Re: [PATCH v3 09/10] mm: thp: always enable mTHP support

[Baolin Wang <baolin.wang@linux.alibaba.com>]

On 4/9/26 4:23 AM, Luiz Capitulino wrote:
> If PMD-sized pages are not supported on an architecture (ie. the
> arch implements arch_has_pmd_leaves() and it returns false) then the
> current code disables all THP, including mTHP.
> 
> This commit fixes this by allowing mTHP to be always enabled for all
> archs. When PMD-sized pages are not supported, its sysfs entry won't be
> created and their mapping will be disallowed at page-fault time.
> 
> Similarly, this commit implements the following changes for shmem:
> 
>   - In shmem_allowable_huge_orders(): drop the pgtable_has_pmd_leaves()
>     check so that mTHP sizes are considered
>   - In shmem_alloc_and_add_folio(): don't consider PMD and PUD orders
>     when PMD-sized pages are not supported by the CPU
> 
> Signed-off-by: Luiz Capitulino <luizcap@redhat.com>
> ---
>   mm/huge_memory.c | 13 ++++++++-----
>   mm/shmem.c       |  4 +++-
>   2 files changed, 11 insertions(+), 6 deletions(-)
> 
> diff --git a/mm/huge_memory.c b/mm/huge_memory.c
> index 86e489c0a150..6de3d8ebc35c 100644
> --- a/mm/huge_memory.c
> +++ b/mm/huge_memory.c
> @@ -118,6 +118,9 @@ unsigned long __thp_vma_allowable_orders(struct vm_area_struct *vma,
>   	else
>   		supported_orders = THP_ORDERS_ALL_FILE_DEFAULT;
>   
> +	if (!pgtable_has_pmd_leaves())
> +		supported_orders &= ~(BIT(PMD_ORDER) | BIT(PUD_ORDER));
> +
>   	orders &= supported_orders;
>   	if (!orders)
>   		return 0;
> @@ -125,7 +128,7 @@ unsigned long __thp_vma_allowable_orders(struct vm_area_struct *vma,
>   	if (!vma->vm_mm)		/* vdso */
>   		return 0;
>   
> -	if (!pgtable_has_pmd_leaves() || vma_thp_disabled(vma, vm_flags, forced_collapse))
> +	if (vma_thp_disabled(vma, vm_flags, forced_collapse))
>   		return 0;
>   
>   	/* khugepaged doesn't collapse DAX vma, but page fault is fine. */
> @@ -787,7 +790,7 @@ static int __init hugepage_init_sysfs(struct kobject **hugepage_kobj)
>   	 * disable all other sizes. powerpc's PMD_ORDER isn't a compile-time
>   	 * constant so we have to do this here.
>   	 */
> -	if (!anon_orders_configured)
> +	if (!anon_orders_configured && pgtable_has_pmd_leaves())
>   		huge_anon_orders_inherit = BIT(PMD_ORDER);
>   
>   	*hugepage_kobj = kobject_create_and_add("transparent_hugepage", mm_kobj);
> @@ -809,6 +812,9 @@ static int __init hugepage_init_sysfs(struct kobject **hugepage_kobj)
>   	}
>   
>   	orders = THP_ORDERS_ALL_ANON | THP_ORDERS_ALL_FILE_DEFAULT;
> +	if (!pgtable_has_pmd_leaves())
> +		orders &= ~(BIT(PMD_ORDER) | BIT(PUD_ORDER));
> +
>   	order = highest_order(orders);
>   	while (orders) {
>   		thpsize = thpsize_create(order, *hugepage_kobj);
> @@ -908,9 +914,6 @@ static int __init hugepage_init(void)
>   	int err;
>   	struct kobject *hugepage_kobj;
>   
> -	if (!pgtable_has_pmd_leaves())
> -		return -EINVAL;
> -
>   	/*
>   	 * hugepages can't be allocated by the buddy allocator
>   	 */
> diff --git a/mm/shmem.c b/mm/shmem.c
> index 613393eae5a9..b49a30475cb0 100644
> --- a/mm/shmem.c
> +++ b/mm/shmem.c
> @@ -1839,7 +1839,7 @@ unsigned long shmem_allowable_huge_orders(struct inode *inode,
>   	vm_flags_t vm_flags = vma ? vma->vm_flags : 0;
>   	unsigned int global_orders;
>   
> -	if (!pgtable_has_pmd_leaves() || (vma && vma_thp_disabled(vma, vm_flags, shmem_huge_force)))
> +	if (vma && vma_thp_disabled(vma, vm_flags, shmem_huge_force))
>   		return 0;
>   
>   	global_orders = shmem_huge_global_enabled(inode, index, write_end,
> @@ -1947,6 +1947,8 @@ static struct folio *shmem_alloc_and_add_folio(struct vm_fault *vmf,
>   
>   	if (!IS_ENABLED(CONFIG_TRANSPARENT_HUGEPAGE))
>   		orders = 0;
> +	else if (!pgtable_has_pmd_leaves())
> +		orders &= ~(BIT(PMD_ORDER) | BIT(PUD_ORDER));
Sorry, I still don't like the changes here, because 
shmem_allowable_huge_orders() is meant to determine which large orders 
are allowed. Something like below (untested):

@@ -1834,6 +1834,7 @@ unsigned long shmem_allowable_huge_orders(struct 
inode *inode,
                                 struct vm_area_struct *vma, pgoff_t index,
                                 loff_t write_end, bool shmem_huge_force)
  {
+       unsigned int filter_orders = pgtable_has_pmd_leaves() ? 
(BIT(PMD_ORDER) | BIT(PUD_ORDER)) : 0;
         unsigned long mask = READ_ONCE(huge_shmem_orders_always);
         unsigned long within_size_orders = 
READ_ONCE(huge_shmem_orders_within_size);
         vm_flags_t vm_flags = vma ? vma->vm_flags : 0;
@@ -1846,7 +1847,7 @@ unsigned long shmem_allowable_huge_orders(struct 
inode *inode,
                                                   shmem_huge_force, 
vma, vm_flags);
         /* Tmpfs huge pages allocation */
         if (!vma || !vma_is_anon_shmem(vma))
-               return global_orders;
+               return global_orders & ~filter_orders;

         /*
          * Following the 'deny' semantics of the top level, force the huge
@@ -1871,6 +1872,7 @@ unsigned long shmem_allowable_huge_orders(struct 
inode *inode,
         if (global_orders > 0)
                 mask |= READ_ONCE(huge_shmem_orders_inherit);

+       mask &= ~filter_orders;
         return THP_ORDERS_ALL_FILE_DEFAULT & mask;
  }


Additionally, we also need a pgtable_has_pmd_leaves() check before 
setting huge_shmem_orders_inherit as well.

@@ -5428,7 +5430,7 @@ void __init shmem_init(void)
          * Default to setting PMD-sized THP to inherit the global 
setting and
          * disable all other multi-size THPs.
          */
-       if (!shmem_orders_configured)
+       if (!shmem_orders_configured && pgtable_has_pmd_leaves())
                 huge_shmem_orders_inherit = BIT(HPAGE_PMD_ORDER);
  #endif
         return;