diff mbox series

[PATCHv3,RESEND,6/8] khugepaged: Allow to collapse PTE-mapped compound pages

Message ID 20200413125220.663-7-kirill.shutemov@linux.intel.com (mailing list archive)
State New, archived
Headers show
Series thp/khugepaged improvements and CoW semantics | expand

Commit Message

kirill.shutemov@linux.intel.com April 13, 2020, 12:52 p.m. UTC 
We can collapse PTE-mapped compound pages. We only need to avoid
handling them more than once: lock/unlock page only once if it's present
in the PMD range multiple times as it handled on compound level. The
same goes for LRU isolation and putback.

Signed-off-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
---
 mm/khugepaged.c | 99 ++++++++++++++++++++++++++++++++-----------------
 1 file changed, 65 insertions(+), 34 deletions(-)

Comments

Yang Shi April 15, 2020, 8:46 p.m. UTC | #1 
On 4/13/20 5:52 AM, Kirill A. Shutemov wrote:
> We can collapse PTE-mapped compound pages. We only need to avoid
> handling them more than once: lock/unlock page only once if it's present
> in the PMD range multiple times as it handled on compound level. The
> same goes for LRU isolation and putback.
>
> Signed-off-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
> ---
>   mm/khugepaged.c | 99 ++++++++++++++++++++++++++++++++-----------------
>   1 file changed, 65 insertions(+), 34 deletions(-)

