Yep, names like p4p1 are soooo much better than eth0 :-(
lowen at pari.edu
Tue Nov 8 14:57:46 UTC 2011
On Monday, October 17, 2011 11:39:29 PM Garry T. Williams wrote:
> On Monday, October 17, 2011 16:57:42 jdow wrote:
> > There is something wrong with ethp2p3? What KIND of device is easier
> > to fathom if it is part of the name, ya know.
> Tell Sun, er, Oracle that. What are hme0, qfe0, and eri0? :-)
Happy Meal Ethernet 0
Haven't seen an ERI in the wild yet, and haven't Googled it, so don't know that one.
The *BSD's also do this sort of thing.
I would love something more consistent, similar to the cisco naming (even though depending upon which BU the device is from, interfaces and slots can either start with 0 or start with 1, but that's a digression).
But PC hardware is so much more variable than cisco stuff is, and motherboards can have different lanes (for PCI-e) out of order relative to the slots, and server motherboards especially (like a SuperMicro P4DP6 to pull one off the top of my head) have multiple buses, so that the fifth slot is actually something like the third bus's second slot or similar, meaning you have to dig out the manual, and that's often of no help at all. Need the ability to 'blink LEDs' at times other than installation, IMO. (yes, you have this already in Fedora; install ethtool, and use ethtool -p $devname )
The current 'ethX' convention breaks in odd ways for different use cases. Especially when you replace a lightning-toasted NIC.
And I have personally seen PCI enumeration order change on seemingly a whim, both due to kernel updates (this was EL4, so not a kernel version upgrade) and due to BIOS updates. And I'm not just talking slot order on a bus, but bus order on the northbridge. (Serverworks chipset in one instance, Intel chipset in another.)
The fact of the matter is that consistent PC device enumeration is a hard problem, and people are working towards making this more consistent and better from the end-users' points of view. And I appreciate the effort, even with the bugs.
Windows has the same problem, just a different flavor, and I've hit that, too, with XP, Vista, Server 2003, 7, and Server 2008.
More information about the users