Hi Fernando,

----- Original Message -----

> From: Fernando Gont <fg...@si6networks.com>
> To: Mark Smith <markzzzsm...@yahoo.com.au>
> Cc: Fernando Gont <ferna...@gont.com.ar>; Brian Haberman 
> <br...@innovationslab.net>; 6man Chairs <6man-cha...@tools.ietf.org>; 6man WG 
> <ipv6@ietf.org>; "draft-ietf-6man-stable-privacy-addres...@tools.ietf.org" 
> <draft-ietf-6man-stable-privacy-addres...@tools.ietf.org>
> Sent: Saturday, 13 April 2013 2:00 PM
> Subject: Re: AD Evaluation: draft-ietf-6man-stable-privacy-addresses
> 
> Hi, Mark,
> 
> On 04/11/2013 05:04 PM, Mark Smith wrote:
>>>  A number of years ago this happened with Linux. IIRC, it was
>>>  related to changing the default order of device driver loading, or
>>>  changing from sequential loading of device drivers to parallel
>>>  loading. That reordered the interface numbering and therefore the
>>>  names, which people didn't like.
> 
> It's not surprising that people didn't like it :-) -- e.g., if you have
> per-interface fw rules, this breaks everything.
> 

I encountered it because I had a box I'd setup for packet capturing with 
multiple interfaces. The sticky labels on the back of it stopped matching up :-(

> 
>>  The solution at the time was to
>>>  use the interface MAC address to assign interfaces persistent names
>>>  by renaming the interfaces late in the initialisation stage.
> 
> mm.. not sure hat you mean...
> 

The Linux kernel has an device event messaging system where kernel events 
generate messages that can be processed by user space. Using the "udev" daemon, 
that can execute various actions and/or scripts. So when a new network 
interface was added, a rule could be set to match on it's MAC address. If the 
MAC address matched, one of the available actions was to rename the interface. 
You can see some examples of these types of udev rules at:

https://wiki.archlinux.org/index.php/Network_Configuration#Udev_rule


> 
> 
>>>  One thing that may also influence this draft is that interfaces may
>>>  be added to the system after it has booted. For example, I carry
>>>  around a couple of USB/Ethernet adaptors that I'll add to my laptop
>>>  if I need them for troubleshooting (and with my new laptop being
>>>  wifi only, just to connect to a wired network), and I may not know
>>>  if I'll need them when I boot my laptop.
> 
> Well, if you add the card once the device has been bootstrapped, that
> shouldn't break the sequence of IID -- i.e., the USB/Ethernet adaptor
> would get an IfInndex that is larger than all existing IfIndexes...
> 

True, however the ifindex may not be the same for that USB/Ethernet adaptor as 
what it was last time it was installed. It's interface name might be though, 
because of e.g. the udev system.

> 
>>>  I also have a 3G mobile
>>>  dongle that I occasionally use. I'm pretty sure Linux allocates
>>>  ifindex values sequentially as the driver is loaded, so what
>>>  ifindex values these dynamic interfaces gets is not going to be
>>>  consistent and persistent.
> 
> If you always connect these cards in the same sequence, why shouldn't them?
> 

I'd expect them to. Of course I don't always connect them in sequence or 
connect them all at once.

> P.S.: In any case, I've tweaked the I-D to essentially say that "the
> IfIndex might change if network interfaces are installed or removed".
> 

Great.

Best regards,
Mark.

> Thanks!
> 
> Best regards,
> -- 
> Fernando Gont
> SI6 Networks
> e-mail: fg...@si6networks.com
> PGP Fingerprint: 6666 31C6 D484 63B2 8FB1 E3C4 AE25 0D55 1D4E 7492
> 
--------------------------------------------------------------------
IETF IPv6 working group mailing list
ipv6@ietf.org
Administrative Requests: https://www.ietf.org/mailman/listinfo/ipv6
--------------------------------------------------------------------

Reply via email to