Re: [PATCH mm-unstable v2 4/5] mm: restart if multiple traversals raced

[Kinsey Ho]

Hi Michal,

> I may be missing (literal) context but I'd suggest not moving the memcg
> assignment and leverage
>         if (memcg != NULL)
>                 css_put(memcg->css)
> so that the is-root comparison needn't be repeated.

I might also be misunderstanding you with respect to the is-root
comparison – the reason the memcg assignment is moved is because it is
possible that on the restart added in this patch, css could be NULL.
In that case, memcg won't be assigned and could be left with a
previous, invalid value. By moving the assignment out, it ensures that
memcg is a valid value.

Best,
Kinsey