Kernel 2054 breaks nvidia.ko loading

[Gilboa Davara]

On 3/16/06, Dave Jones <davej at redhat.com> wrote:
> On Wed, Mar 15, 2006 at 07:11:26PM -0600, Jonathan Berry wrote:
>  > On 3/15/06, Jim Bevier <jim at jbsys.com> wrote:
>  > > It seems the kernel from today, 2054, breaks the loading of the nvidia.ko
>  > > module.  I get the message "nvidia: Unknown symbol print_tainted".  Building
>  > > of VMware module does not have any problem.  The last good kernel that
>  > > worked was 2045.  Was the symbol deleted?
>  >
>  > I am seeing a problem with loading the nvidia module as well.  Builds
>  > just fine from a src.rpm acquired from:
>  > http://www.leemhuis.info/files/fedorarpms/MISC.lvn/graphics/  Running
>  > x86_64, as I think you are.  I cannot seem to find the problem here.
>  > I downloaded kernel-2.6.15-1.2054_FC5.src.rpm and did an rpmbuild -bp
>  > to get the source tree.  I found print_tainted seemingly defined just
>  > fine in kernel/panic.c:
>  > const char *print_tainted(void)
>  > {
>  >         static char buf[20];
>  >         if (tainted) {
>  >                 snprintf(buf, sizeof(buf), "Tainted: %c%c%c%c%c%c",
>  >                         tainted & TAINT_PROPRIETARY_MODULE ? 'P' : 'G',
>  >                         tainted & TAINT_FORCED_MODULE ? 'F' : ' ',
>  >                         tainted & TAINT_UNSAFE_SMP ? 'S' : ' ',
>  >                         tainted & TAINT_FORCED_RMMOD ? 'R' : ' ',
>  >                         tainted & TAINT_MACHINE_CHECK ? 'M' : ' ',
>  >                         tainted & TAINT_BAD_PAGE ? 'B' : ' ');
>  >         }
>  >         else
>  >                 snprintf(buf, sizeof(buf), "Not tainted");
>  >         return(buf);
>  > }
>  > EXPORT_SYMBOL_GPL(print_tainted);
>  > Looks like the symbol is there and exported.  Granted, I'm not an that
>  > knowledgeable about such things, but that looks okay to me.  Anyone
>  > have any ideas what is going on here?
>
> An oversight on my part.
> print_tainted is used in the spinlock macros, so essentially,
> the macro is made GPL-only too.  I've reverted that change
> in cvs, but it's too late for the final FC5 image, which is
> already being pushed out to mirrors.
>
> It'll be fixed in the very first update kernel for FC5, which
> will likely be within the first few days of release.
>
>                 Dave
>
> --

Ohhhh! Gasp!!!!
Not again?!?!?

Dave,

This is beyond *bad*.
This effectively kills each and every non-GPL kernel module out there.
Instead of having a fun release-day, we'll be playing catch-up with
numerous "Fedora-5-ate-my-baby" threads.

There's got to be a way to fix it. Either by stopping the
ISO-mastering or changing the release-notes to acknowledge the problem
and pushing a new kernel ASAP.

Gilboa