via rhine vs D3

[Chris Rouch]

On 2/7/07, Chris Rouch <chris.rouch at gmail.com> wrote:
> I have a diskless via epia m10K with a via-rhine network chip. This
> chip had buggy firmware, so that when the system is shut down cleanly
> (into D3 state according to the kernel code, though I'm not sure what
> that means), the firmware is switched off and does not become active
> again until re-started from the OS. On a diskfull setup this probably
> doesn't matter, but on a diskless client trying to boot from the
> network this is a significant problem.
>
> Fortunately the kernel developers have realised that via aren't going
> to fix their firmware and have added a patch that stops the kernel
> going to D3 state when given the appropriate parameter. From the code
> I see this:
>
> /* Work-around for broken BIOSes: they are unable to get the chip back out of
>    power state D3 so PXE booting fails. bootparam(7): via-rhine.avoid_D3=1 */
>
> My question is: where do I supply this parameter. I've tried adding
> "via-rhine.avoid_D3=1" to the boot stanza, but it is explicitly
> ignored when the system boots. Similarly I've tried adding
>
> options via-rhine avoid_D3=1
>
> to /etc/modprobe.conf
>
> This doesn't generate any error messages but doesn't seem to help.
>
> Of course, as this is diskless I suspect the option has to go into the
> initrd.img.
>

>From the overwhelming response, either I asked this question really
badly, or else no-one is doing this. In any case, and for the benefit
of anyone who googles this later, the solution seems to be to add

options via-rhine avoid_D3=1
to /etc/modprobe.conf, but the copy inside initrd.img, not the one on
the running system.

Setup like this, my diskless box has been powered down and netbooted
half a dozen times successfully.

Chris