starting Fedora Server SIG
Rahul Sundaram
sundaram at fedoraproject.org
Thu Nov 20 03:47:24 UTC 2008
Les Mikesell wrote:
> Jesse Keating wrote:
>> On Wed, 2008-11-19 at 13:24 -0600, Les Mikesell wrote:
>>> Agreed, but when the kernel hardware detection order was predictable,
>>> this was simple. Now it isn't.
>>
>> When was it predictable? Even in the 2.4 era, we'd get one chassis
>> barebones from a vender like supermicro and the nic order would be one
>> way, then next month we'd order the same chassis barebones and the nics
>> would be picked up in a different order. Even more fun is when they'd
>> change with a kernel update, so that the kernel we installed with had
>> one order, and the kernel we updated to and rebooted to had it in a
>> different order.
>
> I'm pretty sure I cloned Centos 3.x across at least 50 IBM 336's with
> the NICs always being chosen in the same order.
That is just being lucky. Even in 2.4 kernel the order wasn't always
predictable. It changed and there was a post from Linus in LKML about
considering making it deliberately random to avoid any misconceptions
that it can always be predictable. It includes details on why it isn't
as well. I am sure you can find good references.
Rahul
More information about the devel
mailing list