Skip to contentSkip to conversation

kvm: Fix page ageing bugs

PATCH New Andres Lagar-Cavilla 4 participants 14 messages Sep 22, 2014 – Sep 24, 2014

Subsystems (13)
lore
Share & tools
lore
#0 lore raw

PATCH 1/1 kvm: Fix page ageing bugs

1. We were calling clear_flush_young_notify in unmap_one, but we are within an mmu notifier invalidate range scope. The spte exists no more (due to range_start) and the accessed bit info has already been propagated (due to kvm_pfn_set_accessed). Simply call clear_flush_young.

2. We clear_flush_young on a primary MMU PMD, but this may be mapped as a collection of PTEs by the secondary MMU (e.g. during log-dirty). This required expanding the interface of the clear_flush_young mmu notifier, so a lot of code has been trivially touched.

3. In the absence of shadow_accessed_mask (e.g. EPT A bit), we emulate the access bit by blowing the spte. This requires proper synchronizing with MMU notifier consumers, like every other removal of spte's does.

---
 arch/arm/include/asm/kvm_host.h     |  3 ++-
 arch/arm64/include/asm/kvm_host.h   |  3 ++-
 arch/powerpc/include/asm/kvm_host.h |  2 +-
 arch/powerpc/kvm/book3s.c           |  4 +--
 arch/powerpc/kvm/book3s.h           |  3 ++-
 arch/powerpc/kvm/book3s_64_mmu_hv.c |  4 +--
 arch/powerpc/kvm/book3s_pr.c        |  3 ++-
 arch/powerpc/kvm/e500_mmu_host.c    |  2 +-
 arch/x86/include/asm/kvm_host.h     |  2 +-
 arch/x86/kvm/mmu.c                  | 50 +++++++++++++++++++++++++------------
 drivers/iommu/amd_iommu_v2.c        |  6 +++--
 include/linux/mmu_notifier.h        | 24 ++++++++++++------
 include/trace/events/kvm.h          | 10 ++++----
 mm/mmu_notifier.c                   |  5 ++--
 mm/rmap.c                           |  6 ++++-
 virt/kvm/kvm_main.c                 |  5 ++--
 16 files changed, 86 insertions(+), 46 deletions(-)
arch/arm/include/asm/kvm_host.h +2 -1
index 032a853..8c3f7eb 100644
@@ -170,7 +170,8 @@ unsigned long kvm_arm_num_regs(struct kvm_vcpu *vcpu);
 int kvm_arm_copy_reg_indices(struct kvm_vcpu *vcpu, u64 __user *indices);
 
 /* We do not have shadow page tables, hence the empty hooks */
-static inline int kvm_age_hva(struct kvm *kvm, unsigned long hva)
+static inline int kvm_age_hva(struct kvm *kvm, unsigned long start,
+			      unsigned long end)
 {
 	return 0;
 }
arch/arm64/include/asm/kvm_host.h +2 -1
index be9970a..a3c671b 100644
@@ -180,7 +180,8 @@ int kvm_unmap_hva_range(struct kvm *kvm,
 void kvm_set_spte_hva(struct kvm *kvm, unsigned long hva, pte_t pte);
 
 /* We do not have shadow page tables, hence the empty hooks */
-static inline int kvm_age_hva(struct kvm *kvm, unsigned long hva)
+static inline int kvm_age_hva(struct kvm *kvm, unsigned long start,
+			      unsigned long end)
 {
 	return 0;
 }
arch/powerpc/include/asm/kvm_host.h +1 -1
index 6040008..d329bc5 100644
@@ -56,7 +56,7 @@
 extern int kvm_unmap_hva(struct kvm *kvm, unsigned long hva);
 extern int kvm_unmap_hva_range(struct kvm *kvm,
 			       unsigned long start, unsigned long end);
-extern int kvm_age_hva(struct kvm *kvm, unsigned long hva);
+extern int kvm_age_hva(struct kvm *kvm, unsigned long start, unsigned long end);
 extern int kvm_test_age_hva(struct kvm *kvm, unsigned long hva);
 extern void kvm_set_spte_hva(struct kvm *kvm, unsigned long hva, pte_t pte);
 
arch/powerpc/kvm/book3s.c +2 -2
index dd03f6b..c16cfbf 100644
@@ -851,9 +851,9 @@ int kvm_unmap_hva_range(struct kvm *kvm, unsigned long start, unsigned long end)
 	return kvm->arch.kvm_ops->unmap_hva_range(kvm, start, end);
 }
 
-int kvm_age_hva(struct kvm *kvm, unsigned long hva)
+int kvm_age_hva(struct kvm *kvm, unsigned long start, unsigned long end)
 {
-	return kvm->arch.kvm_ops->age_hva(kvm, hva);
+	return kvm->arch.kvm_ops->age_hva(kvm, start, end);
 }
 
 int kvm_test_age_hva(struct kvm *kvm, unsigned long hva)
arch/powerpc/kvm/book3s.h +2 -1
index 4bf956c..d2b3ec0 100644
@@ -17,7 +17,8 @@ extern void kvmppc_core_flush_memslot_hv(struct kvm *kvm,
 extern int kvm_unmap_hva_hv(struct kvm *kvm, unsigned long hva);
 extern int kvm_unmap_hva_range_hv(struct kvm *kvm, unsigned long start,
 				  unsigned long end);
-extern int kvm_age_hva_hv(struct kvm *kvm, unsigned long hva);
+extern int kvm_age_hva_hv(struct kvm *kvm, unsigned long start,
+			  unsigned long end);
 extern int kvm_test_age_hva_hv(struct kvm *kvm, unsigned long hva);
 extern void kvm_set_spte_hva_hv(struct kvm *kvm, unsigned long hva, pte_t pte);
 
arch/powerpc/kvm/book3s_64_mmu_hv.c +2 -2
index 72c20bb..81460c5 100644
@@ -1002,11 +1002,11 @@ static int kvm_age_rmapp(struct kvm *kvm, unsigned long *rmapp,
 	return ret;
 }
 
-int kvm_age_hva_hv(struct kvm *kvm, unsigned long hva)
+int kvm_age_hva_hv(struct kvm *kvm, unsigned long start, unsigned long end)
 {
 	if (!kvm->arch.using_mmu_notifiers)
 		return 0;
-	return kvm_handle_hva(kvm, hva, kvm_age_rmapp);
+	return kvm_handle_hva_range(kvm, start, end, kvm_age_rmapp);
 }
 
 static int kvm_test_age_rmapp(struct kvm *kvm, unsigned long *rmapp,
arch/powerpc/kvm/book3s_pr.c +2 -1
index faffb27..852fcd8 100644
@@ -295,7 +295,8 @@ static int kvm_unmap_hva_range_pr(struct kvm *kvm, unsigned long start,
 	return 0;
 }
 
-static int kvm_age_hva_pr(struct kvm *kvm, unsigned long hva)
+static int kvm_age_hva_pr(struct kvm *kvm, unsigned long start,
+			  unsigned long end)
 {
 	/* XXX could be more clever ;) */
 	return 0;
arch/powerpc/kvm/e500_mmu_host.c +1 -1
index 08f14bb..b1f3f63 100644
@@ -732,7 +732,7 @@ int kvm_unmap_hva_range(struct kvm *kvm, unsigned long start, unsigned long end)
 	return 0;
 }
 
-int kvm_age_hva(struct kvm *kvm, unsigned long hva)
+int kvm_age_hva(struct kvm *kvm, unsigned long start, unsigned long end)
 {
 	/* XXX could be more clever ;) */
 	return 0;
arch/x86/include/asm/kvm_host.h +1 -1
index 028df8d..56d6839 100644
@@ -1036,7 +1036,7 @@ asmlinkage void kvm_spurious_fault(void);
 #define KVM_ARCH_WANT_MMU_NOTIFIER
 int kvm_unmap_hva(struct kvm *kvm, unsigned long hva);
 int kvm_unmap_hva_range(struct kvm *kvm, unsigned long start, unsigned long end);
-int kvm_age_hva(struct kvm *kvm, unsigned long hva);
+int kvm_age_hva(struct kvm *kvm, unsigned long start, end);
 int kvm_test_age_hva(struct kvm *kvm, unsigned long hva);
 void kvm_set_spte_hva(struct kvm *kvm, unsigned long hva, pte_t pte);
 int cpuid_maxphyaddr(struct kvm_vcpu *vcpu);
arch/x86/kvm/mmu.c +34 -16

75 lines not on this page · raw message

drivers/iommu/amd_iommu_v2.c +4 -2
index 5f578e8..90d734b 100644
@@ -402,9 +402,11 @@ static void __mn_flush_page(struct mmu_notifier *mn,
 
 static int mn_clear_flush_young(struct mmu_notifier *mn,
 				struct mm_struct *mm,
-				unsigned long address)
+				unsigned long start,
+				unsigned long end)
 {
-	__mn_flush_page(mn, address);
+	for (; start < end; start += PAGE_SIZE)
+		__mn_flush_page(mn, start);
 
 	return 0;
 }
include/linux/mmu_notifier.h +17 -7
index 2728869..f09cb27 100644
@@ -57,10 +57,13 @@ struct mmu_notifier_ops {
 	 * pte. This way the VM will provide proper aging to the
 	 * accesses to the page through the secondary MMUs and not
 	 * only to the ones through the Linux pte.
+	 * Start-end is necessary in case the secondary MMU is mapping the page
+	 * at a different granularity than the primary MMU.
 	 */
 	int (*clear_flush_young)(struct mmu_notifier *mn,
 				 struct mm_struct *mm,
-				 unsigned long address);
+				 unsigned long start,
+				 unsigned long end);
 
