Re: [patch] x86, tlb: switch cr3 in leave_mm() only when needed

[Linus Torvalds <torvalds@linux-foundation.org>]

On Thu, Mar 22, 2012 at 4:33 PM, Suresh Siddha
<suresh.b.siddha@intel.com> wrote:
>
> Currently leave_mm() unconditionally switches the cr3 to swapper_pg_dir.
> But there is no need to change the cr3, if we already left that mm.
>
> intel_idle() for example calls leave_mm() on every deep c-state entry where
> the CPU flushes the TLB for us. Similarly flush_tlb_all() was also calling
> leave_mm() whenever the TLB is in LAZY state. Both these paths will be
> improved with this change.

Hmm. If this is reasonably common (and intel_idle() certainly is),
maybe we shouldn't even do the "test_and_clear" RMW cycle.

We could do it with a read-only bit test (no races I can see - if it's
clear, it will stay clear), so we could do this with

    if (cpumask_test_cpu(cpu, mm_cpumask(active_mm))) {
        cpumask_clear_cpu(cpu,mm_cpumask(active_mm));
        load_cr3(swapper_pg_dir);
    }

instead? And avoid touching that "mm_cpumask" (and the atomic
serializing instruction) when not necessary?

                       Linus