Skip to contentSkip to conversation

mm: thp: use generic THP migration for NUMA hinting fault

RFC PATCH Superseded Yang Shi 4 participants 26 messages Apr 13, 2021 – May 18, 2021

Subsystems (5)
Memory management MM core s390 THP +1 more subsystems

v2 of 3 Superseded Latest is v3 →

lore
Share & tools
lore

Yang Shi

author
#0 lore raw

v2 cover mm: thp: use generic THP migration for NUMA hinting fault

Changelog:
v1 --> v2:
* Adopted the suggestion from Gerald Schaefer to skip huge PMD for S390 for now.
* Used PageTransHuge to distinguish base page or THP instead of a new parameter for migrate_misplaced_page() per Huang Ying.
* Restored PMD lazily to avoid unnecessary TLB shootdown per Huang Ying.
* Skipped shared THP.
* Updated counters correctly.
* Rebased to linux-next (next-20210412).

When the THP NUMA fault support was added THP migration was not supported yet. So the ad hoc THP migration was implemented in NUMA fault handling. Since v4.14 THP migration has been supported so it doesn't make too much sense to still keep another THP migration implementation rather than using the generic migration code. It is definitely a maintenance burden to keep two THP migration implementation for different code paths and it is more error prone. Using the generic THP migration implementation allows us remove the duplicate code and some hacks needed by the old ad hoc implementation.

A quick grep shows x86_64, PowerPC (book3s), ARM64 ans S390 support both THP and NUMA balancing. The most of them support THP migration except for S390. Zi Yan tried to add THP migration support for S390 before but it was not accepted due to the design of S390 PMD. For the discussion, please see:
https://lkml.org/lkml/2018/4/27/953.

Per the discussion with Gerald Schaefer in v1 it is acceptible to skip huge PMD for S390 for now.

I saw there were some hacks about gup from git history, but I didn't figure out if they have been removed or not since I just found FOLL_NUMA code in the current gup implementation and they seems useful.

I'm trying to keep the behavior as consistent as possible between before and after. But there is still some minor disparity. For example, file THP won't get migrated at all in old implementation due to the anon_vma check, but the new implementation doesn't need acquire anon_vma lock anymore, so file THP might get migrated. Not sure if this behavior needs to be kept.

Patch #1 ~ #2 are preparation patches.
Patch #3 is the real meat.
Patch #4 ~ #6 keep consistent counters and behaviors with before. Patch #7 skips change huge PMD to prot_none if thp migration is not supported.

Yang Shi (7):
mm: memory: add orig_pmd to struct vm_fault
mm: memory: make numa_migrate_prep() non-static
mm: thp: refactor NUMA fault handling
mm: migrate: account THP NUMA migration counters correctly
mm: migrate: don't split THP for misplaced NUMA page
mm: migrate: check mapcount for THP instead of ref count
mm: thp: skip make PMD PROT_NONE if THP migration is not supported

 include/linux/huge_mm.h |   9 ++---
 include/linux/migrate.h |  23 -----------
 include/linux/mm.h      |   3 ++
 mm/huge_memory.c        | 156 +++++++++++++++++++++++++-----------------------------------------------
 mm/internal.h           |  21 ++--------
 mm/memory.c             |  31 +++++++--------
 mm/migrate.c            | 204 +++++++++++++++++++++--------------------------------------------------------------------------
 7 files changed, 123 insertions(+), 324 deletions(-)
author
#1 lore raw

PATCH v2 1/7 mm: memory: add orig_pmd to struct vm_fault

Add orig_pmd to struct vm_fault so the "orig_pmd" parameter used by huge page fault could be removed, just like its PTE counterpart does.

---
 include/linux/huge_mm.h |  9 ++++-----
 include/linux/mm.h      |  3 +++
 mm/huge_memory.c        |  9 ++++++---
 mm/memory.c             | 26 +++++++++++++-------------
 4 files changed, 26 insertions(+), 21 deletions(-)
include/linux/huge_mm.h +4 -5
index 9626fda5efce..3b38070f4337 100644
@@ -11,7 +11,7 @@ vm_fault_t do_huge_pmd_anonymous_page(struct vm_fault *vmf);
 int copy_huge_pmd(struct mm_struct *dst_mm, struct mm_struct *src_mm,
 		  pmd_t *dst_pmd, pmd_t *src_pmd, unsigned long addr,
 		  struct vm_area_struct *vma);
-void huge_pmd_set_accessed(struct vm_fault *vmf, pmd_t orig_pmd);
+void huge_pmd_set_accessed(struct vm_fault *vmf);
 int copy_huge_pud(struct mm_struct *dst_mm, struct mm_struct *src_mm,
 		  pud_t *dst_pud, pud_t *src_pud, unsigned long addr,
 		  struct vm_area_struct *vma);
@@ -24,7 +24,7 @@ static inline void huge_pud_set_accessed(struct vm_fault *vmf, pud_t orig_pud)
 }
 #endif
 
-vm_fault_t do_huge_pmd_wp_page(struct vm_fault *vmf, pmd_t orig_pmd);
+vm_fault_t do_huge_pmd_wp_page(struct vm_fault *vmf);
 struct page *follow_trans_huge_pmd(struct vm_area_struct *vma,
 				   unsigned long addr, pmd_t *pmd,
 				   unsigned int flags);
@@ -283,7 +283,7 @@ struct page *follow_devmap_pmd(struct vm_area_struct *vma, unsigned long addr,
 struct page *follow_devmap_pud(struct vm_area_struct *vma, unsigned long addr,
 		pud_t *pud, int flags, struct dev_pagemap **pgmap);
 
-vm_fault_t do_huge_pmd_numa_page(struct vm_fault *vmf, pmd_t orig_pmd);
+vm_fault_t do_huge_pmd_numa_page(struct vm_fault *vmf);
 
 extern struct page *huge_zero_page;
 
@@ -429,8 +429,7 @@ static inline spinlock_t *pud_trans_huge_lock(pud_t *pud,
 	return NULL;
 }
 
-static inline vm_fault_t do_huge_pmd_numa_page(struct vm_fault *vmf,
-		pmd_t orig_pmd)
+static inline vm_fault_t do_huge_pmd_numa_page(struct vm_fault *vmf)
 {
 	return 0;
 }
include/linux/mm.h +3 -0
index 25b9041f9925..9c5856f8cc81 100644
@@ -547,6 +547,9 @@ struct vm_fault {
 					 * the 'address'
 					 */
 	pte_t orig_pte;			/* Value of PTE at the time of fault */
+	pmd_t orig_pmd;			/* Value of PMD at the time of fault,
+					 * used by PMD fault only.
+					 */
 
 	struct page *cow_page;		/* Page handler may use for COW fault */
 	struct page *page;		/* ->fault handlers should return a
mm/huge_memory.c +6 -3
index 63ed6b25deaa..35cac4aeaf68 100644
@@ -1251,11 +1251,12 @@ void huge_pud_set_accessed(struct vm_fault *vmf, pud_t orig_pud)
 }
 #endif /* CONFIG_HAVE_ARCH_TRANSPARENT_HUGEPAGE_PUD */
 
-void huge_pmd_set_accessed(struct vm_fault *vmf, pmd_t orig_pmd)
+void huge_pmd_set_accessed(struct vm_fault *vmf)
 {
 	pmd_t entry;
 	unsigned long haddr;
 	bool write = vmf->flags & FAULT_FLAG_WRITE;
+	pmd_t orig_pmd = vmf->orig_pmd;
 
 	vmf->ptl = pmd_lock(vmf->vma->vm_mm, vmf->pmd);
 	if (unlikely(!pmd_same(*vmf->pmd, orig_pmd)))
@@ -1272,11 +1273,12 @@ void huge_pmd_set_accessed(struct vm_fault *vmf, pmd_t orig_pmd)
 	spin_unlock(vmf->ptl);
 }
 
-vm_fault_t do_huge_pmd_wp_page(struct vm_fault *vmf, pmd_t orig_pmd)
+vm_fault_t do_huge_pmd_wp_page(struct vm_fault *vmf)
 {
 	struct vm_area_struct *vma = vmf->vma;
 	struct page *page;
 	unsigned long haddr = vmf->address & HPAGE_PMD_MASK;
+	pmd_t orig_pmd = vmf->orig_pmd;
 
 	vmf->ptl = pmd_lockptr(vma->vm_mm, vmf->pmd);
 	VM_BUG_ON_VMA(!vma->anon_vma, vma);
@@ -1412,9 +1414,10 @@ struct page *follow_trans_huge_pmd(struct vm_area_struct *vma,
 }
 
 /* NUMA hinting page fault entry point for trans huge pmds */
-vm_fault_t do_huge_pmd_numa_page(struct vm_fault *vmf, pmd_t pmd)
+vm_fault_t do_huge_pmd_numa_page(struct vm_fault *vmf)
 {
 	struct vm_area_struct *vma = vmf->vma;
+	pmd_t pmd = vmf->orig_pmd;
 	struct anon_vma *anon_vma = NULL;
 	struct page *page;
 	unsigned long haddr = vmf->address & HPAGE_PMD_MASK;
mm/memory.c +13 -13
index 4e358601c5d6..e0cbffeceb0b 100644
@@ -4232,12 +4232,12 @@ static inline vm_fault_t create_huge_pmd(struct vm_fault *vmf)
 }
 