 	/*
 	 * test_young is called to check the young/accessed bitflag in
@@ -175,7 +178,8 @@ extern void mmu_notifier_unregister_no_release(struct mmu_notifier *mn,
 extern void __mmu_notifier_mm_destroy(struct mm_struct *mm);
 extern void __mmu_notifier_release(struct mm_struct *mm);
 extern int __mmu_notifier_clear_flush_young(struct mm_struct *mm,
-					  unsigned long address);
+					  unsigned long start,
+					  unsigned long end);
 extern int __mmu_notifier_test_young(struct mm_struct *mm,
 				     unsigned long address);
 extern void __mmu_notifier_change_pte(struct mm_struct *mm,
@@ -194,10 +198,11 @@ static inline void mmu_notifier_release(struct mm_struct *mm)
 }
 
 static inline int mmu_notifier_clear_flush_young(struct mm_struct *mm,
-					  unsigned long address)
+					  unsigned long start,
+					  unsigned long end)
 {
 	if (mm_has_notifiers(mm))
-		return __mmu_notifier_clear_flush_young(mm, address);
+		return __mmu_notifier_clear_flush_young(mm, start, end);
 	return 0;
 }
 
@@ -255,7 +260,9 @@ static inline void mmu_notifier_mm_destroy(struct mm_struct *mm)
 	unsigned long ___address = __address;				\
 	__young = ptep_clear_flush_young(___vma, ___address, __ptep);	\
 	__young |= mmu_notifier_clear_flush_young(___vma->vm_mm,	\
-						  ___address);		\
+						  ___address,		\
+						  ___address +		\
+							PAGE_SIZE);	\
 	__young;							\
 })
 
@@ -266,7 +273,9 @@ static inline void mmu_notifier_mm_destroy(struct mm_struct *mm)
 	unsigned long ___address = __address;				\
 	__young = pmdp_clear_flush_young(___vma, ___address, __pmdp);	\
 	__young |= mmu_notifier_clear_flush_young(___vma->vm_mm,	\
-						  ___address);		\
+						  ___address,		\
+						  ___address +		\
+							PMD_PAGE_SIZE);	\
 	__young;							\
 })
 
@@ -301,7 +310,8 @@ static inline void mmu_notifier_release(struct mm_struct *mm)
 }
 
 static inline int mmu_notifier_clear_flush_young(struct mm_struct *mm,
-					  unsigned long address)
+					  unsigned long start,
+					  unsigned long end)
 {
 	return 0;
 }
include/trace/events/kvm.h +5 -5
index ab679c3..f58ef59 100644
@@ -225,8 +225,8 @@ TRACE_EVENT(kvm_fpu,
 );
 
 TRACE_EVENT(kvm_age_page,
-	TP_PROTO(ulong hva, struct kvm_memory_slot *slot, int ref),
-	TP_ARGS(hva, slot, ref),
+	TP_PROTO(ulong gfn, struct kvm_memory_slot *slot, int ref),
+	TP_ARGS(gfn, slot, ref),
 
 	TP_STRUCT__entry(
 		__field(	u64,	hva		)
@@ -235,9 +235,9 @@ TRACE_EVENT(kvm_age_page,
 	),
 
 	TP_fast_assign(
-		__entry->hva		= hva;
-		__entry->gfn		=
-		  slot->base_gfn + ((hva - slot->userspace_addr) >> PAGE_SHIFT);
+		__entry->gfn		= gfn;
+		__entry->hva		= ((gfn - slot->base_gfn) >>
+					    PAGE_SHIFT) + slot->userspace_addr;
 		__entry->referenced	= ref;
 	),
 
mm/mmu_notifier.c +3 -2
index 950813b..2c8da98 100644
@@ -107,7 +107,8 @@ void __mmu_notifier_release(struct mm_struct *mm)
  * existed or not.
  */
 int __mmu_notifier_clear_flush_young(struct mm_struct *mm,
-					unsigned long address)
+					unsigned long start,
+					unsigned long end)
 {
 	struct mmu_notifier *mn;
 	int young = 0, id;
@@ -115,7 +116,7 @@ int __mmu_notifier_clear_flush_young(struct mm_struct *mm,
 	id = srcu_read_lock(&srcu);
 	hlist_for_each_entry_rcu(mn, &mm->mmu_notifier_mm->list, hlist) {
 		if (mn->ops->clear_flush_young)
-			young |= mn->ops->clear_flush_young(mn, mm, address);
+			young |= mn->ops->clear_flush_young(mn, mm, start, end);
 	}
 	srcu_read_unlock(&srcu, id);
 
mm/rmap.c +5 -1
index 3e8491c..66b075f 100644
@@ -1355,7 +1355,11 @@ static int try_to_unmap_cluster(unsigned long cursor, unsigned int *mapcount,
 			continue;	/* don't unmap */
 		}
 
-		if (ptep_clear_flush_young_notify(vma, address, pte))
+		/*
+		 * No need for _notify because we're called within an
+		 * mmu_notifier_invalidate_range_ {start|end} scope.
+		 */
+		if (ptep_clear_flush_young(vma, address, pte))
 			continue;
 
 		/* Nuke the page table entry. */
virt/kvm/kvm_main.c +3 -2
index db57363..ea327f4 100644
@@ -367,7 +367,8 @@ static void kvm_mmu_notifier_invalidate_range_end(struct mmu_notifier *mn,
 
 static int kvm_mmu_notifier_clear_flush_young(struct mmu_notifier *mn,
 					      struct mm_struct *mm,
-					      unsigned long address)
+					      unsigned long start,
+					      unsigned long end)
 {
 	struct kvm *kvm = mmu_notifier_to_kvm(mn);
 	int young, idx;
@@ -375,7 +376,7 @@ static int kvm_mmu_notifier_clear_flush_young(struct mmu_notifier *mn,
 	idx = srcu_read_lock(&kvm->srcu);
 	spin_lock(&kvm->mmu_lock);
 
-	young = kvm_age_hva(kvm, address);
+	young = kvm_age_hva(kvm, start, end);
 	if (young)
 		kvm_flush_remote_tlbs(kvm);
 
Signature · 7 lines
2.1.0.rc2.206.gedb03e5

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to [email protected].  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"[email protected]"> [email protected] </a>
#1 lore raw

PATCH v2 1/1 kvm: Fix page ageing bugs

1. We were calling clear_flush_young_notify in unmap_one, but we are within an mmu notifier invalidate range scope. The spte exists no more (due to range_start) and the accessed bit info has already been propagated (due to kvm_pfn_set_accessed). Simply call clear_flush_young.

2. We clear_flush_young on a primary MMU PMD, but this may be mapped as a collection of PTEs by the secondary MMU (e.g. during log-dirty). This required expanding the interface of the clear_flush_young mmu notifier, so a lot of code has been trivially touched.

3. In the absence of shadow_accessed_mask (e.g. EPT A bit), we emulate the access bit by blowing the spte. This requires proper synchronizing with MMU notifier consumers, like every other removal of spte's does.

---
v1 -> v2: * Small cosmetic changes
* Remove unnecessary handling of mmu_notifier_seq in mmu_lock protected region
* Func prototypes fixed
---

---
 arch/arm/include/asm/kvm_host.h     |  3 ++-
 arch/arm64/include/asm/kvm_host.h   |  3 ++-
 arch/powerpc/include/asm/kvm_host.h |  2 +-
 arch/powerpc/kvm/book3s.c           |  4 ++--
 arch/powerpc/kvm/book3s.h           |  3 ++-
 arch/powerpc/kvm/book3s_64_mmu_hv.c |  4 ++--
 arch/powerpc/kvm/book3s_pr.c        |  3 ++-
 arch/powerpc/kvm/e500_mmu_host.c    |  2 +-
 arch/x86/include/asm/kvm_host.h     |  2 +-
 arch/x86/kvm/mmu.c                  | 46 +++++++++++++++++++++++--------------
 drivers/iommu/amd_iommu_v2.c        |  6 +++--
 include/linux/mmu_notifier.h        | 24 +++++++++++++------
 include/trace/events/kvm.h          | 10 ++++----
 mm/mmu_notifier.c                   |  5 ++--
 mm/rmap.c                           |  6 ++++-
 virt/kvm/kvm_main.c                 |  5 ++--
 16 files changed, 81 insertions(+), 47 deletions(-)
arch/arm/include/asm/kvm_host.h +2 -1
index 032a853..8c3f7eb 100644
@@ -170,7 +170,8 @@ unsigned long kvm_arm_num_regs(struct kvm_vcpu *vcpu);
 int kvm_arm_copy_reg_indices(struct kvm_vcpu *vcpu, u64 __user *indices);
 
 /* We do not have shadow page tables, hence the empty hooks */
-static inline int kvm_age_hva(struct kvm *kvm, unsigned long hva)
+static inline int kvm_age_hva(struct kvm *kvm, unsigned long start,
+			      unsigned long end)
 {
 	return 0;
 }
arch/arm64/include/asm/kvm_host.h +2 -1
index be9970a..a3c671b 100644
@@ -180,7 +180,8 @@ int kvm_unmap_hva_range(struct kvm *kvm,
 void kvm_set_spte_hva(struct kvm *kvm, unsigned long hva, pte_t pte);
 
 /* We do not have shadow page tables, hence the empty hooks */
-static inline int kvm_age_hva(struct kvm *kvm, unsigned long hva)
+static inline int kvm_age_hva(struct kvm *kvm, unsigned long start,
+			      unsigned long end)
 {
 	return 0;
 }
arch/powerpc/include/asm/kvm_host.h +1 -1
index 6040008..d329bc5 100644
@@ -56,7 +56,7 @@
 extern int kvm_unmap_hva(struct kvm *kvm, unsigned long hva);
 extern int kvm_unmap_hva_range(struct kvm *kvm,
 			       unsigned long start, unsigned long end);
-extern int kvm_age_hva(struct kvm *kvm, unsigned long hva);
+extern int kvm_age_hva(struct kvm *kvm, unsigned long start, unsigned long end);
 extern int kvm_test_age_hva(struct kvm *kvm, unsigned long hva);
 extern void kvm_set_spte_hva(struct kvm *kvm, unsigned long hva, pte_t pte);
 
arch/powerpc/kvm/book3s.c +2 -2
index dd03f6b..c16cfbf 100644
@@ -851,9 +851,9 @@ int kvm_unmap_hva_range(struct kvm *kvm, unsigned long start, unsigned long end)
 	return kvm->arch.kvm_ops->unmap_hva_range(kvm, start, end);
 }
 
