How to rename a network interface from e.g. enx00e04c885302 to something more human like lan0? The preferred method is to create a file /etc/systemd/network/10-persistent-lan0.link (one file for each interface to be renamed) with something like this:
#/etc/systemd/network/10-persistent-lan0.link [Match] MACAddress=00:e0:4c:88:53:02 [Link] Name=lan0
WARNING: You cannot use this method to safely assign the default kernel names (eth0
, eth1
, …) in your preferred order; this is because if systemd tries to rename an interface using a name which is already assigned by the kernel (e.g. it tries to swap eth0
and eth1
), it fails. So the target Name must not potentially conflict with the names assigned by the kernel.
WARNING: This method has an additional problem if you need to create a VLAN. When the system creates e.g. the lan.602, its MAC address matches the systemd snippet, so systemd tries to assign to it the name (lan0 in the example above) but the name already exists, so the interface remains with a temporary name e.g. rename5.
A workaround can be the following: tell the kernel to use the old ethX
schema (passing the net.ifnames=0 option), then instruct systemd to act only on eth* network interfaces:
[Match] MACAddress=00:e0:4c:88:53:02 OriginalName=eth* [Link] Name=lan0
If you have some files in /etc/systemd/network/ to rename network interfaces and you update the initramfs image (e.g. by running the update-initramfs tool), a copy of these files are placed into the image itself (under the path usr/lib/systemd/network/).
These files are considered by init
before it considers the content of /etc/systemd/network/. When the latter files are processed, some renaming may have already occurred and rules may not longer apply or may fail.
So it is always advisable to update the initramfs whenever you update the content of /etc/systemd/network/.
update-initramfs -k all -u
In Debian 12 the iniramfs image may be composed by two appended cpio archives. The first containing the Intel microcode in ASCII cpio archive format, and the second containing the actual initramfs as a Zstandard compressed cpio archive. To extract both, do the following:
Extract the microcode part (take note of the number of blocks printed on stderr):
cat /boot/initrd.img-6.1.0-21-amd64 | cpio --extract --make-directories --no-absolute-filenames 14080 blocks
Make a copy of the initramfs part stripping the microcode and decompress it:
dd if=/boot/initrd.img-6.1.0-21-amd64 of=initramfs.img.zst bs=512 skip=14080 unzstd initramfs.img.zst --stdout | cpio --extract --make-directories --no-absolute-filenames
Both methods seen above seem to work, unfortunately on my XFCE environment, the NetworkManager stepped into the play, reverting back the name:
NetworkManager[581]: <info> [1620916630.3558] device (eth0): interface index 10 renamed iface from 'eth0' to 'enx00e04c885302'
If you want to manage the interface using NetworkManager from the user session, be sure that the interface is not referred by /etc/network/interfaces or any snippet file into /etc/network/interfaces.d/
. If the interface is configured by /etc/network/interfaces
, NetworkManger will ignore it, labelling as unmanaged.
If you want to use the old naming schema (eth0, eth1, …) enable the kernel override option in /etc/default/grub:
GRUB_CMDLINE_LINUX="net.ifnames=0 biosdevname=0"
Followed by update-grub and reboot.
Beware that by using that schema the card's assigned name depends upon the time when the kernel discover the interfaces, in some cases you may find the names swapped after a reboot, even if the Ethernet cards are fixed built-in.
In a modern Linux system, network interface drivers are compiled as modules. Generally there is a subsytem that provides loading kernel modules automatically. Debian GNU/Linux Etch uses udev
for that.
Interface names (eth0, eth1, …) are assigned as they are discovered and kernel modules are loaded, unfortunately the order of discovering can sligtly vary across reboots. So you can find that after a reboot your system has swapped eth0 and eth1, with all the nasty consequences you can imagine.
I suggest two ways to assign interface names in a consisten way.
In a Debian Etch box, check the /etc/udev/rules.d/z25_persistent-net.rules
. This file is updated at bootstrap when new network interfaces are found. It contains a persistent mapping of MAC addresses to interface names.
In general, add a custom configuration file /etc/udev/rules.d/010_netinterfaces.rules
with lines like this:
KERNEL=="eth*", SYSFS{address}=="00:12:34:fe:dc:ba", NAME="eth0"
In the above example we used the SYSFS{address} as a match for the rule, use this command to print all sysfs properties of the specified device that can be used in udev rules to match the specified device:
udevinfo -a -p /sys/class/net/eth0
You configure this userspace program writing the configuration file/etc/iftab
:
# Example /etc/iftab file eth2 mac 08:00:09:DE:82:0E eth3 driver wavelan interrupt 15 baseaddress 0x390 eth4 driver pcnet32 businfo 0000:02:05.0 # wildcard name: pick the lowest available name of air0, air1, air2, etc. air* mac 00:07:0E:* arp 1
The good thing of this program is that you can select intefaces by the name of the kernel module, by MAC address, by bus position and other methods. But beware that the standard Debian startup of ifrename
runs the program once, when all the interfaces are already discovered and named. In this case it is not possible to swap two interface names, because when you try to rename the first interface, an error occurr: Error: cannot change name of eth1 to eth0: File exists
.
The udev cache file is named now /etc/udev/rules.d/70-persistent-net.rules
, sometimes only one Ethernet card is automatically added to this file, some other times the cache file is not created at all.
To force the creation of that file just execute the command:
udevadm trigger --action=add
Esempio: enp3s0f1
Per disabilitare questa funzione, chiamata Predictable Network Interface Names è possibile aggiungere l'opzione kernel net.ifnames=0 ad esempio in /etc/default/grub
, opzione GRUB_CMDLINE_LINUX.
Se è installato la helper utility di udev chiamata biosdevname, è opportuno passare al kernel anche il parametro biosdevname=0.
Avendo attivato l'opzione net.ifnames=0 può essere comunque desiderabile una persistenza nell'assegnare il nome all'interfaccia , considerato che il kernel assegna i nomi legacy eth0, eth1, etc. in base all'ordine di rilevamento hardware, che è non deterministrico. È possibile quindi creare un file /etc/udev/rules.d/70-network.rules con l'assegnamento statico in base al MAC address, ad esempio
SUBSYSTEM=="net", ACTION=="add", ATTR{address}=="00:30:67:9a:75:0b", NAME="eth0" SUBSYSTEM=="net", ACTION=="add", ATTR{address}=="00:18:e7:22:0b:f7", NAME="eth1"
ATTENZIONE: Pare che il file con le regole udev viene ignorato se manca l'opzione net.ifnames=0 nella cmdline del kernel! Questo almeno è accaduto con Debian 11.7 e kernel 5.10.0-23-amd64.
ATTENZIONE: Il file con le regole udev viene ignorato comunque in Debian 11.