Bug#996557: linux-image-5.14.0-3-686-pae-unsigned: just about every sse2 program crashes with a SIGFPE

[Salvatore Bonaccorso <carnil@debian.org>]

Control: tags -1 + unreproducible

Hi,

On Sat, Nov 13, 2021 at 11:09:33AM +0100, Salvatore Bonaccorso wrote:
> Hi,
> 
> On Sun, Oct 31, 2021 at 09:19:05AM +0100, Salvatore Bonaccorso wrote:
> > Control: tags -1 + moreinfo
> > 
> > Hi
> > 
> > On Fri, Oct 15, 2021 at 02:52:17PM +0200, Jiri Palecek wrote:
> > > Package: src:linux
> > > Version: 5.14.12-1
> > > Severity: serious
> > > X-Debbugs-Cc: Jiri Palecek <jpalecek@web.de>
> > > 
> > > Dear Maintainer,
> > > 
> > > when I booted my system with the latest kernel from unstable, many of
> > > the applications didn't work. That includes plasmashell (so I couldn't
> > > log in into plasma) and chromium.
> > > 
> > > I ran xterm (which worked) and some of these failing programs under gdb
> > > and found they all tripped on some sse2 instructions with a
> > > SIGFPE. However, all the programs work under the kernel from
> > > linux-image-5.14.0-2-xxxx.
> > 
> > This is interesting as looking at the commits between 5.14.9 and
> > 5.14.12 there does not seem to be much related. As pointed out by Ben
> > offlist, the only plausible commit in this area might be 0723d4f8b179d
> > "x86/fpu: Restore the masking out of reserved MXCSR bits".
> > 
> > Can you thus please: Verify the issue is still present in 5.14.15 and
> > furthermore try to bisect the commit which introduces the problem for
> > you. In particular check if the 0723d4f8b179d commit is the culprit.
> 
> Were you able to bisect the issue?

furthermore, can you add the reports 'reportbug kernel' will
accordingly add to a bugreport. This helps to get a better overview.

Still i would like to know if you were able to test the above, as we
have not seen any other related reports so far.

Regards,
Salvatore