[oe] [Angstrom-devel] udev problem for zaurus machine solved

[Marcin Juszkiewicz]

On Tuesday 23 of September 2008 17:44:26 Koen Kooi wrote:

> I decided to sit down and debug the udev problem on my c700.
>
> Summary:
>
>   It's not a udev problem, it's pretty much a problem with zaurus
> initscripts.

Currently it is more then Zaurus devices - thats why I am Cc: OE ml.

There is standard script and specific "devices" script are present for:

- all Zaurus devices
- iPaQ devices: h3600, h3900
- Jornada devices: 56x, 6xx, 7xx
- Simpad
- FooNAS
- Openprotium
- UML

> Background:
>
>   Udev mount a tmpfs over /dev and starts adding nodes with the info
> from /sys and creates a tarball when its done to make succesive boots
> faster
>
>   /etc/init.d/devices is a script that attemps to fix up devfs errors
> and for zaurus machines creates some kernel 2.4 specific symlinks
> when udev isn't running
>
>   However instead of doing a runtime check (e.g by looking at
> /proc/cpuinfo) zaurus machines opted to have their own devices script
> in packages/initscripts/initscripts-1.0/<MACHINE>/devices, which, as
> it turns out, have a broken check for udev. It doesn't detect udev
> and mounts and _empty_ ramfs over /dev and fills in a few dev nodes.
> This is hidden by the fact that udevsettle/trigger is still running
> in the background creating extra dev nodes.
>
> Solution:
>
>   Don't install the 'devices' script for angstrom.

I would like to drop that script totally (like it is done in Poky). We 
nearly do not have 2.4 powered devices which would use devfs - if 
someone maintain such one then please rise hand - I think that 
do_install_append() for such devices would be enough - rest can work 
without 'devices' script.


-- 
JID: hrw at jabber.org