 /* `inline' is required to avoid gcc 4.1.2 build error */
-static inline vm_fault_t wp_huge_pmd(struct vm_fault *vmf, pmd_t orig_pmd)
+static inline vm_fault_t wp_huge_pmd(struct vm_fault *vmf)
 {
 	if (vma_is_anonymous(vmf->vma)) {
-		if (userfaultfd_huge_pmd_wp(vmf->vma, orig_pmd))
+		if (userfaultfd_huge_pmd_wp(vmf->vma, vmf->orig_pmd))
 			return handle_userfault(vmf, VM_UFFD_WP);
-		return do_huge_pmd_wp_page(vmf, orig_pmd);
+		return do_huge_pmd_wp_page(vmf);
 	}
 	if (vmf->vma->vm_ops->huge_fault) {
 		vm_fault_t ret = vmf->vma->vm_ops->huge_fault(vmf, PE_SIZE_PMD);
@@ -4464,26 +4464,26 @@ static vm_fault_t __handle_mm_fault(struct vm_area_struct *vma,
 		if (!(ret & VM_FAULT_FALLBACK))
 			return ret;
 	} else {
-		pmd_t orig_pmd = *vmf.pmd;
+		vmf.orig_pmd = *vmf.pmd;
 
 		barrier();
-		if (unlikely(is_swap_pmd(orig_pmd))) {
+		if (unlikely(is_swap_pmd(vmf.orig_pmd))) {
 			VM_BUG_ON(thp_migration_supported() &&
-					  !is_pmd_migration_entry(orig_pmd));
-			if (is_pmd_migration_entry(orig_pmd))
+					  !is_pmd_migration_entry(vmf.orig_pmd));
+			if (is_pmd_migration_entry(vmf.orig_pmd))
 				pmd_migration_entry_wait(mm, vmf.pmd);
 			return 0;
 		}
-		if (pmd_trans_huge(orig_pmd) || pmd_devmap(orig_pmd)) {
-			if (pmd_protnone(orig_pmd) && vma_is_accessible(vma))
-				return do_huge_pmd_numa_page(&vmf, orig_pmd);
+		if (pmd_trans_huge(vmf.orig_pmd) || pmd_devmap(vmf.orig_pmd)) {
+			if (pmd_protnone(vmf.orig_pmd) && vma_is_accessible(vma))
+				return do_huge_pmd_numa_page(&vmf);
 
-			if (dirty && !pmd_write(orig_pmd)) {
-				ret = wp_huge_pmd(&vmf, orig_pmd);
+			if (dirty && !pmd_write(vmf.orig_pmd)) {
+				ret = wp_huge_pmd(&vmf);
 				if (!(ret & VM_FAULT_FALLBACK))
 					return ret;
 			} else {
-				huge_pmd_set_accessed(&vmf, orig_pmd);
+				huge_pmd_set_accessed(&vmf);
 				return 0;
 			}
 		}
Signature · 1 line
2.26.2
#2 lore raw

Re: [v2 PATCH 1/7] mm: memory: add orig_pmd to struct vm_fault

Quoted 12 lines · Yang Shi

Add orig_pmd to struct vm_fault so the "orig_pmd" parameter used by huge page fault could be removed, just like its PTE counterpart does.

<SNIP>

include/linux/mm.h +0 -0
index 25b9041f9925..9c5856f8cc81 100644

include/linux/mm.h struct vm_fault { · line 555

Quoted 9 lines
include/linux/mm.h +3 -0
@@ -547,6 +547,9 @@ struct vm_fault {
 					 * the 'address'
 					 */
 	pte_t orig_pte;			/* Value of PTE at the time of fault */
+	pmd_t orig_pmd;			/* Value of PMD at the time of fault,
+					 * used by PMD fault only.
+					 */
 
 	struct page *cow_page;		/* Page handler may use for COW fault */
 	struct page *page;		/* ->fault handlers should return a

Could this be a union?

Signature · 2 lines
Mel Gorman
SUSE Labs
author
#3 lore raw

Re: [v2 PATCH 1/7] mm: memory: add orig_pmd to struct vm_fault

Quoted 21 lines · Mel Gorman
Quoted 19 lines · Yang Shi

Add orig_pmd to struct vm_fault so the "orig_pmd" parameter used by huge page fault could be removed, just like its PTE counterpart does.

<SNIP>

include/linux/mm.h +3 -0
index 25b9041f9925..9c5856f8cc81 100644
@@ -547,6 +547,9 @@ struct vm_fault {
                                       * the 'address'
                                       */
      pte_t orig_pte;                 /* Value of PTE at the time of fault */
+     pmd_t orig_pmd;                 /* Value of PMD at the time of fault,
+                                      * used by PMD fault only.
+                                      */
 
Quoted 4 lines

include/linux/mm.h struct vm_fault { · line 555

Quoted 2 lines
include/linux/mm.h +0 -0
      struct page *cow_page;          /* Page handler may use for COW fault */
      struct page *page;              /* ->fault handlers should return a

Could this be a union?

Do you mean orig_pte and orig_pmd, or cow_page and page? I think orig_pte and orig_pmd could be union, I don't see that they are used at the same time from vm_fault.

But cow_page and page can't since do_cow_fault() needs copy data from page to cow_page.

Quoted 4 lines

--
Mel Gorman
SUSE Labs

#4 lore raw

Re: [v2 PATCH 1/7] mm: memory: add orig_pmd to struct vm_fault

Quoted 22 lines · Yang Shi
Quoted 21 lines · Mel Gorman
Quoted 19 lines · Yang Shi

Add orig_pmd to struct vm_fault so the "orig_pmd" parameter used by huge page fault could be removed, just like its PTE counterpart does.

<SNIP>

include/linux/mm.h +3 -0
index 25b9041f9925..9c5856f8cc81 100644
@@ -547,6 +547,9 @@ struct vm_fault {
                                       * the 'address'
                                       */
      pte_t orig_pte;                 /* Value of PTE at the time of fault */
+     pmd_t orig_pmd;                 /* Value of PMD at the time of fault,
+                                      * used by PMD fault only.
+                                      */
 
Quoted 6 lines
Quoted 4 lines

include/linux/mm.h struct vm_fault { · line 555

Quoted 2 lines
include/linux/mm.h +0 -0
      struct page *cow_page;          /* Page handler may use for COW fault */
      struct page *page;              /* ->fault handlers should return a

Could this be a union?

Do you mean orig_pte and orig_pmd, or cow_page and page?

orig_pte and orig_pmd given that one for PTE faults and one is for PMD faults and it's very unlikely they would both need to be considered during a single fault.

Signature · 2 lines
Mel Gorman
SUSE Labs
author
#5 lore raw

Re: [v2 PATCH 1/7] mm: memory: add orig_pmd to struct vm_fault

Quoted 29 lines · Mel Gorman
Quoted 27 lines · Yang Shi
Quoted 25 lines · Mel Gorman
Quoted 12 lines · Yang Shi

Add orig_pmd to struct vm_fault so the "orig_pmd" parameter used by huge page fault could be removed, just like its PTE counterpart does.

<SNIP>

include/linux/mm.h +0 -0
index 25b9041f9925..9c5856f8cc81 100644

include/linux/mm.h struct vm_fault { · line 555

Quoted 9 lines
include/linux/mm.h +3 -0
@@ -547,6 +547,9 @@ struct vm_fault {
                                       * the 'address'
                                       */
      pte_t orig_pte;                 /* Value of PTE at the time of fault */
+     pmd_t orig_pmd;                 /* Value of PMD at the time of fault,
+                                      * used by PMD fault only.
+                                      */
 
      struct page *cow_page;          /* Page handler may use for COW fault */
      struct page *page;              /* ->fault handlers should return a

Could this be a union?

Quoted 5 lines
Quoted 1 line

Do you mean orig_pte and orig_pmd, or cow_page and page?

orig_pte and orig_pmd given that one for PTE faults and one is for PMD faults and it's very unlikely they would both need to be considered during a single fault.

Yes, agreed.

Quoted 4 lines

--
Mel Gorman
SUSE Labs

author
#6 lore raw

PATCH v2 2/7 mm: memory: make numa_migrate_prep() non-static

In reply to Yang Shi

The numa_migrate_prep() will be used by huge NUMA fault as well in the following patch, make it non-static.

