Re: [PATCH v2 2/2] mm: madvise: no-op for MADV_NOHUGEPAGE if THP is disabled

[Andrew Morton <akpm@linux-foundation.org>]

On Tue, 06 May 2025 15:44:33 +0200 Ignacio Moreno Gonzalez via B4 Relay <devnull+Ignacio.MorenoGonzalez.kuka.com@kernel.org> wrote:

> From: Ignacio Moreno Gonzalez <Ignacio.MorenoGonzalez@kuka.com>
> 
> VM_NOHUGEPAGE is a no-op in a kernel without THP. So it makes no sense
> to return an error when calling madvise() with MADV_NOHUGEPAGE.

The patch looks rather odd.

> --- a/include/linux/huge_mm.h
> +++ b/include/linux/huge_mm.h
> @@ -7,6 +7,10 @@
>  #include <linux/fs.h> /* only for vma_is_dax() */
>  #include <linux/kobject.h>
>  
> +#ifndef CONFIG_TRANSPARENT_HUGEPAGE
> +#include <uapi/asm-generic/mman-common.h>
> +#endif

Why is the #ifndef here?

This is the only file under include/linux which directly includes
something from uapi/asm-generic.  Indicates that we're doing something
wrong.

If this hunk is truly the correct approach then I think we need a
comment here fully explaining what is going on.   Because it looks odd!

>  vm_fault_t do_huge_pmd_anonymous_page(struct vm_fault *vmf);
>  int copy_huge_pmd(struct mm_struct *dst_mm, struct mm_struct *src_mm,
>  		  pmd_t *dst_pmd, pmd_t *src_pmd, unsigned long addr,
> @@ -598,6 +602,8 @@ static inline bool unmap_huge_pmd_locked(struct vm_area_struct *vma,
>  static inline int hugepage_madvise(struct vm_area_struct *vma,
>  				   unsigned long *vm_flags, int advice)
>  {
> +	if (advice == MADV_NOHUGEPAGE)
> +		return 0;

Also a comment here which explains why we're doing this?

>  	return -EINVAL;
>  }
>  
> 
> -- 
> 2.39.5
>