misc: support upstart within an initramfs

Reported by ceg on 2009-12-02
18
This bug affects 3 people
Affects Status Importance Assigned to Milestone
upstart
Wishlist
Unassigned
initramfs-tools (Ubuntu)
Wishlist
Unassigned
upstart (Ubuntu)
Undecided
Unassigned

Bug Description

Binary package hint: upstart

Many problems arise from the current initramfs scripts not being event driven or a flexible state machine.

Might it be reasonable to use upstart code and configs within the initramfs to get the rootfs up instead of reimplementing and maintaining initramfs scripts with this functionality?

ceg (ceg) on 2009-12-02
affects: upstart (Ubuntu) → initramfs-tools (Ubuntu)

This is one possible future for initramfs-tools. yes

Colin Watson (cjwatson) on 2010-03-19
Changed in initramfs-tools (Ubuntu):
status: New → Triaged
importance: Undecided → Wishlist
ceg (ceg) wrote :

First step to allow using upstart in initramfs would probably be an option to copy/run upstart into initramfs by adding an initramfs-hook script to the upstart package.

http://bazaar.launchpad.net/~ubuntu-branches/ubuntu/lucid/initramfs-tools/lucid/annotate/head%3A/docs/example_hook

Bear in mind that Upstart would have to be the initramfs /init just as its the real system /sbin/init - it has to be PID 1

One of the main pre-requisites for this is state passing; so that the upstart in the initramfs would be able to re-exec itself once the root is mounted and then pass its state to the new instance

Changed in upstart (Ubuntu):
status: New → Invalid
Changed in upstart:
status: New → Triaged
importance: Undecided → Wishlist
summary: - upstart init within initramfs
+ misc: support upstart within an initramfs
ceg (ceg) wrote :

If we would have a defined state at the and of the initramfs boot, say "rootfs fully set up" implying all its dependencies would we still need state passing? Or could the main upstart take it from there.

On Sun, 2010-04-25 at 08:21 +0000, ceg wrote:

> If we would have a defined state at the and of the initramfs boot, say
> "rootfs fully set up" implying all its dependencies would we still need
> state passing? Or could the main upstart take it from there.
>
I think we should do it properly and pass all state, there may be
services running from the initramfs that match services from the real
system - udev, plymouth, bootchart, etc. are all examples *today* where
we use a hack to pass the pid into Upstart

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

ceg (ceg) wrote :

Thanks for the insight!

To post a comment you must log in.
This report contains Public information  Edit
Everyone can see this information.

Other bug subscribers

Related blueprints