Re: [patch 51/60] x86/mm: Allow flushing for future ASID switches

[Peter Zijlstra]

On Mon, Dec 04, 2017 at 02:22:54PM -0800, Andy Lutomirski wrote:

> > +static inline void invalidate_pcid_other(void)
> > +{
> > +       /*
> > +        * With global pages, all of the shared kenel page tables
> > +        * are set as _PAGE_GLOBAL.  We have no shared nonglobals
> > +        * and nothing to do here.
> > +        */
> > +       if (!static_cpu_has_bug(X86_BUG_CPU_SECURE_MODE_KPTI))
> > +               return;
> 
> I think I'd be more comfortable if this check were in the caller, not
> here.  Shouldn't a function called invalidate_pcid_other() do what the
> name says?

Yeah, you're probably right. The thing is course that we only ever need
that operation for kpti (as of now). But me renaming this stuff made
this problem :/

> > +       this_cpu_write(cpu_tlbstate.invalidate_other, true);
> 
> Why do we need this extra variable instead of just looping over all
> other ASIDs and invalidating them?  It would be something like:
> 
>         for (i = 1; i < TLB_NR_DYN_ASIDS; i++) {
>                 if (i != this_cpu_read(cpu_tlbstate.loaded_mm_asid))
>                        this_cpu_write(cpu_tlbstate.ctxs[i].ctx_id, 0);
>         }
> 
> modulo epic whitespace damage and possible typos.

I think the point is that we can do many invalidate_other's before we
ever do a switch_mm(). The above would be more expensive.

Not sure it would matter in practise though.

> >  static inline void __flush_tlb_one(unsigned long addr)
> >  {
> >         count_vm_tlb_event(NR_TLB_LOCAL_FLUSH_ONE);
> >         __flush_tlb_single(addr);
> > +       /*
> > +        * Invalidate other address spaces inaccessible to single-page
> > +        * invalidation:
> > +        */
> 
> Ugh.  If I'm reading this right, __flush_tlb_single() means "flush one
> user address" and __flush_tlb_one() means "flush one kernel address".

That would make sense, woulnd't it? :-) But afaict the __flush_tlb_one()
user in tlb_uv.c is in fact for userspace and should be
__flush_tlb_single().

Andrew, Mike, can either of you shed light on what exactly you need
invalidated there?

> That's, um, not exactly obvious.  Could this be at least commented
> better?

As is __flush_tlb_single() does user and __flush_tlb_one() does
user+kernel.