On 11/29/24 18:18, David Hildenbrand wrote: > On 29.11.24 17:29, Zi Yan wrote: >> On 29 Nov 2024, at 9:44, David Hildenbrand wrote: >> >>> On 25.11.24 22:01, Matthew Wilcox (Oracle) wrote: >>> >>> What I don't quite like is that we now have a bit of an inconsistency, that makes it harder to understand what gives you frozen and what doesn't give you frozen pages. >>> >>> alloc_pages(): non-frozen/refcounted >>> alloc_frozen_pages(): frozen >>> folio_alloc_mpol_noprof(): non-frozen/refcounted >>> alloc_pages_mpol(): ... frozen pages? >>> >>> >>> Ideally, it would be "alloc_pages": non-frozen, "alloc_frozen_pages": frozen. I don't mind it that much as long as it concerns only static functions. We didn't add _frozen to all those in mm/page_alloc.c either. The exported ones seem consistent, AFAICS? >>> The same concern applies to other internal functions, like "__alloc_pages_cpuset_fallback". >> >> Or all internal functions (with "__" prefix) give frozen pages and external >> ones give non-frozen ones, unless with frozen in its name, i.e., alloc_frozen_pages()? > > If we go that path, we should make it consistent for __free_pages() as > well I am afraid. That one seems to have many callers so not worth the churn.
