1 |
On 10/07/2011 12:16 PM, Jonas de Buhr wrote: |
2 |
> |
3 |
> out of interest: why do you have different configs? even if you have |
4 |
> different hardware you could still build a "one fits all"-kernel. or |
5 |
> are they that specialized? |
6 |
> |
7 |
|
8 |
We share kernel config whenever possible, but there are a few cases |
9 |
where they have to diverge. Basically any option that can't be compiled |
10 |
as a module is a candidate. Off the top of my head, |
11 |
|
12 |
* We've got x86/amd64 |
13 |
* Intel/AMD |
14 |
* A couple with RAID hardware that can't have its module installed. |
15 |
* One server with a tulip NIC that can't use a particular driver. |
16 |
* A set where hyperthreading needs to be disabled |
17 |
* A virtual machine host that needs certain hardening features |
18 |
disabled |
19 |
* A separate config for VM guests |
20 |
* Headless vs. GUI requires more grsec/pax tweaking |
21 |
* Different HZ settings. Power management in general depends on what |
22 |
the box will be doing. |
23 |
* A firewall with no non-essential modules available |
24 |
|
25 |
We keep the configs in git, so if two are similar I can usually just |
26 |
pull the last changeset (after a make oldconfig) over. What sucks is |
27 |
testing, and of course driving to work to reboot everything off-hours. |
28 |
|
29 |
|
30 |
> lets just agree on that. im kinda tired of this discussion. there's |
31 |
> nothing we can do about it anyway. |
32 |
> |
33 |
|
34 |
Agreed. |