Re: Dynamic Debug broken on 2.6.35-rc3?

[Jason Baron <jbaron@redhat.com>]

On Fri, Jul 09, 2010 at 01:03:08PM +0200, Thomas Renninger wrote:
> Hi,
> 
> I can confirm that this patch fixes the issue for me.
> 
> On Thursday 08 July 2010 23:53:00 Andrew Morton wrote:
> > On Thu, 8 Jul 2010 17:39:28 -0400
> > Jason Baron <jbaron@redhat.com> wrote:
> > 
> > > Make sure we properly call ddebug_remove_module() when a module fails to
> > > load. In addition, pass the pointer to the "debug table", to both
> > > ddebug_add_module(), and ddebug_remove_module() so that we can uniquely
> > > identify each set of debug statements. In this way even modules with the
> > > same name can be properly identified and removed.
> > > 
> > > 
> > > Signed-off-by: Jason Baron <jbaron@redhat.com>
> > 
> > It'd be nice to track the Reported-by:s.  And the Tested-by:s if/when
> > they arrive.  SighIllDoIt.
> > 
> > The patch (almost) applies to 2.6.34.  So are we missing a Cc:stable tag
> > as well?
> I'll resubmit with some more meta info and will include stable@kernel.org.
> 
> Could it be that this isn't a regression, but a bug that was always present,
> but only gets exposed if you add modules with a specific implementation,
> e.g. specific declarations of functions missing, etc.?
> 

Hi Thomas,

yes, this race has likely been present for a while (i'd have to look at
specific kernel versions to verify). I suspect its getting exposed now
due to more usage of this feature, and the proliferation of kernel
modules...

> I tried to patch this into a 2.6.32.X kernel. While some hunks did not
> succeed, it looks like an adjusted patch should get submitted for older
> stable kernels as well?:
> 

if nobody else has done the 2.6.32 stable patch, I can do it, just let
me know.

thanks again for reporting this to me.

thanks,

-Jason