On Mon, Sep 19, 2022 at 04:46:39PM +0200, Peter Zijlstra wrote: > On Fri, Sep 16, 2022 at 12:57:23PM -0700, Andrew Morton wrote: > > Why is this so complicated. > > > > There's virtually zero value in running all those debug checks from within > > copy_from_user_nmi(). > > > > --- a/arch/x86/lib/usercopy.c~a > > +++ a/arch/x86/lib/usercopy.c > > @@ -44,7 +44,7 @@ copy_from_user_nmi(void *to, const void > > * called from other contexts. > > */ > > pagefault_disable(); > > - ret = __copy_from_user_inatomic(to, from, n); > > + ret = raw_copy_from_user(to, from, n); > > pagefault_enable(); > > > > return ret; > > I'm with Andrew here; this looks a *LOT* saner than all the other stuff. Yeah, I'd agree -- it's a special case of a special case. I'll send a new patch. Thanks! -- Kees Cook
