Skip to contentSkip to conversation

mm: split PTE/PMD PT table Kconfig cleanups+clarifications

PATCH Applied David Hildenbrand 4 participants 8 messages Jul 26, 2024 – Jul 29, 2024

Subsystems (10)
lore
Share & tools
lore
#0 lore raw

cover mm: split PTE/PMD PT table Kconfig cleanups+clarifications

This series is a follow up to the fixes:
"[PATCH v1 0/2] mm/hugetlb: fix hugetlb vs. core-mm PT locking"

When working on the fixes, I wondered why 8xx is fine (-> never uses split PT locks) and how PT locking even works properly with PMD page table sharing (-> always requires split PMD PT locks).

Let's improve the split PT lock detection, make hugetlb properly depend on it and make 8xx bail out if it would ever get enabled by accident.

As an alternative to patch #3 we could extend the Kconfig SPLIT_PTE_PTLOCKS option from patch #2 -- but enforcing it closer to the code that actually implements it feels a bit nicer for documentation purposes, and there is no need to actually disable it because it should always be disabled (!SMP).

Did a bunch of cross-compilations to make sure that split PTE/PMD PT locks are still getting used where we would expect them.

[1] https://lkml.kernel.org/r/[email protected]

Cc Andrew Morton <[email protected]>
Cc Oscar Salvador <[email protected]>
Cc Peter Xu <[email protected]>
Cc Muchun Song <[email protected]>
Cc Russell King <[email protected]>
Cc Michael Ellerman <[email protected]>
Cc Nicholas Piggin <[email protected]>
Cc Christophe Leroy <[email protected]>
Cc "Naveen N. Rao" <[email protected]>
Cc Juergen Gross <[email protected]>
Cc Boris Ostrovsky <[email protected]>
Cc Thomas Gleixner <[email protected]>
Cc Ingo Molnar <[email protected]>
Cc Borislav Petkov <[email protected]>
Cc Dave Hansen <[email protected]>
Cc "H. Peter Anvin" <[email protected]>
Cc Alexander Viro <[email protected]>
Cc Christian Brauner <[email protected]>

David Hildenbrand (3):
mm: turn USE_SPLIT_PTE_PTLOCKS / USE_SPLIT_PTE_PTLOCKS into Kconfig
options
mm/hugetlb: enforce that PMD PT sharing has split PMD PT locks powerpc/8xx: document and enforce that split PT locks are not used

 arch/arm/mm/fault-armv.c      |  6 +++---
 arch/powerpc/mm/pgtable.c     |  6 ++++++
 arch/x86/xen/mmu_pv.c         |  7 ++++---
 fs/Kconfig                    |  4 ++++
 include/linux/hugetlb.h       |  5 ++---
 include/linux/mm.h            |  8 ++++----
 include/linux/mm_types.h      |  2 +-
 include/linux/mm_types_task.h |  3 ---
 kernel/fork.c                 |  4 ++--
 mm/Kconfig                    | 18 +++++++++++-------
 mm/hugetlb.c                  |  8 ++++----
 mm/memory.c                   |  2 +-
 12 files changed, 42 insertions(+), 31 deletions(-)
Signature · 1 line
2.45.2
#1 lore raw

PATCH 1/3 mm: turn USE_SPLIT_PTE_PTLOCKS / USE_SPLIT_PTE_PTLOCKS into Kconfig options

Let's clean that up a bit and prepare for depending on CONFIG_SPLIT_PMD_PTLOCKS in other Kconfig options.

More cleanups would be reasonable (like the arch-specific "depends on" for CONFIG_SPLIT_PTE_PTLOCKS), but we'll leave that for another day.

---
 arch/arm/mm/fault-armv.c      |  6 +++---
 arch/x86/xen/mmu_pv.c         |  7 ++++---
 include/linux/mm.h            |  8 ++++----
 include/linux/mm_types.h      |  2 +-
 include/linux/mm_types_task.h |  3 ---
 kernel/fork.c                 |  4 ++--
 mm/Kconfig                    | 18 +++++++++++-------
 mm/memory.c                   |  2 +-
 8 files changed, 26 insertions(+), 24 deletions(-)
arch/arm/mm/fault-armv.c +3 -3
index 2286c2ea60ec4..831793cd6ff94 100644
@@ -61,7 +61,7 @@ static int do_adjust_pte(struct vm_area_struct *vma, unsigned long address,
 	return ret;
 }
 
-#if USE_SPLIT_PTE_PTLOCKS
+#if defined(CONFIG_SPLIT_PTE_PTLOCKS)
 /*
  * If we are using split PTE locks, then we need to take the page
  * lock here.  Otherwise we are using shared mm->page_table_lock
@@ -80,10 +80,10 @@ static inline void do_pte_unlock(spinlock_t *ptl)
 {
 	spin_unlock(ptl);
 }
-#else /* !USE_SPLIT_PTE_PTLOCKS */
+#else /* !defined(CONFIG_SPLIT_PTE_PTLOCKS) */
 static inline void do_pte_lock(spinlock_t *ptl) {}
 static inline void do_pte_unlock(spinlock_t *ptl) {}
-#endif /* USE_SPLIT_PTE_PTLOCKS */
+#endif /* defined(CONFIG_SPLIT_PTE_PTLOCKS) */
 
 static int adjust_pte(struct vm_area_struct *vma, unsigned long address,
 	unsigned long pfn)
arch/x86/xen/mmu_pv.c +4 -3
index f1ce39d6d32cb..f4a316894bbb4 100644
@@ -665,7 +665,7 @@ static spinlock_t *xen_pte_lock(struct page *page, struct mm_struct *mm)
 {
 	spinlock_t *ptl = NULL;
 
-#if USE_SPLIT_PTE_PTLOCKS
+#if defined(CONFIG_SPLIT_PTE_PTLOCKS)
 	ptl = ptlock_ptr(page_ptdesc(page));
 	spin_lock_nest_lock(ptl, &mm->page_table_lock);
 #endif
@@ -1553,7 +1553,8 @@ static inline void xen_alloc_ptpage(struct mm_struct *mm, unsigned long pfn,
 
