Re: [PATCH RESEND v3 1/2] mm/tlb: skip redundant IPI when TLB flush already synchronized

[Lance Yang]

On 2026/1/6 23:19, David Hildenbrand (Red Hat) wrote:
> >   static void tlb_table_flush(struct mmu_gather *tlb)
> > @@ -367,7 +378,7 @@ void tlb_remove_table(struct mmu_gather *tlb, void 
> > *table)
> >           *batch = (struct mmu_table_batch *)__get_free_page(GFP_NOWAIT);
> >           if (*batch == NULL) {
> >               tlb_table_invalidate(tlb);
> > -            tlb_remove_table_one(table);
> > +            tlb_remove_table_one(table, tlb);
> >               return;
> >           }
> >           (*batch)->nr = 0;
> > @@ -427,6 +438,7 @@ static void __tlb_gather_mmu(struct mmu_gather 
> > *tlb, struct mm_struct *mm,
> >       tlb->vma_pfn = 0;
> >       tlb->fully_unshared_tables = 0;
> > +    tlb->tlb_flush_sent_ipi = 0;
> >       __tlb_reset_range(tlb);
> >       inc_tlb_flush_pending(tlb->mm);
> >   }
>
> But when would we have to reset tlb->tlb_flush_sent_ipi = 0 later? 
> That's where it gets tricky. Just imagine the MMU gather gets reused later.
>
> Also,
>
>      +    if (info->freed_tables && info->tlb)
>      +        info->tlb->tlb_flush_sent_ipi = true;
>
> in native_flush_tlb_multi() misses the fact that we have different 
> flushing types for removed/unshared tables vs. other flush.
>
> So this approach more here certainly gets more complicated and error prone.

Agreed. Tracking the flag through mmu_gather lifecycle does get
more complicated and error-prone ...

> tlb_table_flush_implies_ipi_broadcast() was clearer in that regard: if 
> you flushed the TLB after removing /unsharing tables, the IPI for 
> handling page tables can be skipped. It's on the code flow to assure that.

v2 was definitely simpler.

> What could work is tracking "tlb_table_flush_sent_ipi" really when we 
> are flushing the TLB for removed/unshared tables, and maybe resetting 
> it ... I don't know when from the top of my head.

Not sure what's the best way forward here :(

> v2 was simpler IMHO.

The main concern Dave raised was that with PV hypercalls or when
INVLPGB is available, we can't tell from a static check whether IPIs
were actually sent.

Maybe that's acceptable, or we could find a simpler way to track that ...

Open to suggestions!