Re: [PATCH] mm/huge_memory: restrict __GFP_ZEROTAGS to HW tagging architectures

["David Hildenbrand (Red Hat)" <davidhildenbrandkernel@gmail.com>]

On 09.11.25 01:36, Jan Polensky wrote:
> The previous change added __GFP_ZEROTAGS when allocating the huge zero
> folio to ensure tag initialization for arm64 with MTE enabled. However,
> on s390 this flag is unnecessary and triggers a regression
> (observed as a crash during repeated 'dnf makecache').
> 
> Restrict the use of __GFP_ZEROTAGS to architectures that support
> hardware memory tagging (currently arm64 with MTE or KASAN HW tags).
> This avoids unintended side effects on other platforms.
> 
> Fixes: 1579227fe0f0 ("mm/huge_memory: initialise the tags of the huge zero folio")
> Link: https://lore.kernel.org/r/20251031170133.280742-1-catalin.marinas@arm.com
> Signed-off-by: Jan Polensky <japo@linux.ibm.com>
> ---
>   mm/huge_memory.c | 9 +++++----
>   1 file changed, 5 insertions(+), 4 deletions(-)
> 
> diff --git a/mm/huge_memory.c b/mm/huge_memory.c
> index aae283b00857..0c1794656d7a 100644
> --- a/mm/huge_memory.c
> +++ b/mm/huge_memory.c
> @@ -209,14 +209,15 @@ unsigned long __thp_vma_allowable_orders(struct vm_area_struct *vma,
> 
>   static bool get_huge_zero_folio(void)
>   {
> +	gfp_t gfp = (GFP_TRANSHUGE | __GFP_ZERO) & ~__GFP_MOVABLE;
>   	struct folio *zero_folio;
>   retry:
>   	if (likely(atomic_inc_not_zero(&huge_zero_refcount)))
>   		return true;
> -
> -	zero_folio = folio_alloc((GFP_TRANSHUGE | __GFP_ZERO | __GFP_ZEROTAGS) &
> -				 ~__GFP_MOVABLE,
> -			HPAGE_PMD_ORDER);
> +#if IS_ENABLED(CONFIG_KASAN_HW_TAGS) || IS_ENABLED(CONFIG_ARM64_MTE)
> +	gfp |= __GFP_ZEROTAGS;
> +#endif

That looks like the wrong approach. If an architecture does not support
__GFP_ZEROTAGS it should not trigger anything. __GFP_ZEROTAGS should be ignored.

I think the problem is that post_alloc_hook() does

if (zero_tags) {
	/* Initialize both memory and memory tags. */
	for (i = 0; i != 1 << order; ++i)
		tag_clear_highpage(page + i);

	/* Take note that memory was initialized by the loop above. */
	init = false;
}

And tag_clear_highpage() is a NOP on other architectures.

Gah.

I wonder if the following would work:


diff --git a/include/linux/gfp_types.h b/include/linux/gfp_types.h
index 65db9349f9053..56b82e116cb79 100644
--- a/include/linux/gfp_types.h
+++ b/include/linux/gfp_types.h
@@ -47,7 +47,9 @@ enum {
         ___GFP_HARDWALL_BIT,
         ___GFP_THISNODE_BIT,
         ___GFP_ACCOUNT_BIT,
+#ifdef __HAVE_ARCH_TAG_CLEAR_HIGHPAGE
         ___GFP_ZEROTAGS_BIT,
+#endif
  #ifdef CONFIG_KASAN_HW_TAGS
         ___GFP_SKIP_ZERO_BIT,
         ___GFP_SKIP_KASAN_BIT,
@@ -85,7 +87,11 @@ enum {
  #define ___GFP_HARDWALL                BIT(___GFP_HARDWALL_BIT)
  #define ___GFP_THISNODE                BIT(___GFP_THISNODE_BIT)
  #define ___GFP_ACCOUNT         BIT(___GFP_ACCOUNT_BIT)
+#ifdef __HAVE_ARCH_TAG_CLEAR_HIGHPAGE
  #define ___GFP_ZEROTAGS                BIT(___GFP_ZEROTAGS_BIT)
+#else
+#define ___GFP_ZEROTAGS                0
+#endif
  #ifdef CONFIG_KASAN_HW_TAGS
  #define ___GFP_SKIP_ZERO       BIT(___GFP_SKIP_ZERO_BIT)
  #define ___GFP_SKIP_KASAN      BIT(___GFP_SKIP_KASAN_BIT)


Likely we'd have to make __HAVE_ARCH_TAG_CLEAR_HIGHPAGE a proper
kconfig option.


Then we could turn the default implementation of
tag_clear_highpage() into a BUILD_BUG.