Re: [RFC PATCH 4/6] x86/mm/tlb: Refactor common code into flush_tlb_on_cpus()

[Peter Zijlstra]

On Mon, May 27, 2019 at 06:59:01PM +0000, Nadav Amit wrote:
> > On May 27, 2019, at 2:24 AM, Peter Zijlstra <peterz@xxxxxxxxxxxxx> wrote:
> > 
> > On Sat, May 25, 2019 at 01:22:01AM -0700, Nadav Amit wrote:
> > 
> >> There is one functional change, which should not affect correctness:
> >> flush_tlb_mm_range compared loaded_mm and the mm to figure out if local
> >> flush is needed. Instead, the common code would look at the mm_cpumask()
> >> which should give the same result.
> > 
> >> @@ -786,18 +804,9 @@ void flush_tlb_mm_range(struct mm_struct *mm, unsigned long start,
> >> 	info = get_flush_tlb_info(mm, start, end, stride_shift, freed_tables,
> >> 				  new_tlb_gen);
> >> 
> >> -	if (mm == this_cpu_read(cpu_tlbstate.loaded_mm)) {
> >> -		lockdep_assert_irqs_enabled();
> >> -		local_irq_disable();
> >> -		flush_tlb_func_local(info, TLB_LOCAL_MM_SHOOTDOWN);
> >> -		local_irq_enable();
> >> -	}
> >> -
> >> -	if (cpumask_any_but(mm_cpumask(mm), cpu) < nr_cpu_ids)
> >> -		flush_tlb_others(mm_cpumask(mm), info);
> > 
> > So if we want to double check that; we'd add:
> > 
> > 	WARN_ON_ONCE(cpumask_test_cpu(smp_processor_id(), mm_cpumask(mm)) ==
> > 		     (mm == this_cpu_read(cpu_tlbstate.loaded_mm)));
> > 
> > right?
> 
> Yes, except the condition should be inverted (â!=â instead of â==â), and I
> would prefer to use VM_WARN_ON_ONCE().
> 
> Unfortunately, this condition does fire when copy_init_mm() calls dup_mm().
> I donât think there is a correctness issue, and I am tempted just check,
> before warning, that (mm != init_mm) .
> 
> What do you say?

Works for me.