Acked-by: Yang Shi <yang.shi@linux.alibaba.com>

>
> diff --git a/mm/khugepaged.c b/mm/khugepaged.c
> index f9864644c3b7..11d500396d85 100644
> --- a/mm/khugepaged.c
> +++ b/mm/khugepaged.c
> @@ -515,17 +515,30 @@ void __khugepaged_exit(struct mm_struct *mm)
>   
>   static void release_pte_page(struct page *page)
>   {
> -	dec_node_page_state(page, NR_ISOLATED_ANON + page_is_file_cache(page));
> +	mod_node_page_state(page_pgdat(page),
> +			NR_ISOLATED_ANON + page_is_file_cache(page),
> +			-compound_nr(page));
>   	unlock_page(page);
>   	putback_lru_page(page);
>   }
>   
> -static void release_pte_pages(pte_t *pte, pte_t *_pte)
> +static void release_pte_pages(pte_t *pte, pte_t *_pte,
> +		struct list_head *compound_pagelist)
>   {
> +	struct page *page, *tmp;
> +
>   	while (--_pte >= pte) {
>   		pte_t pteval = *_pte;
> -		if (!pte_none(pteval) && !is_zero_pfn(pte_pfn(pteval)))
> -			release_pte_page(pte_page(pteval));
> +
> +		page = pte_page(pteval);
> +		if (!pte_none(pteval) && !is_zero_pfn(pte_pfn(pteval)) &&
> +				!PageCompound(page))
> +			release_pte_page(page);
> +	}
> +
> +	list_for_each_entry_safe(page, tmp, compound_pagelist, lru) {
> +		list_del(&page->lru);
> +		release_pte_page(page);
>   	}
>   }
>   
> @@ -549,7 +562,8 @@ static bool is_refcount_suitable(struct page *page)
>   
>   static int __collapse_huge_page_isolate(struct vm_area_struct *vma,
>   					unsigned long address,
> -					pte_t *pte)
> +					pte_t *pte,
> +					struct list_head *compound_pagelist)
>   {
>   	struct page *page = NULL;
>   	pte_t *_pte;
> @@ -579,13 +593,21 @@ static int __collapse_huge_page_isolate(struct vm_area_struct *vma,
>   			goto out;
>   		}
>   
> -		/* TODO: teach khugepaged to collapse THP mapped with pte */
> +		VM_BUG_ON_PAGE(!PageAnon(page), page);
> +
>   		if (PageCompound(page)) {
> -			result = SCAN_PAGE_COMPOUND;
> -			goto out;
> -		}
> +			struct page *p;
> +			page = compound_head(page);
>   
> -		VM_BUG_ON_PAGE(!PageAnon(page), page);
> +			/*
> +			 * Check if we have dealt with the compound page
> +			 * already
> +			 */
> +			list_for_each_entry(p, compound_pagelist, lru) {
> +				if (page == p)
> +					goto next;
> +			}
> +		}
>   
>   		/*
>   		 * We can do it before isolate_lru_page because the
> @@ -614,19 +636,15 @@ static int __collapse_huge_page_isolate(struct vm_area_struct *vma,
>   			result = SCAN_PAGE_COUNT;
>   			goto out;
>   		}
> -		if (pte_write(pteval)) {
> -			writable = true;
> -		} else {
> -			if (PageSwapCache(page) &&
> -			    !reuse_swap_page(page, NULL)) {
> -				unlock_page(page);
> -				result = SCAN_SWAP_CACHE_PAGE;
> -				goto out;
> -			}
> +		if (!pte_write(pteval) && PageSwapCache(page) &&
> +				!reuse_swap_page(page, NULL)) {
>   			/*
> -			 * Page is not in the swap cache. It can be collapsed
> -			 * into a THP.
> +			 * Page is in the swap cache and cannot be re-used.
> +			 * It cannot be collapsed into a THP.
>   			 */
> +			unlock_page(page);
> +			result = SCAN_SWAP_CACHE_PAGE;
> +			goto out;
>   		}
>   
>   		/*
> @@ -638,16 +656,23 @@ static int __collapse_huge_page_isolate(struct vm_area_struct *vma,
>   			result = SCAN_DEL_PAGE_LRU;
>   			goto out;
>   		}
> -		inc_node_page_state(page,
> -				NR_ISOLATED_ANON + page_is_file_cache(page));
> +		mod_node_page_state(page_pgdat(page),
> +				NR_ISOLATED_ANON + page_is_file_cache(page),
> +				compound_nr(page));
>   		VM_BUG_ON_PAGE(!PageLocked(page), page);
>   		VM_BUG_ON_PAGE(PageLRU(page), page);
>   
> +		if (PageCompound(page))
> +			list_add_tail(&page->lru, compound_pagelist);
> +next:
>   		/* There should be enough young pte to collapse the page */
>   		if (pte_young(pteval) ||
>   		    page_is_young(page) || PageReferenced(page) ||
>   		    mmu_notifier_test_young(vma->vm_mm, address))
>   			referenced++;
> +
> +		if (pte_write(pteval))
> +			writable = true;
>   	}
>   	if (likely(writable)) {
>   		if (likely(referenced)) {
> @@ -661,7 +686,7 @@ static int __collapse_huge_page_isolate(struct vm_area_struct *vma,
>   	}
>   
>   out:
> -	release_pte_pages(pte, _pte);
> +	release_pte_pages(pte, _pte, compound_pagelist);
>   	trace_mm_collapse_huge_page_isolate(page, none_or_zero,
>   					    referenced, writable, result);
>   	return 0;
> @@ -670,13 +695,14 @@ static int __collapse_huge_page_isolate(struct vm_area_struct *vma,
>   static void __collapse_huge_page_copy(pte_t *pte, struct page *page,
>   				      struct vm_area_struct *vma,
>   				      unsigned long address,
> -				      spinlock_t *ptl)
> +				      spinlock_t *ptl,
> +				      struct list_head *compound_pagelist)
>   {
> +	struct page *src_page, *tmp;
>   	pte_t *_pte;
>   	for (_pte = pte; _pte < pte + HPAGE_PMD_NR;
>   				_pte++, page++, address += PAGE_SIZE) {
>   		pte_t pteval = *_pte;
> -		struct page *src_page;
>   
>   		if (pte_none(pteval) || is_zero_pfn(pte_pfn(pteval))) {
>   			clear_user_highpage(page, address);
> @@ -696,7 +722,8 @@ static void __collapse_huge_page_copy(pte_t *pte, struct page *page,
>   		} else {
>   			src_page = pte_page(pteval);
>   			copy_user_highpage(page, src_page, address, vma);
> -			release_pte_page(src_page);
> +			if (!PageCompound(src_page))
> +				release_pte_page(src_page);
>   			/*
>   			 * ptl mostly unnecessary, but preempt has to
>   			 * be disabled to update the per-cpu stats
> @@ -713,6 +740,11 @@ static void __collapse_huge_page_copy(pte_t *pte, struct page *page,
>   			free_page_and_swap_cache(src_page);
>   		}
>   	}
> +
> +	list_for_each_entry_safe(src_page, tmp, compound_pagelist, lru) {
> +		list_del(&src_page->lru);
> +		release_pte_page(src_page);
> +	}
>   }
>   
>   static void khugepaged_alloc_sleep(void)
> @@ -971,6 +1003,7 @@ static void collapse_huge_page(struct mm_struct *mm,
>   				   struct page **hpage,
>   				   int node, int referenced, int unmapped)
>   {
> +	LIST_HEAD(compound_pagelist);
>   	pmd_t *pmd, _pmd;
>   	pte_t *pte;
>   	pgtable_t pgtable;
> @@ -1071,7 +1104,8 @@ static void collapse_huge_page(struct mm_struct *mm,
>   	mmu_notifier_invalidate_range_end(&range);
>   
>   	spin_lock(pte_ptl);
> -	isolated = __collapse_huge_page_isolate(vma, address, pte);
> +	isolated = __collapse_huge_page_isolate(vma, address, pte,
> +			&compound_pagelist);
>   	spin_unlock(pte_ptl);
>   
>   	if (unlikely(!isolated)) {
> @@ -1096,7 +1130,8 @@ static void collapse_huge_page(struct mm_struct *mm,
>   	 */
>   	anon_vma_unlock_write(vma->anon_vma);
>   
> -	__collapse_huge_page_copy(pte, new_page, vma, address, pte_ptl);
> +	__collapse_huge_page_copy(pte, new_page, vma, address, pte_ptl,
> +			&compound_pagelist);
>   	pte_unmap(pte);
>   	__SetPageUptodate(new_page);
>   	pgtable = pmd_pgtable(_pmd);
> @@ -1193,11 +1228,7 @@ static int khugepaged_scan_pmd(struct mm_struct *mm,
>   			goto out_unmap;
>   		}
>   
> -		/* TODO: teach khugepaged to collapse THP mapped with pte */
> -		if (PageCompound(page)) {
> -			result = SCAN_PAGE_COMPOUND;
> -			goto out_unmap;
> -		}
> +		page = compound_head(page);
>   
>   		/*
>   		 * Record which node the original page is from and save this
Andrew Morton April 15, 2020, 9:44 p.m. UTC | #2 
On Mon, 13 Apr 2020 15:52:18 +0300 "Kirill A. Shutemov" <kirill.shutemov@linux.intel.com> wrote:

