Re: [PATCH] x86: entry_32.S: change ESPFIX test to not touch PT_OLDSS(%esp)

[Ingo Molnar <mingo@kernel.org>]

* Andy Lutomirski <luto@amacapital.net> wrote:

> On Mon, Mar 9, 2015 at 11:16 AM, Linus Torvalds
> <torvalds@linux-foundation.org> wrote:
> > On Mon, Mar 9, 2015 at 11:04 AM, Andy Lutomirski <luto@amacapital.net> wrote:
> >>
> >> One option would be to change the NMI entry code to move itself down 8
> >> bytes if this happens (came from kernel mode or sp == sp0 - 12,
> >> perhaps).
> >
> > Hmm. That whole code currently depends on the stack setup being just a
> > single instruction (the move to esp). And that simplifies things, I'd
> > like to keep it that way.
> >
> > I'd *much* rather just keep the 8-byte padding. What was so
> > problematic with that? It worked. It's been around forever. Removing
> > it is the bug.
> 
> Let's at least fix it, then.  processor.h has:
> 
> #define INIT_TSS  {                              \
>     .x86_tss = {                              \
>         .sp0        = sizeof(init_stack) + (long)&init_stack, \
> 
> (moved in -tip)
> 
> That's bogus, and this bogosity is why I broke 32-bit -next in the
> first place: I assumed it was correct.
> 
> I'll get it if no one beats me to it.

Please do and please post patches ASAP so that I can move tip:x86/asm 
back into a correct state again.

Thanks,

	Ingo