I leave the project

Matthew Burgess matthew at linuxfromscratch.org
Thu Jul 8 11:47:48 PDT 2004


On Thu, 08 Jul 2004 10:04:59 +0200
Jeroen Coumans <jeroen at linuxfromscratch.org> wrote:

> Kevin P. Fleming said the following on 08-07-2004 02:00:
> > Jim Gifford wrote:
> > 
> >> Were exactly does it say to build a kernel with modules in the
> >book??
> > 
> > 
> > Not telling them _to_ build a kernel with modules is completely 
> > different IMHO from not telling them _not_ to build a kernel with 
> > modules. In plenty of other places in the book we warn them about 
> > particular decisions they make make that could have unexpected side 
> > effects, I see no reason why using modules/non-modules for their
> > kernel has to be treated differently.
> > 
> > If there are issues with using modules under certain circumstances
> > (that are not rare), why is there so much opposition to making the
> > reader aware that those issues exist? I've never suggested actually
> > documenting the issues in the book, just making them aware that
> > there are known issues and if they want to learn about them they can
> > read the list archives and anything else we may have to offer.
> 
> As FAQ maintainer, I totally agree with this. If problems are known 
> *before* a release, it should be incorporated in the book. If problems
> show up *after* a release, it should be in the FAQ.

As a long-time follower of this project, and now a somewhat more active
member of the project I have to agree as well.  This was one of those
"documentation" items I mentioned in my brief (too brief in fact!)
summary of the testing branch last night.  I'll provide a better
description of what I did last night & why later on once I've caught up
with the related emails.

I decided to pull hotplug out because it could never become stable
enough for a release in the short timeframe we'll be working to.  Having
said that, it hasn't solved all our problems with regard to udev &
modules.  These are inherent with udev itself and it doesn't matter how
you load the modules, it is possible for the module to be loaded and
then access to the device being attempted *before* udev has created the
device.

This is a pretty damn big problem when it happens (although from what I
understand it's a fairly uncommon situation).  So, the only fair and
reasonable thing to do with this knowledge is pass it on to our readers.
 At least then users won't be surprised if this happens to them, and
know how to avoid it (by compiling the driver in to the kernel if at all
possible).

Cheers,

Matt.



More information about the lfs-dev mailing list