---
 mm/internal.h | 3 +++
 mm/memory.c   | 5 ++---
 2 files changed, 5 insertions(+), 3 deletions(-)
mm/internal.h +3 -0
index f469f69309de..b6f889d9c22d 100644
@@ -659,4 +659,7 @@ int vmap_pages_range_noflush(unsigned long addr, unsigned long end,
 
 void vunmap_range_noflush(unsigned long start, unsigned long end);
 
+int numa_migrate_prep(struct page *page, struct vm_area_struct *vma,
+		      unsigned long addr, int page_nid, int *flags);
+
 #endif	/* __MM_INTERNAL_H */
mm/memory.c +2 -3
index e0cbffeceb0b..bbb38066021a 100644
@@ -4119,9 +4119,8 @@ static vm_fault_t do_fault(struct vm_fault *vmf)
 	return ret;
 }
 
-static int numa_migrate_prep(struct page *page, struct vm_area_struct *vma,
-				unsigned long addr, int page_nid,
-				int *flags)
+int numa_migrate_prep(struct page *page, struct vm_area_struct *vma,
+		      unsigned long addr, int page_nid, int *flags)
 {
 	get_page(page);
 
Signature · 1 line
2.26.2
author
#8 lore raw

PATCH v2 3/7 mm: thp: refactor NUMA fault handling

In reply to Yang Shi

When the THP NUMA fault support was added THP migration was not supported yet. So the ad hoc THP migration was implemented in NUMA fault handling. Since v4.14 THP migration has been supported so it doesn't make too much sense to still keep another THP migration implementation rather than using the generic migration code.

This patch reworked the NUMA fault handling to use generic migration implementation to migrate misplaced page. There is no functional change.

After the refactor the flow of NUMA fault handling looks just like its PTE counterpart:
Acquire ptl
Prepare for migration (elevate page refcount)
Release ptl
Isolate page from lru and elevate page refcount
Migrate the misplaced THP

If migration is failed just restore the old normal PMD.

In the old code anon_vma lock was needed to serialize THP migration against THP split, but since then the THP code has been reworked a lot, it seems anon_vma lock is not required anymore to avoid the race.

The page refcount elevation when holding ptl should prevent from THP split.

Use migrate_misplaced_page() for both base page and THP NUMA hinting fault and remove all the dead and duplicate code.

---
 include/linux/migrate.h |  23 ------
 mm/huge_memory.c        | 143 ++++++++++----------------------
 mm/internal.h           |  18 ----
 mm/migrate.c            | 177 ++++++++--------------------------------
 4 files changed, 77 insertions(+), 284 deletions(-)
include/linux/migrate.h +0 -23
index 4bb4e519e3f5..163d6f2b03d1 100644
@@ -95,14 +95,9 @@ static inline void __ClearPageMovable(struct page *page)
 #endif
 
 #ifdef CONFIG_NUMA_BALANCING
-extern bool pmd_trans_migrating(pmd_t pmd);
 extern int migrate_misplaced_page(struct page *page,
 				  struct vm_area_struct *vma, int node);
 #else
-static inline bool pmd_trans_migrating(pmd_t pmd)
-{
-	return false;
-}
 static inline int migrate_misplaced_page(struct page *page,
 					 struct vm_area_struct *vma, int node)
 {
@@ -110,24 +105,6 @@ static inline int migrate_misplaced_page(struct page *page,
 }
 #endif /* CONFIG_NUMA_BALANCING */
 
-#if defined(CONFIG_NUMA_BALANCING) && defined(CONFIG_TRANSPARENT_HUGEPAGE)
-extern int migrate_misplaced_transhuge_page(struct mm_struct *mm,
-			struct vm_area_struct *vma,
-			pmd_t *pmd, pmd_t entry,
-			unsigned long address,
-			struct page *page, int node);
-#else
-static inline int migrate_misplaced_transhuge_page(struct mm_struct *mm,
-			struct vm_area_struct *vma,
-			pmd_t *pmd, pmd_t entry,
-			unsigned long address,
-			struct page *page, int node)
-{
-	return -EAGAIN;
-}
-#endif /* CONFIG_NUMA_BALANCING && CONFIG_TRANSPARENT_HUGEPAGE*/
-
-
 #ifdef CONFIG_MIGRATION
 
 /*
mm/huge_memory.c +44 -99

180 lines not on this page · raw message

mm/internal.h +0 -18
index b6f889d9c22d..c060551d6ebf 100644
@@ -381,23 +381,6 @@ extern int mlock_future_check(struct mm_struct *mm, unsigned long flags,
  */
 extern void clear_page_mlock(struct page *page);
 
-/*
- * mlock_migrate_page - called only from migrate_misplaced_transhuge_page()
- * (because that does not go through the full procedure of migration ptes):
- * to migrate the Mlocked page flag; update statistics.
- */
-static inline void mlock_migrate_page(struct page *newpage, struct page *page)
-{
-	if (TestClearPageMlocked(page)) {
-		int nr_pages = thp_nr_pages(page);
-
-		/* Holding pmd lock, no change in irq context: __mod is safe */
-		__mod_zone_page_state(page_zone(page), NR_MLOCK, -nr_pages);
-		SetPageMlocked(newpage);
-		__mod_zone_page_state(page_zone(newpage), NR_MLOCK, nr_pages);
-	}
-}
-
 extern pmd_t maybe_pmd_mkwrite(pmd_t pmd, struct vm_area_struct *vma);
 
 /*
@@ -448,7 +431,6 @@ static inline struct file *maybe_unlock_mmap_for_io(struct vm_fault *vmf,
 #else /* !CONFIG_MMU */
 static inline void clear_page_mlock(struct page *page) { }
 static inline void mlock_vma_page(struct page *page) { }
-static inline void mlock_migrate_page(struct page *new, struct page *old) { }
 static inline void vunmap_range_noflush(unsigned long start, unsigned long end)
 {
 }
mm/migrate.c +33 -144

207 lines not on this page · raw message

Signature · 1 line
2.26.2
#9 lore raw

Re: [v2 PATCH 3/7] mm: thp: refactor NUMA fault handling

Quoted 86 lines · Yang Shi

When the THP NUMA fault support was added THP migration was not supported yet. So the ad hoc THP migration was implemented in NUMA fault handling. Since v4.14 THP migration has been supported so it doesn't make too much sense to still keep another THP migration implementation rather than using the generic migration code.

This patch reworked the NUMA fault handling to use generic migration implementation to migrate misplaced page. There is no functional change.

After the refactor the flow of NUMA fault handling looks just like its PTE counterpart:
Acquire ptl
Prepare for migration (elevate page refcount)
Release ptl
Isolate page from lru and elevate page refcount
Migrate the misplaced THP

If migration is failed just restore the old normal PMD.

In the old code anon_vma lock was needed to serialize THP migration against THP split, but since then the THP code has been reworked a lot, it seems anon_vma lock is not required anymore to avoid the race.

The page refcount elevation when holding ptl should prevent from THP split.

Use migrate_misplaced_page() for both base page and THP NUMA hinting fault and remove all the dead and duplicate code.

---
 include/linux/migrate.h |  23 ------
 mm/huge_memory.c        | 143 ++++++++++----------------------
 mm/internal.h           |  18 ----
 mm/migrate.c            | 177 ++++++++--------------------------------
 4 files changed, 77 insertions(+), 284 deletions(-)
include/linux/migrate.h +0 -23
index 4bb4e519e3f5..163d6f2b03d1 100644
@@ -95,14 +95,9 @@ static inline void __ClearPageMovable(struct page *page)
 #endif
 
 #ifdef CONFIG_NUMA_BALANCING
-extern bool pmd_trans_migrating(pmd_t pmd);
 extern int migrate_misplaced_page(struct page *page,
 				  struct vm_area_struct *vma, int node);
 #else
-static inline bool pmd_trans_migrating(pmd_t pmd)
-{
-	return false;
-}
 static inline int migrate_misplaced_page(struct page *page,
 					 struct vm_area_struct *vma, int node)
 {
@@ -110,24 +105,6 @@ static inline int migrate_misplaced_page(struct page *page,
 }
 #endif /* CONFIG_NUMA_BALANCING */
 
-#if defined(CONFIG_NUMA_BALANCING) && defined(CONFIG_TRANSPARENT_HUGEPAGE)
-extern int migrate_misplaced_transhuge_page(struct mm_struct *mm,
-			struct vm_area_struct *vma,
-			pmd_t *pmd, pmd_t entry,
-			unsigned long address,
-			struct page *page, int node);
-#else
-static inline int migrate_misplaced_transhuge_page(struct mm_struct *mm,
-			struct vm_area_struct *vma,
-			pmd_t *pmd, pmd_t entry,
-			unsigned long address,
-			struct page *page, int node)
-{
-	return -EAGAIN;
-}
-#endif /* CONFIG_NUMA_BALANCING && CONFIG_TRANSPARENT_HUGEPAGE*/
-
-
 #ifdef CONFIG_MIGRATION
 
 /*
mm/huge_memory.c +0 -0
index 35cac4aeaf68..94981907fd4c 100644

mm/huge_memory.c vm_fault_t do_huge_pmd_numa_page(struct vm_fault *vmf) · line 1421

Quoted 5 lines
mm/huge_memory.c +1 -1
@@ -1418,93 +1418,21 @@ vm_fault_t do_huge_pmd_numa_page(struct vm_fault *vmf)
 {
 	struct vm_area_struct *vma = vmf->vma;
 	pmd_t pmd = vmf->orig_pmd;
-	struct anon_vma *anon_vma = NULL;
+	pmd_t oldpmd;

nit: the usage of oldpmd and pmd in the function appears not very consistent. How about make oldpmd == vmf->orig_pmd always. While make pmd the changed one?

