Re: [PATCH 4.19 000/323] 4.19.291-rc1 review

[Greg Kroah-Hartman]

On Thu, Aug 10, 2023 at 12:36:41PM -0700, Linus Torvalds wrote:
> On Thu, 10 Aug 2023 at 12:28, Daniel Díaz <daniel.diaz@xxxxxxxxxx> wrote:
> >
> > Two new warnings are introduced on x86_64 with GCC-8 (defconfig):
> >
> > -----8<-----
> > drivers/gpu/drm/drm_edid.o: warning: objtool:
> > drm_mode_std.isra.34()+0xbc: return with modified stack frame
> > drivers/gpu/drm/drm_edid.o: warning: objtool:
> > drm_mode_std.isra.34()+0x0: stack state mismatch: cfa1=7+112 cfa2=7+8
> > ----->8-----
> >
> > Bisection points to the quoted commit ("drm/edid: Fix uninitialized
> > variable in drm_cvt_modes()"), 991fcb77f490 upstream. Reverting makes
> > the warnings disappear.
> 
> Bah. Stable should pick up commit d652d5f1eeeb ("drm/edid: fix objtool
> warning in drm_cvt_modes()") from mainline too.
> 
> Sadly it didn't have a 'Fixes:' tag, so it didn't get picked up
> automatically. My bad.
> 
> Although it's not like the commits it fixes were actually ever marked
> for stable either. I guess commit 3f649ab728cd ("treewide: Remove
> uninitialized_var() usage") got picked up as some kind of "make it
> easier to apply other patches" thing.

I applied the uninitialized_var patch here because I was trying to
attempt to reduce the number of build warnings I am getting using gcc-12
on 4.19.y  It helped out a tiny bit, but we still have lots of warnings
left.

I'll pick up d652d5f1eeeb as well now, thanks for pointing it out.

greg k-h