-int kvm_age_hva(struct kvm *kvm, unsigned long hva)
+int kvm_age_hva(struct kvm *kvm, unsigned long start, unsigned long end)
 {
-	return kvm->arch.kvm_ops->age_hva(kvm, hva);
+	return kvm->arch.kvm_ops->age_hva(kvm, start, end);
 }
 
 int kvm_test_age_hva(struct kvm *kvm, unsigned long hva)
arch/powerpc/kvm/book3s.h +2 -1
index 4bf956c..d2b3ec0 100644
@@ -17,7 +17,8 @@ extern void kvmppc_core_flush_memslot_hv(struct kvm *kvm,
 extern int kvm_unmap_hva_hv(struct kvm *kvm, unsigned long hva);
 extern int kvm_unmap_hva_range_hv(struct kvm *kvm, unsigned long start,
 				  unsigned long end);
-extern int kvm_age_hva_hv(struct kvm *kvm, unsigned long hva);
+extern int kvm_age_hva_hv(struct kvm *kvm, unsigned long start,
+			  unsigned long end);
 extern int kvm_test_age_hva_hv(struct kvm *kvm, unsigned long hva);
 extern void kvm_set_spte_hva_hv(struct kvm *kvm, unsigned long hva, pte_t pte);
 
arch/powerpc/kvm/book3s_64_mmu_hv.c +2 -2
index 72c20bb..81460c5 100644
@@ -1002,11 +1002,11 @@ static int kvm_age_rmapp(struct kvm *kvm, unsigned long *rmapp,
 	return ret;
 }
 
-int kvm_age_hva_hv(struct kvm *kvm, unsigned long hva)
+int kvm_age_hva_hv(struct kvm *kvm, unsigned long start, unsigned long end)
 {
 	if (!kvm->arch.using_mmu_notifiers)
 		return 0;
-	return kvm_handle_hva(kvm, hva, kvm_age_rmapp);
+	return kvm_handle_hva_range(kvm, start, end, kvm_age_rmapp);
 }
 
 static int kvm_test_age_rmapp(struct kvm *kvm, unsigned long *rmapp,
arch/powerpc/kvm/book3s_pr.c +2 -1
index faffb27..852fcd8 100644
@@ -295,7 +295,8 @@ static int kvm_unmap_hva_range_pr(struct kvm *kvm, unsigned long start,
 	return 0;
 }
 
-static int kvm_age_hva_pr(struct kvm *kvm, unsigned long hva)
+static int kvm_age_hva_pr(struct kvm *kvm, unsigned long start,
+			  unsigned long end)
 {
 	/* XXX could be more clever ;) */
 	return 0;
arch/powerpc/kvm/e500_mmu_host.c +1 -1
index 08f14bb..b1f3f63 100644
@@ -732,7 +732,7 @@ int kvm_unmap_hva_range(struct kvm *kvm, unsigned long start, unsigned long end)
 	return 0;
 }
 
-int kvm_age_hva(struct kvm *kvm, unsigned long hva)
+int kvm_age_hva(struct kvm *kvm, unsigned long start, unsigned long end)
 {
 	/* XXX could be more clever ;) */
 	return 0;
arch/x86/include/asm/kvm_host.h +1 -1
index 028df8d..a96c61b 100644
@@ -1036,7 +1036,7 @@ asmlinkage void kvm_spurious_fault(void);
 #define KVM_ARCH_WANT_MMU_NOTIFIER
 int kvm_unmap_hva(struct kvm *kvm, unsigned long hva);
 int kvm_unmap_hva_range(struct kvm *kvm, unsigned long start, unsigned long end);
-int kvm_age_hva(struct kvm *kvm, unsigned long hva);
+int kvm_age_hva(struct kvm *kvm, unsigned long start, unsigned long end);
 int kvm_test_age_hva(struct kvm *kvm, unsigned long hva);
 void kvm_set_spte_hva(struct kvm *kvm, unsigned long hva, pte_t pte);
 int cpuid_maxphyaddr(struct kvm_vcpu *vcpu);
arch/x86/kvm/mmu.c +29 -17

70 lines not on this page · raw message

drivers/iommu/amd_iommu_v2.c +4 -2
index 5f578e8..90d734b 100644
@@ -402,9 +402,11 @@ static void __mn_flush_page(struct mmu_notifier *mn,
 
 static int mn_clear_flush_young(struct mmu_notifier *mn,
 				struct mm_struct *mm,
-				unsigned long address)
+				unsigned long start,
+				unsigned long end)
 {
-	__mn_flush_page(mn, address);
+	for (; start < end; start += PAGE_SIZE)
+		__mn_flush_page(mn, start);
 
 	return 0;
 }
include/linux/mmu_notifier.h +17 -7
index 2728869..4269f2c 100644
@@ -57,10 +57,13 @@ struct mmu_notifier_ops {
 	 * pte. This way the VM will provide proper aging to the
 	 * accesses to the page through the secondary MMUs and not
 	 * only to the ones through the Linux pte.
+	 * Start-end is necessary in case the secondary MMU is mapping the page
+	 * at a smaller granularity than the primary MMU.
 	 */
 	int (*clear_flush_young)(struct mmu_notifier *mn,
 				 struct mm_struct *mm,
-				 unsigned long address);
+				 unsigned long start,
+				 unsigned long end);
 
 	/*
 	 * test_young is called to check the young/accessed bitflag in
@@ -175,7 +178,8 @@ extern void mmu_notifier_unregister_no_release(struct mmu_notifier *mn,
 extern void __mmu_notifier_mm_destroy(struct mm_struct *mm);
 extern void __mmu_notifier_release(struct mm_struct *mm);
 extern int __mmu_notifier_clear_flush_young(struct mm_struct *mm,
-					  unsigned long address);
+					  unsigned long start,
+					  unsigned long end);
 extern int __mmu_notifier_test_young(struct mm_struct *mm,
 				     unsigned long address);
 extern void __mmu_notifier_change_pte(struct mm_struct *mm,
@@ -194,10 +198,11 @@ static inline void mmu_notifier_release(struct mm_struct *mm)
 }
 
 static inline int mmu_notifier_clear_flush_young(struct mm_struct *mm,
-					  unsigned long address)
+					  unsigned long start,
+					  unsigned long end)
 {
 	if (mm_has_notifiers(mm))
-		return __mmu_notifier_clear_flush_young(mm, address);
+		return __mmu_notifier_clear_flush_young(mm, start, end);
 	return 0;
 }
 
@@ -255,7 +260,9 @@ static inline void mmu_notifier_mm_destroy(struct mm_struct *mm)
 	unsigned long ___address = __address;				\
 	__young = ptep_clear_flush_young(___vma, ___address, __ptep);	\
 	__young |= mmu_notifier_clear_flush_young(___vma->vm_mm,	\
-						  ___address);		\
+						  ___address,		\
+						  ___address +		\
+							PAGE_SIZE);	\
 	__young;							\
 })
 
@@ -266,7 +273,9 @@ static inline void mmu_notifier_mm_destroy(struct mm_struct *mm)
 	unsigned long ___address = __address;				\
 	__young = pmdp_clear_flush_young(___vma, ___address, __pmdp);	\
 	__young |= mmu_notifier_clear_flush_young(___vma->vm_mm,	\
-						  ___address);		\
+						  ___address,		\
+						  ___address +		\
+							PMD_PAGE_SIZE);	\
 	__young;							\
 })
 
@@ -301,7 +310,8 @@ static inline void mmu_notifier_release(struct mm_struct *mm)
 }
 
 static inline int mmu_notifier_clear_flush_young(struct mm_struct *mm,
-					  unsigned long address)
+					  unsigned long start,
+					  unsigned long end)
 {
 	return 0;
 }
include/trace/events/kvm.h +5 -5
index ab679c3..f58ef59 100644
@@ -225,8 +225,8 @@ TRACE_EVENT(kvm_fpu,
 );
 
 TRACE_EVENT(kvm_age_page,
-	TP_PROTO(ulong hva, struct kvm_memory_slot *slot, int ref),
-	TP_ARGS(hva, slot, ref),
+	TP_PROTO(ulong gfn, struct kvm_memory_slot *slot, int ref),
+	TP_ARGS(gfn, slot, ref),
 
 	TP_STRUCT__entry(
 		__field(	u64,	hva		)
@@ -235,9 +235,9 @@ TRACE_EVENT(kvm_age_page,
 	),
 
 	TP_fast_assign(
-		__entry->hva		= hva;
-		__entry->gfn		=
-		  slot->base_gfn + ((hva - slot->userspace_addr) >> PAGE_SHIFT);
+		__entry->gfn		= gfn;
+		__entry->hva		= ((gfn - slot->base_gfn) >>
+					    PAGE_SHIFT) + slot->userspace_addr;
 		__entry->referenced	= ref;
 	),
 
mm/mmu_notifier.c +3 -2
index 950813b..2c8da98 100644
@@ -107,7 +107,8 @@ void __mmu_notifier_release(struct mm_struct *mm)
  * existed or not.
  */
 int __mmu_notifier_clear_flush_young(struct mm_struct *mm,
-					unsigned long address)
+					unsigned long start,
+					unsigned long end)
 {
 	struct mmu_notifier *mn;
 	int young = 0, id;
@@ -115,7 +116,7 @@ int __mmu_notifier_clear_flush_young(struct mm_struct *mm,
 	id = srcu_read_lock(&srcu);
 	hlist_for_each_entry_rcu(mn, &mm->mmu_notifier_mm->list, hlist) {
 		if (mn->ops->clear_flush_young)
-			young |= mn->ops->clear_flush_young(mn, mm, address);
+			young |= mn->ops->clear_flush_young(mn, mm, start, end);
 	}
 	srcu_read_unlock(&srcu, id);
 
mm/rmap.c +5 -1
index 3e8491c..66b075f 100644
@@ -1355,7 +1355,11 @@ static int try_to_unmap_cluster(unsigned long cursor, unsigned int *mapcount,
 			continue;	/* don't unmap */
 		}
 
-		if (ptep_clear_flush_young_notify(vma, address, pte))
+		/*
+		 * No need for _notify because we're called within an
+		 * mmu_notifier_invalidate_range_ {start|end} scope.
+		 */
+		if (ptep_clear_flush_young(vma, address, pte))
 			continue;
 
