the /usr/src/linux symlink

Jeremy Utley jeremy at
Tue Jan 4 04:23:12 PST 2005

Bryan Kadzban wrote:

> Jeremy Utley wrote:
>> However, given the frequency that this problem crops up, a note in
>> the book might be warranted, explaining the situation and why it's
>> incorrect.
>> Just looking for comments from the community, as the developer staff
>> is of mixed opinions on this.
> I'd like to see some explanation in the book also.  I know the symlink
> is totally unnecessary -- and I was bitten by the cdrecord breakage,
> back when I had the symlink pointing to 2.6.1-rc1, before I started
> building kernels in my home directory -- but it'd be nice if we can come
> up with some reasons it's wrong.  It'd be better if we included links to
> LKML archives, if anyone knows of any relevant threads (or is good at
> Googling for that kind of thing).

It's wrong because anything that runs in userspace shouldn't be touching 
kernel-internal headers.  The only things that need kernel-internal 
headers is the kernel itself, and any 3rd party modules not in the 
kernel - and those are provided for by 
/lib/modules/{kernel-version}/build.  It's the same thread that started 
the whole debate about raw headers vs sanitized headers.


More information about the lfs-dev mailing list