Re: [Intel-gfx] [PATCH 1/3] treewide: Lift switch variables out of switches

[Kees Cook <keescook@chromium.org>]

On Thu, Jan 24, 2019 at 8:18 AM Matthew Wilcox <willy@infradead.org> wrote:
>
> On Wed, Jan 23, 2019 at 04:17:30PM +0200, Jani Nikula wrote:
> > Can't have:
> >
> >       switch (i) {
> >               int j;
> >       case 0:
> >               /* ... */
> >       }
> >
> > because it can't be turned into:
> >
> >       switch (i) {
> >               int j = 0; /* not valid C */
> >       case 0:
> >               /* ... */
> >       }
> >
> > but can have e.g.:
> >
> >       switch (i) {
> >       case 0:
> >               {
> >                       int j = 0;
> >                       /* ... */
> >               }
> >       }
> >
> > I think Kees' approach of moving such variable declarations to the
> > enclosing block scope is better than adding another nesting block.
>
> Another nesting level would be bad, but I think this is OK:
>
>         switch (i) {
>         case 0: {
>                 int j = 0;
>                 /* ... */
>         }
>         case 1: {
>                 void *p = q;
>                 /* ... */
>         }
>         }
>
> I can imagine Kees' patch might have a bad effect on stack consumption,
> unless GCC can be relied on to be smart enough to notice the
> non-overlapping liveness of the vriables and use the same stack slots
> for both.

GCC is reasonable at this. The main issue, though, was most of these
places were using the variables in multiple case statements, so they
couldn't be limited to a single block (or they'd need to be manually
repeated in each block, which is even more ugly, IMO).

Whatever the consensus, I'm happy to tweak the patch.

Thanks!

-- 
Kees Cook