[v4,5/7] mm: thp: split huge page to any lower order pages (except order-1).

  From: Zi Yan <ziy@nvidia.com>

To split a THP to any lower order (except order-1) pages, we need to
reform THPs on subpages at given order and add page refcount based on the
new page order. Also we need to reinitialize page_deferred_list after
removing the page from the split_queue, otherwise a subsequent split will
see list corruption when checking the page_deferred_list again.

It has many uses, like minimizing the number of pages after
truncating a huge pagecache page. For anonymous THPs, we can only split
them to order-0 like before until we add support for any size anonymous
THPs.

Order-1 folio is not supported because _deferred_list, which is used by
partially mapped folios, is stored in subpage 2 and an order-1 folio only
has subpage 0 and 1.

Signed-off-by: Zi Yan <ziy@nvidia.com>
---
 include/linux/huge_mm.h |  21 +++++---
 mm/huge_memory.c        | 114 +++++++++++++++++++++++++++++++---------
 2 files changed, 101 insertions(+), 34 deletions(-)
  

On Tue, Feb 13, 2024 at 04:55:18PM -0500, Zi Yan wrote:
> From: Zi Yan <ziy@nvidia.com>
> Order-1 folio is not supported because _deferred_list, which is used by
> partially mapped folios, is stored in subpage 2 and an order-1 folio only
> has subpage 0 and 1.

The LBS patches has the patch from Matthew which enables and allowed us
to successfully test order 1. So this restriction could be dropped if
that gets merged.

  Luis
  

On 13.02.24 23:05, Luis Chamberlain wrote:
> On Tue, Feb 13, 2024 at 04:55:18PM -0500, Zi Yan wrote:
>> From: Zi Yan <ziy@nvidia.com>
>> Order-1 folio is not supported because _deferred_list, which is used by
>> partially mapped folios, is stored in subpage 2 and an order-1 folio only
>> has subpage 0 and 1.
> 
> The LBS patches has the patch from Matthew which enables and allowed us
> to successfully test order 1. So this restriction could be dropped if
> that gets merged.

For anon folios it will still be in place, so the restriction will only 
be dropped for !anon.
  

On 13 Feb 2024, at 17:05, Luis Chamberlain wrote:

> On Tue, Feb 13, 2024 at 04:55:18PM -0500, Zi Yan wrote:
>> From: Zi Yan <ziy@nvidia.com>
>> Order-1 folio is not supported because _deferred_list, which is used by
>> partially mapped folios, is stored in subpage 2 and an order-1 folio only
>> has subpage 0 and 1.
>
> The LBS patches has the patch from Matthew which enables and allowed us
> to successfully test order 1. So this restriction could be dropped if
> that gets merged.

OK. But it only applies to file-backed folios IIUC. Anonymous folios still
cannot be split to order-1.

--
Best Regards,
Yan, Zi
  

On 13 Feb 2024, at 17:15, Zi Yan wrote:

> On 13 Feb 2024, at 17:05, Luis Chamberlain wrote:
>
>> On Tue, Feb 13, 2024 at 04:55:18PM -0500, Zi Yan wrote:
>>> From: Zi Yan <ziy@nvidia.com>
>>> Order-1 folio is not supported because _deferred_list, which is used by
>>> partially mapped folios, is stored in subpage 2 and an order-1 folio only
>>> has subpage 0 and 1.
>>
>> The LBS patches has the patch from Matthew which enables and allowed us
>> to successfully test order 1. So this restriction could be dropped if
>> that gets merged.
>
> OK. But it only applies to file-backed folios IIUC. Anonymous folios still
> cannot be split to order-1.

Something like this would lift the restriction:

diff --git a/mm/huge_memory.c b/mm/huge_memory.c
index 0564b007cbd1..7ee7f1906617 100644
--- a/mm/huge_memory.c
+++ b/mm/huge_memory.c
@@ -3045,9 +3045,9 @@ int split_huge_page_to_list_to_order(struct page *page, struct list_head *list,
        VM_BUG_ON_FOLIO(!folio_test_locked(folio), folio);
        VM_BUG_ON_FOLIO(!folio_test_large(folio), folio);

-       /* Cannot split THP to order-1 (no order-1 THPs) */
-       if (new_order == 1) {
-               VM_WARN_ONCE(1, "Cannot split to order-1 folio");
+       /* Cannot split anonymous folios to order-1 (no order-1 anon folios) */
+       if (new_order == 1 && folio_test_anon(folio)) {
+               VM_WARN_ONCE(1, "Cannot split to order-1 anonymous folio");
                return -EINVAL;
        }

--
Best Regards,
Yan, Zi
  

On 13 Feb 2024, at 17:19, Zi Yan wrote:

> On 13 Feb 2024, at 17:15, Zi Yan wrote:
>
>> On 13 Feb 2024, at 17:05, Luis Chamberlain wrote:
>>
>>> On Tue, Feb 13, 2024 at 04:55:18PM -0500, Zi Yan wrote:
>>>> From: Zi Yan <ziy@nvidia.com>
>>>> Order-1 folio is not supported because _deferred_list, which is used by
>>>> partially mapped folios, is stored in subpage 2 and an order-1 folio only
>>>> has subpage 0 and 1.
>>>
>>> The LBS patches has the patch from Matthew which enables and allowed us
>>> to successfully test order 1. So this restriction could be dropped if
>>> that gets merged.
>>
>> OK. But it only applies to file-backed folios IIUC. Anonymous folios still
>> cannot be split to order-1.
>
> Something like this would lift the restriction:

Actually this, since folio_prep_large_rmappable() is changed by that patch:

diff --git a/mm/huge_memory.c b/mm/huge_memory.c
index 0564b007cbd1..05eeeafaa9dc 100644
--- a/mm/huge_memory.c
+++ b/mm/huge_memory.c
@@ -2852,7 +2852,7 @@ static void __split_huge_page_tail(struct folio *folio, int tail,
        clear_compound_head(page_tail);
        if (new_order) {
                prep_compound_page(page_tail, new_order);
-               folio_prep_large_rmappable(page_folio(page_tail));
+               new_folio = folio_prep_large_rmappable(new_folio);
        }

        /* Finally unfreeze refcount. Additional reference from page cache. */
@@ -3045,9 +3045,9 @@ int split_huge_page_to_list_to_order(struct page *page, struct list_head *list,
        VM_BUG_ON_FOLIO(!folio_test_locked(folio), folio);
        VM_BUG_ON_FOLIO(!folio_test_large(folio), folio);

-       /* Cannot split THP to order-1 (no order-1 THPs) */
-       if (new_order == 1) {
-               VM_WARN_ONCE(1, "Cannot split to order-1 folio");
+       /* Cannot split anonymous folios to order-1 (no order-1 anon folios) */
+       if (new_order == 1 && folio_test_anon(folio)) {
+               VM_WARN_ONCE(1, "Cannot split to order-1 anonymous folio");
                return -EINVAL;
        }

--
Best Regards,
Yan, Zi
  

On 13/02/2024 21:55, Zi Yan wrote:
> From: Zi Yan <ziy@nvidia.com>
> 
> To split a THP to any lower order (except order-1) pages, we need to
> reform THPs on subpages at given order and add page refcount based on the
> new page order. Also we need to reinitialize page_deferred_list after
> removing the page from the split_queue, otherwise a subsequent split will
> see list corruption when checking the page_deferred_list again.
> 
> It has many uses, like minimizing the number of pages after
> truncating a huge pagecache page. For anonymous THPs, we can only split
> them to order-0 like before until we add support for any size anonymous
> THPs.

multi-size THP is now upstream. Not sure if this comment still makes sense.
Still its not completely clear to me how you would integrate this new machinery
and decide what non-zero order to split anon THP to?