 		__set_pfn_prot(pfn, PAGE_KERNEL_RO);
 
-		if (level == PT_PTE && USE_SPLIT_PTE_PTLOCKS && !pinned)
+		if (level == PT_PTE && IS_ENABLED(CONFIG_SPLIT_PTE_PTLOCKS) &&
+		    !pinned)
 			__pin_pagetable_pfn(MMUEXT_PIN_L1_TABLE, pfn);
 
 		xen_mc_issue(XEN_LAZY_MMU);
@@ -1581,7 +1582,7 @@ static inline void xen_release_ptpage(unsigned long pfn, unsigned level)
 	if (pinned) {
 		xen_mc_batch();
 
-		if (level == PT_PTE && USE_SPLIT_PTE_PTLOCKS)
+		if (level == PT_PTE && IS_ENABLED(CONFIG_SPLIT_PTE_PTLOCKS))
 			__pin_pagetable_pfn(MMUEXT_UNPIN_TABLE, pfn);
 
 		__set_pfn_prot(pfn, PAGE_KERNEL);
include/linux/mm.h +4 -4
index 0472a5090b180..dff43101572ec 100644
@@ -2843,7 +2843,7 @@ static inline void pagetable_free(struct ptdesc *pt)
 	__free_pages(page, compound_order(page));
 }
 
-#if USE_SPLIT_PTE_PTLOCKS
+#if defined(CONFIG_SPLIT_PTE_PTLOCKS)
 #if ALLOC_SPLIT_PTLOCKS
 void __init ptlock_cache_init(void);
 bool ptlock_alloc(struct ptdesc *ptdesc);
@@ -2895,7 +2895,7 @@ static inline bool ptlock_init(struct ptdesc *ptdesc)
 	return true;
 }
 
-#else	/* !USE_SPLIT_PTE_PTLOCKS */
+#else	/* !defined(CONFIG_SPLIT_PTE_PTLOCKS) */
 /*
  * We use mm->page_table_lock to guard all pagetable pages of the mm.
  */
@@ -2906,7 +2906,7 @@ static inline spinlock_t *pte_lockptr(struct mm_struct *mm, pte_t *pte)
 static inline void ptlock_cache_init(void) {}
 static inline bool ptlock_init(struct ptdesc *ptdesc) { return true; }
 static inline void ptlock_free(struct ptdesc *ptdesc) {}
-#endif /* USE_SPLIT_PTE_PTLOCKS */
+#endif /* defined(CONFIG_SPLIT_PTE_PTLOCKS) */
 
 static inline bool pagetable_pte_ctor(struct ptdesc *ptdesc)
 {
@@ -2966,7 +2966,7 @@ pte_t *pte_offset_map_nolock(struct mm_struct *mm, pmd_t *pmd,
 	((unlikely(pmd_none(*(pmd))) && __pte_alloc_kernel(pmd))? \
 		NULL: pte_offset_kernel(pmd, address))
 
-#if USE_SPLIT_PMD_PTLOCKS
+#if defined(CONFIG_SPLIT_PMD_PTLOCKS)
 
 static inline struct page *pmd_pgtable_page(pmd_t *pmd)
 {
include/linux/mm_types.h +1 -1
index 4854249792545..165c58b12ccc9 100644
@@ -947,7 +947,7 @@ struct mm_struct {
 #ifdef CONFIG_MMU_NOTIFIER
 		struct mmu_notifier_subscriptions *notifier_subscriptions;
 #endif
-#if defined(CONFIG_TRANSPARENT_HUGEPAGE) && !USE_SPLIT_PMD_PTLOCKS
+#if defined(CONFIG_TRANSPARENT_HUGEPAGE) && !defined(CONFIG_SPLIT_PMD_PTLOCKS)
 		pgtable_t pmd_huge_pte; /* protected by page_table_lock */
 #endif
 #ifdef CONFIG_NUMA_BALANCING
include/linux/mm_types_task.h +0 -3
index a2f6179b672b8..bff5706b76e14 100644
@@ -16,9 +16,6 @@
 #include <asm/tlbbatch.h>
 #endif
 
-#define USE_SPLIT_PTE_PTLOCKS	(NR_CPUS >= CONFIG_SPLIT_PTLOCK_CPUS)
-#define USE_SPLIT_PMD_PTLOCKS	(USE_SPLIT_PTE_PTLOCKS && \
-		IS_ENABLED(CONFIG_ARCH_ENABLE_SPLIT_PMD_PTLOCK))
 #define ALLOC_SPLIT_PTLOCKS	(SPINLOCK_SIZE > BITS_PER_LONG/8)
 
 /*
kernel/fork.c +2 -2
index a8362c26ebcb0..216ce9ba4f4e6 100644
@@ -832,7 +832,7 @@ static void check_mm(struct mm_struct *mm)
 		pr_alert("BUG: non-zero pgtables_bytes on freeing mm: %ld\n",
 				mm_pgtables_bytes(mm));
 