Best Regards,
Huang, Ying

Quoted 175 lines
mm/huge_memory.c +43 -98
 	struct page *page;
 	unsigned long haddr = vmf->address & HPAGE_PMD_MASK;
-	int page_nid = NUMA_NO_NODE, this_nid = numa_node_id();
+	int page_nid = NUMA_NO_NODE;
 	int target_nid, last_cpupid = -1;
-	bool page_locked;
 	bool migrated = false;
-	bool was_writable;
+	bool was_writable = pmd_savedwrite(pmd);
 	int flags = 0;
 
 	vmf->ptl = pmd_lock(vma->vm_mm, vmf->pmd);
-	if (unlikely(!pmd_same(pmd, *vmf->pmd)))
-		goto out_unlock;
-
-	/*
-	 * If there are potential migrations, wait for completion and retry
-	 * without disrupting NUMA hinting information. Do not relock and
-	 * check_same as the page may no longer be mapped.
-	 */
-	if (unlikely(pmd_trans_migrating(*vmf->pmd))) {
-		page = pmd_page(*vmf->pmd);
-		if (!get_page_unless_zero(page))
-			goto out_unlock;
+	if (unlikely(!pmd_same(pmd, *vmf->pmd))) {
 		spin_unlock(vmf->ptl);
-		put_and_wait_on_page_locked(page, TASK_UNINTERRUPTIBLE);
 		goto out;
 	}
 
-	page = pmd_page(pmd);
-	BUG_ON(is_huge_zero_page(page));
-	page_nid = page_to_nid(page);
-	last_cpupid = page_cpupid_last(page);
-	count_vm_numa_event(NUMA_HINT_FAULTS);
-	if (page_nid == this_nid) {
-		count_vm_numa_event(NUMA_HINT_FAULTS_LOCAL);
-		flags |= TNF_FAULT_LOCAL;
-	}
-
-	/* See similar comment in do_numa_page for explanation */
-	if (!pmd_savedwrite(pmd))
-		flags |= TNF_NO_GROUP;
-
-	/*
-	 * Acquire the page lock to serialise THP migrations but avoid dropping
-	 * page_table_lock if at all possible
-	 */
-	page_locked = trylock_page(page);
-	target_nid = mpol_misplaced(page, vma, haddr);
-	/* Migration could have started since the pmd_trans_migrating check */
-	if (!page_locked) {
-		page_nid = NUMA_NO_NODE;
-		if (!get_page_unless_zero(page))
-			goto out_unlock;
-		spin_unlock(vmf->ptl);
-		put_and_wait_on_page_locked(page, TASK_UNINTERRUPTIBLE);
-		goto out;
-	} else if (target_nid == NUMA_NO_NODE) {
-		/* There are no parallel migrations and page is in the right
-		 * node. Clear the numa hinting info in this pmd.
-		 */
-		goto clear_pmdnuma;
-	}
-
-	/*
-	 * Page is misplaced. Page lock serialises migrations. Acquire anon_vma
-	 * to serialises splits
-	 */
-	get_page(page);
-	spin_unlock(vmf->ptl);
-	anon_vma = page_lock_anon_vma_read(page);
-
-	/* Confirm the PMD did not change while page_table_lock was released */
-	spin_lock(vmf->ptl);
-	if (unlikely(!pmd_same(pmd, *vmf->pmd))) {
-		unlock_page(page);
-		put_page(page);
-		page_nid = NUMA_NO_NODE;
-		goto out_unlock;
-	}
-
-	/* Bail if we fail to protect against THP splits for any reason */
-	if (unlikely(!anon_vma)) {
-		put_page(page);
-		page_nid = NUMA_NO_NODE;
-		goto clear_pmdnuma;
-	}
-
 	/*
 	 * Since we took the NUMA fault, we must have observed the !accessible
 	 * bit. Make sure all other CPUs agree with that, to avoid them
@@ -1531,43 +1459,60 @@ vm_fault_t do_huge_pmd_numa_page(struct vm_fault *vmf)
 					      haddr + HPAGE_PMD_SIZE);
 	}
 
-	/*
-	 * Migrate the THP to the requested node, returns with page unlocked
-	 * and access rights restored.
-	 */
+	oldpmd = pmd_modify(pmd, vma->vm_page_prot);
+	page = vm_normal_page_pmd(vma, haddr, oldpmd);
+	if (!page) {
+		spin_unlock(vmf->ptl);
+		goto out_map;
+	}
+
+	/* See similar comment in do_numa_page for explanation */
+	if (!was_writable)
+		flags |= TNF_NO_GROUP;
+
+	page_nid = page_to_nid(page);
+	last_cpupid = page_cpupid_last(page);
+	target_nid = numa_migrate_prep(page, vma, haddr, page_nid,
+				       &flags);
+
+	if (target_nid == NUMA_NO_NODE) {
+		put_page(page);
+		goto out_map;
+	}
+
 	spin_unlock(vmf->ptl);
 
-	migrated = migrate_misplaced_transhuge_page(vma->vm_mm, vma,
-				vmf->pmd, pmd, vmf->address, page, target_nid);
+	migrated = migrate_misplaced_page(page, vma, target_nid);
 	if (migrated) {
 		flags |= TNF_MIGRATED;
 		page_nid = target_nid;
-	} else
+	} else {
 		flags |= TNF_MIGRATE_FAIL;
+		vmf->ptl = pmd_lock(vma->vm_mm, vmf->pmd);
+		if (unlikely(!pmd_same(pmd, *vmf->pmd))) {
+			spin_unlock(vmf->ptl);
+			goto out;
+		}
+		goto out_map;
+	}
 
-	goto out;
-clear_pmdnuma:
-	BUG_ON(!PageLocked(page));
-	was_writable = pmd_savedwrite(pmd);
+out:
+	if (page_nid != NUMA_NO_NODE)
+		task_numa_fault(last_cpupid, page_nid, HPAGE_PMD_NR,
+				flags);
+
+	return 0;
+
+out_map:
+	/* Restore the PMD */
 	pmd = pmd_modify(pmd, vma->vm_page_prot);
 	pmd = pmd_mkyoung(pmd);
 	if (was_writable)
 		pmd = pmd_mkwrite(pmd);
 	set_pmd_at(vma->vm_mm, haddr, vmf->pmd, pmd);
 	update_mmu_cache_pmd(vma, vmf->address, vmf->pmd);
-	unlock_page(page);
-out_unlock:
 	spin_unlock(vmf->ptl);
-
-out:
-	if (anon_vma)
-		page_unlock_anon_vma_read(anon_vma);
-
-	if (page_nid != NUMA_NO_NODE)
-		task_numa_fault(last_cpupid, page_nid, HPAGE_PMD_NR,
-				flags);
-
-	return 0;
+	goto out;
 }
 

[snip]

author
#10 lore raw

Re: [v2 PATCH 3/7] mm: thp: refactor NUMA fault handling

Quoted 83 lines · Ying
Quoted 80 lines · Yang Shi

When the THP NUMA fault support was added THP migration was not supported yet. So the ad hoc THP migration was implemented in NUMA fault handling. Since v4.14 THP migration has been supported so it doesn't make too much sense to still keep another THP migration implementation rather than using the generic migration code.

This patch reworked the NUMA fault handling to use generic migration implementation to migrate misplaced page. There is no functional change.

After the refactor the flow of NUMA fault handling looks just like its PTE counterpart:
Acquire ptl
Prepare for migration (elevate page refcount)
Release ptl
Isolate page from lru and elevate page refcount
Migrate the misplaced THP

If migration is failed just restore the old normal PMD.

In the old code anon_vma lock was needed to serialize THP migration against THP split, but since then the THP code has been reworked a lot, it seems anon_vma lock is not required anymore to avoid the race.

The page refcount elevation when holding ptl should prevent from THP split.

Use migrate_misplaced_page() for both base page and THP NUMA hinting fault and remove all the dead and duplicate code.