> 
> Order-1 folio is not supported because _deferred_list, which is used by
> partially mapped folios, is stored in subpage 2 and an order-1 folio only
> has subpage 0 and 1.
> 
> Signed-off-by: Zi Yan <ziy@nvidia.com>
> ---
>  include/linux/huge_mm.h |  21 +++++---
>  mm/huge_memory.c        | 114 +++++++++++++++++++++++++++++++---------
>  2 files changed, 101 insertions(+), 34 deletions(-)
> 
> diff --git a/include/linux/huge_mm.h b/include/linux/huge_mm.h
> index 5adb86af35fc..de0c89105076 100644
> --- a/include/linux/huge_mm.h
> +++ b/include/linux/huge_mm.h
> @@ -265,10 +265,11 @@ unsigned long thp_get_unmapped_area(struct file *filp, unsigned long addr,
>  
>  void folio_prep_large_rmappable(struct folio *folio);
>  bool can_split_folio(struct folio *folio, int *pextra_pins);
> -int split_huge_page_to_list(struct page *page, struct list_head *list);
> +int split_huge_page_to_list_to_order(struct page *page, struct list_head *list,
> +		unsigned int new_order);
>  static inline int split_huge_page(struct page *page)
>  {
> -	return split_huge_page_to_list(page, NULL);
> +	return split_huge_page_to_list_to_order(page, NULL, 0);
>  }
>  void deferred_split_folio(struct folio *folio);
>  
> @@ -422,7 +423,8 @@ can_split_folio(struct folio *folio, int *pextra_pins)
>  	return false;
>  }
>  static inline int
> -split_huge_page_to_list(struct page *page, struct list_head *list)
> +split_huge_page_to_list_to_order(struct page *page, struct list_head *list,
> +		unsigned int new_order)
>  {
>  	return 0;
>  }
> @@ -519,17 +521,20 @@ static inline bool thp_migration_supported(void)
>  }
>  #endif /* CONFIG_TRANSPARENT_HUGEPAGE */
>  
> -static inline int split_folio_to_list(struct folio *folio,
> -		struct list_head *list)
> +static inline int split_folio_to_list_to_order(struct folio *folio,
> +		struct list_head *list, int new_order)
>  {
> -	return split_huge_page_to_list(&folio->page, list);
> +	return split_huge_page_to_list_to_order(&folio->page, list, new_order);
>  }
>  
> -static inline int split_folio(struct folio *folio)
> +static inline int split_folio_to_order(struct folio *folio, int new_order)
>  {
> -	return split_folio_to_list(folio, NULL);
> +	return split_folio_to_list_to_order(folio, NULL, new_order);
>  }
>  
> +#define split_folio_to_list(f, l) split_folio_to_list_to_order(f, l, 0)
> +#define split_folio(f) split_folio_to_order(f, 0)
> +
>  /*
>   * archs that select ARCH_WANTS_THP_SWAP but don't support THP_SWP due to
>   * limitations in the implementation like arm64 MTE can override this to
> diff --git a/mm/huge_memory.c b/mm/huge_memory.c
> index ad7133c97428..d0e555a8ea98 100644
> --- a/mm/huge_memory.c
> +++ b/mm/huge_memory.c
> @@ -2718,11 +2718,14 @@ void vma_adjust_trans_huge(struct vm_area_struct *vma,
>  
>  static void unmap_folio(struct folio *folio)
>  {
> -	enum ttu_flags ttu_flags = TTU_RMAP_LOCKED | TTU_SPLIT_HUGE_PMD |
> -		TTU_SYNC | TTU_BATCH_FLUSH;
> +	enum ttu_flags ttu_flags = TTU_RMAP_LOCKED | TTU_SYNC |
> +		TTU_BATCH_FLUSH;
>  
>  	VM_BUG_ON_FOLIO(!folio_test_large(folio), folio);
>  
> +	if (folio_test_pmd_mappable(folio))
> +		ttu_flags |= TTU_SPLIT_HUGE_PMD;

Should we split this change out? I think it makes sense independent of this series?

> +
>  	/*
>  	 * Anon pages need migration entries to preserve them, but file
>  	 * pages can simply be left unmapped, then faulted back on demand.
> @@ -2756,7 +2759,6 @@ static void lru_add_page_tail(struct page *head, struct page *tail,
>  		struct lruvec *lruvec, struct list_head *list)
>  {
>  	VM_BUG_ON_PAGE(!PageHead(head), head);
> -	VM_BUG_ON_PAGE(PageCompound(tail), head);
>  	VM_BUG_ON_PAGE(PageLRU(tail), head);
>  	lockdep_assert_held(&lruvec->lru_lock);
>  
> @@ -2777,7 +2779,8 @@ static void lru_add_page_tail(struct page *head, struct page *tail,
>  }
>  
>  static void __split_huge_page_tail(struct folio *folio, int tail,
> -		struct lruvec *lruvec, struct list_head *list)
> +		struct lruvec *lruvec, struct list_head *list,
> +		unsigned int new_order)
>  {
>  	struct page *head = &folio->page;
>  	struct page *page_tail = head + tail;
> @@ -2847,10 +2850,15 @@ static void __split_huge_page_tail(struct folio *folio, int tail,
>  	 * which needs correct compound_head().
>  	 */
>  	clear_compound_head(page_tail);
> +	if (new_order) {
> +		prep_compound_page(page_tail, new_order);
> +		folio_prep_large_rmappable(page_folio(page_tail));
> +	}
>  
>  	/* Finally unfreeze refcount. Additional reference from page cache. */
> -	page_ref_unfreeze(page_tail, 1 + (!folio_test_anon(folio) ||
> -					  folio_test_swapcache(folio)));
> +	page_ref_unfreeze(page_tail,
> +		1 + ((!folio_test_anon(folio) || folio_test_swapcache(folio)) ?
> +			     folio_nr_pages(page_folio(page_tail)) : 0));
>  
>  	if (folio_test_young(folio))
>  		folio_set_young(new_folio);
> @@ -2868,7 +2876,7 @@ static void __split_huge_page_tail(struct folio *folio, int tail,
>  }
>  
>  static void __split_huge_page(struct page *page, struct list_head *list,
> -		pgoff_t end)
> +		pgoff_t end, unsigned int new_order)
>  {
>  	struct folio *folio = page_folio(page);
>  	struct page *head = &folio->page;
> @@ -2877,10 +2885,11 @@ static void __split_huge_page(struct page *page, struct list_head *list,
>  	unsigned long offset = 0;
>  	unsigned int nr = thp_nr_pages(head);
>  	int i, nr_dropped = 0;
> +	unsigned int new_nr = 1 << new_order;
>  	int order = folio_order(folio);
>  
>  	/* complete memcg works before add pages to LRU */
> -	split_page_memcg(head, order, 0);
> +	split_page_memcg(head, order, new_order);
>  
>  	if (folio_test_anon(folio) && folio_test_swapcache(folio)) {
>  		offset = swp_offset(folio->swap);
> @@ -2893,8 +2902,8 @@ static void __split_huge_page(struct page *page, struct list_head *list,
>  
>  	ClearPageHasHWPoisoned(head);
>  
> -	for (i = nr - 1; i >= 1; i--) {
> -		__split_huge_page_tail(folio, i, lruvec, list);
> +	for (i = nr - new_nr; i >= new_nr; i -= new_nr) {
> +		__split_huge_page_tail(folio, i, lruvec, list, new_order);
>  		/* Some pages can be beyond EOF: drop them from page cache */
>  		if (head[i].index >= end) {
>  			struct folio *tail = page_folio(head + i);
> @@ -2910,29 +2919,41 @@ static void __split_huge_page(struct page *page, struct list_head *list,
>  			__xa_store(&head->mapping->i_pages, head[i].index,
>  					head + i, 0);
>  		} else if (swap_cache) {
> +			/*
> +			 * split anonymous THPs (including swapped out ones) to
> +			 * non-zero order not supported
> +			 */
> +			VM_WARN_ONCE(new_order,
> +				"Split swap-cached anon folio to non-0 order not supported");

Why isn't it supported? Even if it's not supported, is this level the right
place to enforce these kinds of policy decisions? I wonder if we should be
leaving that to the higher level to decide?