-#if defined(CONFIG_TRANSPARENT_HUGEPAGE) && !USE_SPLIT_PMD_PTLOCKS
+#if defined(CONFIG_TRANSPARENT_HUGEPAGE) && !defined(CONFIG_SPLIT_PMD_PTLOCKS)
 	VM_BUG_ON_MM(mm->pmd_huge_pte, mm);
 #endif
 }
@@ -1276,7 +1276,7 @@ static struct mm_struct *mm_init(struct mm_struct *mm, struct task_struct *p,
 	RCU_INIT_POINTER(mm->exe_file, NULL);
 	mmu_notifier_subscriptions_init(mm);
 	init_tlb_flush_pending(mm);
-#if defined(CONFIG_TRANSPARENT_HUGEPAGE) && !USE_SPLIT_PMD_PTLOCKS
+#if defined(CONFIG_TRANSPARENT_HUGEPAGE) && !defined(CONFIG_SPLIT_PMD_PTLOCKS)
 	mm->pmd_huge_pte = NULL;
 #endif
 	mm_init_uprobes_state(mm);
mm/Kconfig +11 -7
index b72e7d040f789..7b716ac802726 100644
@@ -585,17 +585,21 @@ config ARCH_MHP_MEMMAP_ON_MEMORY_ENABLE
 # at the same time (e.g. copy_page_range()).
 # DEBUG_SPINLOCK and DEBUG_LOCK_ALLOC spinlock_t also enlarge struct page.
 #
-config SPLIT_PTLOCK_CPUS
-	int
-	default "999999" if !MMU
-	default "999999" if ARM && !CPU_CACHE_VIPT
-	default "999999" if PARISC && !PA20
-	default "999999" if SPARC32
-	default "4"
+config SPLIT_PTE_PTLOCKS
+	def_bool y
+	depends on MMU
+	depends on NR_CPUS >= 4
+	depends on !ARM || CPU_CACHE_VIPT
+	depends on !PARISC || PA20
+	depends on !SPARC32
 
 config ARCH_ENABLE_SPLIT_PMD_PTLOCK
 	bool
 
+config SPLIT_PMD_PTLOCKS
+	def_bool y
+	depends on SPLIT_PTE_PTLOCKS && ARCH_ENABLE_SPLIT_PMD_PTLOCK
+
 #
 # support for memory balloon
 config MEMORY_BALLOON
mm/memory.c +1 -1
index 833d2cad6eb29..714589582fe15 100644
@@ -6559,7 +6559,7 @@ long copy_folio_from_user(struct folio *dst_folio,
 }
 #endif /* CONFIG_TRANSPARENT_HUGEPAGE || CONFIG_HUGETLBFS */
 
-#if USE_SPLIT_PTE_PTLOCKS && ALLOC_SPLIT_PTLOCKS
+#if defined(CONFIG_SPLIT_PTE_PTLOCKS) && ALLOC_SPLIT_PTLOCKS
 
 static struct kmem_cache *page_ptl_cachep;
 
Signature · 1 line
2.45.2
reviewer Acked-by on v1 1/3
#2 lore raw

Re: [PATCH v1 1/3] mm: turn USE_SPLIT_PTE_PTLOCKS / USE_SPLIT_PTE_PTLOCKS into Kconfig options

Quoted 7 lines · David Hildenbrand

Let's clean that up a bit and prepare for depending on CONFIG_SPLIT_PMD_PTLOCKS in other Kconfig options.

More cleanups would be reasonable (like the arch-specific "depends on" for CONFIG_SPLIT_PTE_PTLOCKS), but we'll leave that for another day.

Quoted 208 lines
---
 arch/arm/mm/fault-armv.c      |  6 +++---
 arch/x86/xen/mmu_pv.c         |  7 ++++---
 include/linux/mm.h            |  8 ++++----
 include/linux/mm_types.h      |  2 +-
 include/linux/mm_types_task.h |  3 ---
 kernel/fork.c                 |  4 ++--
 mm/Kconfig                    | 18 +++++++++++-------
 mm/memory.c                   |  2 +-
 8 files changed, 26 insertions(+), 24 deletions(-)
arch/arm/mm/fault-armv.c +3 -3
index 2286c2ea60ec4..831793cd6ff94 100644
@@ -61,7 +61,7 @@ static int do_adjust_pte(struct vm_area_struct *vma, unsigned long address,
 	return ret;
 }
 
-#if USE_SPLIT_PTE_PTLOCKS
+#if defined(CONFIG_SPLIT_PTE_PTLOCKS)
 /*
  * If we are using split PTE locks, then we need to take the page
  * lock here.  Otherwise we are using shared mm->page_table_lock
@@ -80,10 +80,10 @@ static inline void do_pte_unlock(spinlock_t *ptl)
 {
 	spin_unlock(ptl);
 }
-#else /* !USE_SPLIT_PTE_PTLOCKS */
+#else /* !defined(CONFIG_SPLIT_PTE_PTLOCKS) */
 static inline void do_pte_lock(spinlock_t *ptl) {}
 static inline void do_pte_unlock(spinlock_t *ptl) {}
-#endif /* USE_SPLIT_PTE_PTLOCKS */
+#endif /* defined(CONFIG_SPLIT_PTE_PTLOCKS) */
 
 static int adjust_pte(struct vm_area_struct *vma, unsigned long address,
 	unsigned long pfn)
arch/x86/xen/mmu_pv.c +4 -3
index f1ce39d6d32cb..f4a316894bbb4 100644
@@ -665,7 +665,7 @@ static spinlock_t *xen_pte_lock(struct page *page, struct mm_struct *mm)
 {
 	spinlock_t *ptl = NULL;
 
-#if USE_SPLIT_PTE_PTLOCKS
+#if defined(CONFIG_SPLIT_PTE_PTLOCKS)
 	ptl = ptlock_ptr(page_ptdesc(page));
 	spin_lock_nest_lock(ptl, &mm->page_table_lock);
 #endif
@@ -1553,7 +1553,8 @@ static inline void xen_alloc_ptpage(struct mm_struct *mm, unsigned long pfn,
 
 		__set_pfn_prot(pfn, PAGE_KERNEL_RO);
 
-		if (level == PT_PTE && USE_SPLIT_PTE_PTLOCKS && !pinned)
+		if (level == PT_PTE && IS_ENABLED(CONFIG_SPLIT_PTE_PTLOCKS) &&
+		    !pinned)
 			__pin_pagetable_pfn(MMUEXT_PIN_L1_TABLE, pfn);
 
 		xen_mc_issue(XEN_LAZY_MMU);
@@ -1581,7 +1582,7 @@ static inline void xen_release_ptpage(unsigned long pfn, unsigned level)
 	if (pinned) {
 		xen_mc_batch();
 
-		if (level == PT_PTE && USE_SPLIT_PTE_PTLOCKS)
+		if (level == PT_PTE && IS_ENABLED(CONFIG_SPLIT_PTE_PTLOCKS))
 			__pin_pagetable_pfn(MMUEXT_UNPIN_TABLE, pfn);
 