---
 include/linux/migrate.h |  23 ------
 mm/huge_memory.c        | 143 ++++++++++----------------------
 mm/internal.h           |  18 ----
 mm/migrate.c            | 177 ++++++++--------------------------------
 4 files changed, 77 insertions(+), 284 deletions(-)
include/linux/migrate.h +0 -23
index 4bb4e519e3f5..163d6f2b03d1 100644
@@ -95,14 +95,9 @@ static inline void __ClearPageMovable(struct page *page)
 #endif
 
 #ifdef CONFIG_NUMA_BALANCING
-extern bool pmd_trans_migrating(pmd_t pmd);
 extern int migrate_misplaced_page(struct page *page,
                                struct vm_area_struct *vma, int node);
 #else
-static inline bool pmd_trans_migrating(pmd_t pmd)
-{
-     return false;
-}
 static inline int migrate_misplaced_page(struct page *page,
                                       struct vm_area_struct *vma, int node)
 {
@@ -110,24 +105,6 @@ static inline int migrate_misplaced_page(struct page *page,
 }
 #endif /* CONFIG_NUMA_BALANCING */
 
-#if defined(CONFIG_NUMA_BALANCING) && defined(CONFIG_TRANSPARENT_HUGEPAGE)
-extern int migrate_misplaced_transhuge_page(struct mm_struct *mm,
-                     struct vm_area_struct *vma,
-                     pmd_t *pmd, pmd_t entry,
-                     unsigned long address,
-                     struct page *page, int node);
-#else
-static inline int migrate_misplaced_transhuge_page(struct mm_struct *mm,
-                     struct vm_area_struct *vma,
-                     pmd_t *pmd, pmd_t entry,
-                     unsigned long address,
-                     struct page *page, int node)
-{
-     return -EAGAIN;
-}
-#endif /* CONFIG_NUMA_BALANCING && CONFIG_TRANSPARENT_HUGEPAGE*/
-
-
 #ifdef CONFIG_MIGRATION
 
Quoted 15 lines

mm/huge_memory.c vm_fault_t do_huge_pmd_numa_page(struct vm_fault *vmf) · line 1421

Quoted 11 lines
include/linux/migrate.h +0 -0
 /*
mm/huge_memory.c +1 -1
index 35cac4aeaf68..94981907fd4c 100644
@@ -1418,93 +1418,21 @@ vm_fault_t do_huge_pmd_numa_page(struct vm_fault *vmf)
 {
      struct vm_area_struct *vma = vmf->vma;
      pmd_t pmd = vmf->orig_pmd;
-     struct anon_vma *anon_vma = NULL;
+     pmd_t oldpmd;

nit: the usage of oldpmd and pmd in the function appears not very consistent. How about make oldpmd == vmf->orig_pmd always. While make pmd the changed one?

Thanks for the suggestion. Yes, it seemed neater. Will fix it in the next version.

Quoted 181 lines

Best Regards,
Huang, Ying

Quoted 175 lines
mm/huge_memory.c +43 -98
      struct page *page;
      unsigned long haddr = vmf->address & HPAGE_PMD_MASK;
-     int page_nid = NUMA_NO_NODE, this_nid = numa_node_id();
+     int page_nid = NUMA_NO_NODE;
      int target_nid, last_cpupid = -1;
-     bool page_locked;
      bool migrated = false;
-     bool was_writable;
+     bool was_writable = pmd_savedwrite(pmd);
      int flags = 0;
 
      vmf->ptl = pmd_lock(vma->vm_mm, vmf->pmd);
-     if (unlikely(!pmd_same(pmd, *vmf->pmd)))
-             goto out_unlock;
-
-     /*
-      * If there are potential migrations, wait for completion and retry
-      * without disrupting NUMA hinting information. Do not relock and
-      * check_same as the page may no longer be mapped.
-      */
-     if (unlikely(pmd_trans_migrating(*vmf->pmd))) {
-             page = pmd_page(*vmf->pmd);
-             if (!get_page_unless_zero(page))
-                     goto out_unlock;
+     if (unlikely(!pmd_same(pmd, *vmf->pmd))) {
              spin_unlock(vmf->ptl);
-             put_and_wait_on_page_locked(page, TASK_UNINTERRUPTIBLE);
              goto out;
      }
 
-     page = pmd_page(pmd);
-     BUG_ON(is_huge_zero_page(page));
-     page_nid = page_to_nid(page);
-     last_cpupid = page_cpupid_last(page);
-     count_vm_numa_event(NUMA_HINT_FAULTS);
-     if (page_nid == this_nid) {
-             count_vm_numa_event(NUMA_HINT_FAULTS_LOCAL);
-             flags |= TNF_FAULT_LOCAL;
-     }
-
-     /* See similar comment in do_numa_page for explanation */
-     if (!pmd_savedwrite(pmd))
-             flags |= TNF_NO_GROUP;
-
-     /*
-      * Acquire the page lock to serialise THP migrations but avoid dropping
-      * page_table_lock if at all possible
-      */
-     page_locked = trylock_page(page);
-     target_nid = mpol_misplaced(page, vma, haddr);
-     /* Migration could have started since the pmd_trans_migrating check */
-     if (!page_locked) {
-             page_nid = NUMA_NO_NODE;
-             if (!get_page_unless_zero(page))
-                     goto out_unlock;
-             spin_unlock(vmf->ptl);
-             put_and_wait_on_page_locked(page, TASK_UNINTERRUPTIBLE);
-             goto out;
-     } else if (target_nid == NUMA_NO_NODE) {
-             /* There are no parallel migrations and page is in the right
-              * node. Clear the numa hinting info in this pmd.
-              */
-             goto clear_pmdnuma;
-     }
-
-     /*
-      * Page is misplaced. Page lock serialises migrations. Acquire anon_vma
-      * to serialises splits
-      */
-     get_page(page);
-     spin_unlock(vmf->ptl);
-     anon_vma = page_lock_anon_vma_read(page);
-
-     /* Confirm the PMD did not change while page_table_lock was released */
-     spin_lock(vmf->ptl);
-     if (unlikely(!pmd_same(pmd, *vmf->pmd))) {
-             unlock_page(page);
-             put_page(page);
-             page_nid = NUMA_NO_NODE;
-             goto out_unlock;
-     }
-
-     /* Bail if we fail to protect against THP splits for any reason */
-     if (unlikely(!anon_vma)) {
-             put_page(page);
-             page_nid = NUMA_NO_NODE;
-             goto clear_pmdnuma;
-     }
-
      /*
       * Since we took the NUMA fault, we must have observed the !accessible
       * bit. Make sure all other CPUs agree with that, to avoid them
@@ -1531,43 +1459,60 @@ vm_fault_t do_huge_pmd_numa_page(struct vm_fault *vmf)
                                            haddr + HPAGE_PMD_SIZE);
      }
 
-     /*
-      * Migrate the THP to the requested node, returns with page unlocked
-      * and access rights restored.
-      */
+     oldpmd = pmd_modify(pmd, vma->vm_page_prot);
+     page = vm_normal_page_pmd(vma, haddr, oldpmd);
+     if (!page) {
+             spin_unlock(vmf->ptl);
+             goto out_map;
+     }
+
+     /* See similar comment in do_numa_page for explanation */
+     if (!was_writable)
+             flags |= TNF_NO_GROUP;
+
+     page_nid = page_to_nid(page);
+     last_cpupid = page_cpupid_last(page);
+     target_nid = numa_migrate_prep(page, vma, haddr, page_nid,
+                                    &flags);
+
+     if (target_nid == NUMA_NO_NODE) {
+             put_page(page);
+             goto out_map;
+     }
+
      spin_unlock(vmf->ptl);
 
-     migrated = migrate_misplaced_transhuge_page(vma->vm_mm, vma,
-                             vmf->pmd, pmd, vmf->address, page, target_nid);
+     migrated = migrate_misplaced_page(page, vma, target_nid);
      if (migrated) {
              flags |= TNF_MIGRATED;
              page_nid = target_nid;
-     } else
+     } else {
              flags |= TNF_MIGRATE_FAIL;
+             vmf->ptl = pmd_lock(vma->vm_mm, vmf->pmd);
+             if (unlikely(!pmd_same(pmd, *vmf->pmd))) {
+                     spin_unlock(vmf->ptl);
+                     goto out;
+             }
+             goto out_map;
+     }
 
-     goto out;
-clear_pmdnuma:
-     BUG_ON(!PageLocked(page));
-     was_writable = pmd_savedwrite(pmd);
+out:
+     if (page_nid != NUMA_NO_NODE)
+             task_numa_fault(last_cpupid, page_nid, HPAGE_PMD_NR,
+                             flags);
+
+     return 0;
+
+out_map:
+     /* Restore the PMD */
      pmd = pmd_modify(pmd, vma->vm_page_prot);
      pmd = pmd_mkyoung(pmd);
      if (was_writable)
              pmd = pmd_mkwrite(pmd);
      set_pmd_at(vma->vm_mm, haddr, vmf->pmd, pmd);
      update_mmu_cache_pmd(vma, vmf->address, vmf->pmd);
-     unlock_page(page);
-out_unlock:
      spin_unlock(vmf->ptl);
-
-out:
-     if (anon_vma)
-             page_unlock_anon_vma_read(anon_vma);
-
-     if (page_nid != NUMA_NO_NODE)
-             task_numa_fault(last_cpupid, page_nid, HPAGE_PMD_NR,
-                             flags);
-
-     return 0;
+     goto out;
 }
 

[snip]

Acked-by on v2 3/7
#11 lore raw

Re: [v2 PATCH 3/7] mm: thp: refactor NUMA fault handling

In reply to Yang Shi
Quoted 23 lines · Yang Shi

When the THP NUMA fault support was added THP migration was not supported yet. So the ad hoc THP migration was implemented in NUMA fault handling. Since v4.14 THP migration has been supported so it doesn't make too much sense to still keep another THP migration implementation rather than using the generic migration code.

