1 |
james wrote: |
2 |
> After deleting the 70-persistent-net.rule file |
3 |
> |
4 |
> udev does not re-create it. All is now fine with rc-status |
5 |
> only showing net.eth0 which is set up how I like it |
6 |
> per /etc/conf.d/net. All services are fine |
7 |
|
8 |
Beware. The automatic persistent net rules generator is intrinsically |
9 |
broken because naming eth* devices within the same namespace (eth* to |
10 |
eth*) can break things horribly. I have had this happen to me on several |
11 |
occasions and was helping someone out in #gentoo who had experienced the |
12 |
same issue only recently. |
13 |
|
14 |
For more information, take a look at this bug: |
15 |
|
16 |
https://bugzilla.redhat.com/show_bug.cgi?id=782145 |
17 |
|
18 |
Because of this, Kay Sievers has removed the rule generator from later |
19 |
versions of udev (*). The current version of udev marked stable in |
20 |
gentoo still has it but it can be disabled by setting |
21 |
USE="-rule_generator". I've started doing exactly that on my systems |
22 |
because it has only ever caused problems instead of providing solutions. |
23 |
Besides which, eventually a version of udev will be stabilized that no |
24 |
longer has it. |
25 |
|
26 |
If you want to use persistent net rules, the only safe way of doing it |
27 |
is to define the rules manually and to rename interfaces in such a |
28 |
fashion that they don't conflict with the existing device namespace. For |
29 |
example, you could rename "eth0" to "lan", "eth1" to "dmz" or whatever |
30 |
is deemed appropriate. |
31 |
|
32 |
This doesn't appear to be common knowledge, so it struck me as worth |
33 |
mentioning. |
34 |
|
35 |
Cheers, |
36 |
|
37 |
--Kerin |
38 |
|
39 |
* I cannot recall the exact version off-hand. |