 		__set_pfn_prot(pfn, PAGE_KERNEL);
include/linux/mm.h +4 -4
index 0472a5090b180..dff43101572ec 100644
@@ -2843,7 +2843,7 @@ static inline void pagetable_free(struct ptdesc *pt)
 	__free_pages(page, compound_order(page));
 }
 
-#if USE_SPLIT_PTE_PTLOCKS
+#if defined(CONFIG_SPLIT_PTE_PTLOCKS)
 #if ALLOC_SPLIT_PTLOCKS
 void __init ptlock_cache_init(void);
 bool ptlock_alloc(struct ptdesc *ptdesc);
@@ -2895,7 +2895,7 @@ static inline bool ptlock_init(struct ptdesc *ptdesc)
 	return true;
 }
 
-#else	/* !USE_SPLIT_PTE_PTLOCKS */
+#else	/* !defined(CONFIG_SPLIT_PTE_PTLOCKS) */
 /*
  * We use mm->page_table_lock to guard all pagetable pages of the mm.
  */
@@ -2906,7 +2906,7 @@ static inline spinlock_t *pte_lockptr(struct mm_struct *mm, pte_t *pte)
 static inline void ptlock_cache_init(void) {}
 static inline bool ptlock_init(struct ptdesc *ptdesc) { return true; }
 static inline void ptlock_free(struct ptdesc *ptdesc) {}
-#endif /* USE_SPLIT_PTE_PTLOCKS */
+#endif /* defined(CONFIG_SPLIT_PTE_PTLOCKS) */
 
 static inline bool pagetable_pte_ctor(struct ptdesc *ptdesc)
 {
@@ -2966,7 +2966,7 @@ pte_t *pte_offset_map_nolock(struct mm_struct *mm, pmd_t *pmd,
 	((unlikely(pmd_none(*(pmd))) && __pte_alloc_kernel(pmd))? \
 		NULL: pte_offset_kernel(pmd, address))
 
-#if USE_SPLIT_PMD_PTLOCKS
+#if defined(CONFIG_SPLIT_PMD_PTLOCKS)
 
 static inline struct page *pmd_pgtable_page(pmd_t *pmd)
 {
include/linux/mm_types.h +1 -1
index 4854249792545..165c58b12ccc9 100644
@@ -947,7 +947,7 @@ struct mm_struct {
 #ifdef CONFIG_MMU_NOTIFIER
 		struct mmu_notifier_subscriptions *notifier_subscriptions;
 #endif
-#if defined(CONFIG_TRANSPARENT_HUGEPAGE) && !USE_SPLIT_PMD_PTLOCKS
+#if defined(CONFIG_TRANSPARENT_HUGEPAGE) && !defined(CONFIG_SPLIT_PMD_PTLOCKS)
 		pgtable_t pmd_huge_pte; /* protected by page_table_lock */
 #endif
 #ifdef CONFIG_NUMA_BALANCING
include/linux/mm_types_task.h +0 -3
index a2f6179b672b8..bff5706b76e14 100644
@@ -16,9 +16,6 @@
 #include <asm/tlbbatch.h>
 #endif
 
-#define USE_SPLIT_PTE_PTLOCKS	(NR_CPUS >= CONFIG_SPLIT_PTLOCK_CPUS)
-#define USE_SPLIT_PMD_PTLOCKS	(USE_SPLIT_PTE_PTLOCKS && \
-		IS_ENABLED(CONFIG_ARCH_ENABLE_SPLIT_PMD_PTLOCK))
 #define ALLOC_SPLIT_PTLOCKS	(SPINLOCK_SIZE > BITS_PER_LONG/8)
 
 /*
kernel/fork.c +2 -2
index a8362c26ebcb0..216ce9ba4f4e6 100644
@@ -832,7 +832,7 @@ static void check_mm(struct mm_struct *mm)
 		pr_alert("BUG: non-zero pgtables_bytes on freeing mm: %ld\n",
 				mm_pgtables_bytes(mm));
 
-#if defined(CONFIG_TRANSPARENT_HUGEPAGE) && !USE_SPLIT_PMD_PTLOCKS
+#if defined(CONFIG_TRANSPARENT_HUGEPAGE) && !defined(CONFIG_SPLIT_PMD_PTLOCKS)
 	VM_BUG_ON_MM(mm->pmd_huge_pte, mm);
 #endif
 }
@@ -1276,7 +1276,7 @@ static struct mm_struct *mm_init(struct mm_struct *mm, struct task_struct *p,
 	RCU_INIT_POINTER(mm->exe_file, NULL);
 	mmu_notifier_subscriptions_init(mm);
 	init_tlb_flush_pending(mm);
-#if defined(CONFIG_TRANSPARENT_HUGEPAGE) && !USE_SPLIT_PMD_PTLOCKS
+#if defined(CONFIG_TRANSPARENT_HUGEPAGE) && !defined(CONFIG_SPLIT_PMD_PTLOCKS)
 	mm->pmd_huge_pte = NULL;
 #endif
 	mm_init_uprobes_state(mm);
mm/Kconfig +11 -7
index b72e7d040f789..7b716ac802726 100644
@@ -585,17 +585,21 @@ config ARCH_MHP_MEMMAP_ON_MEMORY_ENABLE
 # at the same time (e.g. copy_page_range()).
 # DEBUG_SPINLOCK and DEBUG_LOCK_ALLOC spinlock_t also enlarge struct page.
 #
-config SPLIT_PTLOCK_CPUS
-	int
-	default "999999" if !MMU
-	default "999999" if ARM && !CPU_CACHE_VIPT
-	default "999999" if PARISC && !PA20
-	default "999999" if SPARC32
-	default "4"
+config SPLIT_PTE_PTLOCKS
+	def_bool y
+	depends on MMU
+	depends on NR_CPUS >= 4
+	depends on !ARM || CPU_CACHE_VIPT
+	depends on !PARISC || PA20
+	depends on !SPARC32
 
 config ARCH_ENABLE_SPLIT_PMD_PTLOCK
 	bool
 
+config SPLIT_PMD_PTLOCKS
+	def_bool y
+	depends on SPLIT_PTE_PTLOCKS && ARCH_ENABLE_SPLIT_PMD_PTLOCK
+
 #
 # support for memory balloon
 config MEMORY_BALLOON
mm/memory.c +1 -1
index 833d2cad6eb29..714589582fe15 100644
@@ -6559,7 +6559,7 @@ long copy_folio_from_user(struct folio *dst_folio,
 }
 #endif /* CONFIG_TRANSPARENT_HUGEPAGE || CONFIG_HUGETLBFS */
 