This patch reworked the NUMA fault handling to use generic migration implementation to migrate misplaced page. There is no functional change.

After the refactor the flow of NUMA fault handling looks just like its PTE counterpart:
Acquire ptl
Prepare for migration (elevate page refcount)
Release ptl
Isolate page from lru and elevate page refcount
Migrate the misplaced THP

If migration is failed just restore the old normal PMD.

In the old code anon_vma lock was needed to serialize THP migration against THP split, but since then the THP code has been reworked a lot, it seems anon_vma lock is not required anymore to avoid the race.

Quoted 7 lines

The page refcount elevation when holding ptl should prevent from THP split.

Use migrate_misplaced_page() for both base page and THP NUMA hinting fault and remove all the dead and duplicate code.

I did not spot any big problems and FWIW, the series overall passed a series of tests that exercise NUMA balancing migrations so...

Signature · 2 lines
Mel Gorman
SUSE Labs
author
#12 lore raw

Re: [v2 PATCH 3/7] mm: thp: refactor NUMA fault handling

Quoted 31 lines · Mel Gorman
Quoted 29 lines · Yang Shi

When the THP NUMA fault support was added THP migration was not supported yet. So the ad hoc THP migration was implemented in NUMA fault handling. Since v4.14 THP migration has been supported so it doesn't make too much sense to still keep another THP migration implementation rather than using the generic migration code.

This patch reworked the NUMA fault handling to use generic migration implementation to migrate misplaced page. There is no functional change.

After the refactor the flow of NUMA fault handling looks just like its PTE counterpart:
Acquire ptl
Prepare for migration (elevate page refcount)
Release ptl
Isolate page from lru and elevate page refcount
Migrate the misplaced THP

If migration is failed just restore the old normal PMD.

In the old code anon_vma lock was needed to serialize THP migration against THP split, but since then the THP code has been reworked a lot, it seems anon_vma lock is not required anymore to avoid the race.

The page refcount elevation when holding ptl should prevent from THP split.

Use migrate_misplaced_page() for both base page and THP NUMA hinting fault and remove all the dead and duplicate code.

Quoted 6 lines
Quoted 1 line

I did not spot any big problems and FWIW, the series overall passed a series of tests that exercise NUMA balancing migrations so...

Thanks a lot for testing and reviewing the series.

Quoted 4 lines

--
Mel Gorman
SUSE Labs

author
#13 lore raw

PATCH v2 4/7 mm: migrate: account THP NUMA migration counters correctly

In reply to Yang Shi

Now both base page and THP NUMA migration is done via migrate_misplaced_page(), keep the counters correctly for THP.

---
 mm/migrate.c | 7 ++++---
 1 file changed, 4 insertions(+), 3 deletions(-)
mm/migrate.c +4 -3
index 333448aa53f1..a473f25fbd01 100644
@@ -2111,6 +2111,7 @@ int migrate_misplaced_page(struct page *page, struct vm_area_struct *vma,
 	LIST_HEAD(migratepages);
 	new_page_t *new;
 	bool compound;
+	unsigned int nr_pages = thp_nr_pages(page);
 
 	/*
 	 * PTE mapped THP or HugeTLB page can't reach here so the page could
@@ -2149,13 +2150,13 @@ int migrate_misplaced_page(struct page *page, struct vm_area_struct *vma,
 	if (nr_remaining) {
 		if (!list_empty(&migratepages)) {
 			list_del(&page->lru);
-			dec_node_page_state(page, NR_ISOLATED_ANON +
-					page_is_file_lru(page));
+			mod_node_page_state(page_pgdat(page), NR_ISOLATED_ANON +
+					page_is_file_lru(page), -nr_pages);
 			putback_lru_page(page);
 		}
 		isolated = 0;
 	} else
-		count_vm_numa_event(NUMA_PAGE_MIGRATE);
+		count_vm_numa_events(NUMA_PAGE_MIGRATE, nr_pages);
 	BUG_ON(!list_empty(&migratepages));
 	return isolated;
 
Signature · 1 line
2.26.2
Acked-by on v2 4/7
#14 lore raw

Re: [v2 PATCH 4/7] mm: migrate: account THP NUMA migration counters correctly

Quoted 4 lines · Yang Shi

Now both base page and THP NUMA migration is done via migrate_misplaced_page(), keep the counters correctly for THP.

Signature · 2 lines
Mel Gorman
SUSE Labs
author
#15 lore raw

PATCH v2 5/7 mm: migrate: don't split THP for misplaced NUMA page

In reply to Yang Shi

The old behavior didn't split THP if migration is failed due to lack of memory on the target node. But the THP migration does split THP, so keep the old behavior for misplaced NUMA page migration.

---
 mm/migrate.c | 4 +++-
 1 file changed, 3 insertions(+), 1 deletion(-)
mm/migrate.c +3 -1
index a473f25fbd01..a72994c68ec6 100644
@@ -1417,6 +1417,7 @@ int migrate_pages(struct list_head *from, new_page_t get_new_page,
 	int swapwrite = current->flags & PF_SWAPWRITE;
 	int rc, nr_subpages;
 	LIST_HEAD(ret_pages);
+	bool nosplit = (reason == MR_NUMA_MISPLACED);
 
 	trace_mm_migrate_pages_start(mode, reason);
 
@@ -1488,8 +1489,9 @@ int migrate_pages(struct list_head *from, new_page_t get_new_page,
 				/*
 				 * When memory is low, don't bother to try to migrate
 				 * other pages, just exit.
+				 * THP NUMA faulting doesn't split THP to retry.
 				 */
-				if (is_thp) {
+				if (is_thp && !nosplit) {
 					if (!try_split_thp(page, &page2, from)) {
 						nr_thp_split++;
 						goto retry;
Signature · 1 line
2.26.2
Acked-by on v2 5/7
#16 lore raw

Re: [v2 PATCH 5/7] mm: migrate: don't split THP for misplaced NUMA page

Quoted 5 lines · Yang Shi

The old behavior didn't split THP if migration is failed due to lack of memory on the target node. But the THP migration does split THP, so keep the old behavior for misplaced NUMA page migration.

Signature · 2 lines
Mel Gorman
SUSE Labs
author
#17 lore raw

PATCH v2 6/7 mm: migrate: check mapcount for THP instead of ref count

In reply to Yang Shi

The generic migration path will check refcount, so no need check refcount here. But the old code actually prevents from migrating shared THP (mapped by multiple processes), so bail out early if mapcount is > 1 to keep the behavior.

---
 mm/migrate.c | 16 ++++------------
 1 file changed, 4 insertions(+), 12 deletions(-)
mm/migrate.c +4 -12
index a72994c68ec6..dc7cc7f3a124 100644
@@ -2067,6 +2067,10 @@ static int numamigrate_isolate_page(pg_data_t *pgdat, struct page *page)
 
 	VM_BUG_ON_PAGE(compound_order(page) && !PageTransHuge(page), page);
 
+	/* Do not migrate THP mapped by multiple processes */
+	if (PageTransHuge(page) && page_mapcount(page) > 1)
+		return 0;
+
 	/* Avoid migrating to a node that is nearly full */
 	if (!migrate_balanced_pgdat(pgdat, compound_nr(page)))
 		return 0;
@@ -2074,18 +2078,6 @@ static int numamigrate_isolate_page(pg_data_t *pgdat, struct page *page)
 	if (isolate_lru_page(page))
 		return 0;
 
-	/*
-	 * migrate_misplaced_transhuge_page() skips page migration's usual
-	 * check on page_count(), so we must do it here, now that the page
-	 * has been isolated: a GUP pin, or any other pin, prevents migration.
-	 * The expected page count is 3: 1 for page's mapcount and 1 for the
-	 * caller's pin and 1 for the reference taken by isolate_lru_page().
-	 */
-	if (PageTransHuge(page) && page_count(page) != 3) {
-		putback_lru_page(page);
-		return 0;
-	}
-
 	page_lru = page_is_file_lru(page);
 	mod_node_page_state(page_pgdat(page), NR_ISOLATED_ANON + page_lru,
 				thp_nr_pages(page));
Signature · 1 line
2.26.2
#18 lore raw

Re: [v2 PATCH 6/7] mm: migrate: check mapcount for THP instead of ref count

Quoted 3 lines · Yang Shi

The generic migration path will check refcount, so no need check refcount here. But the old code actually prevents from migrating shared THP (mapped by multiple processes), so bail out early if mapcount is > 1 to keep the behavior.

What prevents us from migrating shared THP? If no, why not just remove the old refcount checking?

