Re: [PATCH] mm: Fix memblock_free_late() when using deferred struct page

[Mike Rapoport <rppt@kernel.org>]

Hi Ben,

On Tue, Feb 03, 2026 at 07:02:08PM +1100, Benjamin Herrenschmidt wrote:
> Currently, when using deferred struct page init, we may end up not
> freeing the pages depending on where they are. Typically this happens
> when efi_free_boot_services() tries to free UEFI Boot Services pages.
> 
> We can hit the !early_page_initialised() test in memblock_free_pages()
> since the deferred initializer hasn't even started yet. As a result we
> drop the pages on the floor.
> 
> Now, memblock_free_late() should only ever be called for pages that
> are reserved, and thus for which the struct page has already been
> initialized by memmap_init_reserved_pages().
> 
> So it should be safe to just free them normally and ignore the deferred
> initializer, which will skip over them as it skips over anything still
> in the memblock reserved list.
> 
> This recovers something like 130MB of RAM on EC2 t3a.nano instances who
> only have 512MB to begin with (as to why UEFI uses that much, that's a
> question for another day).
> 
> Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
> ---
>  mm/internal.h | 2 +-
>  mm/memblock.c | 6 ++++--
>  mm/mm_init.c  | 4 ++--
>  3 files changed, 7 insertions(+), 5 deletions(-)
 
...

> @@ -2440,9 +2440,9 @@ void *__init alloc_large_system_hash(const char
> *tablename,
>  }
>  
>  void __init memblock_free_pages(struct page *page, unsigned long pfn,
> -							unsigned int
> order)
> +				unsigned int order, bool reserved)

I've been thinking about after more coffee after our chat on IRC, and I
believe we don't need the bool reserved here.

Since the assumption that memblock_free_late() should be called only after
buddy is initialized, all the reserved pages should have their memmap setup
with PG_Reserved set. So we can use PageReserved() instead of passing the
boolean.

>  {
> -	if (IS_ENABLED(CONFIG_DEFERRED_STRUCT_PAGE_INIT)) {
> +	if (IS_ENABLED(CONFIG_DEFERRED_STRUCT_PAGE_INIT) && !reserved)
> {
>  		int nid = early_pfn_to_nid(pfn);
>  
>  		if (!early_page_initialised(pfn, nid))

-- 
Sincerely yours,
Mike.