-#if USE_SPLIT_PTE_PTLOCKS && ALLOC_SPLIT_PTLOCKS
+#if defined(CONFIG_SPLIT_PTE_PTLOCKS) && ALLOC_SPLIT_PTLOCKS
 
 static struct kmem_cache *page_ptl_cachep;
 
Signature · 1 line
2.45.2
Signature · 2 lines
Sincerely yours,
Mike.
Reviewed-by on v1 1/3
#3 lore raw

Re: [PATCH v1 1/3] mm: turn USE_SPLIT_PTE_PTLOCKS / USE_SPLIT_PTE_PTLOCKS into Kconfig options

In reply to David Hildenbrand
Quoted 7 lines · David Hildenbrand

Let's clean that up a bit and prepare for depending on CONFIG_SPLIT_PMD_PTLOCKS in other Kconfig options.

More cleanups would be reasonable (like the arch-specific "depends on" for CONFIG_SPLIT_PTE_PTLOCKS), but we'll leave that for another day.

Quoted 10 lines
---
  arch/arm/mm/fault-armv.c      |  6 +++---
  arch/x86/xen/mmu_pv.c         |  7 ++++---
  include/linux/mm.h            |  8 ++++----
  include/linux/mm_types.h      |  2 +-
  include/linux/mm_types_task.h |  3 ---
  kernel/fork.c                 |  4 ++--
  mm/Kconfig                    | 18 +++++++++++-------
  mm/memory.c                   |  2 +-
  8 files changed, 26 insertions(+), 24 deletions(-)

That's great. Thanks!

Quoted 194 lines
arch/arm/mm/fault-armv.c +3 -3
index 2286c2ea60ec4..831793cd6ff94 100644
@@ -61,7 +61,7 @@ static int do_adjust_pte(struct vm_area_struct *vma, unsigned long address,
  	return ret;
  }
 
-#if USE_SPLIT_PTE_PTLOCKS
+#if defined(CONFIG_SPLIT_PTE_PTLOCKS)
  /*
   * If we are using split PTE locks, then we need to take the page
   * lock here.  Otherwise we are using shared mm->page_table_lock
@@ -80,10 +80,10 @@ static inline void do_pte_unlock(spinlock_t *ptl)
  {
  	spin_unlock(ptl);
  }
-#else /* !USE_SPLIT_PTE_PTLOCKS */
+#else /* !defined(CONFIG_SPLIT_PTE_PTLOCKS) */
  static inline void do_pte_lock(spinlock_t *ptl) {}
  static inline void do_pte_unlock(spinlock_t *ptl) {}
-#endif /* USE_SPLIT_PTE_PTLOCKS */
+#endif /* defined(CONFIG_SPLIT_PTE_PTLOCKS) */
 
  static int adjust_pte(struct vm_area_struct *vma, unsigned long address,
  	unsigned long pfn)
arch/x86/xen/mmu_pv.c +4 -3
index f1ce39d6d32cb..f4a316894bbb4 100644
@@ -665,7 +665,7 @@ static spinlock_t *xen_pte_lock(struct page *page, struct mm_struct *mm)
  {
  	spinlock_t *ptl = NULL;
 
-#if USE_SPLIT_PTE_PTLOCKS
+#if defined(CONFIG_SPLIT_PTE_PTLOCKS)
  	ptl = ptlock_ptr(page_ptdesc(page));
  	spin_lock_nest_lock(ptl, &mm->page_table_lock);
  #endif
@@ -1553,7 +1553,8 @@ static inline void xen_alloc_ptpage(struct mm_struct *mm, unsigned long pfn,
 
  		__set_pfn_prot(pfn, PAGE_KERNEL_RO);
 
-		if (level == PT_PTE && USE_SPLIT_PTE_PTLOCKS && !pinned)
+		if (level == PT_PTE && IS_ENABLED(CONFIG_SPLIT_PTE_PTLOCKS) &&
+		    !pinned)
  			__pin_pagetable_pfn(MMUEXT_PIN_L1_TABLE, pfn);
 
  		xen_mc_issue(XEN_LAZY_MMU);
@@ -1581,7 +1582,7 @@ static inline void xen_release_ptpage(unsigned long pfn, unsigned level)
  	if (pinned) {
  		xen_mc_batch();
 
-		if (level == PT_PTE && USE_SPLIT_PTE_PTLOCKS)
+		if (level == PT_PTE && IS_ENABLED(CONFIG_SPLIT_PTE_PTLOCKS))
  			__pin_pagetable_pfn(MMUEXT_UNPIN_TABLE, pfn);
 
  		__set_pfn_prot(pfn, PAGE_KERNEL);
include/linux/mm.h +4 -4
index 0472a5090b180..dff43101572ec 100644
@@ -2843,7 +2843,7 @@ static inline void pagetable_free(struct ptdesc *pt)
  	__free_pages(page, compound_order(page));
  }
 
-#if USE_SPLIT_PTE_PTLOCKS
+#if defined(CONFIG_SPLIT_PTE_PTLOCKS)
  #if ALLOC_SPLIT_PTLOCKS
  void __init ptlock_cache_init(void);
  bool ptlock_alloc(struct ptdesc *ptdesc);
@@ -2895,7 +2895,7 @@ static inline bool ptlock_init(struct ptdesc *ptdesc)
  	return true;
  }
 
