upstart as the 'init' executable

On Thu, 2009-10-08 at 23:19 +0000, SaintDanBert wrote:

> It seems that there are two instances of an 'init' executable -- once as part of the initrd.img file contents
> and once read from the mounted, real root file system as /sbin/init. Further, I understand that what we call
> "upstart" is simply an implementation of /sbin/init configuration and processing.
>
> QUESTION: Is it reasonable to expect that "upstart" as 'init' runs in both places?
>
Yes, though nobody's done that yet.

In order to run Upstart in the initramfs, ideally you'd have
that /sbin/init re-exec the /sbin/init in the real filesystem and
transfer its state across.

You also need to take care of "cleaning up" the initramfs when done.

> QUESTION: Does it make sense that "upstart" learns or gets told or otherwise knows or cares
> which context is in play?
>
Maybe.

> QUESTION: Does it make sense that jobs and tasks and events know or care which context
> is in play? Regardless of the value, can a job or task or event learn which context is happening?
>
Probably not. You'd want things to work if you skipped the initramfs,
which is one of the compelling reasons why using Upstart inside it would
be good.

Scott
--
Have you ever, ever felt like this?
Had strange things happen? Are you going round the twist?