Re: [RFC PATCH 5/9] arm64: mm: Permit permissions changes on huge vmappings

Hi Ard

On Sat, Aug 22, 2026 at 03:53:27PM +0200, Ard Biesheuvel wrote:
>From: Ard Biesheuvel <[email protected]>
>
>Allow permission changes on huge vmappings in cases where no splitting
>is needed (i.e., the region is aligned sufficiently), or when the system
>has support for splitting live mappings.
>
>Signed-off-by: Ard Biesheuvel <[email protected]>
>---
> arch/arm64/mm/pageattr.c | 13 ++++++++++---
> 1 file changed, 10 insertions(+), 3 deletions(-)
>
>diff --git a/arch/arm64/mm/pageattr.c b/arch/arm64/mm/pageattr.c
>index bbe98ac9ad8c..20ff9cb273c1 100644
>--- a/arch/arm64/mm/pageattr.c
>+++ b/arch/arm64/mm/pageattr.c
>@@ -169,8 +169,6 @@ static int change_memory_common(unsigned long addr, int numpages,
> 	 * we are operating on does not result in such splitting.
> 	 *
> 	 * Let's restrict ourselves to mappings created by vmalloc (or vmap).
>-	 * Disallow VM_ALLOW_HUGE_VMAP mappings to guarantee that only page
>-	 * mappings are updated and splitting is never needed.
> 	 *
> 	 * So check whether the [addr, addr + size) interval is entirely
> 	 * covered by precisely one VM area that has the VM_ALLOC flag set.
>@@ -179,7 +177,16 @@ static int change_memory_common(unsigned long addr, int numpages,
> 	if (!area ||
> 	    ((unsigned long)kasan_reset_tag((void *)end) >
> 	     (unsigned long)kasan_reset_tag(area->addr) + area->size) ||
>-	    ((area->flags & (VM_ALLOC | VM_ALLOW_HUGE_VMAP)) != VM_ALLOC))
>+	    !(area->flags & VM_ALLOC))
>+               return -EINVAL;
>+
>+	/*
>+	 * Disallow VM_ALLOW_HUGE_VMAP mappings unless the region is PMD
>+	 * aligned, or splitting live huge mappings is supported.
>+	 */
>+	if ((area->flags & VM_ALLOW_HUGE_VMAP) &&
>+	   ((start % PMD_SIZE) || (size % PMD_SIZE)) &&

If I understand the intention here correctly, I don't think it is valid. Even   
if it is PMD-sized and PMD-aligned, it would still cause a split because  
the loop below is not using page order, but performs attribute changes 
page by page.                                                                     
                                                                             
Best regards,                                                               
Adrian