Re: [PATCH v5 14/14] mm: thp: always enable mTHP support

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

On 5/29/26 10:55 PM, 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
>   - Filter out PMD and PUD orders from allowable orders when
>     PMD-sized pages are not supported by the CPU
> 
> Signed-off-by: Luiz Capitulino <luizcap@redhat.com>
> ---
>   mm/huge_memory.c | 25 ++++++++++++++++++++-----
>   mm/shmem.c       | 14 +++++++++-----
>   2 files changed, 29 insertions(+), 10 deletions(-)
> 
> diff --git a/mm/huge_memory.c b/mm/huge_memory.c
> index 32254febe097..059901a8c6cb 100644
> --- a/mm/huge_memory.c
> +++ b/mm/huge_memory.c
> @@ -126,6 +126,15 @@ unsigned long __thp_vma_allowable_orders(struct vm_area_struct *vma,
>   	else
>   		supported_orders = THP_ORDERS_ALL_FILE_DEFAULT;
>   
> +	if (!pgtable_has_pmd_leaves()) {
> +		/*
> +		 * If the CPU does not support PMD leaves, assume for
> +		 * now that it does not support PUD leaves and disable
> +		 * both folio orders.
> +		 */
> +		supported_orders &= ~(BIT(PMD_ORDER) | BIT(PUD_ORDER));
> +	}
> +
>   	orders &= supported_orders;
>   	if (!orders)
>   		return 0;
> @@ -133,7 +142,7 @@ 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. */
> @@ -848,7 +857,7 @@ 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);
> @@ -870,6 +879,15 @@ 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()) {
> +		/*
> +		 * If the CPU does not support PMD leaves, assume for
> +		 * now that it does not support PUD leaves and disable
> +		 * both folio orders.
> +		 */
> +		orders &= ~(BIT(PMD_ORDER) | BIT(PUD_ORDER));
> +	}
> +
>   	order = highest_order(orders);
>   	while (orders) {
>   		thpsize = thpsize_create(order, *hugepage_kobj);
> @@ -969,9 +987,6 @@ 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 079e299ea789..c15dffd0eb41 100644
> --- a/mm/shmem.c
> +++ b/mm/shmem.c
> @@ -1844,16 +1844,19 @@ unsigned long shmem_allowable_huge_orders(struct inode *inode,
>   	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;
> -	unsigned int global_orders;
> +	unsigned int global_orders, disabled_orders = 0;
>   
> -	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;
>   
> +	if (!pgtable_has_pmd_leaves())
> +		disabled_orders = BIT(PMD_ORDER);
> +
>   	global_orders = shmem_huge_global_enabled(inode, index, write_end,
>   						  shmem_huge_force, vma, vm_flags);
>   	/* Tmpfs huge pages allocation */
>   	if (!vma || !vma_is_anon_shmem(vma))
> -		return global_orders;
> +		return global_orders & ~disabled_orders;
>   
>   	/*
>   	 * Following the 'deny' semantics of the top level, force the huge
> @@ -1867,7 +1870,7 @@ unsigned long shmem_allowable_huge_orders(struct inode *inode,
>   	 * means non-PMD sized THP can not override 'huge' mount option now.
>   	 */
>   	if (shmem_huge == SHMEM_HUGE_FORCE)
> -		return READ_ONCE(huge_shmem_orders_inherit);
> +		return READ_ONCE(huge_shmem_orders_inherit) & ~disabled_orders;
>   
>   	/* Allow mTHP that will be fully within i_size. */
>   	mask |= shmem_get_orders_within_size(inode, within_size_orders, index, 0);
> @@ -1878,6 +1881,7 @@ unsigned long shmem_allowable_huge_orders(struct inode *inode,
>   	if (global_orders > 0)
>   		mask |= READ_ONCE(huge_shmem_orders_inherit);
>   
> +	mask &= ~disabled_orders;
>   	return THP_ORDERS_ALL_FILE_DEFAULT & mask;
>   }
>   
> @@ -5461,7 +5465,7 @@ 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;

Seems we should also filter the PMD-sized order in shmem_getattr() if 
pgtable_has_pmd_leaves() returns false, like I mentioned in the previous 
link[1].

[1] 
https://lore.kernel.org/linux-mm/6591a74c-7ef9-4614-9ae9-cb2fbed86ebf@linux.alibaba.com/