[SOLVED] kernel module loading does not work with binutils-gold (was Re: [Bug #14626] oops on boot starting udev)

[Soeren Sonnenburg <sonne-8fiUuRrzOP0dnm+yROfE0A@public.gmane.org>]

On Sat, 2009-11-21 at 09:56 +0100, Soeren Sonnenburg wrote:
> On Wed, 2009-11-18 at 18:59 -0800, Dmitry Torokhov wrote:
> > On Tue, Nov 17, 2009 at 05:06:47AM +0100, Soeren Sonnenburg wrote:
> > > On Mon, 2009-11-16 at 20:01 -0800, Dmitry Torokhov wrote:
> > > > On Tue, Nov 17, 2009 at 03:59:03AM +0100, Soeren Sonnenburg wrote:
> > > > > On Mon, 2009-11-16 at 18:04 -0800, Dmitry Torokhov wrote:
> > > > > > On Mon, Nov 16, 2009 at 05:14:55PM -0800, Greg KH wrote:
> > > > > > > On Mon, Nov 16, 2009 at 11:37:48PM +0100, Rafael J. Wysocki wrote:
> > > > > > > > This message has been generated automatically as a part of a report
> > > > > > > > of recent regressions.
> > > > > > > > 
> > > > > > > > The following bug entry is on the current list of known regressions
> > > > > > > > from 2.6.31.  Please verify if it still should be listed and let me know
> > > > > > > > (either way).
> > > > > > > > 
> > > > > > > > 
> > > > > > > > Bug-Entry	: http://bugzilla.kernel.org/show_bug.cgi?id=14626
> > > > > > > > Subject		: oops on boot starting udev
> > > > > > > > Submitter	: Soeren Sonnenburg <sonne-8fiUuRrzOP0dnm+yROfE0A@public.gmane.org>
> > > > > > > > Date		: 2009-11-14 10:16 (3 days old)
> > > > > > > > References	: http://marc.info/?l=linux-kernel&m=125819380206800&w=4
> > > > > > > 
> > > > > > > This looks like an input core problem, as the evdev module was just
> > > > > > > loaded and died.
> > > > > > > 
> > > > > > > Any input developers have any ideas?
> > > > > > > 
> > > > > > 
> > > > > > 
> > > > > > Hmm, evdev does:
> > > > > > 
> > > > > > 	dev_set_name(&evdev->dev, "event%d", minor);
> > > > > > 
> > > > > > Not sure how it can go wrong...
> > > > > 
> > > > > Anything I should/could do to narrow it down a bit (apart from
> > > > > bisecting?).
> > > > > 
> > > > 
> > > > Umm, I looked through the changes between -rc6 and 7 but nothing jumped
> > > > out at me... You don't happen to have any local changes in your tree?
> > > 
> > > Well only the mouse button #1 emulation - though I don't see what could
> > > go wrong there.
> > > 
> > 
> > I have been looking through the changes and I really don't see anything
> > suspicious. I am also not hittign this oops on any of my boxes. Any
> > chance you could bisect?
> > 
> > Thanks.
> 
> Alright so I tried to do a bisect when I noticed that building a knwon
> to work -rc5 did no longer work either. Thought it might be a gcc
> problem (gcc-4.3 here) so upgraded to 4.4 - same thing.
> Then I recognized that it crashes on loading basically *any* module,
> tried tun and applesmc. Attaching the crashes...
> 
> I am starting to run out of ideas...

OK, I've found the culprit: binutils-gold

I build all kernels upto and including -rc6 with the old binutils and
since then have upgraded to binutils gold 2.20-4 which - in contrast to
the old binutils - uses --no-add-needed per default.

So I suspect it triggers an error(?) in the way how the kernel links
modules: It is now required to provide all needed libraries to the
linker when building the modules. I guess this problem could be worked
around by adding --add-needed to the LDFLAGS_MODULE ...

Soeren
-- 
For the one fact about the future of which we can be certain is that it
will be utterly fantastic. -- Arthur C. Clarke, 1962