On Sat, May 27, 2023 at 04:09:31PM +0100, Matthew Wilcox wrote:
On Sat, May 27, 2023 at 01:41:44PM +0300, Mike Rapoport wrote:
Sorry if I wasn't clear, by "page table page" I meant the page (or memory
for that matter) for actual page table rather than struct page describing
that memory.
So what we allocate here is the actual memory for the page tables and not
the memory for the metadata. That's why I think the name ptdesc_alloc is
confusing.
But that's going to be the common pattern in the Glorious Future.
You allocate a folio and that includes both the folio memory descriptor
and the 2^n pages of memory described by that folio. Similarly for all
the other memory descriptors.
I'm not arguing with that, I'm not happy about the naming. IMO, the name
should reflect that we allocate memory for page tables rather than for the
descriptor of that memory, say pgtable_alloc() or page_table_alloc().
--
Sincerely yours,
Mike.
