free_pages() is supposed to be called when we only have a virtual address.
__free_pages() is supposed to be called when we have a page.

There are a number of callers that use page_address() to get a page's
virtual address then call free_pages() on it when they should just call
__free_pages() directly.

Add kernel-docs for free_pages() to help callers better understand which
function they should be calling, and replace the obvious cases of
misuse.

-----------------
Based on mm-new, I intend to have all of these taken through the mm tree.

I've split the patches into separate subsystems to make it easier to
resolve conflicts, but there aren't any functional changes.

v3:
  - Collect some Reviewed-by Tags
  - Replace remaining free_page() calls in patch 7 (all other patches
    are unchanged from v2)
  - Add all appropriate mailing lists that were missing from v2

v2:
  - Reference __get_free_pages() instead of alloc_pages() in the
  free_pages() kernel-doc
  - Get some Reviewed-by tags
  - cc the subsystem maintainers related to specific patches

Vishal Moola (Oracle) (7):
  mm/page_alloc: Add kernel-docs for free_pages()
  aoe: Stop calling page_address() in free_page()
  x86: Stop calling page_address() in free_pages()
  riscv: Stop calling page_address() in free_pages()
  powerpc: Stop calling page_address() in free_pages()
  arm64: Stop calling page_address() in free_pages()
  virtio_balloon: Stop calling page_address() in free_pages()

 arch/arm64/mm/mmu.c                      | 2 +-
 arch/powerpc/mm/book3s64/radix_pgtable.c | 2 +-
 arch/riscv/mm/init.c                     | 4 ++--
 arch/x86/mm/init_64.c                    | 2 +-
 arch/x86/platform/efi/memmap.c           | 2 +-
 drivers/block/aoe/aoecmd.c               | 2 +-
 drivers/virtio/virtio_balloon.c          | 8 +++-----
 mm/page_alloc.c                          | 9 +++++++++
 8 files changed, 19 insertions(+), 12 deletions(-)

-- 
2.51.0

On Wed, Sep 03, 2025 at 11:59:14AM -0700, Vishal Moola (Oracle) wrote:
> free_pages() is supposed to be called when we only have a virtual address.
> __free_pages() is supposed to be called when we have a page.
> 
> There are a number of callers that use page_address() to get a page's
> virtual address then call free_pages() on it when they should just call
> __free_pages() directly.
> 
> Add kernel-docs for free_pages() to help callers better understand which
> function they should be calling, and replace the obvious cases of
> misuse.
> 
> Vishal Moola (Oracle) (7):
>   mm/page_alloc: Add kernel-docs for free_pages()
>   aoe: Stop calling page_address() in free_page()
>   x86: Stop calling page_address() in free_pages()
>   riscv: Stop calling page_address() in free_pages()
>   powerpc: Stop calling page_address() in free_pages()
>   arm64: Stop calling page_address() in free_pages()
>   virtio_balloon: Stop calling page_address() in free_pages()

Acked-by: Mike Rapoport (Microsoft) <rppt@kernel.org>

>  arch/arm64/mm/mmu.c                      | 2 +-
>  arch/powerpc/mm/book3s64/radix_pgtable.c | 2 +-
>  arch/riscv/mm/init.c                     | 4 ++--
>  arch/x86/mm/init_64.c                    | 2 +-
>  arch/x86/platform/efi/memmap.c           | 2 +-
>  drivers/block/aoe/aoecmd.c               | 2 +-
>  drivers/virtio/virtio_balloon.c          | 8 +++-----
>  mm/page_alloc.c                          | 9 +++++++++
>  8 files changed, 19 insertions(+), 12 deletions(-)
> 
> -- 
> 2.51.0
> 
> 

-- 
Sincerely yours,
Mike.