Thanks,
Ryan

>  			__xa_store(&swap_cache->i_pages, offset + i,
>  					head + i, 0);
>  		}
>  	}
>  
> -	ClearPageCompound(head);
> +	if (!new_order)
> +		ClearPageCompound(head);
> +	else {
> +		struct folio *new_folio = (struct folio *)head;
> +
> +		folio_set_order(new_folio, new_order);
> +	}
>  	unlock_page_lruvec(lruvec);
>  	/* Caller disabled irqs, so they are still disabled here */
>  
> -	split_page_owner(head, order, 0);
> +	split_page_owner(head, order, new_order);
>  
>  	/* See comment in __split_huge_page_tail() */
>  	if (PageAnon(head)) {
>  		/* Additional pin to swap cache */
>  		if (PageSwapCache(head)) {
> -			page_ref_add(head, 2);
> +			page_ref_add(head, 1 + new_nr);
>  			xa_unlock(&swap_cache->i_pages);
>  		} else {
>  			page_ref_inc(head);
>  		}
>  	} else {
>  		/* Additional pin to page cache */
> -		page_ref_add(head, 2);
> +		page_ref_add(head, 1 + new_nr);
>  		xa_unlock(&head->mapping->i_pages);
>  	}
>  	local_irq_enable();
> @@ -2944,7 +2965,15 @@ static void __split_huge_page(struct page *page, struct list_head *list,
>  	if (folio_test_swapcache(folio))
>  		split_swap_cluster(folio->swap);
>  
> -	for (i = 0; i < nr; i++) {
> +	/*
> +	 * set page to its compound_head when split to non order-0 pages, so
> +	 * we can skip unlocking it below, since PG_locked is transferred to
> +	 * the compound_head of the page and the caller will unlock it.
> +	 */
> +	if (new_order)
> +		page = compound_head(page);
> +
> +	for (i = 0; i < nr; i += new_nr) {
>  		struct page *subpage = head + i;
>  		if (subpage == page)
>  			continue;
> @@ -2978,29 +3007,35 @@ bool can_split_folio(struct folio *folio, int *pextra_pins)
>  }
>  
>  /*
> - * This function splits huge page into normal pages. @page can point to any
> - * subpage of huge page to split. Split doesn't change the position of @page.
> + * This function splits huge page into pages in @new_order. @page can point to
> + * any subpage of huge page to split. Split doesn't change the position of
> + * @page.
> + *
> + * NOTE: order-1 folio is not supported because _deferred_list, which is used
> + * by partially mapped folios, is stored in subpage 2 and an order-1 folio
> + * only has subpage 0 and 1.
>   *
>   * Only caller must hold pin on the @page, otherwise split fails with -EBUSY.
>   * The huge page must be locked.
>   *
>   * If @list is null, tail pages will be added to LRU list, otherwise, to @list.
>   *
> - * Both head page and tail pages will inherit mapping, flags, and so on from
> - * the hugepage.
> + * Pages in new_order will inherit mapping, flags, and so on from the hugepage.
>   *
> - * GUP pin and PG_locked transferred to @page. Rest subpages can be freed if
> - * they are not mapped.
> + * GUP pin and PG_locked transferred to @page or the compound page @page belongs
> + * to. Rest subpages can be freed if they are not mapped.
>   *
>   * Returns 0 if the hugepage is split successfully.
>   * Returns -EBUSY if the page is pinned or if anon_vma disappeared from under
>   * us.
>   */
> -int split_huge_page_to_list(struct page *page, struct list_head *list)
> +int split_huge_page_to_list_to_order(struct page *page, struct list_head *list,
> +				     unsigned int new_order)
>  {
>  	struct folio *folio = page_folio(page);
>  	struct deferred_split *ds_queue = get_deferred_split_queue(folio);
> -	XA_STATE(xas, &folio->mapping->i_pages, folio->index);
> +	/* reset xarray order to new order after split */
> +	XA_STATE_ORDER(xas, &folio->mapping->i_pages, folio->index, new_order);
>  	struct anon_vma *anon_vma = NULL;
>  	struct address_space *mapping = NULL;
>  	int extra_pins, ret;
> @@ -3010,6 +3045,26 @@ int split_huge_page_to_list(struct page *page, struct list_head *list)
>  	VM_BUG_ON_FOLIO(!folio_test_locked(folio), folio);
>  	VM_BUG_ON_FOLIO(!folio_test_large(folio), folio);
>  
> +	/* Cannot split THP to order-1 (no order-1 THPs) */
> +	if (new_order == 1) {
> +		VM_WARN_ONCE(1, "Cannot split to order-1 folio");
> +		return -EINVAL;
> +	}
> +
> +	if (new_order) {
> +		/* Split shmem folio to non-zero order not supported */
> +		if (shmem_mapping(folio->mapping)) {
> +			VM_WARN_ONCE(1, "Split shmem folio to non-0 order not support");
> +			return -EINVAL;
> +		}
> +		/* No split if the file system does not support large folio */
> +		if (!mapping_large_folio_support(folio->mapping)) {
> +			VM_WARN_ONCE(1, "Split file folio to non-0 order not support");
> +			return -EINVAL;
> +		}
> +	}
> +
> +
>  	is_hzp = is_huge_zero_page(&folio->page);
>  	if (is_hzp) {
>  		pr_warn_ratelimited("Called split_huge_page for huge zero page\n");
> @@ -3105,14 +3160,21 @@ int split_huge_page_to_list(struct page *page, struct list_head *list)
>  	if (folio_ref_freeze(folio, 1 + extra_pins)) {
>  		if (!list_empty(&folio->_deferred_list)) {
>  			ds_queue->split_queue_len--;
> -			list_del(&folio->_deferred_list);
> +			/*
> +			 * Reinitialize page_deferred_list after removing the
> +			 * page from the split_queue, otherwise a subsequent
> +			 * split will see list corruption when checking the
> +			 * page_deferred_list.
> +			 */
> +			list_del_init(&folio->_deferred_list);
>  		}
>  		spin_unlock(&ds_queue->split_queue_lock);
>  		if (mapping) {
>  			int nr = folio_nr_pages(folio);
>  
>  			xas_split(&xas, folio, folio_order(folio));
> -			if (folio_test_pmd_mappable(folio)) {
> +			if (folio_test_pmd_mappable(folio) &&
> +			    new_order < HPAGE_PMD_ORDER) {
>  				if (folio_test_swapbacked(folio)) {
>  					__lruvec_stat_mod_folio(folio,
>  							NR_SHMEM_THPS, -nr);
> @@ -3124,7 +3186,7 @@ int split_huge_page_to_list(struct page *page, struct list_head *list)
>  			}
>  		}
>  
> -		__split_huge_page(page, list, end);
> +		__split_huge_page(page, list, end, new_order);
>  		ret = 0;
>  	} else {
>  		spin_unlock(&ds_queue->split_queue_lock);
  

On 14 Feb 2024, at 5:38, Ryan Roberts wrote:

> On 13/02/2024 21:55, Zi Yan wrote:
>> From: Zi Yan <ziy@nvidia.com>
>>
>> To split a THP to any lower order (except order-1) pages, we need to
>> reform THPs on subpages at given order and add page refcount based on the
>> new page order. Also we need to reinitialize page_deferred_list after
>> removing the page from the split_queue, otherwise a subsequent split will
>> see list corruption when checking the page_deferred_list again.
>>
>> It has many uses, like minimizing the number of pages after
>> truncating a huge pagecache page. For anonymous THPs, we can only split
>> them to order-0 like before until we add support for any size anonymous
>> THPs.
>
> multi-size THP is now upstream. Not sure if this comment still makes sense.
Will change it to reflect the fact that multi-size THP is already upstream.

> Still its not completely clear to me how you would integrate this new machinery
> and decide what non-zero order to split anon THP to?

Originally, it was developed along with my 1GB THP support. So it was intended
to split order-18 to order-9. But for now, like you and David said in the cover
letter email thread, we might not want to use it for anonymous large folios
until we find a necessary use case.