-#else	/* !USE_SPLIT_PTE_PTLOCKS */
+#else	/* !defined(CONFIG_SPLIT_PTE_PTLOCKS) */
  /*
   * We use mm->page_table_lock to guard all pagetable pages of the mm.
   */
@@ -2906,7 +2906,7 @@ static inline spinlock_t *pte_lockptr(struct mm_struct *mm, pte_t *pte)
  static inline void ptlock_cache_init(void) {}
  static inline bool ptlock_init(struct ptdesc *ptdesc) { return true; }
  static inline void ptlock_free(struct ptdesc *ptdesc) {}
-#endif /* USE_SPLIT_PTE_PTLOCKS */
+#endif /* defined(CONFIG_SPLIT_PTE_PTLOCKS) */
 
  static inline bool pagetable_pte_ctor(struct ptdesc *ptdesc)
  {
@@ -2966,7 +2966,7 @@ pte_t *pte_offset_map_nolock(struct mm_struct *mm, pmd_t *pmd,
  	((unlikely(pmd_none(*(pmd))) && __pte_alloc_kernel(pmd))? \
  		NULL: pte_offset_kernel(pmd, address))
 
-#if USE_SPLIT_PMD_PTLOCKS
+#if defined(CONFIG_SPLIT_PMD_PTLOCKS)
 
  static inline struct page *pmd_pgtable_page(pmd_t *pmd)
  {
include/linux/mm_types.h +1 -1
index 4854249792545..165c58b12ccc9 100644
@@ -947,7 +947,7 @@ struct mm_struct {
  #ifdef CONFIG_MMU_NOTIFIER
  		struct mmu_notifier_subscriptions *notifier_subscriptions;
  #endif
-#if defined(CONFIG_TRANSPARENT_HUGEPAGE) && !USE_SPLIT_PMD_PTLOCKS
+#if defined(CONFIG_TRANSPARENT_HUGEPAGE) && !defined(CONFIG_SPLIT_PMD_PTLOCKS)
  		pgtable_t pmd_huge_pte; /* protected by page_table_lock */
  #endif
  #ifdef CONFIG_NUMA_BALANCING
include/linux/mm_types_task.h +0 -3
index a2f6179b672b8..bff5706b76e14 100644
@@ -16,9 +16,6 @@
  #include <asm/tlbbatch.h>
  #endif
 
-#define USE_SPLIT_PTE_PTLOCKS	(NR_CPUS >= CONFIG_SPLIT_PTLOCK_CPUS)
-#define USE_SPLIT_PMD_PTLOCKS	(USE_SPLIT_PTE_PTLOCKS && \
-		IS_ENABLED(CONFIG_ARCH_ENABLE_SPLIT_PMD_PTLOCK))
  #define ALLOC_SPLIT_PTLOCKS	(SPINLOCK_SIZE > BITS_PER_LONG/8)
 
  /*
kernel/fork.c +2 -2
index a8362c26ebcb0..216ce9ba4f4e6 100644
@@ -832,7 +832,7 @@ static void check_mm(struct mm_struct *mm)
  		pr_alert("BUG: non-zero pgtables_bytes on freeing mm: %ld\n",
  				mm_pgtables_bytes(mm));
 
-#if defined(CONFIG_TRANSPARENT_HUGEPAGE) && !USE_SPLIT_PMD_PTLOCKS
+#if defined(CONFIG_TRANSPARENT_HUGEPAGE) && !defined(CONFIG_SPLIT_PMD_PTLOCKS)
  	VM_BUG_ON_MM(mm->pmd_huge_pte, mm);
  #endif
  }
@@ -1276,7 +1276,7 @@ static struct mm_struct *mm_init(struct mm_struct *mm, struct task_struct *p,
  	RCU_INIT_POINTER(mm->exe_file, NULL);
  	mmu_notifier_subscriptions_init(mm);
  	init_tlb_flush_pending(mm);
-#if defined(CONFIG_TRANSPARENT_HUGEPAGE) && !USE_SPLIT_PMD_PTLOCKS
+#if defined(CONFIG_TRANSPARENT_HUGEPAGE) && !defined(CONFIG_SPLIT_PMD_PTLOCKS)
  	mm->pmd_huge_pte = NULL;
  #endif
  	mm_init_uprobes_state(mm);
mm/Kconfig +11 -7
index b72e7d040f789..7b716ac802726 100644
@@ -585,17 +585,21 @@ config ARCH_MHP_MEMMAP_ON_MEMORY_ENABLE
  # at the same time (e.g. copy_page_range()).
  # DEBUG_SPINLOCK and DEBUG_LOCK_ALLOC spinlock_t also enlarge struct page.
  #
-config SPLIT_PTLOCK_CPUS
-	int
-	default "999999" if !MMU
-	default "999999" if ARM && !CPU_CACHE_VIPT
-	default "999999" if PARISC && !PA20
-	default "999999" if SPARC32
-	default "4"
+config SPLIT_PTE_PTLOCKS
+	def_bool y
+	depends on MMU
+	depends on NR_CPUS >= 4
+	depends on !ARM || CPU_CACHE_VIPT
+	depends on !PARISC || PA20
+	depends on !SPARC32
 
  config ARCH_ENABLE_SPLIT_PMD_PTLOCK
  	bool
 
+config SPLIT_PMD_PTLOCKS
+	def_bool y
+	depends on SPLIT_PTE_PTLOCKS && ARCH_ENABLE_SPLIT_PMD_PTLOCK
+
  #
  # support for memory balloon
  config MEMORY_BALLOON
mm/memory.c +1 -1
index 833d2cad6eb29..714589582fe15 100644
@@ -6559,7 +6559,7 @@ long copy_folio_from_user(struct folio *dst_folio,
  }
  #endif /* CONFIG_TRANSPARENT_HUGEPAGE || CONFIG_HUGETLBFS */
 
-#if USE_SPLIT_PTE_PTLOCKS && ALLOC_SPLIT_PTLOCKS
+#if defined(CONFIG_SPLIT_PTE_PTLOCKS) && ALLOC_SPLIT_PTLOCKS
 
  static struct kmem_cache *page_ptl_cachep;
 
maintainer Reviewed-by on v1 1/3
#4 lore raw

Re: [PATCH v1 1/3] mm: turn USE_SPLIT_PTE_PTLOCKS / USE_SPLIT_PTE_PTLOCKS into Kconfig options

In reply to David Hildenbrand
Quoted 7 lines · David Hildenbrand

Let's clean that up a bit and prepare for depending on CONFIG_SPLIT_PMD_PTLOCKS in other Kconfig options.

