[PATCH V2 0/3] powerpc: Fix random application crashes with NUMA_BALANCING enabled

[PATCH V2 0/3] powerpc: Fix random application crashes with NUMA_BALANCING enabled

[Aneesh Kumar K.V]

Hello,

This patch series fix random application crashes observed on ppc64 with numa
balancing enabled. Without the patch we see crashes like

anacron[14551]: unhandled signal 11 at 0000000000000041 nip 000000003cfd54b4 lr 000000003cfd5464 code 30001
anacron[14599]: unhandled signal 11 at 0000000000000041 nip 000000003efc54b4 lr 000000003efc5464 code 30001

Changes from V1:
* Build fix for CONFIG_NUMA_BALANCING disabled

-aneesh

[PATCH V2 1/3] powerpc: mm: Add new set flag argument to pte/pmd update function

[Aneesh Kumar K.V]

From: "Aneesh Kumar K.V" <aneesh.kumar@linux.vnet.ibm.com>

We will use this later to set the _PAGE_NUMA bit.

Acked-by: Mel Gorman <mgorman@suse.de>
Acked-by: Rik van Riel <riel@redhat.com>
Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
---
 arch/powerpc/include/asm/hugetlb.h       |  2 +-
 arch/powerpc/include/asm/pgtable-ppc64.h | 26 +++++++++++++++-----------
 arch/powerpc/mm/pgtable_64.c             | 12 +++++++-----
 arch/powerpc/mm/subpage-prot.c           |  2 +-
 4 files changed, 24 insertions(+), 18 deletions(-)

diff --git a/arch/powerpc/include/asm/hugetlb.h b/arch/powerpc/include/asm/hugetlb.h
index d750336b171d..623f2971ce0e 100644
--- a/arch/powerpc/include/asm/hugetlb.h
+++ b/arch/powerpc/include/asm/hugetlb.h
@@ -127,7 +127,7 @@ static inline pte_t huge_ptep_get_and_clear(struct mm_struct *mm,
 					    unsigned long addr, pte_t *ptep)
 {
 #ifdef CONFIG_PPC64
-	return __pte(pte_update(mm, addr, ptep, ~0UL, 1));
+	return __pte(pte_update(mm, addr, ptep, ~0UL, 0, 1));
 #else
 	return __pte(pte_update(ptep, ~0UL, 0));
 #endif
diff --git a/arch/powerpc/include/asm/pgtable-ppc64.h b/arch/powerpc/include/asm/pgtable-ppc64.h
index bc141c950b1e..eb9261024f51 100644
--- a/arch/powerpc/include/asm/pgtable-ppc64.h
+++ b/arch/powerpc/include/asm/pgtable-ppc64.h
@@ -195,6 +195,7 @@ extern void hpte_need_flush(struct mm_struct *mm, unsigned long addr,
 static inline unsigned long pte_update(struct mm_struct *mm,
 				       unsigned long addr,
 				       pte_t *ptep, unsigned long clr,
+				       unsigned long set,
 				       int huge)
 {
 #ifdef PTE_ATOMIC_UPDATES
@@ -205,14 +206,15 @@ static inline unsigned long pte_update(struct mm_struct *mm,
 	andi.	%1,%0,%6\n\
 	bne-	1b \n\
 	andc	%1,%0,%4 \n\
+	or	%1,%1,%7\n\
 	stdcx.	%1,0,%3 \n\
 	bne-	1b"
 	: "=&r" (old), "=&r" (tmp), "=m" (*ptep)
-	: "r" (ptep), "r" (clr), "m" (*ptep), "i" (_PAGE_BUSY)
+	: "r" (ptep), "r" (clr), "m" (*ptep), "i" (_PAGE_BUSY), "r" (set)
 	: "cc" );
 #else
 	unsigned long old = pte_val(*ptep);
-	*ptep = __pte(old & ~clr);
+	*ptep = __pte((old & ~clr) | set);
 #endif
 	/* huge pages use the old page table lock */
 	if (!huge)
@@ -231,9 +233,9 @@ static inline int __ptep_test_and_clear_young(struct mm_struct *mm,
 {
 	unsigned long old;
 
-       	if ((pte_val(*ptep) & (_PAGE_ACCESSED | _PAGE_HASHPTE)) == 0)
+	if ((pte_val(*ptep) & (_PAGE_ACCESSED | _PAGE_HASHPTE)) == 0)
 		return 0;
-	old = pte_update(mm, addr, ptep, _PAGE_ACCESSED, 0);
+	old = pte_update(mm, addr, ptep, _PAGE_ACCESSED, 0, 0);
 	return (old & _PAGE_ACCESSED) != 0;
 }
 #define __HAVE_ARCH_PTEP_TEST_AND_CLEAR_YOUNG
@@ -252,7 +254,7 @@ static inline void ptep_set_wrprotect(struct mm_struct *mm, unsigned long addr,
 	if ((pte_val(*ptep) & _PAGE_RW) == 0)
 		return;
 
