Re: [PATCH v1 1/3] mm: pass meminit_context to __free_pages_core()

[David Hildenbrand]

On 10.06.24 06:03, Oscar Salvador wrote:
> On Fri, Jun 07, 2024 at 11:09:36AM +0200, David Hildenbrand wrote:
> > In preparation for further changes, let's teach __free_pages_core()
> > about the differences of memory hotplug handling.
> >
> > Move the memory hotplug specific handling from generic_online_page() to
> > __free_pages_core(), use adjust_managed_page_count() on the memory
> > hotplug path, and spell out why memory freed via memblock
> > cannot currently use adjust_managed_page_count().
> >
> > Signed-off-by: David Hildenbrand <david@xxxxxxxxxx>
>
> All looks good but I am puzzled with something.
>
> > +	} else {
> > +		/* memblock adjusts totalram_pages() ahead of time. */
> > +		atomic_long_add(nr_pages, &page_zone(page)->managed_pages);
> > +	}
>
> You say that memblock adjusts totalram_pages ahead of time, and I guess
> you mean in memblock_free_all()

And memblock_free_late(), which uses atomic_long_inc().

>   pages = free_low_memory_core_early()
>   totalram_pages_add(pages);
>
> but that is not ahead, it looks like it is upading __after__ sending
> them to buddy?

Right (it's suboptimal, but not really problematic so far. Hopefully Wei 
can clean it up and move it in here as well)

For the time being

"/* memblock adjusts totalram_pages() manually. */"

?

Thanks!

--
Cheers,

David / dhildenb