> We can collapse PTE-mapped compound pages. We only need to avoid
> handling them more than once: lock/unlock page only once if it's present
> in the PMD range multiple times as it handled on compound level. The
> same goes for LRU isolation and putback.
> 
> ...
>
> --- a/mm/khugepaged.c
> +++ b/mm/khugepaged.c
> @@ -515,17 +515,30 @@ void __khugepaged_exit(struct mm_struct *mm)
>  
>  static void release_pte_page(struct page *page)
>  {
> -	dec_node_page_state(page, NR_ISOLATED_ANON + page_is_file_cache(page));

I have

	dec_node_page_state(page, NR_ISOLATED_ANON + page_is_file_lru(page));

here.  Is there some prerequisite which I wasn't able to find?
Kirill A . Shutemov April 15, 2020, 9:52 p.m. UTC | #3 
On Wed, Apr 15, 2020 at 02:44:26PM -0700, Andrew Morton wrote:
> On Mon, 13 Apr 2020 15:52:18 +0300 "Kirill A. Shutemov" <kirill.shutemov@linux.intel.com> wrote:
> 
> > We can collapse PTE-mapped compound pages. We only need to avoid
> > handling them more than once: lock/unlock page only once if it's present
> > in the PMD range multiple times as it handled on compound level. The
> > same goes for LRU isolation and putback.
> > 
> > ...
> >
> > --- a/mm/khugepaged.c
> > +++ b/mm/khugepaged.c
> > @@ -515,17 +515,30 @@ void __khugepaged_exit(struct mm_struct *mm)
> >  
> >  static void release_pte_page(struct page *page)
> >  {
> > -	dec_node_page_state(page, NR_ISOLATED_ANON + page_is_file_cache(page));
> 
> I have
> 
> 	dec_node_page_state(page, NR_ISOLATED_ANON + page_is_file_lru(page));
> 
> here.  Is there some prerequisite which I wasn't able to find?