 		/* Nuke the page table entry. */
virt/kvm/kvm_main.c +3 -2
index db57363..ea327f4 100644
@@ -367,7 +367,8 @@ static void kvm_mmu_notifier_invalidate_range_end(struct mmu_notifier *mn,
 
 static int kvm_mmu_notifier_clear_flush_young(struct mmu_notifier *mn,
 					      struct mm_struct *mm,
-					      unsigned long address)
+					      unsigned long start,
+					      unsigned long end)
 {
 	struct kvm *kvm = mmu_notifier_to_kvm(mn);
 	int young, idx;
@@ -375,7 +376,7 @@ static int kvm_mmu_notifier_clear_flush_young(struct mmu_notifier *mn,
 	idx = srcu_read_lock(&kvm->srcu);
 	spin_lock(&kvm->mmu_lock);
 
-	young = kvm_age_hva(kvm, address);
+	young = kvm_age_hva(kvm, start, end);
 	if (young)
 		kvm_flush_remote_tlbs(kvm);
 
Signature · 7 lines
2.1.0.rc2.206.gedb03e5

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to [email protected].  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"[email protected]"> [email protected] </a>
#2 lore raw

PATCH v3 1/1 kvm: Fix page ageing bugs

In reply to Andres Lagar-Cavilla

1. We were calling clear_flush_young_notify in unmap_one, but we are within an mmu notifier invalidate range scope. The spte exists no more (due to range_start) and the accessed bit info has already been propagated (due to kvm_pfn_set_accessed). Simply call clear_flush_young.

2. We clear_flush_young on a primary MMU PMD, but this may be mapped as a collection of PTEs by the secondary MMU (e.g. during log-dirty). This required expanding the interface of the clear_flush_young mmu notifier, so a lot of code has been trivially touched.

3. In the absence of shadow_accessed_mask (e.g. EPT A bit), we emulate the access bit by blowing the spte. This requires proper synchronizing with MMU notifier consumers, like every other removal of spte's does.

---
v1 -> v2: * Small cosmetic changes
* Remove unnecessary handling of mmu_notifier_seq in mmu_lock protected region
* Func prototypes fixed

v2 -> v3 * Added Acked-by
* Fix compilation in PowerPC
---

---
 arch/arm/include/asm/kvm_host.h     |  3 ++-
 arch/arm64/include/asm/kvm_host.h   |  3 ++-
 arch/powerpc/include/asm/kvm_host.h |  2 +-
 arch/powerpc/kvm/book3s.c           |  4 ++--
 arch/powerpc/kvm/book3s.h           |  3 ++-
 arch/powerpc/kvm/book3s_64_mmu_hv.c |  4 ++--
 arch/powerpc/kvm/book3s_pr.c        |  3 ++-
 arch/powerpc/kvm/e500_mmu_host.c    |  2 +-
 arch/x86/include/asm/kvm_host.h     |  2 +-
 arch/x86/kvm/mmu.c                  | 46 +++++++++++++++++++++++--------------
 drivers/iommu/amd_iommu_v2.c        |  6 +++--
 include/linux/mmu_notifier.h        | 24 +++++++++++++------
 include/trace/events/kvm.h          | 10 ++++----
 mm/mmu_notifier.c                   |  5 ++--
 mm/rmap.c                           |  6 ++++-
 virt/kvm/kvm_main.c                 |  5 ++--
 16 files changed, 81 insertions(+), 47 deletions(-)
arch/arm/include/asm/kvm_host.h +2 -1
index 032a853..8c3f7eb 100644
@@ -170,7 +170,8 @@ unsigned long kvm_arm_num_regs(struct kvm_vcpu *vcpu);
 int kvm_arm_copy_reg_indices(struct kvm_vcpu *vcpu, u64 __user *indices);
 
 /* We do not have shadow page tables, hence the empty hooks */
-static inline int kvm_age_hva(struct kvm *kvm, unsigned long hva)
+static inline int kvm_age_hva(struct kvm *kvm, unsigned long start,
+			      unsigned long end)
 {
 	return 0;
 }
arch/arm64/include/asm/kvm_host.h +2 -1
index be9970a..a3c671b 100644
@@ -180,7 +180,8 @@ int kvm_unmap_hva_range(struct kvm *kvm,
 void kvm_set_spte_hva(struct kvm *kvm, unsigned long hva, pte_t pte);
 
 /* We do not have shadow page tables, hence the empty hooks */
-static inline int kvm_age_hva(struct kvm *kvm, unsigned long hva)
+static inline int kvm_age_hva(struct kvm *kvm, unsigned long start,
+			      unsigned long end)
 {
 	return 0;
 }
arch/powerpc/include/asm/kvm_host.h +1 -1
index 6040008..d329bc5 100644
@@ -56,7 +56,7 @@
 extern int kvm_unmap_hva(struct kvm *kvm, unsigned long hva);
 extern int kvm_unmap_hva_range(struct kvm *kvm,
 			       unsigned long start, unsigned long end);
-extern int kvm_age_hva(struct kvm *kvm, unsigned long hva);
+extern int kvm_age_hva(struct kvm *kvm, unsigned long start, unsigned long end);
 extern int kvm_test_age_hva(struct kvm *kvm, unsigned long hva);
 extern void kvm_set_spte_hva(struct kvm *kvm, unsigned long hva, pte_t pte);
 
arch/powerpc/kvm/book3s.c +2 -2
index dd03f6b..c16cfbf 100644
@@ -851,9 +851,9 @@ int kvm_unmap_hva_range(struct kvm *kvm, unsigned long start, unsigned long end)
 	return kvm->arch.kvm_ops->unmap_hva_range(kvm, start, end);
 }
 
-int kvm_age_hva(struct kvm *kvm, unsigned long hva)
+int kvm_age_hva(struct kvm *kvm, unsigned long start, unsigned long end)
 {
-	return kvm->arch.kvm_ops->age_hva(kvm, hva);
+	return kvm->arch.kvm_ops->age_hva(kvm, start, end);
 }
 
 int kvm_test_age_hva(struct kvm *kvm, unsigned long hva)
arch/powerpc/kvm/book3s.h +2 -1
index 4bf956c..d2b3ec0 100644
@@ -17,7 +17,8 @@ extern void kvmppc_core_flush_memslot_hv(struct kvm *kvm,
 extern int kvm_unmap_hva_hv(struct kvm *kvm, unsigned long hva);
 extern int kvm_unmap_hva_range_hv(struct kvm *kvm, unsigned long start,
 				  unsigned long end);
-extern int kvm_age_hva_hv(struct kvm *kvm, unsigned long hva);
+extern int kvm_age_hva_hv(struct kvm *kvm, unsigned long start,
+			  unsigned long end);
 extern int kvm_test_age_hva_hv(struct kvm *kvm, unsigned long hva);
 extern void kvm_set_spte_hva_hv(struct kvm *kvm, unsigned long hva, pte_t pte);
 
arch/powerpc/kvm/book3s_64_mmu_hv.c +2 -2
index 72c20bb..81460c5 100644
@@ -1002,11 +1002,11 @@ static int kvm_age_rmapp(struct kvm *kvm, unsigned long *rmapp,
 	return ret;
 }
 
-int kvm_age_hva_hv(struct kvm *kvm, unsigned long hva)
+int kvm_age_hva_hv(struct kvm *kvm, unsigned long start, unsigned long end)
 {
 	if (!kvm->arch.using_mmu_notifiers)
 		return 0;
-	return kvm_handle_hva(kvm, hva, kvm_age_rmapp);
+	return kvm_handle_hva_range(kvm, start, end, kvm_age_rmapp);
 }
 
 static int kvm_test_age_rmapp(struct kvm *kvm, unsigned long *rmapp,
arch/powerpc/kvm/book3s_pr.c +2 -1
index faffb27..852fcd8 100644
@@ -295,7 +295,8 @@ static int kvm_unmap_hva_range_pr(struct kvm *kvm, unsigned long start,
 	return 0;
 }
 
-static int kvm_age_hva_pr(struct kvm *kvm, unsigned long hva)
+static int kvm_age_hva_pr(struct kvm *kvm, unsigned long start,
+			  unsigned long end)
 {
 	/* XXX could be more clever ;) */
 	return 0;
arch/powerpc/kvm/e500_mmu_host.c +1 -1
index 08f14bb..b1f3f63 100644
@@ -732,7 +732,7 @@ int kvm_unmap_hva_range(struct kvm *kvm, unsigned long start, unsigned long end)
 	return 0;
 }
 
-int kvm_age_hva(struct kvm *kvm, unsigned long hva)
+int kvm_age_hva(struct kvm *kvm, unsigned long start, unsigned long end)
 {
 	/* XXX could be more clever ;) */
 	return 0;
arch/x86/include/asm/kvm_host.h +1 -1
index 028df8d..a96c61b 100644
@@ -1036,7 +1036,7 @@ asmlinkage void kvm_spurious_fault(void);
 #define KVM_ARCH_WANT_MMU_NOTIFIER
 int kvm_unmap_hva(struct kvm *kvm, unsigned long hva);
 int kvm_unmap_hva_range(struct kvm *kvm, unsigned long start, unsigned long end);
-int kvm_age_hva(struct kvm *kvm, unsigned long hva);
+int kvm_age_hva(struct kvm *kvm, unsigned long start, unsigned long end);
 int kvm_test_age_hva(struct kvm *kvm, unsigned long hva);
 void kvm_set_spte_hva(struct kvm *kvm, unsigned long hva, pte_t pte);
 int cpuid_maxphyaddr(struct kvm_vcpu *vcpu);
arch/x86/kvm/mmu.c +29 -17

