SLUG Mailing List Archives
RE: [SLUG] FC5 Kernel headers
- To: "Peter Rundle" <prundle@xxxxxxxxxxxxxx>, <slug@xxxxxxxxxxx>
- Subject: RE: [SLUG] FC5 Kernel headers
- From: "Visser, Martin" <martin.visser@xxxxxx>
- Date: Mon, 3 Apr 2006 16:32:26 +1000
- Thread-index: AcZW40p6B/qhjSRjRYOsy11v4zXTTwAAfWhg
- Thread-topic: [SLUG] FC5 Kernel headers
I can share your frustration, many many times - yet I persevere knowing
that the end mostly justifies the means.
As far as packaging goes, however I have learned in the last year or so
(having been a Linux hobbyist for about a 14 years now) that Debian
distros (in particular Ubuntu) have this thing pretty well sorted out.
(I previously shied away from Debian because of the incredibly painful
installation process - Ubuntu has fixed this maginficently).
To do this exact process (installing Vmware server) a few weeks ago
using Ubuntu I did
1. "apt-get install kernel-headers" which actually just returns a
virtual package prompting to select the correct architecture
2. so then "apt-get install kernel-headers-2.6.12-something" grabs the
package and nice-ly puts it under /usr/src/
3. Then running the vmware install, which all works without much
coercion until you find that the gcc that the kernel was built with
(3.4) doesn't match the standard gcc installed (4.0)
4. Again this is a pretty simple "apt-get install gcc-3.4"
5. Running vmware install again and it is all good.
I suppose that this isn't "automagic", but Ubuntu certainly does a good
job of solving dependencies on the fly. Because pretty well all of the
known OSS software universe exists as Ubuntu/Debian packages then
apt-get removes much of the heartache of trying to pull together
As far as the issue of kernel versions, CPU architectures and module
interaction I do think this is a pain. I imagine that a kernel module
could be built that is less strict on having kernel modules build
specifically against the headers for the kernel. That is I should
imagine "late-binding" of the modules should be possible in 90% or more
occasions. Any idea why this hasn't be explored for the Linux kernel. I
can't believe that the hooks that bind modules to the kernel really
change all that often, and if they do it should be done exception and
probably could be handled by some manner that is less painful than
having to recompile the module.
Consulting & Integration
Technology Solutions Group - HP Services
410 Concord Road
Rhodes NSW 2138
This email (including any attachments) is intended only for the use of
the individual or entity named above and may contain information that is
confidential, proprietary or privileged. If you are not the intended
recipient, please notify HP immediately by return email and then delete
the email, destroy any printed copy and do not disclose or use the
information in it.
From: slug-bounces@xxxxxxxxxxx [mailto:slug-bounces@xxxxxxxxxxx] On
Behalf Of Peter Rundle
Sent: Monday, 3 April 2006 3:55 PM
Subject: [SLUG] FC5 Kernel headers
I've just installed FC5 onto a single processor P4 and it all went
fairly smoothly, except that it claims to have detected my sound card
but no sound comes out, (shrugs, par for the course with Linux).
However, now I want to install VMware and compile the modules. Of course
this requires the kernel headers, which, even though I ticked all the
development check boxes, weren't installed by default and I can't find
them on any of the distribution CD's. So a quick uname -a reveals that
I'm running kernel 2.6.15-1.2054_FC5smp. So we download and install
kernel-smp-devel which creates the directory;
I now try to run vmware-config.pl and when it asks for the kernel header
source, I point it at said directory and of course being Linux it
doesn't work, but returns the error;
The kernel defined by this directory of header files does not have
address space size as your running kernel.
Any takers before I run screaming back to MS and admit that Linux was a
mistake and that I'll never ever doubt the software from a monopolistic
corporate giant ever again? Seriously though, why why why can't Linux
ever just work? After many years of using Linux my bucket of tolerance
for it's lack of polish is just about empty. I simply can't be bothered
with Linux anymore precisely because of this sort of thing. How hard can
it be to deliver the kernel headers for the kernel that you deliver on a
supposedly polished distribution.
P.S sorry for the spray but really if Linux is ever gonna be taken
seriously this sort of crap has to stop.
SLUG - Sydney Linux User's Group Mailing List - http://slug.org.au/
Subscription info and FAQs: http://slug.org.au/faq/mailinglists.html