Re: [RFC 11/26] mm, slub: move disabling irqs closer to get_partial() in ___slab_alloc()

[Jann Horn]

On Tue, May 25, 2021 at 1:40 AM Vlastimil Babka <vbabka@xxxxxxx> wrote:
> Continue reducing the irq disabled scope. Check for per-cpu partial slabs with
> first with irqs enabled and then recheck with irqs disabled before grabbing
> the slab page. Mostly preparatory for the following patches.
[...]
> diff --git a/mm/slub.c b/mm/slub.c
[...]
>         if (slub_percpu_partial(c)) {
> +               local_irq_save(flags);
> +               if (unlikely(c->page)) {
> +                       local_irq_restore(flags);
> +                       goto reread_page;
> +               }
> +               if (unlikely(!slub_percpu_partial(c))) /* stolen by IRQ? */
> +                       goto new_objects;

nit: I think this comment is wrong by the end of the patch series,
since at that point, in RT configurations, it could also be stolen by
another task, if I understand correctly what migrate_disable() means?

Similarly the comment above ___slab_alloc() still talks about
disabling preemption for bulk allocation.