Re: [PATCH v2 00/12] clean-up the migration target allocation functions

[Joonsoo Kim]

2020ë 5ì 29ì (ê) ìì 4:25, Vlastimil Babka <vbabka@xxxxxxx>ëì ìì:
>
> On 5/27/20 8:44 AM, js1304@xxxxxxxxx wrote:
> > From: Joonsoo Kim <iamjoonsoo.kim@xxxxxxx>
> >
> > This patchset clean-up the migration target allocation functions.
> >
> > * Changes on v2
> > - add acked-by tags
> > - fix missing compound_head() call for the patch #3
> > - remove thisnode field on alloc_control and use __GFP_THISNODE directly
> > - fix missing __gfp_mask setup for the patch
> > "mm/hugetlb: do not modify user provided gfp_mask"
> >
> > * Cover-letter
> >
> > Contributions of this patchset are:
> > 1. unify two hugetlb alloc functions. As a result, one is remained.
> > 2. make one external hugetlb alloc function to internal one.
> > 3. unify three functions for migration target allocation.
> >
> > The patchset is based on next-20200526.
> > The patchset is available on:
>
> I went through the series and I'd like to make some high-level suggestions
> first, that should hopefully simplify the code a bit more and reduce churn:

Thanks for review!
I have not enough time today to check your suggestions.
I will check on next week and then reply again.

Thanks.

> - in the series, alloc_huge_page_nodemask() becomes the only caller of
> alloc_migrate_huge_page(). So you can inline the code there, and it's one less
> function out of many with similar name :)
>
> - after that, alloc_huge_page_nodemask(ac) uses ac mostly just to extract
> individual fields, and only pass it as a whole to dequeue_huge_page_nodemask().
> The only other caller of dequeue...() is dequeue_huge_page_vma() who has to
> construct ac from scratch. It might be probably simpler not to introduce struct
> alloc_control into hugetlb code at all, and only keep it for
> alloc_migrate_target(), at which point it can have a more specific name as
> discussed and there's less churn
>
> - I'd suggest not change signature of migrate_pages(), free_page_t and
> new_page_t, keeping the opaque private field is fine as not all callbacks use
> struct alloc_context pointer, and then e.g. compaction_alloc has to use the
> private field etc. alloc_migration_target() can simply cast the private to
> struct alloc_control *ac as the first thing