** Reply to message from Michael Fratoni <[EMAIL PROTECTED]> on Thu, 27 Mar 2003 
18:58:14 -0500

> -----BEGIN PGP SIGNED MESSAGE-----
> Hash: SHA1
> 
> On Thursday 27 March 2003 03:05 am, Jack Bowling wrote:
> > ** Reply to message from Michael Fratoni 
> > on Wed, 26 Mar 2003 23:08:22 -0500
> 
> > > On Wednesday 26 March 2003 04:14 pm, Robert P. J. Day wrote:
> [...]
> > > >   so, before i spend a lot of time on this, what is the function
> > > > of that symlink, and could having it point at a non-existent
> > > > kernel source directory be causing the rebuild of my NVIDIA
> > > > kernel src rpm to blow up with dozens of parse errors from
> > > > include files?
> [...]
> > > >   (not having the kernel source directory for the current running
> > > > kernel doesn't affect anything else -- system runs fine otherwise.)
> > >
> > > But building kernel modules will fail, as modules require the headers
> > > from the kernel tree, not the headers installed in
> > > /usr/include/linux/, (which are the headers glibc was compiled with,
> > > and are provided by the glibc-kernheaders package). Any kernel module
> > > that tries to include system headers should generate warning messages
> > > similar to:
> > > #warning Using kernel header in userland!
> > > #warning Using kernel headers in userspace.
> > >
> > > In the case above, if /lib/modules/`uname -r`/build is a broken link,
> > > the build will fail with the above warnings followed by many parse
> > > errors.
> > >
> > > So, the short version is reinstall the kernel-source package if you
> > > want to build kernel modules.
> >
> > And most drivers will not be able to build until you do a "make
> > oldconfig" and "make dep" in the kernel source tree.
> 
> I used to think so as well. And I'm sure I had to do both in the past at 
> some point. However on the Pheobe list, this was disputed by 
> Arjan van de Ven. 
> (https://listman.redhat.com/pipermail/phoebe-list/2003-January/000762.html)
> 
> I'll quote the relevent parts of the mail message:
> Begin quote:
> 
> > make oldconfig
> > make dep 
> > 
> > ** THEN you can compile stuff against that kenel (like spca-50x driver
> > or wlan-ng drivers, vmware, etc etc.. but I have had better luck just
> > modifying the config file and using the built in wireless drivers :)
> > (peruse the psyche-list for examples)
> 
> (To which Arjan replied:)
> 
> I'm sorry to say it but you're full of shit here ;)
> All you just achieved is that you've blown away the proper kernel symbol
> version setup you need to compile against the kernel.
> 
> 1) The *ONLY* place the kernel headers of the current kernel live is
> /lib/modules/`uname -r`/build/include
> not /usr/include/linux, not /usr/include/linux-2.4 not anything else
> this is per Linus' decree fwiw

Interesting. Take a look at this:

 /usr/lib/modules/2.4.18-14]$ ll
total 336
lrwxrwxrwx    1 root     root           32 Feb 16 23:42 build -> 
../../../usr/src/linux-2.4.18-14
drwxr-xr-x    9 root     root         4096 Feb 16 23:42 kernel
-rw-r--r--    1 root     root        94610 Feb 17 01:10 modules.dep
-rw-r--r--    1 root     root           31 Feb 17 01:10 modules.generic_string
-rw-r--r--    1 root     root          147 Feb 17 01:10 modules.ieee1394map
-rw-r--r--    1 root     root         8257 Feb 17 01:10 modules.isapnpmap
-rw-r--r--    1 root     root           29 Feb 17 01:10 modules.parportmap
-rw-r--r--    1 root     root        94767 Feb 17 01:10 modules.pcimap
-rw-r--r--    1 root     root           24 Feb 17 01:10 modules.pnpbiosmap
-rw-r--r--    1 root     root        95129 Feb 17 01:10 modules.usbmap
drwxr-xr-x    2 root     root         4096 Feb 16 23:42 pcmcia

> 
> 2) You do NOT need all these steps; the headers RHL ships by default
> Just Work(tm) and will generate a module for the currently running
> kernel.

The key being "that RHL ships". Lots of people roll their own kernels on RH boxes.

jb

--
Jack Bowling                mailto:[EMAIL PROTECTED]
Prince George, BC



-- 
redhat-list mailing list
unsubscribe mailto:[EMAIL PROTECTED]
https://listman.redhat.com/mailman/listinfo/redhat-list

Reply via email to