70 lines not on this page · raw message

drivers/iommu/amd_iommu_v2.c +4 -2
index 5f578e8..90d734b 100644
@@ -402,9 +402,11 @@ static void __mn_flush_page(struct mmu_notifier *mn,
 
 static int mn_clear_flush_young(struct mmu_notifier *mn,
 				struct mm_struct *mm,
-				unsigned long address)
+				unsigned long start,
+				unsigned long end)
 {
-	__mn_flush_page(mn, address);
+	for (; start < end; start += PAGE_SIZE)
+		__mn_flush_page(mn, start);
 
 	return 0;
 }
include/linux/mmu_notifier.h +17 -7
index 2728869..88787bb 100644
@@ -57,10 +57,13 @@ struct mmu_notifier_ops {
 	 * pte. This way the VM will provide proper aging to the
 	 * accesses to the page through the secondary MMUs and not
 	 * only to the ones through the Linux pte.
+	 * Start-end is necessary in case the secondary MMU is mapping the page
+	 * at a smaller granularity than the primary MMU.
 	 */
 	int (*clear_flush_young)(struct mmu_notifier *mn,
 				 struct mm_struct *mm,
-				 unsigned long address);
+				 unsigned long start,
+				 unsigned long end);
 
 	/*
 	 * test_young is called to check the young/accessed bitflag in
@@ -175,7 +178,8 @@ extern void mmu_notifier_unregister_no_release(struct mmu_notifier *mn,
 extern void __mmu_notifier_mm_destroy(struct mm_struct *mm);
 extern void __mmu_notifier_release(struct mm_struct *mm);
 extern int __mmu_notifier_clear_flush_young(struct mm_struct *mm,
-					  unsigned long address);
+					  unsigned long start,
+					  unsigned long end);
 extern int __mmu_notifier_test_young(struct mm_struct *mm,
 				     unsigned long address);
 extern void __mmu_notifier_change_pte(struct mm_struct *mm,
@@ -194,10 +198,11 @@ static inline void mmu_notifier_release(struct mm_struct *mm)
 }
 
 static inline int mmu_notifier_clear_flush_young(struct mm_struct *mm,
-					  unsigned long address)
+					  unsigned long start,
+					  unsigned long end)
 {
 	if (mm_has_notifiers(mm))
-		return __mmu_notifier_clear_flush_young(mm, address);
+		return __mmu_notifier_clear_flush_young(mm, start, end);
 	return 0;
 }
 
@@ -255,7 +260,9 @@ static inline void mmu_notifier_mm_destroy(struct mm_struct *mm)
 	unsigned long ___address = __address;				\
 	__young = ptep_clear_flush_young(___vma, ___address, __ptep);	\
 	__young |= mmu_notifier_clear_flush_young(___vma->vm_mm,	\
-						  ___address);		\
+						  ___address,		\
+						  ___address +		\
+							PAGE_SIZE);	\
 	__young;							\
 })
 
@@ -266,7 +273,9 @@ static inline void mmu_notifier_mm_destroy(struct mm_struct *mm)
 	unsigned long ___address = __address;				\
 	__young = pmdp_clear_flush_young(___vma, ___address, __pmdp);	\
 	__young |= mmu_notifier_clear_flush_young(___vma->vm_mm,	\
-						  ___address);		\
+						  ___address,		\
+						  ___address +		\
+							PMD_SIZE);	\
 	__young;							\
 })
 
@@ -301,7 +310,8 @@ static inline void mmu_notifier_release(struct mm_struct *mm)
 }
 
 static inline int mmu_notifier_clear_flush_young(struct mm_struct *mm,
-					  unsigned long address)
+					  unsigned long start,
+					  unsigned long end)
 {
 	return 0;
 }
include/trace/events/kvm.h +5 -5
index ab679c3..f58ef59 100644
@@ -225,8 +225,8 @@ TRACE_EVENT(kvm_fpu,
 );
 
 TRACE_EVENT(kvm_age_page,
-	TP_PROTO(ulong hva, struct kvm_memory_slot *slot, int ref),
-	TP_ARGS(hva, slot, ref),
+	TP_PROTO(ulong gfn, struct kvm_memory_slot *slot, int ref),
+	TP_ARGS(gfn, slot, ref),
 
 	TP_STRUCT__entry(
 		__field(	u64,	hva		)
@@ -235,9 +235,9 @@ TRACE_EVENT(kvm_age_page,
 	),
 
 	TP_fast_assign(
-		__entry->hva		= hva;
-		__entry->gfn		=
-		  slot->base_gfn + ((hva - slot->userspace_addr) >> PAGE_SHIFT);
+		__entry->gfn		= gfn;
+		__entry->hva		= ((gfn - slot->base_gfn) >>
+					    PAGE_SHIFT) + slot->userspace_addr;
 		__entry->referenced	= ref;
 	),
 
mm/mmu_notifier.c +3 -2
index 950813b..2c8da98 100644
@@ -107,7 +107,8 @@ void __mmu_notifier_release(struct mm_struct *mm)
  * existed or not.
  */
 int __mmu_notifier_clear_flush_young(struct mm_struct *mm,
-					unsigned long address)
+					unsigned long start,
+					unsigned long end)
 {
 	struct mmu_notifier *mn;
 	int young = 0, id;
@@ -115,7 +116,7 @@ int __mmu_notifier_clear_flush_young(struct mm_struct *mm,
 	id = srcu_read_lock(&srcu);
 	hlist_for_each_entry_rcu(mn, &mm->mmu_notifier_mm->list, hlist) {
 		if (mn->ops->clear_flush_young)
-			young |= mn->ops->clear_flush_young(mn, mm, address);
+			young |= mn->ops->clear_flush_young(mn, mm, start, end);
 	}
 	srcu_read_unlock(&srcu, id);
 
mm/rmap.c +5 -1
index 3e8491c..66b075f 100644
@@ -1355,7 +1355,11 @@ static int try_to_unmap_cluster(unsigned long cursor, unsigned int *mapcount,
 			continue;	/* don't unmap */
 		}
 
-		if (ptep_clear_flush_young_notify(vma, address, pte))
+		/*
+		 * No need for _notify because we're called within an
+		 * mmu_notifier_invalidate_range_ {start|end} scope.
+		 */
+		if (ptep_clear_flush_young(vma, address, pte))
 			continue;
 
 		/* Nuke the page table entry. */
virt/kvm/kvm_main.c +3 -2
index db57363..ea327f4 100644
@@ -367,7 +367,8 @@ static void kvm_mmu_notifier_invalidate_range_end(struct mmu_notifier *mn,
 
 static int kvm_mmu_notifier_clear_flush_young(struct mmu_notifier *mn,
 					      struct mm_struct *mm,
-					      unsigned long address)
+					      unsigned long start,
+					      unsigned long end)
 {
 	struct kvm *kvm = mmu_notifier_to_kvm(mn);
 	int young, idx;
@@ -375,7 +376,7 @@ static int kvm_mmu_notifier_clear_flush_young(struct mmu_notifier *mn,
 	idx = srcu_read_lock(&kvm->srcu);
 	spin_lock(&kvm->mmu_lock);
 
-	young = kvm_age_hva(kvm, address);
+	young = kvm_age_hva(kvm, start, end);
 	if (young)
 		kvm_flush_remote_tlbs(kvm);
 
Signature · 7 lines
2.1.0.rc2.206.gedb03e5

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to [email protected].  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"[email protected]"> [email protected] </a>
maintainer
#3 lore raw

Il 22/09/2014 22:26, Andres Lagar-Cavilla ha scritto:

include/trace/events/kvm.h TRACE_EVENT(kvm_age_page, · line 239

Quoted 2 lines
include/trace/events/kvm.h +2 -0
+		__entry->gfn		= gfn;
+		__entry->hva		= ((gfn - slot->base_gfn) >>

This must be <<.

mm/rmap.c static int try_to_unmap_cluster(unsigned long cursor, unsigned int *mapcount, · line 1361

Quoted 6 lines
mm/rmap.c +5 -0
+					    PAGE_SHIFT) + slot->userspace_addr;
 
+		/*
+		 * No need for _notify because we're called within an
+		 * mmu_notifier_invalidate_range_ {start|end} scope.
+		 */

Why "called within"? It is try_to_unmap_cluster itself that calls mmu_notifier_invalidate_range_*, so "we're within an mmu_notifier_invalidate_range_start/end scope" sounds better, and it's also what you use in the commit message.

Paolo

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in the body to [email protected]. For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"[email protected]"> [email protected] </a>

#4 lore raw
Quoted 5 lines · Paolo Bonzini

Il 22/09/2014 22:26, Andres Lagar-Cavilla ha scritto:

include/trace/events/kvm.h TRACE_EVENT(kvm_age_page, · line 239

Quoted 2 lines
include/trace/events/kvm.h +2 -0
+             __entry->gfn            = gfn;
+             __entry->hva            = ((gfn - slot->base_gfn) >>

This must be <<.

Correct, thanks.

Quoted 8 lines

mm/rmap.c static int try_to_unmap_cluster(unsigned long cursor, unsigned int *mapcount, · line 1361

Quoted 7 lines
mm/rmap.c +5 -0
+                                         PAGE_SHIFT) + slot->userspace_addr;
 
+             /*
+              * No need for _notify because we're called within an
+              * mmu_notifier_invalidate_range_ {start|end} scope.
+              */
Quoted 4 lines

Why "called within"? It is try_to_unmap_cluster itself that calls mmu_notifier_invalidate_range_*, so "we're within an mmu_notifier_invalidate_range_start/end scope" sounds better, and it's also what you use in the commit message.

Also correct. V4...
Andres

Quoted 2 lines

Paolo

Signature · 7 lines
Andres Lagar-Cavilla | Google Kernel Team | [email protected]

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to [email protected].  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"[email protected]"> [email protected] </a>
#5 lore raw

PATCH v4 1/1 kvm: Fix page ageing bugs

In reply to Andres Lagar-Cavilla

1. We were calling clear_flush_young_notify in unmap_one, but we are within an mmu notifier invalidate range scope. The spte exists no more (due to range_start) and the accessed bit info has already been propagated (due to kvm_pfn_set_accessed). Simply call clear_flush_young.

2. We clear_flush_young on a primary MMU PMD, but this may be mapped as a collection of PTEs by the secondary MMU (e.g. during log-dirty). This required expanding the interface of the clear_flush_young mmu notifier, so a lot of code has been trivially touched.

3. In the absence of shadow_accessed_mask (e.g. EPT A bit), we emulate the access bit by blowing the spte. This requires proper synchronizing with MMU notifier consumers, like every other removal of spte's does.