-	pte_update(mm, addr, ptep, _PAGE_RW, 0);
+	pte_update(mm, addr, ptep, _PAGE_RW, 0, 0);
 }
 
 static inline void huge_ptep_set_wrprotect(struct mm_struct *mm,
@@ -261,7 +263,7 @@ static inline void huge_ptep_set_wrprotect(struct mm_struct *mm,
 	if ((pte_val(*ptep) & _PAGE_RW) == 0)
 		return;
 
-	pte_update(mm, addr, ptep, _PAGE_RW, 1);
+	pte_update(mm, addr, ptep, _PAGE_RW, 0, 1);
 }
 
 /*
@@ -284,14 +286,14 @@ static inline void huge_ptep_set_wrprotect(struct mm_struct *mm,
 static inline pte_t ptep_get_and_clear(struct mm_struct *mm,
 				       unsigned long addr, pte_t *ptep)
 {
-	unsigned long old = pte_update(mm, addr, ptep, ~0UL, 0);
+	unsigned long old = pte_update(mm, addr, ptep, ~0UL, 0, 0);
 	return __pte(old);
 }
 
 static inline void pte_clear(struct mm_struct *mm, unsigned long addr,
 			     pte_t * ptep)
 {
-	pte_update(mm, addr, ptep, ~0UL, 0);
+	pte_update(mm, addr, ptep, ~0UL, 0, 0);
 }
 
 
@@ -506,7 +508,9 @@ extern int pmdp_set_access_flags(struct vm_area_struct *vma,
 
 extern unsigned long pmd_hugepage_update(struct mm_struct *mm,
 					 unsigned long addr,
-					 pmd_t *pmdp, unsigned long clr);
+					 pmd_t *pmdp,
+					 unsigned long clr,
+					 unsigned long set);
 
 static inline int __pmdp_test_and_clear_young(struct mm_struct *mm,
 					      unsigned long addr, pmd_t *pmdp)
@@ -515,7 +519,7 @@ static inline int __pmdp_test_and_clear_young(struct mm_struct *mm,
 
 	if ((pmd_val(*pmdp) & (_PAGE_ACCESSED | _PAGE_HASHPTE)) == 0)
 		return 0;
-	old = pmd_hugepage_update(mm, addr, pmdp, _PAGE_ACCESSED);
+	old = pmd_hugepage_update(mm, addr, pmdp, _PAGE_ACCESSED, 0);
 	return ((old & _PAGE_ACCESSED) != 0);
 }
 
@@ -542,7 +546,7 @@ static inline void pmdp_set_wrprotect(struct mm_struct *mm, unsigned long addr,
 	if ((pmd_val(*pmdp) & _PAGE_RW) == 0)
 		return;
 
-	pmd_hugepage_update(mm, addr, pmdp, _PAGE_RW);
+	pmd_hugepage_update(mm, addr, pmdp, _PAGE_RW, 0);
 }
 
 #define __HAVE_ARCH_PMDP_SPLITTING_FLUSH
diff --git a/arch/powerpc/mm/pgtable_64.c b/arch/powerpc/mm/pgtable_64.c
index 65b7b65e8708..62bf5e8e78da 100644
--- a/arch/powerpc/mm/pgtable_64.c
+++ b/arch/powerpc/mm/pgtable_64.c
@@ -510,7 +510,8 @@ int pmdp_set_access_flags(struct vm_area_struct *vma, unsigned long address,
 }
 
 unsigned long pmd_hugepage_update(struct mm_struct *mm, unsigned long addr,
-				  pmd_t *pmdp, unsigned long clr)
+				  pmd_t *pmdp, unsigned long clr,
+				  unsigned long set)
 {
 
