Re: [PATCH] drm: panel: fix excessive stack usage in td028ttec1_prepare

[Laurent Pinchart]

Hi Arnd,

On Tue, Jan 07, 2020 at 11:09:13PM +0100, Arnd Bergmann wrote:
> On Tue, Jan 7, 2020 at 11:00 PM Laurent Pinchart wrote:
> > On Tue, Jan 07, 2020 at 10:27:33PM +0100, Arnd Bergmann wrote:
> > > With gcc -O3, the compiler can inline very aggressively,
> > > leading to rather large stack usage:
> > >
> > > drivers/gpu/drm/panel/panel-tpo-td028ttec1.c: In function 'td028ttec1_prepare':
> > > drivers/gpu/drm/panel/panel-tpo-td028ttec1.c:233:1: error: the frame size of 2768 bytes is larger than 2048 bytes [-Werror=frame-larger-than=]
> > >  }
> > >
> > > Marking jbt_reg_write_1() as noinline avoids the case where
> > > multiple instances of this function get inlined into the same
> > > stack frame and each one adds a copy of 'tx_buf'.
> > >
> > > Fixes: mmtom ("init/Kconfig: enable -O3 for all arches")
> > > Signed-off-by: Arnd Bergmann <arnd@xxxxxxxx>
> >
> > Isn't this something that should be fixed at the compiler level ?
> 
> I suspect but have not verified that structleak gcc plugin is partly at
> fault here as well, it has caused similar problems elsewhere.
> 
> If you like I can try to dig deeper before that patch gets merged,
> and explain more in the changelog or open a gcc bug if necessary.

I think we'll need to merge this in the meantime, but if gcc is able to
detect too large frame sizes, I think it should have the ability to take
a frame size limit into account when optimizing. I haven't checked if
this is already possible and just not honoured here (possibly due to a
bug) or if the feature is entirely missing. In any case we'll likely
have to live with this compiler issue for quite some time.

-- 
Regards,

Laurent Pinchart