---
v1 -> v2: * Small cosmetic changes
* Remove unnecessary handling of mmu_notifier_seq in mmu_lock protected region
* Func prototypes fixed

v2 -> v3 * Added Acked-by
* Fix compilation in PowerPC

v3 -> v4 * Fix page shift direction in trace routine
* Clarified comment in rmap.c
---

---
 arch/arm/include/asm/kvm_host.h     |  3 ++-
 arch/arm64/include/asm/kvm_host.h   |  3 ++-
 arch/powerpc/include/asm/kvm_host.h |  2 +-
 arch/powerpc/kvm/book3s.c           |  4 ++--
 arch/powerpc/kvm/book3s.h           |  3 ++-
 arch/powerpc/kvm/book3s_64_mmu_hv.c |  4 ++--
 arch/powerpc/kvm/book3s_pr.c        |  3 ++-
 arch/powerpc/kvm/e500_mmu_host.c    |  2 +-
 arch/x86/include/asm/kvm_host.h     |  2 +-
 arch/x86/kvm/mmu.c                  | 46 +++++++++++++++++++++++--------------
 drivers/iommu/amd_iommu_v2.c        |  6 +++--
 include/linux/mmu_notifier.h        | 24 +++++++++++++------
 include/trace/events/kvm.h          | 10 ++++----
 mm/mmu_notifier.c                   |  5 ++--
 mm/rmap.c                           |  6 ++++-
 virt/kvm/kvm_main.c                 |  5 ++--
 16 files changed, 81 insertions(+), 47 deletions(-)
arch/arm/include/asm/kvm_host.h +2 -1
index 032a853..8c3f7eb 100644
@@ -170,7 +170,8 @@ unsigned long kvm_arm_num_regs(struct kvm_vcpu *vcpu);
 int kvm_arm_copy_reg_indices(struct kvm_vcpu *vcpu, u64 __user *indices);
 
 /* We do not have shadow page tables, hence the empty hooks */
-static inline int kvm_age_hva(struct kvm *kvm, unsigned long hva)
+static inline int kvm_age_hva(struct kvm *kvm, unsigned long start,
+			      unsigned long end)
 {
 	return 0;
 }
arch/arm64/include/asm/kvm_host.h +2 -1
index be9970a..a3c671b 100644
@@ -180,7 +180,8 @@ int kvm_unmap_hva_range(struct kvm *kvm,
 void kvm_set_spte_hva(struct kvm *kvm, unsigned long hva, pte_t pte);
 
 /* We do not have shadow page tables, hence the empty hooks */
-static inline int kvm_age_hva(struct kvm *kvm, unsigned long hva)
+static inline int kvm_age_hva(struct kvm *kvm, unsigned long start,
+			      unsigned long end)
 {
 	return 0;
 }
arch/powerpc/include/asm/kvm_host.h +1 -1
index 6040008..d329bc5 100644
@@ -56,7 +56,7 @@
 extern int kvm_unmap_hva(struct kvm *kvm, unsigned long hva);
 extern int kvm_unmap_hva_range(struct kvm *kvm,
 			       unsigned long start, unsigned long end);
-extern int kvm_age_hva(struct kvm *kvm, unsigned long hva);
+extern int kvm_age_hva(struct kvm *kvm, unsigned long start, unsigned long end);
 extern int kvm_test_age_hva(struct kvm *kvm, unsigned long hva);
 extern void kvm_set_spte_hva(struct kvm *kvm, unsigned long hva, pte_t pte);
 
arch/powerpc/kvm/book3s.c +2 -2
index dd03f6b..c16cfbf 100644
@@ -851,9 +851,9 @@ int kvm_unmap_hva_range(struct kvm *kvm, unsigned long start, unsigned long end)
 	return kvm->arch.kvm_ops->unmap_hva_range(kvm, start, end);
 }
 
-int kvm_age_hva(struct kvm *kvm, unsigned long hva)
+int kvm_age_hva(struct kvm *kvm, unsigned long start, unsigned long end)
 {
-	return kvm->arch.kvm_ops->age_hva(kvm, hva);
+	return kvm->arch.kvm_ops->age_hva(kvm, start, end);
 }
 
 int kvm_test_age_hva(struct kvm *kvm, unsigned long hva)
arch/powerpc/kvm/book3s.h +2 -1
index 4bf956c..d2b3ec0 100644
@@ -17,7 +17,8 @@ extern void kvmppc_core_flush_memslot_hv(struct kvm *kvm,
 extern int kvm_unmap_hva_hv(struct kvm *kvm, unsigned long hva);
 extern int kvm_unmap_hva_range_hv(struct kvm *kvm, unsigned long start,
 				  unsigned long end);
-extern int kvm_age_hva_hv(struct kvm *kvm, unsigned long hva);
+extern int kvm_age_hva_hv(struct kvm *kvm, unsigned long start,
+			  unsigned long end);
 extern int kvm_test_age_hva_hv(struct kvm *kvm, unsigned long hva);
 extern void kvm_set_spte_hva_hv(struct kvm *kvm, unsigned long hva, pte_t pte);
 
arch/powerpc/kvm/book3s_64_mmu_hv.c +2 -2
index 72c20bb..81460c5 100644
@@ -1002,11 +1002,11 @@ static int kvm_age_rmapp(struct kvm *kvm, unsigned long *rmapp,
 	return ret;
 }
 
-int kvm_age_hva_hv(struct kvm *kvm, unsigned long hva)
+int kvm_age_hva_hv(struct kvm *kvm, unsigned long start, unsigned long end)
 {
 	if (!kvm->arch.using_mmu_notifiers)
 		return 0;
-	return kvm_handle_hva(kvm, hva, kvm_age_rmapp);
+	return kvm_handle_hva_range(kvm, start, end, kvm_age_rmapp);
 }
 
 static int kvm_test_age_rmapp(struct kvm *kvm, unsigned long *rmapp,
arch/powerpc/kvm/book3s_pr.c +2 -1
index faffb27..852fcd8 100644
@@ -295,7 +295,8 @@ static int kvm_unmap_hva_range_pr(struct kvm *kvm, unsigned long start,
 	return 0;
 }
 
-static int kvm_age_hva_pr(struct kvm *kvm, unsigned long hva)
+static int kvm_age_hva_pr(struct kvm *kvm, unsigned long start,
+			  unsigned long end)
 {
 	/* XXX could be more clever ;) */
 	return 0;
arch/powerpc/kvm/e500_mmu_host.c +1 -1
index 08f14bb..b1f3f63 100644
@@ -732,7 +732,7 @@ int kvm_unmap_hva_range(struct kvm *kvm, unsigned long start, unsigned long end)
 	return 0;
 }
 
-int kvm_age_hva(struct kvm *kvm, unsigned long hva)
+int kvm_age_hva(struct kvm *kvm, unsigned long start, unsigned long end)
 {
 	/* XXX could be more clever ;) */
 	return 0;
arch/x86/include/asm/kvm_host.h +1 -1
index 028df8d..a96c61b 100644
@@ -1036,7 +1036,7 @@ asmlinkage void kvm_spurious_fault(void);
 #define KVM_ARCH_WANT_MMU_NOTIFIER
 int kvm_unmap_hva(struct kvm *kvm, unsigned long hva);
 int kvm_unmap_hva_range(struct kvm *kvm, unsigned long start, unsigned long end);
-int kvm_age_hva(struct kvm *kvm, unsigned long hva);
+int kvm_age_hva(struct kvm *kvm, unsigned long start, unsigned long end);
 int kvm_test_age_hva(struct kvm *kvm, unsigned long hva);
 void kvm_set_spte_hva(struct kvm *kvm, unsigned long hva, pte_t pte);
 int cpuid_maxphyaddr(struct kvm_vcpu *vcpu);
arch/x86/kvm/mmu.c +29 -17

70 lines not on this page · raw message

drivers/iommu/amd_iommu_v2.c +4 -2
index 5f578e8..90d734b 100644
@@ -402,9 +402,11 @@ static void __mn_flush_page(struct mmu_notifier *mn,
 
 static int mn_clear_flush_young(struct mmu_notifier *mn,
 				struct mm_struct *mm,
-				unsigned long address)
+				unsigned long start,
+				unsigned long end)
 {
-	__mn_flush_page(mn, address);
+	for (; start < end; start += PAGE_SIZE)
+		__mn_flush_page(mn, start);
 
 	return 0;
 }
include/linux/mmu_notifier.h +17 -7
index 2728869..88787bb 100644
@@ -57,10 +57,13 @@ struct mmu_notifier_ops {
 	 * pte. This way the VM will provide proper aging to the
 	 * accesses to the page through the secondary MMUs and not
 	 * only to the ones through the Linux pte.
+	 * Start-end is necessary in case the secondary MMU is mapping the page
+	 * at a smaller granularity than the primary MMU.
 	 */
 	int (*clear_flush_young)(struct mmu_notifier *mn,
 				 struct mm_struct *mm,
-				 unsigned long address);
+				 unsigned long start,
+				 unsigned long end);
 
 	/*
 	 * test_young is called to check the young/accessed bitflag in
@@ -175,7 +178,8 @@ extern void mmu_notifier_unregister_no_release(struct mmu_notifier *mn,
 extern void __mmu_notifier_mm_destroy(struct mm_struct *mm);
 extern void __mmu_notifier_release(struct mm_struct *mm);
 extern int __mmu_notifier_clear_flush_young(struct mm_struct *mm,
-					  unsigned long address);
+					  unsigned long start,
+					  unsigned long end);
 extern int __mmu_notifier_test_young(struct mm_struct *mm,
 				     unsigned long address);
 extern void __mmu_notifier_change_pte(struct mm_struct *mm,
@@ -194,10 +198,11 @@ static inline void mmu_notifier_release(struct mm_struct *mm)
 }
 
 static inline int mmu_notifier_clear_flush_young(struct mm_struct *mm,
-					  unsigned long address)
+					  unsigned long start,
+					  unsigned long end)
 {
 	if (mm_has_notifiers(mm))
-		return __mmu_notifier_clear_flush_young(mm, address);
+		return __mmu_notifier_clear_flush_young(mm, start, end);
 	return 0;
 }
 
