET_NAME_ALLOW=0
  ID_NET_NAME_ALLOW_ACPI_INDEX=1
  ID_NET_NAME_ALLOW_ADDR_ASSIGN_TYPE=1
  ID_NET_NAME_ALLOW_ADDRESS=1
  ID_NET_NAME_ALLOW_ARI_ENABLED=1
  ID_NET_NAME_ALLOW_DEV_PORT=1
  ID_NET_NAME_ALLOW_FUNCTION_ID=1
  ID_NET_NAME_ALLOW_IFLINK=1
  ID_NET_NAME_ALLOW_INDEX=1
  ID_NET_NAME_ALLOW_LABEL=1
  ID_NET_NAME_ALLOW_PHYS_PORT_NAME=1
  ID_NET_NAME_ALLOW_TYPE=1

so the plot twist here is ID_NET_NAME_ALLOW=0 — that's the default saying nope to everything — and then each individual attribute gets flipped back to 1 to let it through. it's giving "deny all, permit by exception". you're basically writing a little bouncer list for virtio_net cards. 🚪💅

heads up bestie: the catch-all <something>_ALLOW sets the baseline for all the attributes, and the per-attribute lines override it one by one. don't skip the catch-all or the whole thing is meaningless.

🌸 Example 11. Set a denylist so that specified sysfs attribute are ignored

/etc/udev/hwdb.d/50-net-naming-denylist.hwdb
net:naming:drvirtio_net:*
  ID_NET_NAME_ALLOW=1
  ID_NET_NAME_ALLOW_DEV_PORT=0
  ID_NET_NAME_ALLOW_PHYS_PORT_NAME=0

opposite energy, same slay. ID_NET_NAME_ALLOW=1 lets everything through by default, and then you personally veto DEV_PORT and PHYS_PORT_NAME with a 0. choose your fighter: allowlist or denylist. both are valid. 🥊

🌈 SEE ALSO 🌈

📝 NOTES 📝

  1. Predictable Network Interface Names https://systemd.io/PREDICTABLE_INTERFACE_NAMES

ok that's the whole page bestie, you made it to the end and you ATE. 💖✨ go forth and name your interfaces with confidence. no more eth0 roulette. 🎀

this was systemd 262.1, page SYSTEMD.N...ING-SCHEME(7). that's the version and the source, periodt. 🫶
📦 where this page comes from
  • arch rolling: systemd 262.1-1 · LGPL-2.1-or-later AND CC0-1.0 AND GPL-2.0-or-later AND MIT-0
  • 💖 girlified by deepseek-ai/DeepSeek-V4.1-Flash on : a language model's rewrite of the original, under its terms