 	unsigned long old, tmp;
@@ -526,14 +527,15 @@ unsigned long pmd_hugepage_update(struct mm_struct *mm, unsigned long addr,
 		andi.	%1,%0,%6\n\
 		bne-	1b \n\
 		andc	%1,%0,%4 \n\
+		or	%1,%1,%7\n\
 		stdcx.	%1,0,%3 \n\
 		bne-	1b"
 	: "=&r" (old), "=&r" (tmp), "=m" (*pmdp)
-	: "r" (pmdp), "r" (clr), "m" (*pmdp), "i" (_PAGE_BUSY)
+	: "r" (pmdp), "r" (clr), "m" (*pmdp), "i" (_PAGE_BUSY), "r" (set)
 	: "cc" );
 #else
 	old = pmd_val(*pmdp);
-	*pmdp = __pmd(old & ~clr);
+	*pmdp = __pmd((old & ~clr) | set);
 #endif
 	if (old & _PAGE_HASHPTE)
 		hpte_do_hugepage_flush(mm, addr, pmdp);
@@ -708,7 +710,7 @@ void set_pmd_at(struct mm_struct *mm, unsigned long addr,
 void pmdp_invalidate(struct vm_area_struct *vma, unsigned long address,
 		     pmd_t *pmdp)
 {
-	pmd_hugepage_update(vma->vm_mm, address, pmdp, _PAGE_PRESENT);
+	pmd_hugepage_update(vma->vm_mm, address, pmdp, _PAGE_PRESENT, 0);
 }
 
 /*
@@ -835,7 +837,7 @@ pmd_t pmdp_get_and_clear(struct mm_struct *mm,
 	unsigned long old;
 	pgtable_t *pgtable_slot;
 
-	old = pmd_hugepage_update(mm, addr, pmdp, ~0UL);
+	old = pmd_hugepage_update(mm, addr, pmdp, ~0UL, 0);
 	old_pmd = __pmd(old);
 	/*
 	 * We have pmd == none and we are holding page_table_lock.
diff --git a/arch/powerpc/mm/subpage-prot.c b/arch/powerpc/mm/subpage-prot.c
index a770df2dae70..6c0b1f5f8d2c 100644
--- a/arch/powerpc/mm/subpage-prot.c
+++ b/arch/powerpc/mm/subpage-prot.c
@@ -78,7 +78,7 @@ static void hpte_flush_range(struct mm_struct *mm, unsigned long addr,
 	pte = pte_offset_map_lock(mm, pmd, addr, &ptl);
 	arch_enter_lazy_mmu_mode();
 	for (; npages > 0; --npages) {
-		pte_update(mm, addr, pte, 0, 0);
+		pte_update(mm, addr, pte, 0, 0, 0);
 		addr += PAGE_SIZE;
 		++pte;
 	}
-- 
1.8.3.2

[PATCH V2 2/3] mm: dirty accountable change only apply to non prot numa case

[Aneesh Kumar K.V]

From: "Aneesh Kumar K.V" <aneesh.kumar@linux.vnet.ibm.com>

So move it within the if loop

Acked-by: Mel Gorman <mgorman@suse.de>
Reviewed-by: Rik van Riel <riel@redhat.com>
Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
---
 mm/mprotect.c | 21 +++++++--------------
 1 file changed, 7 insertions(+), 14 deletions(-)

diff --git a/mm/mprotect.c b/mm/mprotect.c
index 7332c1785744..33eab902f10e 100644
--- a/mm/mprotect.c
+++ b/mm/mprotect.c
@@ -58,6 +58,13 @@ static unsigned long change_pte_range(struct vm_area_struct *vma, pmd_t *pmd,
 				if (pte_numa(ptent))
 					ptent = pte_mknonnuma(ptent);
 				ptent = pte_modify(ptent, newprot);
+				/*
+				 * Avoid taking write faults for pages we
+				 * know to be dirty.
+				 */
+				if (dirty_accountable && pte_dirty(ptent))
+					ptent = pte_mkwrite(ptent);
+				ptep_modify_prot_commit(mm, addr, pte, ptent);
 				updated = true;
 			} else {
 				struct page *page;
@@ -72,22 +79,8 @@ static unsigned long change_pte_range(struct vm_area_struct *vma, pmd_t *pmd,
 					}
 				}
 			}
-
-			/*
-			 * Avoid taking write faults for pages we know to be
-			 * dirty.
-			 */
-			if (dirty_accountable && pte_dirty(ptent)) {
-				ptent = pte_mkwrite(ptent);
-				updated = true;
-			}
-
 			if (updated)
 				pages++;
-
-			/* Only !prot_numa always clears the pte */
-			if (!prot_numa)
-				ptep_modify_prot_commit(mm, addr, pte, ptent);
 		} else if (IS_ENABLED(CONFIG_MIGRATION) && !pte_file(oldpte)) {
 			swp_entry_t entry = pte_to_swp_entry(oldpte);
 
-- 
1.8.3.2

[PATCH V2 3/3] mm: Use ptep/pmdp_set_numa for updating _PAGE_NUMA bit

[Aneesh Kumar K.V]