Best Regards,
Huang, Ying

Quoted 39 lines
---
 mm/migrate.c | 16 ++++------------
 1 file changed, 4 insertions(+), 12 deletions(-)
mm/migrate.c +4 -12
index a72994c68ec6..dc7cc7f3a124 100644
@@ -2067,6 +2067,10 @@ static int numamigrate_isolate_page(pg_data_t *pgdat, struct page *page)
 
 	VM_BUG_ON_PAGE(compound_order(page) && !PageTransHuge(page), page);
 
+	/* Do not migrate THP mapped by multiple processes */
+	if (PageTransHuge(page) && page_mapcount(page) > 1)
+		return 0;
+
 	/* Avoid migrating to a node that is nearly full */
 	if (!migrate_balanced_pgdat(pgdat, compound_nr(page)))
 		return 0;
@@ -2074,18 +2078,6 @@ static int numamigrate_isolate_page(pg_data_t *pgdat, struct page *page)
 	if (isolate_lru_page(page))
 		return 0;
 
-	/*
-	 * migrate_misplaced_transhuge_page() skips page migration's usual
-	 * check on page_count(), so we must do it here, now that the page
-	 * has been isolated: a GUP pin, or any other pin, prevents migration.
-	 * The expected page count is 3: 1 for page's mapcount and 1 for the
-	 * caller's pin and 1 for the reference taken by isolate_lru_page().
-	 */
-	if (PageTransHuge(page) && page_count(page) != 3) {
-		putback_lru_page(page);
-		return 0;
-	}
-
 	page_lru = page_is_file_lru(page);
 	mod_node_page_state(page_pgdat(page), NR_ISOLATED_ANON + page_lru,
 				thp_nr_pages(page));
reviewer
#19 lore raw

Re: [v2 PATCH 6/7] mm: migrate: check mapcount for THP instead of ref count

Quoted 8 lines · Ying
Quoted 3 lines · Yang Shi

The generic migration path will check refcount, so no need check refcount here. But the old code actually prevents from migrating shared THP (mapped by multiple processes), so bail out early if mapcount is > 1 to keep the behavior.

What prevents us from migrating shared THP? If no, why not just remove the old refcount checking?

If two or more processes are in different NUMA nodes, a THP shared by them can be migrated back and forth between NUMA nodes, which is quite costly. Unless we have a better way of figuring out a good location for such pages to reduce the number of migration, it might be better not to move them, right?

Quoted 28 lines

Best Regards,
Huang, Ying

Quoted 24 lines
---
 mm/migrate.c | 16 ++++------------
 1 file changed, 4 insertions(+), 12 deletions(-)
mm/migrate.c +4 -0
index a72994c68ec6..dc7cc7f3a124 100644
@@ -2067,6 +2067,10 @@ static int numamigrate_isolate_page(pg_data_t *pgdat, struct page *page)
 
 	VM_BUG_ON_PAGE(compound_order(page) && !PageTransHuge(page), page);
 
+	/* Do not migrate THP mapped by multiple processes */
+	if (PageTransHuge(page) && page_mapcount(page) > 1)
+		return 0;
+
 	/* Avoid migrating to a node that is nearly full */
 	if (!migrate_balanced_pgdat(pgdat, compound_nr(page)))
 		return 0;
@@ -2074,18 +2078,6 @@ static int numamigrate_isolate_page(pg_data_t *pgdat, struct page *page)
 	if (isolate_lru_page(page))
 		return 0;
 
Quoted 15 lines
Quoted 5 lines
mm/migrate.c +0 -5
-	/*
-	 * migrate_misplaced_transhuge_page() skips page migration's usual
-	 * check on page_count(), so we must do it here, now that the page
-	 * has been isolated: a GUP pin, or any other pin, prevents migration.
-	 * The expected page count is 3: 1 for page's mapcount and 1 for the

mm/migrate.c static int numamigrate_isolate_page(pg_data_t *pgdat, struct page *page) · line 2083

Quoted 10 lines
mm/migrate.c +0 -7
-	 * caller's pin and 1 for the reference taken by isolate_lru_page().
-	 */
-	if (PageTransHuge(page) && page_count(page) != 3) {
-		putback_lru_page(page);
-		return 0;
-	}
-
 	page_lru = page_is_file_lru(page);
 	mod_node_page_state(page_pgdat(page), NR_ISOLATED_ANON + page_lru,
 				thp_nr_pages(page));

—
Best Regards,
Yan Zi

#20 lore raw

Re: [v2 PATCH 6/7] mm: migrate: check mapcount for THP instead of ref count

Quoted 8 lines · Zi Yan
Quoted 6 lines · Ying
Quoted 3 lines · Yang Shi

The generic migration path will check refcount, so no need check refcount here. But the old code actually prevents from migrating shared THP (mapped by multiple processes), so bail out early if mapcount is > 1 to keep the behavior.

Quoted 8 lines
Quoted 2 lines

What prevents us from migrating shared THP? If no, why not just remove the old refcount checking?

If two or more processes are in different NUMA nodes, a THP shared by them can be migrated back and forth between NUMA nodes, which is quite costly. Unless we have a better way of figuring out a good location for such pages to reduce the number of migration, it might be better not to move them, right?

Some mechanism has been provided in should_numa_migrate_memory() to identify the shared pages from the private pages. Do you find it doesn't work well in some situations?

The multiple threads in one process which run on different NUMA nodes may share pages too. So it isn't a good solution to exclude pages shared by multiple processes.

Best Regards,
Huang, Ying

reviewer
#21 lore raw

Re: [v2 PATCH 6/7] mm: migrate: check mapcount for THP instead of ref count

Quoted 19 lines · Ying
Quoted 8 lines · Zi Yan
Quoted 6 lines · Ying
Quoted 3 lines · Yang Shi

The generic migration path will check refcount, so no need check refcount here. But the old code actually prevents from migrating shared THP (mapped by multiple processes), so bail out early if mapcount is > 1 to keep the behavior.

Quoted 8 lines
Quoted 2 lines

What prevents us from migrating shared THP? If no, why not just remove the old refcount checking?

If two or more processes are in different NUMA nodes, a THP shared by them can be migrated back and forth between NUMA nodes, which is quite costly. Unless we have a better way of figuring out a good location for such pages to reduce the number of migration, it might be better not to move them, right?

Quoted 7 lines

Some mechanism has been provided in should_numa_migrate_memory() to identify the shared pages from the private pages. Do you find it doesn't work well in some situations?

The multiple threads in one process which run on different NUMA nodes may share pages too. So it isn't a good solution to exclude pages shared by multiple processes.

After recheck the patch, it seems that no shared THP migration here is a side effect of the original page_count check, which might not be intended and be worth fixing. But Yang just want to solve one problem, simplifying THP NUMA migration,
at a time. Maybe a separate patch would be better for both discussing and fixing this problem.

—
Best Regards,
Yan Zi

author
#22 lore raw

Re: [v2 PATCH 6/7] mm: migrate: check mapcount for THP instead of ref count

In reply to Huang, Ying
Quoted 9 lines · Ying
Quoted 3 lines · Yang Shi

The generic migration path will check refcount, so no need check refcount here. But the old code actually prevents from migrating shared THP (mapped by multiple processes), so bail out early if mapcount is > 1 to keep the behavior.

What prevents us from migrating shared THP? If no, why not just remove the old refcount checking?

We could migrate shared THP if we don't care about the bounce back and forth between nodes as Zi Yan described. The other reason is, as I mentioned in the cover letter, I'd like to keep the behavior as consistent as possible between before and after for now. The old behavior does prevent migrating shared THP, so I did so in this series. We definitely could optimize the behavior later on.

Quoted 43 lines

Best Regards,
Huang, Ying

Quoted 39 lines
---
 mm/migrate.c | 16 ++++------------
 1 file changed, 4 insertions(+), 12 deletions(-)
mm/migrate.c +4 -12
index a72994c68ec6..dc7cc7f3a124 100644
@@ -2067,6 +2067,10 @@ static int numamigrate_isolate_page(pg_data_t *pgdat, struct page *page)
 
      VM_BUG_ON_PAGE(compound_order(page) && !PageTransHuge(page), page);
 
+     /* Do not migrate THP mapped by multiple processes */
+     if (PageTransHuge(page) && page_mapcount(page) > 1)
+             return 0;
+
      /* Avoid migrating to a node that is nearly full */
      if (!migrate_balanced_pgdat(pgdat, compound_nr(page)))
              return 0;
@@ -2074,18 +2078,6 @@ static int numamigrate_isolate_page(pg_data_t *pgdat, struct page *page)
      if (isolate_lru_page(page))
              return 0;
 
-     /*
-      * migrate_misplaced_transhuge_page() skips page migration's usual
-      * check on page_count(), so we must do it here, now that the page
-      * has been isolated: a GUP pin, or any other pin, prevents migration.
-      * The expected page count is 3: 1 for page's mapcount and 1 for the
-      * caller's pin and 1 for the reference taken by isolate_lru_page().
-      */
-     if (PageTransHuge(page) && page_count(page) != 3) {
-             putback_lru_page(page);
-             return 0;
-     }
-
      page_lru = page_is_file_lru(page);
      mod_node_page_state(page_pgdat(page), NR_ISOLATED_ANON + page_lru,
                              thp_nr_pages(page));
author
#23 lore raw

PATCH v2 7/7 mm: thp: skip make PMD PROT_NONE if THP migration is not supported

In reply to Yang Shi

A quick grep shows x86_64, PowerPC (book3s), ARM64 and S390 support both NUMA balancing and THP. But S390 doesn't support THP migration so NUMA balancing actually can't migrate any misplaced pages.