>>
>> Order-1 folio is not supported because _deferred_list, which is used by
>> partially mapped folios, is stored in subpage 2 and an order-1 folio only
>> has subpage 0 and 1.
>>
>> Signed-off-by: Zi Yan <ziy@nvidia.com>
>> ---
>>  include/linux/huge_mm.h |  21 +++++---
>>  mm/huge_memory.c        | 114 +++++++++++++++++++++++++++++++---------
>>  2 files changed, 101 insertions(+), 34 deletions(-)
>>
>> diff --git a/include/linux/huge_mm.h b/include/linux/huge_mm.h
>> index 5adb86af35fc..de0c89105076 100644
>> --- a/include/linux/huge_mm.h
>> +++ b/include/linux/huge_mm.h
>> @@ -265,10 +265,11 @@ unsigned long thp_get_unmapped_area(struct file *filp, unsigned long addr,
>>
>>  void folio_prep_large_rmappable(struct folio *folio);
>>  bool can_split_folio(struct folio *folio, int *pextra_pins);
>> -int split_huge_page_to_list(struct page *page, struct list_head *list);
>> +int split_huge_page_to_list_to_order(struct page *page, struct list_head *list,
>> +		unsigned int new_order);
>>  static inline int split_huge_page(struct page *page)
>>  {
>> -	return split_huge_page_to_list(page, NULL);
>> +	return split_huge_page_to_list_to_order(page, NULL, 0);
>>  }
>>  void deferred_split_folio(struct folio *folio);
>>
>> @@ -422,7 +423,8 @@ can_split_folio(struct folio *folio, int *pextra_pins)
>>  	return false;
>>  }
>>  static inline int
>> -split_huge_page_to_list(struct page *page, struct list_head *list)
>> +split_huge_page_to_list_to_order(struct page *page, struct list_head *list,
>> +		unsigned int new_order)
>>  {
>>  	return 0;
>>  }
>> @@ -519,17 +521,20 @@ static inline bool thp_migration_supported(void)
>>  }
>>  #endif /* CONFIG_TRANSPARENT_HUGEPAGE */
>>
>> -static inline int split_folio_to_list(struct folio *folio,
>> -		struct list_head *list)
>> +static inline int split_folio_to_list_to_order(struct folio *folio,
>> +		struct list_head *list, int new_order)
>>  {
>> -	return split_huge_page_to_list(&folio->page, list);
>> +	return split_huge_page_to_list_to_order(&folio->page, list, new_order);
>>  }
>>
>> -static inline int split_folio(struct folio *folio)
>> +static inline int split_folio_to_order(struct folio *folio, int new_order)
>>  {
>> -	return split_folio_to_list(folio, NULL);
>> +	return split_folio_to_list_to_order(folio, NULL, new_order);
>>  }
>>
>> +#define split_folio_to_list(f, l) split_folio_to_list_to_order(f, l, 0)
>> +#define split_folio(f) split_folio_to_order(f, 0)
>> +
>>  /*
>>   * archs that select ARCH_WANTS_THP_SWAP but don't support THP_SWP due to
>>   * limitations in the implementation like arm64 MTE can override this to
>> diff --git a/mm/huge_memory.c b/mm/huge_memory.c
>> index ad7133c97428..d0e555a8ea98 100644
>> --- a/mm/huge_memory.c
>> +++ b/mm/huge_memory.c
>> @@ -2718,11 +2718,14 @@ void vma_adjust_trans_huge(struct vm_area_struct *vma,
>>
>>  static void unmap_folio(struct folio *folio)
>>  {
>> -	enum ttu_flags ttu_flags = TTU_RMAP_LOCKED | TTU_SPLIT_HUGE_PMD |
>> -		TTU_SYNC | TTU_BATCH_FLUSH;
>> +	enum ttu_flags ttu_flags = TTU_RMAP_LOCKED | TTU_SYNC |
>> +		TTU_BATCH_FLUSH;
>>
>>  	VM_BUG_ON_FOLIO(!folio_test_large(folio), folio);
>>
>> +	if (folio_test_pmd_mappable(folio))
>> +		ttu_flags |= TTU_SPLIT_HUGE_PMD;
>
> Should we split this change out? I think it makes sense independent of this series?
>

Sure. Since multi-size THP is upstream, this avoid unnecessary code path if
the THP is not PMD-mapped.

>> +
>>  	/*
>>  	 * Anon pages need migration entries to preserve them, but file
>>  	 * pages can simply be left unmapped, then faulted back on demand.
>> @@ -2756,7 +2759,6 @@ static void lru_add_page_tail(struct page *head, struct page *tail,
>>  		struct lruvec *lruvec, struct list_head *list)
>>  {
>>  	VM_BUG_ON_PAGE(!PageHead(head), head);
>> -	VM_BUG_ON_PAGE(PageCompound(tail), head);
>>  	VM_BUG_ON_PAGE(PageLRU(tail), head);
>>  	lockdep_assert_held(&lruvec->lru_lock);
>>
>> @@ -2777,7 +2779,8 @@ static void lru_add_page_tail(struct page *head, struct page *tail,
>>  }
>>
>>  static void __split_huge_page_tail(struct folio *folio, int tail,
>> -		struct lruvec *lruvec, struct list_head *list)
>> +		struct lruvec *lruvec, struct list_head *list,
>> +		unsigned int new_order)
>>  {
>>  	struct page *head = &folio->page;
>>  	struct page *page_tail = head + tail;
>> @@ -2847,10 +2850,15 @@ static void __split_huge_page_tail(struct folio *folio, int tail,
>>  	 * which needs correct compound_head().
>>  	 */
>>  	clear_compound_head(page_tail);
>> +	if (new_order) {
>> +		prep_compound_page(page_tail, new_order);
>> +		folio_prep_large_rmappable(page_folio(page_tail));
>> +	}
>>
>>  	/* Finally unfreeze refcount. Additional reference from page cache. */
>> -	page_ref_unfreeze(page_tail, 1 + (!folio_test_anon(folio) ||
>> -					  folio_test_swapcache(folio)));
>> +	page_ref_unfreeze(page_tail,
>> +		1 + ((!folio_test_anon(folio) || folio_test_swapcache(folio)) ?
>> +			     folio_nr_pages(page_folio(page_tail)) : 0));
>>
>>  	if (folio_test_young(folio))
>>  		folio_set_young(new_folio);
>> @@ -2868,7 +2876,7 @@ static void __split_huge_page_tail(struct folio *folio, int tail,
>>  }
>>
>>  static void __split_huge_page(struct page *page, struct list_head *list,
>> -		pgoff_t end)
>> +		pgoff_t end, unsigned int new_order)
>>  {
>>  	struct folio *folio = page_folio(page);
>>  	struct page *head = &folio->page;
>> @@ -2877,10 +2885,11 @@ static void __split_huge_page(struct page *page, struct list_head *list,
>>  	unsigned long offset = 0;
>>  	unsigned int nr = thp_nr_pages(head);
>>  	int i, nr_dropped = 0;
>> +	unsigned int new_nr = 1 << new_order;
>>  	int order = folio_order(folio);
>>
>>  	/* complete memcg works before add pages to LRU */
>> -	split_page_memcg(head, order, 0);
>> +	split_page_memcg(head, order, new_order);
>>
>>  	if (folio_test_anon(folio) && folio_test_swapcache(folio)) {
>>  		offset = swp_offset(folio->swap);
>> @@ -2893,8 +2902,8 @@ static void __split_huge_page(struct page *page, struct list_head *list,
>>
>>  	ClearPageHasHWPoisoned(head);
>>
>> -	for (i = nr - 1; i >= 1; i--) {
>> -		__split_huge_page_tail(folio, i, lruvec, list);
>> +	for (i = nr - new_nr; i >= new_nr; i -= new_nr) {
>> +		__split_huge_page_tail(folio, i, lruvec, list, new_order);
>>  		/* Some pages can be beyond EOF: drop them from page cache */
>>  		if (head[i].index >= end) {
>>  			struct folio *tail = page_folio(head + i);
>> @@ -2910,29 +2919,41 @@ static void __split_huge_page(struct page *page, struct list_head *list,
>>  			__xa_store(&head->mapping->i_pages, head[i].index,
>>  					head + i, 0);
>>  		} else if (swap_cache) {
>> +			/*
>> +			 * split anonymous THPs (including swapped out ones) to
>> +			 * non-zero order not supported
>> +			 */
>> +			VM_WARN_ONCE(new_order,
>> +				"Split swap-cached anon folio to non-0 order not supported");
>
> Why isn't it supported? Even if it's not supported, is this level the right
> place to enforce these kinds of policy decisions? I wonder if we should be
> leaving that to the higher level to decide?

Is the swap-out small-size THP without splitting merged? This needs that patchset.
You are right that a warning here is not appropriate. I will fail the splitting
if the folio is swapcached and going to be split into >0 order.

>>  			__xa_store(&swap_cache->i_pages, offset + i,
>>  					head + i, 0);
>>  		}
>>  	}
>>