@@ -255,7 +260,9 @@ static inline void mmu_notifier_mm_destroy(struct mm_struct *mm)
 	unsigned long ___address = __address;				\
 	__young = ptep_clear_flush_young(___vma, ___address, __ptep);	\
 	__young |= mmu_notifier_clear_flush_young(___vma->vm_mm,	\
-						  ___address);		\
+						  ___address,		\
+						  ___address +		\
+							PAGE_SIZE);	\
 	__young;							\
 })
 
@@ -266,7 +273,9 @@ static inline void mmu_notifier_mm_destroy(struct mm_struct *mm)
 	unsigned long ___address = __address;				\
 	__young = pmdp_clear_flush_young(___vma, ___address, __pmdp);	\
 	__young |= mmu_notifier_clear_flush_young(___vma->vm_mm,	\
-						  ___address);		\
+						  ___address,		\
+						  ___address +		\
+							PMD_SIZE);	\
 	__young;							\
 })
 
@@ -301,7 +310,8 @@ static inline void mmu_notifier_release(struct mm_struct *mm)
 }
 
 static inline int mmu_notifier_clear_flush_young(struct mm_struct *mm,
-					  unsigned long address)
+					  unsigned long start,
+					  unsigned long end)
 {
 	return 0;
 }
include/trace/events/kvm.h +5 -5
index ab679c3..0d2de78 100644
@@ -225,8 +225,8 @@ TRACE_EVENT(kvm_fpu,
 );
 
 TRACE_EVENT(kvm_age_page,
-	TP_PROTO(ulong hva, struct kvm_memory_slot *slot, int ref),
-	TP_ARGS(hva, slot, ref),
+	TP_PROTO(ulong gfn, struct kvm_memory_slot *slot, int ref),
+	TP_ARGS(gfn, slot, ref),
 
 	TP_STRUCT__entry(
 		__field(	u64,	hva		)
@@ -235,9 +235,9 @@ TRACE_EVENT(kvm_age_page,
 	),
 
 	TP_fast_assign(
-		__entry->hva		= hva;
-		__entry->gfn		=
-		  slot->base_gfn + ((hva - slot->userspace_addr) >> PAGE_SHIFT);
+		__entry->gfn		= gfn;
+		__entry->hva		= ((gfn - slot->base_gfn) <<
+					    PAGE_SHIFT) + slot->userspace_addr;
 		__entry->referenced	= ref;
 	),
 
mm/mmu_notifier.c +3 -2
index 950813b..2c8da98 100644
@@ -107,7 +107,8 @@ void __mmu_notifier_release(struct mm_struct *mm)
  * existed or not.
  */
 int __mmu_notifier_clear_flush_young(struct mm_struct *mm,
-					unsigned long address)
+					unsigned long start,
+					unsigned long end)
 {
 	struct mmu_notifier *mn;
 	int young = 0, id;
@@ -115,7 +116,7 @@ int __mmu_notifier_clear_flush_young(struct mm_struct *mm,
 	id = srcu_read_lock(&srcu);
 	hlist_for_each_entry_rcu(mn, &mm->mmu_notifier_mm->list, hlist) {
 		if (mn->ops->clear_flush_young)
-			young |= mn->ops->clear_flush_young(mn, mm, address);
+			young |= mn->ops->clear_flush_young(mn, mm, start, end);
 	}
 	srcu_read_unlock(&srcu, id);
 
mm/rmap.c +5 -1
index 3e8491c..bc74e00 100644
@@ -1355,7 +1355,11 @@ static int try_to_unmap_cluster(unsigned long cursor, unsigned int *mapcount,
 			continue;	/* don't unmap */
 		}
 
-		if (ptep_clear_flush_young_notify(vma, address, pte))
+		/*
+		 * No need for _notify because we're within an
+		 * mmu_notifier_invalidate_range_ {start|end} scope.
+		 */
+		if (ptep_clear_flush_young(vma, address, pte))
 			continue;
 
 		/* Nuke the page table entry. */
virt/kvm/kvm_main.c +3 -2
index db57363..ea327f4 100644
@@ -367,7 +367,8 @@ static void kvm_mmu_notifier_invalidate_range_end(struct mmu_notifier *mn,
 
 static int kvm_mmu_notifier_clear_flush_young(struct mmu_notifier *mn,
 					      struct mm_struct *mm,
-					      unsigned long address)
+					      unsigned long start,
+					      unsigned long end)
 {
 	struct kvm *kvm = mmu_notifier_to_kvm(mn);
 	int young, idx;
@@ -375,7 +376,7 @@ static int kvm_mmu_notifier_clear_flush_young(struct mmu_notifier *mn,
 	idx = srcu_read_lock(&kvm->srcu);
 	spin_lock(&kvm->mmu_lock);
 
-	young = kvm_age_hva(kvm, address);
+	young = kvm_age_hva(kvm, start, end);
 	if (young)
 		kvm_flush_remote_tlbs(kvm);
 
Signature · 7 lines
2.1.0.rc2.206.gedb03e5

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to [email protected].  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"[email protected]"> [email protected] </a>
maintainer
#6 lore raw

Il 22/09/2014 23:54, Andres Lagar-Cavilla ha scritto:

Quoted 23 lines
arch/x86/kvm/mmu.c +3 -12
@@ -1406,32 +1406,24 @@ static int kvm_age_rmapp(struct kvm *kvm, unsigned long *rmapp,
 	struct rmap_iterator uninitialized_var(iter);
 	int young = 0;
 
-	/*
-	 * In case of absence of EPT Access and Dirty Bits supports,
-	 * emulate the accessed bit for EPT, by checking if this page has
-	 * an EPT mapping, and clearing it if it does. On the next access,
-	 * a new EPT mapping will be established.
-	 * This has some overhead, but not as much as the cost of swapping
-	 * out actively used pages or breaking up actively used hugepages.
-	 */
-	if (!shadow_accessed_mask) {
-		young = kvm_unmap_rmapp(kvm, rmapp, slot, data);
-		goto out;
-	}
+	BUG_ON(!shadow_accessed_mask);
 
 	for (sptep = rmap_get_first(*rmapp, &iter); sptep;
 	     sptep = rmap_get_next(&iter)) {
+		struct kvm_mmu_page *sp;
+		gfn_t gfn;
 		BUG_ON(!is_shadow_present_pte(*sptep));

arch/x86/kvm/mmu.c static int kvm_age_rmapp(struct kvm *kvm, unsigned long *rmapp, · line 1425

Quoted 10 lines
arch/x86/kvm/mmu.c +4 -0
+		/* From spte to gfn. */
+		sp = page_header(__pa(sptep));
+		gfn = kvm_mmu_page_get_gfn(sp, sptep - sp->spt);
 
 		if (*sptep & shadow_accessed_mask) {
 			young = 1;
 			clear_bit((ffs(shadow_accessed_mask) - 1),
 				 (unsigned long *)sptep);
 		}
+		trace_kvm_age_page(gfn, slot, young);

Yesterday I couldn't think of a way to avoid the page_header/kvm_mmu_page_get_gfn on every iteration, but it's actually not hard. Instead of passing hva as datum, you can pass (unsigned long) &start. Then you can add PAGE_SIZE to it at the end of every call to kvm_age_rmapp, and keep the old tracing logic.

Paolo

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in the body to [email protected]. For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"[email protected]"> [email protected] </a>

#7 lore raw
Quoted 28 lines · Paolo Bonzini

Il 22/09/2014 23:54, Andres Lagar-Cavilla ha scritto:

Quoted 27 lines
arch/x86/kvm/mmu.c +6 -12
@@ -1406,32 +1406,24 @@ static int kvm_age_rmapp(struct kvm *kvm, unsigned long *rmapp,
      struct rmap_iterator uninitialized_var(iter);
      int young = 0;
 
-     /*
-      * In case of absence of EPT Access and Dirty Bits supports,
-      * emulate the accessed bit for EPT, by checking if this page has
-      * an EPT mapping, and clearing it if it does. On the next access,
-      * a new EPT mapping will be established.
-      * This has some overhead, but not as much as the cost of swapping
-      * out actively used pages or breaking up actively used hugepages.
-      */
-     if (!shadow_accessed_mask) {
-             young = kvm_unmap_rmapp(kvm, rmapp, slot, data);
-             goto out;
-     }
+     BUG_ON(!shadow_accessed_mask);
 
      for (sptep = rmap_get_first(*rmapp, &iter); sptep;
           sptep = rmap_get_next(&iter)) {
+             struct kvm_mmu_page *sp;
+             gfn_t gfn;
              BUG_ON(!is_shadow_present_pte(*sptep));
+             /* From spte to gfn. */
+             sp = page_header(__pa(sptep));
+             gfn = kvm_mmu_page_get_gfn(sp, sptep - sp->spt);
 
Quoted 12 lines

arch/x86/kvm/mmu.c static int kvm_age_rmapp(struct kvm *kvm, unsigned long *rmapp, · line 1425

Quoted 6 lines
arch/x86/kvm/mmu.c +1 -0
              if (*sptep & shadow_accessed_mask) {
                      young = 1;
                      clear_bit((ffs(shadow_accessed_mask) - 1),
                               (unsigned long *)sptep);
              }
+             trace_kvm_age_page(gfn, slot, young);

Yesterday I couldn't think of a way to avoid the page_header/kvm_mmu_page_get_gfn on every iteration, but it's actually not hard. Instead of passing hva as datum, you can pass (unsigned long) &start. Then you can add PAGE_SIZE to it at the end of every call to kvm_age_rmapp, and keep the old tracing logic.

I'm not sure. The addition is not always by PAGE_SIZE, since it depends on the current level we are iterating at in the outer kvm_handle_hva_range(). IOW, could be PMD_SIZE or even PUD_SIZE, and is_large_pte() enough to tell?

This is probably worth a general fix, I can see all the callbacks benefiting from knowing the gfn (passed down by kvm_handle_hva_range()) without any additional computation, and adding that to a tracing call if they don't already.