The patchset is on top of v5.6. It has been changed since. See
9de4f22a60f7 ("mm: code cleanup for MADV_FREE").

Look like a trivial fixup is required.
Andrew Morton April 15, 2020, 10:52 p.m. UTC | #4 
On Thu, 16 Apr 2020 00:52:05 +0300 "Kirill A. Shutemov" <kirill@shutemov.name> wrote:

> On Wed, Apr 15, 2020 at 02:44:26PM -0700, Andrew Morton wrote:
> > On Mon, 13 Apr 2020 15:52:18 +0300 "Kirill A. Shutemov" <kirill.shutemov@linux.intel.com> wrote:
> > 
> > > We can collapse PTE-mapped compound pages. We only need to avoid
> > > handling them more than once: lock/unlock page only once if it's present
> > > in the PMD range multiple times as it handled on compound level. The
> > > same goes for LRU isolation and putback.
> > > 
> > > ...
> > >
> > > --- a/mm/khugepaged.c
> > > +++ b/mm/khugepaged.c
> > > @@ -515,17 +515,30 @@ void __khugepaged_exit(struct mm_struct *mm)
> > >  
> > >  static void release_pte_page(struct page *page)
> > >  {
> > > -	dec_node_page_state(page, NR_ISOLATED_ANON + page_is_file_cache(page));
> > 
> > I have
> > 
> > 	dec_node_page_state(page, NR_ISOLATED_ANON + page_is_file_lru(page));
> > 
> > here.  Is there some prerequisite which I wasn't able to find?
> 
> The patchset is on top of v5.6. It has been changed since. See
> 9de4f22a60f7 ("mm: code cleanup for MADV_FREE").
> 
> Look like a trivial fixup is required.

[7/8] makes a big mess.  Can we please have a v4?
Kirill A . Shutemov April 15, 2020, 10:59 p.m. UTC | #5 
On Wed, Apr 15, 2020 at 03:52:59PM -0700, Andrew Morton wrote:
> On Thu, 16 Apr 2020 00:52:05 +0300 "Kirill A. Shutemov" <kirill@shutemov.name> wrote:
> 
> > On Wed, Apr 15, 2020 at 02:44:26PM -0700, Andrew Morton wrote:
> > > On Mon, 13 Apr 2020 15:52:18 +0300 "Kirill A. Shutemov" <kirill.shutemov@linux.intel.com> wrote:
> > > 
> > > > We can collapse PTE-mapped compound pages. We only need to avoid
> > > > handling them more than once: lock/unlock page only once if it's present
> > > > in the PMD range multiple times as it handled on compound level. The
> > > > same goes for LRU isolation and putback.
> > > > 
> > > > ...
> > > >
> > > > --- a/mm/khugepaged.c
> > > > +++ b/mm/khugepaged.c
> > > > @@ -515,17 +515,30 @@ void __khugepaged_exit(struct mm_struct *mm)
> > > >  
> > > >  static void release_pte_page(struct page *page)
> > > >  {
> > > > -	dec_node_page_state(page, NR_ISOLATED_ANON + page_is_file_cache(page));
> > > 
> > > I have
> > > 
> > > 	dec_node_page_state(page, NR_ISOLATED_ANON + page_is_file_lru(page));
> > > 
> > > here.  Is there some prerequisite which I wasn't able to find?
> > 
> > The patchset is on top of v5.6. It has been changed since. See
> > 9de4f22a60f7 ("mm: code cleanup for MADV_FREE").
> > 
> > Look like a trivial fixup is required.
> 
> [7/8] makes a big mess.  Can we please have a v4?

