[oe] machine-specific udev packages

[pHilipp Zabel]

On 9/13/06, Richard Purdie <rpurdie at rpsys.net> wrote:
> On Wed, 2006-09-13 at 08:10 +0200, pHilipp Zabel wrote:
> > I only recently noticed that udev packages are machine specific now
> > due to the machine specific mount.blacklist file. It looks like most
> > machines blacklist the same devices, namely:
> >
> > /dev/mtdblock*
> > /dev/loop*
> > /dev/ram*
> >
> > Only spitz and slugos differ significantly.
> > (/dev/hda* on spitz, hd[a-z], sd[a-z] on slugos)
> >
> > I'd like to add /dev/mtdblock to the default mount.blacklist and
> > remove the specialization for akita, amsdelta, c7x0, h2200,
> > ipaq-pxa270 and tosa.
> >
> > Is there any good way to avoid spitz and slugos specific udev packages
> > that are only due mount.blacklist? Maybe this could be part of
> > base-files or some other package?
>
> Personally, I'd like to see several things happen. Firstly, if these
> lists are going to be machine specific, lets move them to something like
> base-files and make sure the scripts can survive without them. This
> means udev stops being machine specific. Alternatively, we create a new
> machine specific package if people don't want the extra file in
> base-files.
>
> Secondly, I'd prefer to see this change to a whitelist approach.

I agree.
It is certainly better to have udev not mount anything dynamically
unless a mount.whitelist (provided by machine specific base-files)
exists than the other way around.
I'm all for including the whitelist in base-files, are there reasons
not to do it? If so, is there perhaps another package already existing
that would fit this purpose?

> We also
> need to be more intelligent about automounting eg, not mounting /dev/hda
> when its partitioned and hda1, hda2 are present etc.

Definitely. As mount.sh grows in size and functionality, aren't we
starting to duplicate pmount?

btw,
http://www.openembedded.org/repo/org.openembedded.dev/packages/udev/files/mount.sh
returns a 403

regards
Philipp