--
Best Regards,
Yan, Zi
  

On 14/02/2024 16:11, Zi Yan wrote:
> On 14 Feb 2024, at 5:38, Ryan Roberts wrote:
> 
>> On 13/02/2024 21:55, Zi Yan wrote:
>>> From: Zi Yan <ziy@nvidia.com>
>>>
>>> To split a THP to any lower order (except order-1) pages, we need to
>>> reform THPs on subpages at given order and add page refcount based on the
>>> new page order. Also we need to reinitialize page_deferred_list after
>>> removing the page from the split_queue, otherwise a subsequent split will
>>> see list corruption when checking the page_deferred_list again.
>>>
>>> It has many uses, like minimizing the number of pages after
>>> truncating a huge pagecache page. For anonymous THPs, we can only split
>>> them to order-0 like before until we add support for any size anonymous
>>> THPs.
>>
>> multi-size THP is now upstream. Not sure if this comment still makes sense.
> Will change it to reflect the fact that multi-size THP is already upstream.
> 
>> Still its not completely clear to me how you would integrate this new machinery
>> and decide what non-zero order to split anon THP to?
> 
> Originally, it was developed along with my 1GB THP support. So it was intended
> to split order-18 to order-9. But for now, like you and David said in the cover
> letter email thread, we might not want to use it for anonymous large folios
> until we find a necessary use case.
> 
>>>
>>> Order-1 folio is not supported because _deferred_list, which is used by
>>> partially mapped folios, is stored in subpage 2 and an order-1 folio only
>>> has subpage 0 and 1.
>>>
>>> Signed-off-by: Zi Yan <ziy@nvidia.com>
>>> ---
>>>  include/linux/huge_mm.h |  21 +++++---
>>>  mm/huge_memory.c        | 114 +++++++++++++++++++++++++++++++---------
>>>  2 files changed, 101 insertions(+), 34 deletions(-)
>>>
>>> diff --git a/include/linux/huge_mm.h b/include/linux/huge_mm.h
>>> index 5adb86af35fc..de0c89105076 100644
>>> --- a/include/linux/huge_mm.h
>>> +++ b/include/linux/huge_mm.h
>>> @@ -265,10 +265,11 @@ unsigned long thp_get_unmapped_area(struct file *filp, unsigned long addr,
>>>
>>>  void folio_prep_large_rmappable(struct folio *folio);
>>>  bool can_split_folio(struct folio *folio, int *pextra_pins);
>>> -int split_huge_page_to_list(struct page *page, struct list_head *list);
>>> +int split_huge_page_to_list_to_order(struct page *page, struct list_head *list,
>>> +		unsigned int new_order);
>>>  static inline int split_huge_page(struct page *page)
>>>  {
>>> -	return split_huge_page_to_list(page, NULL);
>>> +	return split_huge_page_to_list_to_order(page, NULL, 0);
>>>  }
>>>  void deferred_split_folio(struct folio *folio);
>>>
>>> @@ -422,7 +423,8 @@ can_split_folio(struct folio *folio, int *pextra_pins)
>>>  	return false;
>>>  }
>>>  static inline int
>>> -split_huge_page_to_list(struct page *page, struct list_head *list)
>>> +split_huge_page_to_list_to_order(struct page *page, struct list_head *list,
>>> +		unsigned int new_order)
>>>  {
>>>  	return 0;
>>>  }
>>> @@ -519,17 +521,20 @@ static inline bool thp_migration_supported(void)
>>>  }
>>>  #endif /* CONFIG_TRANSPARENT_HUGEPAGE */
>>>
>>> -static inline int split_folio_to_list(struct folio *folio,
>>> -		struct list_head *list)
>>> +static inline int split_folio_to_list_to_order(struct folio *folio,
>>> +		struct list_head *list, int new_order)
>>>  {
>>> -	return split_huge_page_to_list(&folio->page, list);
>>> +	return split_huge_page_to_list_to_order(&folio->page, list, new_order);
>>>  }
>>>
>>> -static inline int split_folio(struct folio *folio)
>>> +static inline int split_folio_to_order(struct folio *folio, int new_order)
>>>  {
>>> -	return split_folio_to_list(folio, NULL);
>>> +	return split_folio_to_list_to_order(folio, NULL, new_order);
>>>  }
>>>
>>> +#define split_folio_to_list(f, l) split_folio_to_list_to_order(f, l, 0)
>>> +#define split_folio(f) split_folio_to_order(f, 0)
>>> +
>>>  /*
>>>   * archs that select ARCH_WANTS_THP_SWAP but don't support THP_SWP due to
>>>   * limitations in the implementation like arm64 MTE can override this to
>>> diff --git a/mm/huge_memory.c b/mm/huge_memory.c
>>> index ad7133c97428..d0e555a8ea98 100644
>>> --- a/mm/huge_memory.c
>>> +++ b/mm/huge_memory.c
>>> @@ -2718,11 +2718,14 @@ void vma_adjust_trans_huge(struct vm_area_struct *vma,
>>>
>>>  static void unmap_folio(struct folio *folio)
>>>  {
>>> -	enum ttu_flags ttu_flags = TTU_RMAP_LOCKED | TTU_SPLIT_HUGE_PMD |
>>> -		TTU_SYNC | TTU_BATCH_FLUSH;
>>> +	enum ttu_flags ttu_flags = TTU_RMAP_LOCKED | TTU_SYNC |
>>> +		TTU_BATCH_FLUSH;
>>>
>>>  	VM_BUG_ON_FOLIO(!folio_test_large(folio), folio);
>>>
>>> +	if (folio_test_pmd_mappable(folio))
>>> +		ttu_flags |= TTU_SPLIT_HUGE_PMD;
>>
>> Should we split this change out? I think it makes sense independent of this series?
>>
> 
> Sure. Since multi-size THP is upstream, this avoid unnecessary code path if
> the THP is not PMD-mapped.
> 
>>> +
>>>  	/*
>>>  	 * Anon pages need migration entries to preserve them, but file
>>>  	 * pages can simply be left unmapped, then faulted back on demand.
>>> @@ -2756,7 +2759,6 @@ static void lru_add_page_tail(struct page *head, struct page *tail,
>>>  		struct lruvec *lruvec, struct list_head *list)
>>>  {
>>>  	VM_BUG_ON_PAGE(!PageHead(head), head);
>>> -	VM_BUG_ON_PAGE(PageCompound(tail), head);
>>>  	VM_BUG_ON_PAGE(PageLRU(tail), head);
>>>  	lockdep_assert_held(&lruvec->lru_lock);
>>>
>>> @@ -2777,7 +2779,8 @@ static void lru_add_page_tail(struct page *head, struct page *tail,
>>>  }
>>>
>>>  static void __split_huge_page_tail(struct folio *folio, int tail,
>>> -		struct lruvec *lruvec, struct list_head *list)
>>> +		struct lruvec *lruvec, struct list_head *list,
>>> +		unsigned int new_order)
>>>  {
>>>  	struct page *head = &folio->page;
>>>  	struct page *page_tail = head + tail;
>>> @@ -2847,10 +2850,15 @@ static void __split_huge_page_tail(struct folio *folio, int tail,
>>>  	 * which needs correct compound_head().
>>>  	 */
>>>  	clear_compound_head(page_tail);
>>> +	if (new_order) {
>>> +		prep_compound_page(page_tail, new_order);
>>> +		folio_prep_large_rmappable(page_folio(page_tail));
>>> +	}
>>>
>>>  	/* Finally unfreeze refcount. Additional reference from page cache. */
>>> -	page_ref_unfreeze(page_tail, 1 + (!folio_test_anon(folio) ||
>>> -					  folio_test_swapcache(folio)));
>>> +	page_ref_unfreeze(page_tail,
>>> +		1 + ((!folio_test_anon(folio) || folio_test_swapcache(folio)) ?
>>> +			     folio_nr_pages(page_folio(page_tail)) : 0));
>>>
>>>  	if (folio_test_young(folio))
>>>  		folio_set_young(new_folio);
>>> @@ -2868,7 +2876,7 @@ static void __split_huge_page_tail(struct folio *folio, int tail,
>>>  }
>>>
>>>  static void __split_huge_page(struct page *page, struct list_head *list,
>>> -		pgoff_t end)
>>> +		pgoff_t end, unsigned int new_order)
>>>  {
>>>  	struct folio *folio = page_folio(page);
>>>  	struct page *head = &folio->page;
>>> @@ -2877,10 +2885,11 @@ static void __split_huge_page(struct page *page, struct list_head *list,
>>>  	unsigned long offset = 0;
>>>  	unsigned int nr = thp_nr_pages(head);
>>>  	int i, nr_dropped = 0;
>>> +	unsigned int new_nr = 1 << new_order;
>>>  	int order = folio_order(folio);
>>>
>>>  	/* complete memcg works before add pages to LRU */
>>> -	split_page_memcg(head, order, 0);
>>> +	split_page_memcg(head, order, new_order);
>>>
>>>  	if (folio_test_anon(folio) && folio_test_swapcache(folio)) {
>>>  		offset = swp_offset(folio->swap);
>>> @@ -2893,8 +2902,8 @@ static void __split_huge_page(struct page *page, struct list_head *list,
>>>
>>>  	ClearPageHasHWPoisoned(head);
>>>
>>> -	for (i = nr - 1; i >= 1; i--) {
>>> -		__split_huge_page_tail(folio, i, lruvec, list);
>>> +	for (i = nr - new_nr; i >= new_nr; i -= new_nr) {
>>> +		__split_huge_page_tail(folio, i, lruvec, list, new_order);
>>>  		/* Some pages can be beyond EOF: drop them from page cache */
>>>  		if (head[i].index >= end) {
>>>  			struct folio *tail = page_folio(head + i);
>>> @@ -2910,29 +2919,41 @@ static void __split_huge_page(struct page *page, struct list_head *list,
>>>  			__xa_store(&head->mapping->i_pages, head[i].index,
>>>  					head + i, 0);
>>>  		} else if (swap_cache) {
>>> +			/*
>>> +			 * split anonymous THPs (including swapped out ones) to
>>> +			 * non-zero order not supported
>>> +			 */
>>> +			VM_WARN_ONCE(new_order,
>>> +				"Split swap-cached anon folio to non-0 order not supported");
>>
>> Why isn't it supported? Even if it's not supported, is this level the right
>> place to enforce these kinds of policy decisions? I wonder if we should be
>> leaving that to the higher level to decide?
> 
> Is the swap-out small-size THP without splitting merged? This needs that patchset.

