[PATCH 1/2] mm/tlb: fix fullmm semantics

[Catalin Marinas]

On Wed, Jan 03, 2024 at 12:26:29PM -0800, Dave Hansen wrote:
> On 1/3/24 10:05, Catalin Marinas wrote:
> >> --- a/mm/mmu_gather.c
> >> +++ b/mm/mmu_gather.c
> >> @@ -384,7 +384,7 @@ void tlb_finish_mmu(struct mmu_gather *tlb)
> >>  		 * On x86 non-fullmm doesn't yield significant difference
> >>  		 * against fullmm.
> >>  		 */
> >> -		tlb->fullmm = 1;
> >> +		tlb->need_flush_all = 1;
> >>  		__tlb_reset_range(tlb);
> >>  		tlb->freed_tables = 1;
> >>  	}
> > The optimisation here was added about a year later in commit
> > 7a30df49f63a ("mm: mmu_gather: remove __tlb_reset_range() for force
> > flush"). Do we still need to keep freed_tables = 1 here? I'd say only
> > __tlb_reset_range().
> 
> I think the __tlb_reset_range() can be dangerous if it clears
> ->freed_tables.  On x86 at least, it might lead to skipping the TLB IPI
> for CPUs that are in lazy TLB mode.  When those wake back up they might
> start using the freed page tables.

You are right, I did not realise freed_tables is reset in
__tlb_reset_range().

-- 
Catalin