More cleanups would be reasonable (like the arch-specific "depends on" for CONFIG_SPLIT_PTE_PTLOCKS), but we'll leave that for another day.

Thanks!

Signature · 2 lines
RMK's Patch system: https://www.armlinux.org.uk/developer/patches/
FTTP is here! 80Mbps down 10Mbps up. Decent connectivity at last!
#5 lore raw

PATCH 2/3 mm/hugetlb: enforce that PMD PT sharing has split PMD PT locks

In reply to David Hildenbrand

Sharing page tables between processes but falling back to per-MM page table locks cannot possibly work.

So, let's make sure that we do have split PMD locks by adding a new Kconfig option and letting that depend on CONFIG_SPLIT_PMD_PTLOCKS.

---
 fs/Kconfig              | 4 ++++
 include/linux/hugetlb.h | 5 ++---
 mm/hugetlb.c            | 8 ++++----
 3 files changed, 10 insertions(+), 7 deletions(-)
fs/Kconfig +4 -0
index a46b0cbc4d8f6..0e4efec1d92e6 100644
@@ -288,6 +288,10 @@ config HUGETLB_PAGE_OPTIMIZE_VMEMMAP
 	depends on ARCH_WANT_OPTIMIZE_HUGETLB_VMEMMAP
 	depends on SPARSEMEM_VMEMMAP
 
+config HUGETLB_PMD_PAGE_TABLE_SHARING
+	def_bool HUGETLB_PAGE
+	depends on ARCH_WANT_HUGE_PMD_SHARE && SPLIT_PMD_PTLOCKS
+
 config ARCH_HAS_GIGANTIC_PAGE
 	bool
 
include/linux/hugetlb.h +2 -3
index da800e56fe590..4d2f3224ff027 100644
@@ -1243,7 +1243,7 @@ static inline __init void hugetlb_cma_reserve(int order)
 }
 #endif
 
-#ifdef CONFIG_ARCH_WANT_HUGE_PMD_SHARE
+#ifdef CONFIG_HUGETLB_PMD_PAGE_TABLE_SHARING
 static inline bool hugetlb_pmd_shared(pte_t *pte)
 {
 	return page_count(virt_to_page(pte)) > 1;
@@ -1279,8 +1279,7 @@ bool __vma_private_lock(struct vm_area_struct *vma);
 static inline pte_t *
 hugetlb_walk(struct vm_area_struct *vma, unsigned long addr, unsigned long sz)
 {
-#if defined(CONFIG_HUGETLB_PAGE) && \
-	defined(CONFIG_ARCH_WANT_HUGE_PMD_SHARE) && defined(CONFIG_LOCKDEP)
+#if defined(CONFIG_HUGETLB_PMD_PAGE_TABLE_SHARING) && defined(CONFIG_LOCKDEP)
 	struct hugetlb_vma_lock *vma_lock = vma->vm_private_data;
 
 	/*
mm/hugetlb.c +4 -4
index 0858a18272073..c4d94e122c41f 100644
@@ -7211,7 +7211,7 @@ long hugetlb_unreserve_pages(struct inode *inode, long start, long end,
 	return 0;
 }
 
-#ifdef CONFIG_ARCH_WANT_HUGE_PMD_SHARE
+#ifdef CONFIG_HUGETLB_PMD_PAGE_TABLE_SHARING
 static unsigned long page_table_shareable(struct vm_area_struct *svma,
 				struct vm_area_struct *vma,
 				unsigned long addr, pgoff_t idx)
@@ -7373,7 +7373,7 @@ int huge_pmd_unshare(struct mm_struct *mm, struct vm_area_struct *vma,
 	return 1;
 }
 
-#else /* !CONFIG_ARCH_WANT_HUGE_PMD_SHARE */
+#else /* !CONFIG_HUGETLB_PMD_PAGE_TABLE_SHARING */
 
 pte_t *huge_pmd_share(struct mm_struct *mm, struct vm_area_struct *vma,
 		      unsigned long addr, pud_t *pud)
@@ -7396,7 +7396,7 @@ bool want_pmd_share(struct vm_area_struct *vma, unsigned long addr)
 {
 	return false;
 }
-#endif /* CONFIG_ARCH_WANT_HUGE_PMD_SHARE */
+#endif /* CONFIG_HUGETLB_PMD_PAGE_TABLE_SHARING */
 
 #ifdef CONFIG_ARCH_WANT_GENERAL_HUGETLB
 pte_t *huge_pte_alloc(struct mm_struct *mm, struct vm_area_struct *vma,
@@ -7494,7 +7494,7 @@ unsigned long hugetlb_mask_last_page(struct hstate *h)
 /* See description above.  Architectures can provide their own version. */
 __weak unsigned long hugetlb_mask_last_page(struct hstate *h)
 {
-#ifdef CONFIG_ARCH_WANT_HUGE_PMD_SHARE
+#ifdef CONFIG_HUGETLB_PMD_PAGE_TABLE_SHARING
 	if (huge_page_size(h) == PMD_SIZE)
 		return PUD_SIZE - PMD_SIZE;
 #endif
Signature · 1 line
2.45.2
reviewer Acked-by on v1 2/3
#6 lore raw

Re: [PATCH v1 2/3] mm/hugetlb: enforce that PMD PT sharing has split PMD PT locks

Quoted 7 lines · David Hildenbrand

Sharing page tables between processes but falling back to per-MM page table locks cannot possibly work.