Skip make PMD PROT_NONE for such case otherwise CPU cycles may be wasted by pointless NUMA hinting faults on S390.

---
 mm/huge_memory.c | 4 ++++
 1 file changed, 4 insertions(+)
mm/huge_memory.c +4 -0
index 94981907fd4c..f63445f3a17d 100644
@@ -1741,6 +1741,7 @@ bool move_huge_pmd(struct vm_area_struct *vma, unsigned long old_addr,
  * Returns
  *  - 0 if PMD could not be locked
  *  - 1 if PMD was locked but protections unchanged and TLB flush unnecessary
+ *      or if prot_numa but THP migration is not supported
  *  - HPAGE_PMD_NR if protections changed and TLB flush necessary
  */
 int change_huge_pmd(struct vm_area_struct *vma, pmd_t *pmd,
@@ -1755,6 +1756,9 @@ int change_huge_pmd(struct vm_area_struct *vma, pmd_t *pmd,
 	bool uffd_wp = cp_flags & MM_CP_UFFD_WP;
 	bool uffd_wp_resolve = cp_flags & MM_CP_UFFD_WP_RESOLVE;
 
+	if (prot_numa && !thp_migration_supported())
+		return 1;
+
 	ptl = __pmd_trans_huge_lock(pmd, vma);
 	if (!ptl)
 		return 0;
Signature · 1 line
2.26.2
Acked-by on v2 7/7
#24 lore raw

Re: [v2 PATCH 7/7] mm: thp: skip make PMD PROT_NONE if THP migration is not supported

Quoted 8 lines · Yang Shi

A quick grep shows x86_64, PowerPC (book3s), ARM64 and S390 support both NUMA balancing and THP. But S390 doesn't support THP migration so NUMA balancing actually can't migrate any misplaced pages.

Skip make PMD PROT_NONE for such case otherwise CPU cycles may be wasted by pointless NUMA hinting faults on S390.

Signature · 2 lines
Mel Gorman
SUSE Labs
author
#25 lore raw
In reply to Yang Shi

Gently ping.

I also did some tests to measure the latency of do_huge_pmd_numa_page. The test VM has 80 vcpus and 64G memory. The test would create 2 processes to consume 128G memory together which would incur memory pressure to cause THP splits. And it also creates 80 processes to hog cpu, and the memory consumer processes are bound to different nodes periodically in order to increase NUMA faults.

The below test script is used:

echo 3 > /proc/sys/vm/drop_caches

# Run stress-ng for 24 hours
./stress-ng/stress-ng --vm 2 --vm-bytes 64G --timeout 24h & PID=$!

./stress-ng/stress-ng --cpu $NR_CPUS --timeout 24h &

# Wait for vm stressors forked
sleep 5

PID_1=pgrep -P $PID | awk 'NR == 1'
PID_2=pgrep -P $PID | awk 'NR == 2'

JOB1=pgrep -P $PID_1
JOB2=pgrep -P $PID_2

# Bind load jobs to different nodes periodically to force generate

# cross node memory access
while [ -d "/proc/$PID" ]
do
        taskset -apc 8 $JOB1
        taskset -apc 8 $JOB2
        sleep 300
        taskset -apc 58 $JOB1
        taskset -apc 58 $JOB2
        sleep 300
done

With the above test the histogram of latency of do_huge_pmd_numa_page is as shown below. Since the number of do_huge_pmd_numa_page varies drastically for each run (should be due to scheduler), so I converted the raw number to percentage.

patched base
@us[stress-ng]:
[0] 3.57% 0.16%
[1] 55.68% 18.36%
[2, 4) 10.46% 40.44%
[4, 8) 7.26% 17.82%
[8, 16) 21.12% 13.41%

[16, 32)                         1.06%                           4.27%
[32, 64)                         0.56%                           4.07%
[64, 128)                       0.16%                            0.35%
[128, 256)                     < 0.1%                          < 0.1%
[256, 512)                     < 0.1%                          < 0.1%
[512, 1K)                       < 0.1%                          < 0.1%
[1K, 2K)                         < 0.1%                          < 0.1%
[2K, 4K)                         < 0.1%                          < 0.1%
[4K, 8K)                         < 0.1%                          < 0.1%
[8K, 16K)                       < 0.1%                          < 0.1%
[16K, 32K)                     < 0.1%                          < 0.1%
[32K, 64K)                     < 0.1%                          < 0.1%

Per the result, patched kernel is even slightly better than the base kernel. I think this is because the lock contention against THP split is less than base kernel due to the refactor.

To exclude the affect from THP split, I also did test w/o memory pressure. No obvious regression is spotted. The below is the test result *w/o* memory pressure.
patched
base
@us[stress-ng]:
[0] 7.97%

        18.4%
[1]                                      69.63%
        58.24%
[2, 4)                                  4.18%
        2.63%
[4, 8)                                  0.22%
        0.17%
[8, 16)                                1.03%
       0.92%
[16, 32)                              0.14%
      < 0.1%
[32, 64)                              < 0.1%
      < 0.1%
[64, 128)                            < 0.1%
     < 0.1%
[128, 256)                          < 0.1%
     < 0.1%
[256, 512)                           0.45%
     1.19%
[512, 1K)                            15.45%
     17.27%
[1K, 2K)                               < 0.1%
       < 0.1%
[2K, 4K)                              < 0.1%
       < 0.1%
[4K, 8K)                              < 0.1%
       < 0.1%
[8K, 16K)                             0.86%
      0.88%
[16K, 32K)                           < 0.1%
     0.15%
[32K, 64K)                           < 0.1%
     < 0.1%
[64K, 128K)                         < 0.1%
    < 0.1%

[128K, 256K) < 0.1% < 0.1%

Quoted 65 lines · Yang Shi

Changelog:
v1 --> v2:
* Adopted the suggestion from Gerald Schaefer to skip huge PMD for S390 for now.
* Used PageTransHuge to distinguish base page or THP instead of a new parameter for migrate_misplaced_page() per Huang Ying.
* Restored PMD lazily to avoid unnecessary TLB shootdown per Huang Ying.
* Skipped shared THP.
* Updated counters correctly.
* Rebased to linux-next (next-20210412).

When the THP NUMA fault support was added THP migration was not supported yet. So the ad hoc THP migration was implemented in NUMA fault handling. Since v4.14 THP migration has been supported so it doesn't make too much sense to still keep another THP migration implementation rather than using the generic migration code. It is definitely a maintenance burden to keep two THP migration implementation for different code paths and it is more error prone. Using the generic THP migration implementation allows us remove the duplicate code and some hacks needed by the old ad hoc implementation.

A quick grep shows x86_64, PowerPC (book3s), ARM64 ans S390 support both THP and NUMA balancing. The most of them support THP migration except for S390. Zi Yan tried to add THP migration support for S390 before but it was not accepted due to the design of S390 PMD. For the discussion, please see:
https://lkml.org/lkml/2018/4/27/953.

Per the discussion with Gerald Schaefer in v1 it is acceptible to skip huge PMD for S390 for now.

I saw there were some hacks about gup from git history, but I didn't figure out if they have been removed or not since I just found FOLL_NUMA code in the current gup implementation and they seems useful.

I'm trying to keep the behavior as consistent as possible between before and after. But there is still some minor disparity. For example, file THP won't get migrated at all in old implementation due to the anon_vma check, but the new implementation doesn't need acquire anon_vma lock anymore, so file THP might get migrated. Not sure if this behavior needs to be kept.

Patch #1 ~ #2 are preparation patches.
Patch #3 is the real meat.
Patch #4 ~ #6 keep consistent counters and behaviors with before. Patch #7 skips change huge PMD to prot_none if thp migration is not supported.

Yang Shi (7):
mm: memory: add orig_pmd to struct vm_fault
mm: memory: make numa_migrate_prep() non-static
mm: thp: refactor NUMA fault handling
mm: migrate: account THP NUMA migration counters correctly
mm: migrate: don't split THP for misplaced NUMA page
mm: migrate: check mapcount for THP instead of ref count
mm: thp: skip make PMD PROT_NONE if THP migration is not supported

 include/linux/huge_mm.h |   9 ++---
 include/linux/migrate.h |  23 -----------
 include/linux/mm.h      |   3 ++
 mm/huge_memory.c        | 156 +++++++++++++++++++++++++-----------------------------------------------
 mm/internal.h           |  21 ++--------
 mm/memory.c             |  31 +++++++--------
 mm/migrate.c            | 204 +++++++++++++++++++++--------------------------------------------------------------------------
 7 files changed, 123 insertions(+), 324 deletions(-)