Re: [PATCH v14 6/8] mm: hugetlb: introduce nr_free_vmemmap_pages in the struct hstate

On 2021/2/5 16:56, Oscar Salvador wrote:

On Fri, Feb 05, 2021 at 04:39:26PM +0800, Miaohe Lin wrote:

quoted

Hi:
On 2021/2/5 16:22, Oscar Salvador wrote:

quoted

On Fri, Feb 05, 2021 at 03:29:43PM +0800, Miaohe Lin wrote:

quoted

quoted

+	if (likely(vmemmap_pages > RESERVE_VMEMMAP_NR))
+		h->nr_free_vmemmap_pages = vmemmap_pages - RESERVE_VMEMMAP_NR;

Not a problem. Should we set h->nr_free_vmemmap_pages to 0 in 'else' case explicitly ?

No, hstate fields are already zeroed.

I know hstate fields are already zeroed. What I mean is should we set nr_free_vmemmap_pages
to 0 _explicitly_ like nr_huge_pages and free_huge_pages in hugetlb_add_hstate() ?
But this is really trival.

We do not anny more [1]

[1] https://patchwork.kernel.org/project/linux-mm/patch/20201119112141.6452-1-osalvador@suse.de/

Oh, I see. I thought we do it deliberately. Many thanks for your patient answer. :)