So, let's make sure that we do have split PMD locks by adding a new Kconfig option and letting that depend on CONFIG_SPLIT_PMD_PTLOCKS.

Quoted 88 lines
---
 fs/Kconfig              | 4 ++++
 include/linux/hugetlb.h | 5 ++---
 mm/hugetlb.c            | 8 ++++----
 3 files changed, 10 insertions(+), 7 deletions(-)
fs/Kconfig +4 -0
index a46b0cbc4d8f6..0e4efec1d92e6 100644
@@ -288,6 +288,10 @@ config HUGETLB_PAGE_OPTIMIZE_VMEMMAP
 	depends on ARCH_WANT_OPTIMIZE_HUGETLB_VMEMMAP
 	depends on SPARSEMEM_VMEMMAP
 
+config HUGETLB_PMD_PAGE_TABLE_SHARING
+	def_bool HUGETLB_PAGE
+	depends on ARCH_WANT_HUGE_PMD_SHARE && SPLIT_PMD_PTLOCKS
+
 config ARCH_HAS_GIGANTIC_PAGE
 	bool
 
include/linux/hugetlb.h +2 -3
index da800e56fe590..4d2f3224ff027 100644
@@ -1243,7 +1243,7 @@ static inline __init void hugetlb_cma_reserve(int order)
 }
 #endif
 
-#ifdef CONFIG_ARCH_WANT_HUGE_PMD_SHARE
+#ifdef CONFIG_HUGETLB_PMD_PAGE_TABLE_SHARING
 static inline bool hugetlb_pmd_shared(pte_t *pte)
 {
 	return page_count(virt_to_page(pte)) > 1;
@@ -1279,8 +1279,7 @@ bool __vma_private_lock(struct vm_area_struct *vma);
 static inline pte_t *
 hugetlb_walk(struct vm_area_struct *vma, unsigned long addr, unsigned long sz)
 {
-#if defined(CONFIG_HUGETLB_PAGE) && \
-	defined(CONFIG_ARCH_WANT_HUGE_PMD_SHARE) && defined(CONFIG_LOCKDEP)
+#if defined(CONFIG_HUGETLB_PMD_PAGE_TABLE_SHARING) && defined(CONFIG_LOCKDEP)
 	struct hugetlb_vma_lock *vma_lock = vma->vm_private_data;
 
 	/*
mm/hugetlb.c +4 -4
index 0858a18272073..c4d94e122c41f 100644
@@ -7211,7 +7211,7 @@ long hugetlb_unreserve_pages(struct inode *inode, long start, long end,
 	return 0;
 }
 
-#ifdef CONFIG_ARCH_WANT_HUGE_PMD_SHARE
+#ifdef CONFIG_HUGETLB_PMD_PAGE_TABLE_SHARING
 static unsigned long page_table_shareable(struct vm_area_struct *svma,
 				struct vm_area_struct *vma,
 				unsigned long addr, pgoff_t idx)
@@ -7373,7 +7373,7 @@ int huge_pmd_unshare(struct mm_struct *mm, struct vm_area_struct *vma,
 	return 1;
 }
 
-#else /* !CONFIG_ARCH_WANT_HUGE_PMD_SHARE */
+#else /* !CONFIG_HUGETLB_PMD_PAGE_TABLE_SHARING */
 
 pte_t *huge_pmd_share(struct mm_struct *mm, struct vm_area_struct *vma,
 		      unsigned long addr, pud_t *pud)
@@ -7396,7 +7396,7 @@ bool want_pmd_share(struct vm_area_struct *vma, unsigned long addr)
 {
 	return false;
 }
-#endif /* CONFIG_ARCH_WANT_HUGE_PMD_SHARE */
+#endif /* CONFIG_HUGETLB_PMD_PAGE_TABLE_SHARING */
 
 #ifdef CONFIG_ARCH_WANT_GENERAL_HUGETLB
 pte_t *huge_pte_alloc(struct mm_struct *mm, struct vm_area_struct *vma,
@@ -7494,7 +7494,7 @@ unsigned long hugetlb_mask_last_page(struct hstate *h)
 /* See description above.  Architectures can provide their own version. */
 __weak unsigned long hugetlb_mask_last_page(struct hstate *h)
 {
-#ifdef CONFIG_ARCH_WANT_HUGE_PMD_SHARE
+#ifdef CONFIG_HUGETLB_PMD_PAGE_TABLE_SHARING
 	if (huge_page_size(h) == PMD_SIZE)
 		return PUD_SIZE - PMD_SIZE;
 #endif
Signature · 1 line
2.45.2
Signature · 2 lines
Sincerely yours,
Mike.
#7 lore raw

PATCH 3/3 powerpc/8xx: document and enforce that split PT locks are not used

In reply to David Hildenbrand

Right now, we cannot have split PT locks because 8xx does not support SMP.

But for the sake of documentation *why* 8xx is fine regarding what we documented in huge_pte_lockptr(), let's just add code to enforce it at the same time as documenting it.

This should also make everybody who wants to copy from the 8xx approach of supporting such unusual ways of mapping hugetlb folios aware that it gets tricky once multiple page tables are involved.

---
 arch/powerpc/mm/pgtable.c | 6 ++++++
 1 file changed, 6 insertions(+)
arch/powerpc/mm/pgtable.c +6 -0
index ab0656115424f..7316396e452d8 100644
@@ -297,6 +297,12 @@ int huge_ptep_set_access_flags(struct vm_area_struct *vma,
 }
 
 #if defined(CONFIG_PPC_8xx)
+
+#if defined(CONFIG_SPLIT_PTE_PTLOCKS) || defined(CONFIG_SPLIT_PMD_PTLOCKS)
+/* We need the same lock to protect the PMD table and the two PTE tables. */
+#error "8M hugetlb folios are incompatible with split page table locks"
+#endif
+
 static void __set_huge_pte_at(pmd_t *pmd, pte_t *ptep, pte_basic_t val)
 {
 	pte_basic_t *entry = (pte_basic_t *)ptep;
Signature · 1 line
2.45.2