No not yet. I have to respin it. Its on my todo list.

I'm not sure I understand the dependency though?

> You are right that a warning here is not appropriate. I will fail the splitting
> if the folio is swapcached and going to be split into >0 order.
> 
>>>  			__xa_store(&swap_cache->i_pages, offset + i,
>>>  					head + i, 0);
>>>  		}
>>>  	}
>>>
> 
> 
> --
> Best Regards,
> Yan, Zi
  

On 14 Feb 2024, at 11:22, Ryan Roberts wrote:

> On 14/02/2024 16:11, Zi Yan wrote:
>> On 14 Feb 2024, at 5:38, Ryan Roberts wrote:
>>
>>> On 13/02/2024 21:55, Zi Yan wrote:
>>>> From: Zi Yan <ziy@nvidia.com>
>>>>
>>>> To split a THP to any lower order (except order-1) pages, we need to
>>>> reform THPs on subpages at given order and add page refcount based on the
>>>> new page order. Also we need to reinitialize page_deferred_list after
>>>> removing the page from the split_queue, otherwise a subsequent split will
>>>> see list corruption when checking the page_deferred_list again.
>>>>
>>>> It has many uses, like minimizing the number of pages after
>>>> truncating a huge pagecache page. For anonymous THPs, we can only split
>>>> them to order-0 like before until we add support for any size anonymous
>>>> THPs.
>>>
>>> multi-size THP is now upstream. Not sure if this comment still makes sense.
>> Will change it to reflect the fact that multi-size THP is already upstream.
>>
>>> Still its not completely clear to me how you would integrate this new machinery
>>> and decide what non-zero order to split anon THP to?
>>
>> Originally, it was developed along with my 1GB THP support. So it was intended
>> to split order-18 to order-9. But for now, like you and David said in the cover
>> letter email thread, we might not want to use it for anonymous large folios
>> until we find a necessary use case.
>>
>>>>
>>>> Order-1 folio is not supported because _deferred_list, which is used by
>>>> partially mapped folios, is stored in subpage 2 and an order-1 folio only
>>>> has subpage 0 and 1.
>>>>
>>>> Signed-off-by: Zi Yan <ziy@nvidia.com>
>>>> ---
>>>>  include/linux/huge_mm.h |  21 +++++---
>>>>  mm/huge_memory.c        | 114 +++++++++++++++++++++++++++++++---------
>>>>  2 files changed, 101 insertions(+), 34 deletions(-)
>>>>
>>>> diff --git a/include/linux/huge_mm.h b/include/linux/huge_mm.h
>>>> index 5adb86af35fc..de0c89105076 100644
>>>> --- a/include/linux/huge_mm.h
>>>> +++ b/include/linux/huge_mm.h
>>>> @@ -265,10 +265,11 @@ unsigned long thp_get_unmapped_area(struct file *filp, unsigned long addr,
>>>>
>>>>  void folio_prep_large_rmappable(struct folio *folio);
>>>>  bool can_split_folio(struct folio *folio, int *pextra_pins);
>>>> -int split_huge_page_to_list(struct page *page, struct list_head *list);
>>>> +int split_huge_page_to_list_to_order(struct page *page, struct list_head *list,
>>>> +		unsigned int new_order);
>>>>  static inline int split_huge_page(struct page *page)
>>>>  {
>>>> -	return split_huge_page_to_list(page, NULL);
>>>> +	return split_huge_page_to_list_to_order(page, NULL, 0);
>>>>  }
>>>>  void deferred_split_folio(struct folio *folio);
>>>>
>>>> @@ -422,7 +423,8 @@ can_split_folio(struct folio *folio, int *pextra_pins)
>>>>  	return false;
>>>>  }
>>>>  static inline int
>>>> -split_huge_page_to_list(struct page *page, struct list_head *list)
>>>> +split_huge_page_to_list_to_order(struct page *page, struct list_head *list,
>>>> +		unsigned int new_order)
>>>>  {
>>>>  	return 0;
>>>>  }
>>>> @@ -519,17 +521,20 @@ static inline bool thp_migration_supported(void)
>>>>  }
>>>>  #endif /* CONFIG_TRANSPARENT_HUGEPAGE */
>>>>
>>>> -static inline int split_folio_to_list(struct folio *folio,
>>>> -		struct list_head *list)
>>>> +static inline int split_folio_to_list_to_order(struct folio *folio,
>>>> +		struct list_head *list, int new_order)
>>>>  {
>>>> -	return split_huge_page_to_list(&folio->page, list);
>>>> +	return split_huge_page_to_list_to_order(&folio->page, list, new_order);
>>>>  }
>>>>
>>>> -static inline int split_folio(struct folio *folio)
>>>> +static inline int split_folio_to_order(struct folio *folio, int new_order)
>>>>  {
>>>> -	return split_folio_to_list(folio, NULL);
>>>> +	return split_folio_to_list_to_order(folio, NULL, new_order);
>>>>  }
>>>>
>>>> +#define split_folio_to_list(f, l) split_folio_to_list_to_order(f, l, 0)
>>>> +#define split_folio(f) split_folio_to_order(f, 0)
>>>> +
>>>>  /*
>>>>   * archs that select ARCH_WANTS_THP_SWAP but don't support THP_SWP due to
>>>>   * limitations in the implementation like arm64 MTE can override this to
>>>> diff --git a/mm/huge_memory.c b/mm/huge_memory.c
>>>> index ad7133c97428..d0e555a8ea98 100644
>>>> --- a/mm/huge_memory.c
>>>> +++ b/mm/huge_memory.c
>>>> @@ -2718,11 +2718,14 @@ void vma_adjust_trans_huge(struct vm_area_struct *vma,
>>>>
>>>>  static void unmap_folio(struct folio *folio)
>>>>  {
>>>> -	enum ttu_flags ttu_flags = TTU_RMAP_LOCKED | TTU_SPLIT_HUGE_PMD |
>>>> -		TTU_SYNC | TTU_BATCH_FLUSH;
>>>> +	enum ttu_flags ttu_flags = TTU_RMAP_LOCKED | TTU_SYNC |
>>>> +		TTU_BATCH_FLUSH;
>>>>
>>>>  	VM_BUG_ON_FOLIO(!folio_test_large(folio), folio);
>>>>
>>>> +	if (folio_test_pmd_mappable(folio))
>>>> +		ttu_flags |= TTU_SPLIT_HUGE_PMD;
>>>
>>> Should we split this change out? I think it makes sense independent of this series?
>>>
>>
>> Sure. Since multi-size THP is upstream, this avoid unnecessary code path if
>> the THP is not PMD-mapped.
>>
>>>> +
>>>>  	/*
>>>>  	 * Anon pages need migration entries to preserve them, but file
>>>>  	 * pages can simply be left unmapped, then faulted back on demand.
>>>> @@ -2756,7 +2759,6 @@ static void lru_add_page_tail(struct page *head, struct page *tail,
>>>>  		struct lruvec *lruvec, struct list_head *list)
>>>>  {
>>>>  	VM_BUG_ON_PAGE(!PageHead(head), head);
>>>> -	VM_BUG_ON_PAGE(PageCompound(tail), head);
>>>>  	VM_BUG_ON_PAGE(PageLRU(tail), head);
>>>>  	lockdep_assert_held(&lruvec->lru_lock);
>>>>
>>>> @@ -2777,7 +2779,8 @@ static void lru_add_page_tail(struct page *head, struct page *tail,
>>>>  }
>>>>
>>>>  static void __split_huge_page_tail(struct folio *folio, int tail,
>>>> -		struct lruvec *lruvec, struct list_head *list)
>>>> +		struct lruvec *lruvec, struct list_head *list,
>>>> +		unsigned int new_order)
>>>>  {
>>>>  	struct page *head = &folio->page;
>>>>  	struct page *page_tail = head + tail;
>>>> @@ -2847,10 +2850,15 @@ static void __split_huge_page_tail(struct folio *folio, int tail,
>>>>  	 * which needs correct compound_head().
>>>>  	 */
>>>>  	clear_compound_head(page_tail);
>>>> +	if (new_order) {
>>>> +		prep_compound_page(page_tail, new_order);
>>>> +		folio_prep_large_rmappable(page_folio(page_tail));
>>>> +	}
>>>>
>>>>  	/* Finally unfreeze refcount. Additional reference from page cache. */
>>>> -	page_ref_unfreeze(page_tail, 1 + (!folio_test_anon(folio) ||
>>>> -					  folio_test_swapcache(folio)));
>>>> +	page_ref_unfreeze(page_tail,
>>>> +		1 + ((!folio_test_anon(folio) || folio_test_swapcache(folio)) ?
>>>> +			     folio_nr_pages(page_folio(page_tail)) : 0));
>>>>
>>>>  	if (folio_test_young(folio))
>>>>  		folio_set_young(new_folio);
>>>> @@ -2868,7 +2876,7 @@ static void __split_huge_page_tail(struct folio *folio, int tail,
>>>>  }
>>>>
>>>>  static void __split_huge_page(struct page *page, struct list_head *list,
>>>> -		pgoff_t end)
>>>> +		pgoff_t end, unsigned int new_order)
>>>>  {
>>>>  	struct folio *folio = page_folio(page);
>>>>  	struct page *head = &folio->page;
>>>> @@ -2877,10 +2885,11 @@ static void __split_huge_page(struct page *page, struct list_head *list,
>>>>  	unsigned long offset = 0;
>>>>  	unsigned int nr = thp_nr_pages(head);
>>>>  	int i, nr_dropped = 0;
>>>> +	unsigned int new_nr = 1 << new_order;
>>>>  	int order = folio_order(folio);
>>>>
>>>>  	/* complete memcg works before add pages to LRU */
>>>> -	split_page_memcg(head, order, 0);
>>>> +	split_page_memcg(head, order, new_order);
>>>>
>>>>  	if (folio_test_anon(folio) && folio_test_swapcache(folio)) {
>>>>  		offset = swp_offset(folio->swap);
>>>> @@ -2893,8 +2902,8 @@ static void __split_huge_page(struct page *page, struct list_head *list,
>>>>
>>>>  	ClearPageHasHWPoisoned(head);
>>>>
>>>> -	for (i = nr - 1; i >= 1; i--) {
>>>> -		__split_huge_page_tail(folio, i, lruvec, list);
>>>> +	for (i = nr - new_nr; i >= new_nr; i -= new_nr) {
>>>> +		__split_huge_page_tail(folio, i, lruvec, list, new_order);
>>>>  		/* Some pages can be beyond EOF: drop them from page cache */
>>>>  		if (head[i].index >= end) {
>>>>  			struct folio *tail = page_folio(head + i);
>>>> @@ -2910,29 +2919,41 @@ static void __split_huge_page(struct page *page, struct list_head *list,
>>>>  			__xa_store(&head->mapping->i_pages, head[i].index,
>>>>  					head + i, 0);
>>>>  		} else if (swap_cache) {
>>>> +			/*
>>>> +			 * split anonymous THPs (including swapped out ones) to
>>>> +			 * non-zero order not supported
>>>> +			 */
>>>> +			VM_WARN_ONCE(new_order,
>>>> +				"Split swap-cached anon folio to non-0 order not supported");
>>>
>>> Why isn't it supported? Even if it's not supported, is this level the right
>>> place to enforce these kinds of policy decisions? I wonder if we should be
>>> leaving that to the higher level to decide?
>>
>> Is the swap-out small-size THP without splitting merged? This needs that patchset.
>
> No not yet. I have to respin it. Its on my todo list.
>
> I'm not sure I understand the dependency though?

