| 1 |
Am Tue, 2 Apr 2013 23:15:40 +0200 |
| 2 |
schrieb Marc Joliet <marcec@×××.de>: |
| 3 |
|
| 4 |
> Am Tue, 2 Apr 2013 20:31:10 +0000 (UTC) |
| 5 |
> schrieb Grant Edwards <grant.b.edwards@×××××.com>: |
| 6 |
> |
| 7 |
> > On 2013-04-02, Alan McKinnon <alan.mckinnon@×××××.com> wrote: |
| 8 |
> > |
| 9 |
> > > No, you are stilling misunderstanding. |
| 10 |
> > |
| 11 |
> > He's not the only one. |
| 12 |
> > |
| 13 |
> > > The news item goes to great lengths to explain that there is a new |
| 14 |
> > > way and it is different from the old way. |
| 15 |
> > |
| 16 |
> > I did grok that much. I had a 70-persistent-net.rules file that named |
| 17 |
> > my three interfaces "eth0" "eth1" and "eth2" based on their MAC |
| 18 |
> > addresses. After reading the news item and flameeyes blog, I was still |
| 19 |
> > pretty much at a loss regarding what I was actually supposed to _do_. |
| 20 |
> |
| 21 |
> AFAIU, as soon as the names in your rules file differ from the in-kernel names |
| 22 |
> (e.g., if the kernel switches eth0 and eth1), bad things can happen during |
| 23 |
> renaming, due to deadlocks or something like that (others will have understood |
| 24 |
> it better and should explain it rather than I). |
| 25 |
|
| 26 |
OK, I should have looked first. Here's a technical explanation from |
| 27 |
http://www.freedesktop.org/wiki/Software/systemd/PredictableNetworkInterfaceNames |
| 28 |
(referenced in the news item, BTW): |
| 29 |
|
| 30 |
For a longer time udev shipped support for assigning permanent "ethX" names to |
| 31 |
certain interfaces based on their MAC addresses. This turned out to have a |
| 32 |
multitude of problems, among them: [snipped various other reasons]. The |
| 33 |
biggest of all however is that the userspace components trying to assign the |
| 34 |
interface name raced against the kernel assigning new names from the same |
| 35 |
"ethX" namespace, a race condition with all kinds of weird effects, among |
| 36 |
them that assignment of names sometimes failed. As a result support for this |
| 37 |
has been removed from systemd/udev a while back. |
| 38 |
|
| 39 |
So not a deadlock, but a race condition. |
| 40 |
|
| 41 |
-- |
| 42 |
Marc Joliet |
| 43 |
-- |
| 44 |
"People who think they know everything really annoy those of us who know we |
| 45 |
don't" - Bjarne Stroustrup |
Archives