Re: [PATCH 1/2] x86: math-emu: check __copy_from_user result

[Kees Cook]

On Wed, Oct 02, 2019 at 09:11:23AM +0200, Arnd Bergmann wrote:
> On Wed, Oct 2, 2019 at 1:39 AM Kees Cook <keescook@xxxxxxxxxxxx> wrote:
> 
> > > diff --git a/arch/x86/math-emu/reg_ld_str.c b/arch/x86/math-emu/reg_ld_str.c
> > > index f3779743d15e..fe6246ff9887 100644
> > > --- a/arch/x86/math-emu/reg_ld_str.c
> > > +++ b/arch/x86/math-emu/reg_ld_str.c
> > > @@ -85,7 +85,7 @@ int FPU_load_extended(long double __user *s, int stnr)
> > >
> > >       RE_ENTRANT_CHECK_OFF;
> > >       FPU_access_ok(s, 10);
> > > -     __copy_from_user(sti_ptr, s, 10);
> > > +     FPU_copy_from_user(sti_ptr, s, 10);
> >
> > These access_ok() checks seem redundant everywhere in this file (after
> > your switch from __copy* to copy*. I mean, I guess, just leave them, but
> > *shrug*
> 
> There have always been duplicate/inconsistent for the get_user/put_user
> case. I considered cleaning it all up but then decided to touch it as little
> as possible.

Yeah, at this point, I'd agree. :)

-- 
Kees Cook