IIUC, swap cache only supports one cluster size, HPAGE_PMD_NR, so splitting
a PMD-size swapcached folio will need to split a cluster to smaller ones, which
needs your patchset support. Let me know if I get it wrong.

>
>> You are right that a warning here is not appropriate. I will fail the splitting
>> if the folio is swapcached and going to be split into >0 order.
>>
>>>>  			__xa_store(&swap_cache->i_pages, offset + i,
>>>>  					head + i, 0);
>>>>  		}
>>>>  	}
>>>>
>>
>>
>> --
>> Best Regards,
>> Yan, Zi


--
Best Regards,
Yan, Zi
  

On 14/02/2024 16:28, Zi Yan wrote:
> On 14 Feb 2024, at 11:22, Ryan Roberts wrote:
> 
>> On 14/02/2024 16:11, Zi Yan wrote:
>>> On 14 Feb 2024, at 5:38, Ryan Roberts wrote:
>>>
>>>> On 13/02/2024 21:55, Zi Yan wrote:
>>>>> From: Zi Yan <ziy@nvidia.com>
>>>>>
>>>>> To split a THP to any lower order (except order-1) pages, we need to
>>>>> reform THPs on subpages at given order and add page refcount based on the
>>>>> new page order. Also we need to reinitialize page_deferred_list after
>>>>> removing the page from the split_queue, otherwise a subsequent split will
>>>>> see list corruption when checking the page_deferred_list again.
>>>>>
>>>>> It has many uses, like minimizing the number of pages after
>>>>> truncating a huge pagecache page. For anonymous THPs, we can only split
>>>>> them to order-0 like before until we add support for any size anonymous
>>>>> THPs.
>>>>
>>>> multi-size THP is now upstream. Not sure if this comment still makes sense.
>>> Will change it to reflect the fact that multi-size THP is already upstream.
>>>
>>>> Still its not completely clear to me how you would integrate this new machinery
>>>> and decide what non-zero order to split anon THP to?
>>>
>>> Originally, it was developed along with my 1GB THP support. So it was intended
>>> to split order-18 to order-9. But for now, like you and David said in the cover
>>> letter email thread, we might not want to use it for anonymous large folios
>>> until we find a necessary use case.
>>>
>>>>>
>>>>> Order-1 folio is not supported because _deferred_list, which is used by
>>>>> partially mapped folios, is stored in subpage 2 and an order-1 folio only
>>>>> has subpage 0 and 1.
>>>>>
>>>>> Signed-off-by: Zi Yan <ziy@nvidia.com>
>>>>> ---
>>>>>  include/linux/huge_mm.h |  21 +++++---
>>>>>  mm/huge_memory.c        | 114 +++++++++++++++++++++++++++++++---------
>>>>>  2 files changed, 101 insertions(+), 34 deletions(-)
>>>>>
>>>>> diff --git a/include/linux/huge_mm.h b/include/linux/huge_mm.h
>>>>> index 5adb86af35fc..de0c89105076 100644
>>>>> --- a/include/linux/huge_mm.h
>>>>> +++ b/include/linux/huge_mm.h
>>>>> @@ -265,10 +265,11 @@ unsigned long thp_get_unmapped_area(struct file *filp, unsigned long addr,
>>>>>
>>>>>  void folio_prep_large_rmappable(struct folio *folio);
>>>>>  bool can_split_folio(struct folio *folio, int *pextra_pins);
>>>>> -int split_huge_page_to_list(struct page *page, struct list_head *list);
>>>>> +int split_huge_page_to_list_to_order(struct page *page, struct list_head *list,
>>>>> +		unsigned int new_order);
>>>>>  static inline int split_huge_page(struct page *page)
>>>>>  {
>>>>> -	return split_huge_page_to_list(page, NULL);
>>>>> +	return split_huge_page_to_list_to_order(page, NULL, 0);
>>>>>  }
>>>>>  void deferred_split_folio(struct folio *folio);
>>>>>
>>>>> @@ -422,7 +423,8 @@ can_split_folio(struct folio *folio, int *pextra_pins)
>>>>>  	return false;
>>>>>  }
>>>>>  static inline int
>>>>> -split_huge_page_to_list(struct page *page, struct list_head *list)
>>>>> +split_huge_page_to_list_to_order(struct page *page, struct list_head *list,
>>>>> +		unsigned int new_order)
>>>>>  {
>>>>>  	return 0;
>>>>>  }
>>>>> @@ -519,17 +521,20 @@ static inline bool thp_migration_supported(void)
>>>>>  }
>>>>>  #endif /* CONFIG_TRANSPARENT_HUGEPAGE */
>>>>>
>>>>> -static inline int split_folio_to_list(struct folio *folio,
>>>>> -		struct list_head *list)
>>>>> +static inline int split_folio_to_list_to_order(struct folio *folio,
>>>>> +		struct list_head *list, int new_order)
>>>>>  {
>>>>> -	return split_huge_page_to_list(&folio->page, list);
>>>>> +	return split_huge_page_to_list_to_order(&folio->page, list, new_order);
>>>>>  }
>>>>>
>>>>> -static inline int split_folio(struct folio *folio)
>>>>> +static inline int split_folio_to_order(struct folio *folio, int new_order)
>>>>>  {
>>>>> -	return split_folio_to_list(folio, NULL);
>>>>> +	return split_folio_to_list_to_order(folio, NULL, new_order);
>>>>>  }
>>>>>
>>>>> +#define split_folio_to_list(f, l) split_folio_to_list_to_order(f, l, 0)
>>>>> +#define split_folio(f) split_folio_to_order(f, 0)
>>>>> +
>>>>>  /*
>>>>>   * archs that select ARCH_WANTS_THP_SWAP but don't support THP_SWP due to
>>>>>   * limitations in the implementation like arm64 MTE can override this to
>>>>> diff --git a/mm/huge_memory.c b/mm/huge_memory.c
>>>>> index ad7133c97428..d0e555a8ea98 100644
>>>>> --- a/mm/huge_memory.c
>>>>> +++ b/mm/huge_memory.c
>>>>> @@ -2718,11 +2718,14 @@ void vma_adjust_trans_huge(struct vm_area_struct *vma,
>>>>>
>>>>>  static void unmap_folio(struct folio *folio)
>>>>>  {
>>>>> -	enum ttu_flags ttu_flags = TTU_RMAP_LOCKED | TTU_SPLIT_HUGE_PMD |
