Re: [PATCH 5.15 746/913] ARM: ftrace: avoid redundant loads or clobbering IP

[Ard Biesheuvel]

On Tue, 5 Apr 2022 at 18:52, Greg Kroah-Hartman
<gregkh@xxxxxxxxxxxxxxxxxxx> wrote:
>
> On Tue, Apr 05, 2022 at 12:01:19PM +0200, Ard Biesheuvel wrote:
> > On Tue, 5 Apr 2022 at 11:54, Greg Kroah-Hartman
> > <gregkh@xxxxxxxxxxxxxxxxxxx> wrote:
> > >
> > > From: Ard Biesheuvel <ardb@xxxxxxxxxx>
> > >
> > > [ Upstream commit d11967870815b5ab89843980e35aab616c97c463 ]
> > >
> > > Tweak the ftrace return paths to avoid redundant loads of SP, as well as
> > > unnecessary clobbering of IP.
> > >
> > > This also fixes the inconsistency of using MOV to perform a function
> > > return, which is sub-optimal on recent micro-architectures but more
> > > importantly, does not perform an interworking return, unlike compiler
> > > generated function returns in Thumb2 builds.
> > >
> > > Let's fix this by popping PC from the stack like most ordinary code
> > > does.
> > >
> > > Signed-off-by: Ard Biesheuvel <ardb@xxxxxxxxxx>
> > > Reviewed-by: Steven Rostedt (Google) <rostedt@xxxxxxxxxxx>
> > > Signed-off-by: Sasha Levin <sashal@xxxxxxxxxx>
> >
> > Please drop all the 32-bit ARM patches authored by me from the stable
> > queues except the ones that have fixes tags. These are highly likely
> > to cause an explosion of regressions, and they should have never been
> > selected, as I don't remember anyone proposing these for stable.
>
> From what I can tell, that is only this commit.  I'll go drop it from
> all trees, thanks.
>

Ah ok, that's not so bad then. But still better to avoid it.