From: "Aneesh Kumar K.V" <aneesh.kumar@linux.vnet.ibm.com>

Archs like ppc64 doesn't do tlb flush in set_pte/pmd functions. ppc64 also doesn't implement
flush_tlb_range. ppc64 require the tlb flushing to be batched within ptl locks. The reason
to do that is to ensure that the hash page table is in sync with linux page table.
We track the hpte index in linux pte and if we clear them without flushing hash and drop the
ptl lock, we can have another cpu update the pte and can end up with double hash. We also want
to keep set_pte_at simpler by not requiring them to do hash flush for performance reason.
Hence cannot use them while updating _PAGE_NUMA bit. Add new functions for marking pte/pmd numa

Acked-by: Mel Gorman <mgorman@suse.de>
Reviewed-by: Rik van Riel <riel@redhat.com>
Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
---
Changes from V1:
 * Build fix for non numa balancing config
 
 arch/powerpc/include/asm/pgtable.h | 22 +++++++++++++++++++++
 include/asm-generic/pgtable.h      | 39 ++++++++++++++++++++++++++++++++++++++
 mm/huge_memory.c                   |  9 ++-------
 mm/mprotect.c                      |  4 +---
 4 files changed, 64 insertions(+), 10 deletions(-)

diff --git a/arch/powerpc/include/asm/pgtable.h b/arch/powerpc/include/asm/pgtable.h
index f83b6f3e1b39..3ebb188c3ff5 100644
--- a/arch/powerpc/include/asm/pgtable.h
+++ b/arch/powerpc/include/asm/pgtable.h
@@ -75,12 +75,34 @@ static inline pte_t pte_mknuma(pte_t pte)
 	return pte;
 }
 
+#define ptep_set_numa ptep_set_numa
+static inline void ptep_set_numa(struct mm_struct *mm, unsigned long addr,
+				 pte_t *ptep)
+{
+	if ((pte_val(*ptep) & _PAGE_PRESENT) == 0)
+		VM_BUG_ON(1);
+
+	pte_update(mm, addr, ptep, _PAGE_PRESENT, _PAGE_NUMA, 0);
+	return;
+}
+
 #define pmd_numa pmd_numa
 static inline int pmd_numa(pmd_t pmd)
 {
 	return pte_numa(pmd_pte(pmd));
 }
 
+#define pmdp_set_numa pmdp_set_numa
+static inline void pmdp_set_numa(struct mm_struct *mm, unsigned long addr,
+				 pmd_t *pmdp)
+{
+	if ((pmd_val(*pmdp) & _PAGE_PRESENT) == 0)
+		VM_BUG_ON(1);
+
+	pmd_hugepage_update(mm, addr, pmdp, _PAGE_PRESENT, _PAGE_NUMA);
+	return;
+}
+
 #define pmd_mknonnuma pmd_mknonnuma
 static inline pmd_t pmd_mknonnuma(pmd_t pmd)
 {
diff --git a/include/asm-generic/pgtable.h b/include/asm-generic/pgtable.h
index 8e4f41d9af4d..34c7bdc06014 100644
--- a/include/asm-generic/pgtable.h
+++ b/include/asm-generic/pgtable.h
@@ -701,6 +701,18 @@ static inline pte_t pte_mknuma(pte_t pte)
 }
 #endif
 
+#ifndef ptep_set_numa
+static inline void ptep_set_numa(struct mm_struct *mm, unsigned long addr,
+				 pte_t *ptep)
+{
+	pte_t ptent = *ptep;
+
+	ptent = pte_mknuma(ptent);
+	set_pte_at(mm, addr, ptep, ptent);
+	return;
+}
+#endif
+
 #ifndef pmd_mknuma
 static inline pmd_t pmd_mknuma(pmd_t pmd)
 {
@@ -708,6 +720,18 @@ static inline pmd_t pmd_mknuma(pmd_t pmd)
 	return pmd_clear_flags(pmd, _PAGE_PRESENT);
 }
 #endif
+
+#ifndef pmdp_set_numa
+static inline void pmdp_set_numa(struct mm_struct *mm, unsigned long addr,
+				 pmd_t *pmdp)
+{
+	pmd_t pmd = *pmdp;
+
+	pmd = pmd_mknuma(pmd);
+	set_pmd_at(mm, addr, pmdp, pmd);
+	return;
+}
+#endif
 #else
 extern int pte_numa(pte_t pte);
 extern int pmd_numa(pmd_t pmd);
@@ -715,6 +739,8 @@ extern pte_t pte_mknonnuma(pte_t pte);
 extern pmd_t pmd_mknonnuma(pmd_t pmd);
 extern pte_t pte_mknuma(pte_t pte);
 extern pmd_t pmd_mknuma(pmd_t pmd);
+extern void ptep_set_numa(struct mm_struct *mm, unsigned long addr, pte_t *ptep);
+extern void pmdp_set_numa(struct mm_struct *mm, unsigned long addr, pmd_t *pmdp);
 #endif /* CONFIG_ARCH_USES_NUMA_PROT_NONE */
 #else
 static inline int pmd_numa(pmd_t pmd)
@@ -742,10 +768,23 @@ static inline pte_t pte_mknuma(pte_t pte)
 	return pte;
 }
 