Sure. Give me a day.
diff mbox series

Patch

diff --git a/mm/khugepaged.c b/mm/khugepaged.c
index f9864644c3b7..11d500396d85 100644
--- a/mm/khugepaged.c
+++ b/mm/khugepaged.c
@@ -515,17 +515,30 @@  void __khugepaged_exit(struct mm_struct *mm)
 
 static void release_pte_page(struct page *page)
 {
-	dec_node_page_state(page, NR_ISOLATED_ANON + page_is_file_cache(page));
+	mod_node_page_state(page_pgdat(page),
+			NR_ISOLATED_ANON + page_is_file_cache(page),
+			-compound_nr(page));
 	unlock_page(page);
 	putback_lru_page(page);
 }
 
-static void release_pte_pages(pte_t *pte, pte_t *_pte)
+static void release_pte_pages(pte_t *pte, pte_t *_pte,
+		struct list_head *compound_pagelist)
 {
+	struct page *page, *tmp;
+
 	while (--_pte >= pte) {
 		pte_t pteval = *_pte;
-		if (!pte_none(pteval) && !is_zero_pfn(pte_pfn(pteval)))
-			release_pte_page(pte_page(pteval));
+
+		page = pte_page(pteval);
+		if (!pte_none(pteval) && !is_zero_pfn(pte_pfn(pteval)) &&
+				!PageCompound(page))
+			release_pte_page(page);
+	}
+
+	list_for_each_entry_safe(page, tmp, compound_pagelist, lru) {
+		list_del(&page->lru);
+		release_pte_page(page);
 	}
 }
 
@@ -549,7 +562,8 @@  static bool is_refcount_suitable(struct page *page)
 
 static int __collapse_huge_page_isolate(struct vm_area_struct *vma,
 					unsigned long address,
-					pte_t *pte)
+					pte_t *pte,
+					struct list_head *compound_pagelist)
 {
 	struct page *page = NULL;
 	pte_t *_pte;
@@ -579,13 +593,21 @@  static int __collapse_huge_page_isolate(struct vm_area_struct *vma,
 			goto out;
 		}
 
-		/* TODO: teach khugepaged to collapse THP mapped with pte */
+		VM_BUG_ON_PAGE(!PageAnon(page), page);
+
 		if (PageCompound(page)) {
-			result = SCAN_PAGE_COMPOUND;
-			goto out;
-		}
+			struct page *p;
+			page = compound_head(page);
 
-		VM_BUG_ON_PAGE(!PageAnon(page), page);
+			/*
+			 * Check if we have dealt with the compound page
+			 * already
+			 */
+			list_for_each_entry(p, compound_pagelist, lru) {
+				if (page == p)
+					goto next;
+			}
+		}
 
 		/*
 		 * We can do it before isolate_lru_page because the
@@ -614,19 +636,15 @@  static int __collapse_huge_page_isolate(struct vm_area_struct *vma,
 			result = SCAN_PAGE_COUNT;
 			goto out;
 		}
-		if (pte_write(pteval)) {
-			writable = true;
-		} else {
-			if (PageSwapCache(page) &&
-			    !reuse_swap_page(page, NULL)) {
-				unlock_page(page);
-				result = SCAN_SWAP_CACHE_PAGE;
-				goto out;
-			}
+		if (!pte_write(pteval) && PageSwapCache(page) &&
+				!reuse_swap_page(page, NULL)) {
 			/*
-			 * Page is not in the swap cache. It can be collapsed
-			 * into a THP.
+			 * Page is in the swap cache and cannot be re-used.
+			 * It cannot be collapsed into a THP.
 			 */
+			unlock_page(page);
+			result = SCAN_SWAP_CACHE_PAGE;
+			goto out;
 		}
 
 		/*
@@ -638,16 +656,23 @@  static int __collapse_huge_page_isolate(struct vm_area_struct *vma,
 			result = SCAN_DEL_PAGE_LRU;
 			goto out;
 		}
-		inc_node_page_state(page,
-				NR_ISOLATED_ANON + page_is_file_cache(page));
+		mod_node_page_state(page_pgdat(page),
+				NR_ISOLATED_ANON + page_is_file_cache(page),
+				compound_nr(page));
 		VM_BUG_ON_PAGE(!PageLocked(page), page);
 		VM_BUG_ON_PAGE(PageLRU(page), page);
 