Even passing the level down to the callback would help by cutting down to one arithmetic op (subtract rmapp from slot rmap base pointer for that level)

Andres

Quoted 3 lines

Paolo

Signature · 7 lines
Andres Lagar-Cavilla | Google Kernel Team | [email protected]

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to [email protected].  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"[email protected]"> [email protected] </a>
maintainer
#8 lore raw

Il 23/09/2014 19:04, Andres Lagar-Cavilla ha scritto:

Quoted 5 lines

I'm not sure. The addition is not always by PAGE_SIZE, since it depends on the current level we are iterating at in the outer kvm_handle_hva_range(). IOW, could be PMD_SIZE or even PUD_SIZE, and is_large_pte() enough to tell?

Quoted 8 lines

This is probably worth a general fix, I can see all the callbacks benefiting from knowing the gfn (passed down by kvm_handle_hva_range()) without any additional computation, and adding that to a tracing call if they don't already.

Even passing the level down to the callback would help by cutting down to one arithmetic op (subtract rmapp from slot rmap base pointer for that level)

You're right. Let's apply this patch and work on that as a follow-up.

Paolo

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in the body to [email protected]. For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"[email protected]"> [email protected] </a>

#9 lore raw
In reply to Andres Lagar-Cavilla

Hi Andres,

Quoted 15 lines · Andres Lagar-Cavilla

1. We were calling clear_flush_young_notify in unmap_one, but we are within an mmu notifier invalidate range scope. The spte exists no more (due to range_start) and the accessed bit info has already been propagated (due to kvm_pfn_set_accessed). Simply call clear_flush_young.

2. We clear_flush_young on a primary MMU PMD, but this may be mapped as a collection of PTEs by the secondary MMU (e.g. during log-dirty). This required expanding the interface of the clear_flush_young mmu notifier, so a lot of code has been trivially touched.

3. In the absence of shadow_accessed_mask (e.g. EPT A bit), we emulate the access bit by blowing the spte. This requires proper synchronizing with MMU notifier consumers, like every other removal of spte's does.

[...]

Quoted 17 lines

---

arch/x86/kvm/mmu.c +1 -0
+	BUG_ON(!shadow_accessed_mask);
	for (sptep = rmap_get_first(*rmapp, &iter); sptep;
	     sptep = rmap_get_next(&iter)) {
arch/x86/kvm/mmu.c +2 -0
+		struct kvm_mmu_page *sp;
+		gfn_t gfn;
		BUG_ON(!is_shadow_present_pte(*sptep));
arch/x86/kvm/mmu.c +3 -0
+		/* From spte to gfn. */
+		sp = page_header(__pa(sptep));
+		gfn = kvm_mmu_page_get_gfn(sp, sptep - sp->spt);
		if (*sptep & shadow_accessed_mask) {
			young = 1;
			clear_bit((ffs(shadow_accessed_mask) - 1),
				 (unsigned long *)sptep);
		}

arch/x86/kvm/mmu.c static int kvm_age_rmapp(struct kvm *kvm, unsigned long *rmapp, · line 1425

Quoted 1 line
arch/x86/kvm/mmu.c +1 -0
+		trace_kvm_age_page(gfn, slot, young);

IIUC, all the rmapps in this for loop are against the same gfn which results in the above trace point dump the message duplicated.

Regards,
Wanpeng Li

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in the body to [email protected]. For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"[email protected]"> [email protected] </a>

maintainer
#10 lore raw

Il 24/09/2014 04:27, Wanpeng Li ha scritto:

Quoted 30 lines

Hi Andres,

Quoted 15 lines · Andres Lagar-Cavilla

1. We were calling clear_flush_young_notify in unmap_one, but we are within an mmu notifier invalidate range scope. The spte exists no more (due to range_start) and the accessed bit info has already been propagated (due to kvm_pfn_set_accessed). Simply call clear_flush_young.

2. We clear_flush_young on a primary MMU PMD, but this may be mapped as a collection of PTEs by the secondary MMU (e.g. during log-dirty). This required expanding the interface of the clear_flush_young mmu notifier, so a lot of code has been trivially touched.

3. In the absence of shadow_accessed_mask (e.g. EPT A bit), we emulate the access bit by blowing the spte. This requires proper synchronizing with MMU notifier consumers, like every other removal of spte's does.

[...]

Quoted 12 lines

---

arch/x86/kvm/mmu.c +1 -0
+	BUG_ON(!shadow_accessed_mask);
	for (sptep = rmap_get_first(*rmapp, &iter); sptep;
	     sptep = rmap_get_next(&iter)) {
arch/x86/kvm/mmu.c +2 -0
+		struct kvm_mmu_page *sp;
+		gfn_t gfn;
		BUG_ON(!is_shadow_present_pte(*sptep));
arch/x86/kvm/mmu.c +3 -0
+		/* From spte to gfn. */
+		sp = page_header(__pa(sptep));
+		gfn = kvm_mmu_page_get_gfn(sp, sptep - sp->spt);
Quoted 9 lines

arch/x86/kvm/mmu.c static int kvm_age_rmapp(struct kvm *kvm, unsigned long *rmapp, · line 1425

Quoted 6 lines
		if (*sptep & shadow_accessed_mask) {
			young = 1;
			clear_bit((ffs(shadow_accessed_mask) - 1),
				 (unsigned long *)sptep);
		}
arch/x86/kvm/mmu.c +1 -0
+		trace_kvm_age_page(gfn, slot, young);

IIUC, all the rmapps in this for loop are against the same gfn which results in the above trace point dump the message duplicated.

You're right; Andres's patch "[PATCH] kvm/x86/mmu: Pass gfn and level to rmapp callback" helps avoiding that.

Paolo

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in the body to [email protected]. For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"[email protected]"> [email protected] </a>

#11 lore raw

Hi Paolo,

Quoted 39 lines · Paolo Bonzini

Il 24/09/2014 04:27, Wanpeng Li ha scritto:

Quoted 30 lines

Hi Andres,

Quoted 15 lines · Andres Lagar-Cavilla

1. We were calling clear_flush_young_notify in unmap_one, but we are within an mmu notifier invalidate range scope. The spte exists no more (due to range_start) and the accessed bit info has already been propagated (due to kvm_pfn_set_accessed). Simply call clear_flush_young.

2. We clear_flush_young on a primary MMU PMD, but this may be mapped as a collection of PTEs by the secondary MMU (e.g. during log-dirty). This required expanding the interface of the clear_flush_young mmu notifier, so a lot of code has been trivially touched.

3. In the absence of shadow_accessed_mask (e.g. EPT A bit), we emulate the access bit by blowing the spte. This requires proper synchronizing with MMU notifier consumers, like every other removal of spte's does.

[...]

Quoted 12 lines

---

arch/x86/kvm/mmu.c +1 -0
+	BUG_ON(!shadow_accessed_mask);
	for (sptep = rmap_get_first(*rmapp, &iter); sptep;
	     sptep = rmap_get_next(&iter)) {
arch/x86/kvm/mmu.c +2 -0
+		struct kvm_mmu_page *sp;
+		gfn_t gfn;
		BUG_ON(!is_shadow_present_pte(*sptep));
arch/x86/kvm/mmu.c +3 -0
+		/* From spte to gfn. */
+		sp = page_header(__pa(sptep));
+		gfn = kvm_mmu_page_get_gfn(sp, sptep - sp->spt);
Quoted 8 lines

arch/x86/kvm/mmu.c static int kvm_age_rmapp(struct kvm *kvm, unsigned long *rmapp, · line 1425

Quoted 6 lines
		if (*sptep & shadow_accessed_mask) {
			young = 1;
			clear_bit((ffs(shadow_accessed_mask) - 1),
				 (unsigned long *)sptep);
		}
arch/x86/kvm/mmu.c +1 -0
+		trace_kvm_age_page(gfn, slot, young);

IIUC, all the rmapps in this for loop are against the same gfn which results in the above trace point dump the message duplicated.

Quoted 2 lines

You're right; Andres's patch "[PATCH] kvm/x86/mmu: Pass gfn and level to rmapp callback" helps avoiding that.

From Andres's patch "[PATCH] kvm/x86/mmu: Pass gfn and level to rmapp callback"

@@ -1410,25 +1421,20 @@ static int kvm_age_rmapp(struct kvm *kvm, unsigned long *rmapp,

for (sptep = rmap_get_first(*rmapp, &iter); sptep; sptep = rmap_get_next(&iter)) {
- struct kvm_mmu_page *sp;
- gfn_t gfn; BUG_ON(!is_shadow_present_pte(*sptep));
- /* From spte to gfn. */
- sp = page_header(__pa(sptep));
- gfn = kvm_mmu_page_get_gfn(sp, sptep - sp->spt); -
if (*sptep & shadow_accessed_mask) {
young = 1; clear_bit((ffs(shadow_accessed_mask) - 1), (unsigned long *)sptep);
}
- trace_kvm_age_page(gfn, slot, young);
+ trace_kvm_age_page(gfn, level, slot, young); }
return young;
}

This trace point still dup duplicated message for the same gfn in the for loop.

Regards,
Wanpeng Li

Quoted 7 lines

Paolo
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to [email protected]
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in the body to [email protected]. For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"[email protected]"> [email protected] </a>

maintainer
#12 lore raw

Il 24/09/2014 09:20, Wanpeng Li ha scritto:

Quoted 3 lines

This trace point still dup duplicated message for the same gfn in the for loop.

Yes, but the gfn argument lets you take it out again.

Note that having the duplicated trace would make sense if you included the accessed bit for each spte, instead of just the "young" variable.

Paolo

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in the body to [email protected]. For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"[email protected]"> [email protected] </a>

#13 lore raw
Quoted 9 lines · Paolo Bonzini

Il 24/09/2014 09:20, Wanpeng Li ha scritto:

Quoted 3 lines

This trace point still dup duplicated message for the same gfn in the for loop.

Yes, but the gfn argument lets you take it out again.

Note that having the duplicated trace would make sense if you included the accessed bit for each spte, instead of just the "young" variable.

FWIW the new arrangement in kvm.git/queue LGTM

Thanks
Andres

Quoted 2 lines

Paolo

Signature · 7 lines
Andres Lagar-Cavilla | Google Kernel Team | [email protected]

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to [email protected].  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"[email protected]"> [email protected] </a>