Yep, names like p4p1 are soooo much better than eth0 :-(

jdow jdow at
Mon Oct 17 23:57:42 UTC 2011

On 2011/10/17 06:50, Marko Vojinovic wrote:
> On Monday 17 October 2011 02:34:47 Tom Horsley wrote:
>> On Sun, 16 Oct 2011 20:20:35 -0500
>> Chris Adams wrote:
>>> Define "first".  Hint: that's not a solvable problem on general PC
>>> hardware (thus, biosdevname).  There have been many problems over the
>>> years where the ordering changed on some hardware from one kernel
>>> release to the next.
>> Well, either biosdevname is also a nonsolvable problem, or a
>> perfectly wonderful definition of "first" is simply "The
>> lexicographic ordering of the biosdevname generated name strings".
> Suppose you have a server with a two-port ethernet PCI card plugged into the
> second PCI slot on the motherboard. The biosdevname names are p2p1 and p2p2.
> By your definition, they would map to eth0 and eth1, right?
> Now, at some point you decide to upgrade your hardware by adding an aditional
> two-port ethernet card in PCI slot 1. The biosdevname names for the new ports
> are p1p1 and p1p2, while p2p1 and p2p2 stay the same, as before. But this
> time, according to your proposal, eth0 and eth1 would not correspond to p2p1
> and p2p2, but to the new ports, p1p1 and p1p2, while the p2p1 and p2p2 would
> be renamed eth2 and eth3. Assuming that you plug all the cables back to their
> original positions, every script that refers to eth0 and eth1 is now broken,
> while every script that refers to p2p1 and p2p2 still works.
> IIUC, this is one of the problems that biosdevname is meant to fix. The concept
> of physical location is a well-defined one, while the concept of linearly
> ordered (lexicographically or otherwise) location is ill-defined. Hence the
> names are given according to the physial location rather than to any
> particular ordering.

There is something wrong with ethp2p3? What KIND of device is easier to fathom
if it is part of the name, ya know.


More information about the users mailing list