+		if (PageCompound(page))
+			list_add_tail(&page->lru, compound_pagelist);
+next:
 		/* There should be enough young pte to collapse the page */
 		if (pte_young(pteval) ||
 		    page_is_young(page) || PageReferenced(page) ||
 		    mmu_notifier_test_young(vma->vm_mm, address))
 			referenced++;
+
+		if (pte_write(pteval))
+			writable = true;
 	}
 	if (likely(writable)) {
 		if (likely(referenced)) {
@@ -661,7 +686,7 @@  static int __collapse_huge_page_isolate(struct vm_area_struct *vma,
 	}
 
 out:
-	release_pte_pages(pte, _pte);
+	release_pte_pages(pte, _pte, compound_pagelist);
 	trace_mm_collapse_huge_page_isolate(page, none_or_zero,
 					    referenced, writable, result);
 	return 0;
@@ -670,13 +695,14 @@  static int __collapse_huge_page_isolate(struct vm_area_struct *vma,
 static void __collapse_huge_page_copy(pte_t *pte, struct page *page,
 				      struct vm_area_struct *vma,
 				      unsigned long address,
-				      spinlock_t *ptl)
+				      spinlock_t *ptl,
+				      struct list_head *compound_pagelist)
 {
+	struct page *src_page, *tmp;
 	pte_t *_pte;
 	for (_pte = pte; _pte < pte + HPAGE_PMD_NR;
 				_pte++, page++, address += PAGE_SIZE) {
 		pte_t pteval = *_pte;
-		struct page *src_page;
 
 		if (pte_none(pteval) || is_zero_pfn(pte_pfn(pteval))) {
 			clear_user_highpage(page, address);
@@ -696,7 +722,8 @@  static void __collapse_huge_page_copy(pte_t *pte, struct page *page,
 		} else {
 			src_page = pte_page(pteval);
 			copy_user_highpage(page, src_page, address, vma);
-			release_pte_page(src_page);
+			if (!PageCompound(src_page))
+				release_pte_page(src_page);
 			/*
 			 * ptl mostly unnecessary, but preempt has to
 			 * be disabled to update the per-cpu stats
@@ -713,6 +740,11 @@  static void __collapse_huge_page_copy(pte_t *pte, struct page *page,
 			free_page_and_swap_cache(src_page);
 		}
 	}
+
+	list_for_each_entry_safe(src_page, tmp, compound_pagelist, lru) {
+		list_del(&src_page->lru);
+		release_pte_page(src_page);
+	}
 }
 
 static void khugepaged_alloc_sleep(void)
@@ -971,6 +1003,7 @@  static void collapse_huge_page(struct mm_struct *mm,
 				   struct page **hpage,
 				   int node, int referenced, int unmapped)
 {
+	LIST_HEAD(compound_pagelist);
 	pmd_t *pmd, _pmd;
 	pte_t *pte;
 	pgtable_t pgtable;
@@ -1071,7 +1104,8 @@  static void collapse_huge_page(struct mm_struct *mm,
 	mmu_notifier_invalidate_range_end(&range);
 
 	spin_lock(pte_ptl);
-	isolated = __collapse_huge_page_isolate(vma, address, pte);
+	isolated = __collapse_huge_page_isolate(vma, address, pte,
+			&compound_pagelist);
 	spin_unlock(pte_ptl);
 
 	if (unlikely(!isolated)) {
@@ -1096,7 +1130,8 @@  static void collapse_huge_page(struct mm_struct *mm,
 	 */
 	anon_vma_unlock_write(vma->anon_vma);
 
-	__collapse_huge_page_copy(pte, new_page, vma, address, pte_ptl);
+	__collapse_huge_page_copy(pte, new_page, vma, address, pte_ptl,
+			&compound_pagelist);
 	pte_unmap(pte);
 	__SetPageUptodate(new_page);
 	pgtable = pmd_pgtable(_pmd);
@@ -1193,11 +1228,7 @@  static int khugepaged_scan_pmd(struct mm_struct *mm,
 			goto out_unmap;
 		}
 
-		/* TODO: teach khugepaged to collapse THP mapped with pte */
-		if (PageCompound(page)) {
-			result = SCAN_PAGE_COMPOUND;
-			goto out_unmap;
-		}
+		page = compound_head(page);
 
 		/*
 		 * Record which node the original page is from and save this