>>>>> -		TTU_SYNC | TTU_BATCH_FLUSH;
>>>>> +	enum ttu_flags ttu_flags = TTU_RMAP_LOCKED | TTU_SYNC |
>>>>> +		TTU_BATCH_FLUSH;
>>>>>
>>>>>  	VM_BUG_ON_FOLIO(!folio_test_large(folio), folio);
>>>>>
>>>>> +	if (folio_test_pmd_mappable(folio))
>>>>> +		ttu_flags |= TTU_SPLIT_HUGE_PMD;
>>>>
>>>> Should we split this change out? I think it makes sense independent of this series?
>>>>
>>>
>>> Sure. Since multi-size THP is upstream, this avoid unnecessary code path if
>>> the THP is not PMD-mapped.
>>>
>>>>> +
>>>>>  	/*
>>>>>  	 * Anon pages need migration entries to preserve them, but file
>>>>>  	 * pages can simply be left unmapped, then faulted back on demand.
>>>>> @@ -2756,7 +2759,6 @@ static void lru_add_page_tail(struct page *head, struct page *tail,
>>>>>  		struct lruvec *lruvec, struct list_head *list)
>>>>>  {
>>>>>  	VM_BUG_ON_PAGE(!PageHead(head), head);
>>>>> -	VM_BUG_ON_PAGE(PageCompound(tail), head);
>>>>>  	VM_BUG_ON_PAGE(PageLRU(tail), head);
>>>>>  	lockdep_assert_held(&lruvec->lru_lock);
>>>>>
>>>>> @@ -2777,7 +2779,8 @@ static void lru_add_page_tail(struct page *head, struct page *tail,
>>>>>  }
>>>>>
>>>>>  static void __split_huge_page_tail(struct folio *folio, int tail,
>>>>> -		struct lruvec *lruvec, struct list_head *list)
>>>>> +		struct lruvec *lruvec, struct list_head *list,
>>>>> +		unsigned int new_order)
>>>>>  {
>>>>>  	struct page *head = &folio->page;
>>>>>  	struct page *page_tail = head + tail;
>>>>> @@ -2847,10 +2850,15 @@ static void __split_huge_page_tail(struct folio *folio, int tail,
>>>>>  	 * which needs correct compound_head().
>>>>>  	 */
>>>>>  	clear_compound_head(page_tail);
>>>>> +	if (new_order) {
>>>>> +		prep_compound_page(page_tail, new_order);
>>>>> +		folio_prep_large_rmappable(page_folio(page_tail));
>>>>> +	}
>>>>>
>>>>>  	/* Finally unfreeze refcount. Additional reference from page cache. */
>>>>> -	page_ref_unfreeze(page_tail, 1 + (!folio_test_anon(folio) ||
>>>>> -					  folio_test_swapcache(folio)));
>>>>> +	page_ref_unfreeze(page_tail,
>>>>> +		1 + ((!folio_test_anon(folio) || folio_test_swapcache(folio)) ?
>>>>> +			     folio_nr_pages(page_folio(page_tail)) : 0));
>>>>>
>>>>>  	if (folio_test_young(folio))
>>>>>  		folio_set_young(new_folio);
>>>>> @@ -2868,7 +2876,7 @@ static void __split_huge_page_tail(struct folio *folio, int tail,
>>>>>  }
>>>>>
>>>>>  static void __split_huge_page(struct page *page, struct list_head *list,
>>>>> -		pgoff_t end)
>>>>> +		pgoff_t end, unsigned int new_order)
>>>>>  {
>>>>>  	struct folio *folio = page_folio(page);
>>>>>  	struct page *head = &folio->page;
>>>>> @@ -2877,10 +2885,11 @@ static void __split_huge_page(struct page *page, struct list_head *list,
>>>>>  	unsigned long offset = 0;
>>>>>  	unsigned int nr = thp_nr_pages(head);
>>>>>  	int i, nr_dropped = 0;
>>>>> +	unsigned int new_nr = 1 << new_order;
>>>>>  	int order = folio_order(folio);
>>>>>
>>>>>  	/* complete memcg works before add pages to LRU */
>>>>> -	split_page_memcg(head, order, 0);
>>>>> +	split_page_memcg(head, order, new_order);
>>>>>
>>>>>  	if (folio_test_anon(folio) && folio_test_swapcache(folio)) {
>>>>>  		offset = swp_offset(folio->swap);
>>>>> @@ -2893,8 +2902,8 @@ static void __split_huge_page(struct page *page, struct list_head *list,
>>>>>
>>>>>  	ClearPageHasHWPoisoned(head);
>>>>>
>>>>> -	for (i = nr - 1; i >= 1; i--) {
>>>>> -		__split_huge_page_tail(folio, i, lruvec, list);
>>>>> +	for (i = nr - new_nr; i >= new_nr; i -= new_nr) {
>>>>> +		__split_huge_page_tail(folio, i, lruvec, list, new_order);
>>>>>  		/* Some pages can be beyond EOF: drop them from page cache */
>>>>>  		if (head[i].index >= end) {
>>>>>  			struct folio *tail = page_folio(head + i);
>>>>> @@ -2910,29 +2919,41 @@ static void __split_huge_page(struct page *page, struct list_head *list,
>>>>>  			__xa_store(&head->mapping->i_pages, head[i].index,
>>>>>  					head + i, 0);
>>>>>  		} else if (swap_cache) {
>>>>> +			/*
>>>>> +			 * split anonymous THPs (including swapped out ones) to
>>>>> +			 * non-zero order not supported
>>>>> +			 */
>>>>> +			VM_WARN_ONCE(new_order,
>>>>> +				"Split swap-cached anon folio to non-0 order not supported");
>>>>
>>>> Why isn't it supported? Even if it's not supported, is this level the right
>>>> place to enforce these kinds of policy decisions? I wonder if we should be
>>>> leaving that to the higher level to decide?
>>>
>>> Is the swap-out small-size THP without splitting merged? This needs that patchset.
>>
>> No not yet. I have to respin it. Its on my todo list.
>>
>> I'm not sure I understand the dependency though?
> 
> IIUC, swap cache only supports one cluster size, HPAGE_PMD_NR, so splitting
> a PMD-size swapcached folio will need to split a cluster to smaller ones, which
> needs your patchset support. Let me know if I get it wrong.

Ahh yeah, sorry, obvious now that you've spelled it out - thanks!

> 
>>
>>> You are right that a warning here is not appropriate. I will fail the splitting
>>> if the folio is swapcached and going to be split into >0 order.
>>>
>>>>>  			__xa_store(&swap_cache->i_pages, offset + i,
>>>>>  					head + i, 0);
>>>>>  		}
>>>>>  	}
>>>>>
>>>
>>>
>>> --
>>> Best Regards,
>>> Yan, Zi
> 
> 
> --
> Best Regards,
> Yan, Zi