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

[Will Deacon]

On Wed, Jan 03, 2024 at 05:50:01PM +0000, Will Deacon wrote:
> On Thu, Dec 28, 2023 at 04:46:41PM +0800, Jisheng Zhang wrote:
> > From: Nadav Amit <namit@xxxxxxxxxx>
> > 
> > fullmm in mmu_gather is supposed to indicate that the mm is torn-down
> > (e.g., on process exit) and can therefore allow certain optimizations.
> > However, tlb_finish_mmu() sets fullmm, when in fact it want to say that
> > the TLB should be fully flushed.
> > 
> > Change tlb_finish_mmu() to set need_flush_all and check this flag in
> > tlb_flush_mmu_tlbonly() when deciding whether a flush is needed.
> > 
> > At the same time, bring the arm64 fullmm on process exit optimization back.
> > 
> > Signed-off-by: Nadav Amit <namit@xxxxxxxxxx>
> > Signed-off-by: Jisheng Zhang <jszhang@xxxxxxxxxx>
> > Cc: Andrea Arcangeli <aarcange@xxxxxxxxxx>
> > Cc: Andrew Morton <akpm@xxxxxxxxxxxxxxxxxxxx>
> > Cc: Andy Lutomirski <luto@xxxxxxxxxx>
> > Cc: Dave Hansen <dave.hansen@xxxxxxxxxxxxxxx>
> > Cc: Peter Zijlstra <peterz@xxxxxxxxxxxxx>
> > Cc: Thomas Gleixner <tglx@xxxxxxxxxxxxx>
> > Cc: Will Deacon <will@xxxxxxxxxx>
> > Cc: Yu Zhao <yuzhao@xxxxxxxxxx>
> > Cc: Nick Piggin <npiggin@xxxxxxxxx>
> > Cc: x86@xxxxxxxxxx
> > ---
> >  arch/arm64/include/asm/tlb.h | 5 ++++-
> >  include/asm-generic/tlb.h    | 2 +-
> >  mm/mmu_gather.c              | 2 +-
> >  3 files changed, 6 insertions(+), 3 deletions(-)
> > 
> > diff --git a/arch/arm64/include/asm/tlb.h b/arch/arm64/include/asm/tlb.h
> > index 846c563689a8..6164c5f3b78f 100644
> > --- a/arch/arm64/include/asm/tlb.h
> > +++ b/arch/arm64/include/asm/tlb.h
> > @@ -62,7 +62,10 @@ static inline void tlb_flush(struct mmu_gather *tlb)
> >  	 * invalidating the walk-cache, since the ASID allocator won't
> >  	 * reallocate our ASID without invalidating the entire TLB.
> >  	 */
> > -	if (tlb->fullmm) {
> > +	if (tlb->fullmm)
> > +		return;
> > +
> > +	if (tlb->need_flush_all) {
> >  		if (!last_level)
> >  			flush_tlb_mm(tlb->mm);
> >  		return;
> 
> Why isn't the 'last_level' check sufficient here? In other words, when do
> we perform a !last_level invalidation with 'fullmm' set outside of teardown?

Sorry, logic inversion typo there. I should've said:

  When do we perform a last_level invalidation with 'fullmm' set outside
  of teardown?

I remember this used to be the case for OOM ages ago, but 687cb0884a71
("mm, oom_reaper: gather each vma to prevent leaking TLB entry") sorted
that out.

I'm not against making this clearer and/or more robust, I'm just trying
to understand whether this is fixing a bug (as implied by the subject)
or not.

Will