+static inline void ptep_set_numa(struct mm_struct *mm, unsigned long addr,
+				 pte_t *ptep)
+{
+	return;
+}
+
+
 static inline pmd_t pmd_mknuma(pmd_t pmd)
 {
 	return pmd;
 }
+
+static inline void pmdp_set_numa(struct mm_struct *mm, unsigned long addr,
+				 pmd_t *pmdp)
+{
+	return ;
+}
 #endif /* CONFIG_NUMA_BALANCING */
 
 #endif /* CONFIG_MMU */
diff --git a/mm/huge_memory.c b/mm/huge_memory.c
index 82166bf974e1..da23eb96779f 100644
--- a/mm/huge_memory.c
+++ b/mm/huge_memory.c
@@ -1545,6 +1545,7 @@ int change_huge_pmd(struct vm_area_struct *vma, pmd_t *pmd,
 				entry = pmd_mknonnuma(entry);
 			entry = pmd_modify(entry, newprot);
 			ret = HPAGE_PMD_NR;
+			set_pmd_at(mm, addr, pmd, entry);
 			BUG_ON(pmd_write(entry));
 		} else {
 			struct page *page = pmd_page(*pmd);
@@ -1557,16 +1558,10 @@ int change_huge_pmd(struct vm_area_struct *vma, pmd_t *pmd,
 			 */
 			if (!is_huge_zero_page(page) &&
 			    !pmd_numa(*pmd)) {
-				entry = *pmd;
-				entry = pmd_mknuma(entry);
+				pmdp_set_numa(mm, addr, pmd);
 				ret = HPAGE_PMD_NR;
 			}
 		}
-
-		/* Set PMD if cleared earlier */
-		if (ret == HPAGE_PMD_NR)
-			set_pmd_at(mm, addr, pmd, entry);
-
 		spin_unlock(ptl);
 	}
 
diff --git a/mm/mprotect.c b/mm/mprotect.c
index 33eab902f10e..769a67a15803 100644
--- a/mm/mprotect.c
+++ b/mm/mprotect.c
@@ -69,12 +69,10 @@ static unsigned long change_pte_range(struct vm_area_struct *vma, pmd_t *pmd,
 			} else {
 				struct page *page;
 
-				ptent = *pte;
 				page = vm_normal_page(vma, addr, oldpte);
 				if (page && !PageKsm(page)) {
 					if (!pte_numa(oldpte)) {
-						ptent = pte_mknuma(ptent);
-						set_pte_at(mm, addr, pte, ptent);
+						ptep_set_numa(mm, addr, pte);
 						updated = true;
 					}
 				}
-- 
1.8.3.2

Re: [PATCH V2 0/3] powerpc: Fix random application crashes with NUMA_BALANCING enabled

[Andrew Morton]

On Wed, 12 Feb 2014 09:13:35 +0530 "Aneesh Kumar K.V" <aneesh.kumar@linux.vnet.ibm.com> wrote:

> Hello,
> 
> This patch series fix random application crashes observed on ppc64 with numa
> balancing enabled. Without the patch we see crashes like
> 
> anacron[14551]: unhandled signal 11 at 0000000000000041 nip 000000003cfd54b4 lr 000000003cfd5464 code 30001
> anacron[14599]: unhandled signal 11 at 0000000000000041 nip 000000003efc54b4 lr 000000003efc5464 code 30001
> 

Random application crashes are bad.  Which kernel version(s) do you think
need fixing here?

I grabbed the patches but would like to hear from Ben (or something
approximating him) before doing anything with them, please.

Re: [PATCH V2 0/3] powerpc: Fix random application crashes with NUMA_BALANCING enabled

[Aneesh Kumar K.V]

Andrew Morton <akpm@linux-foundation.org> writes:

> On Wed, 12 Feb 2014 09:13:35 +0530 "Aneesh Kumar K.V" <aneesh.kumar@linux.vnet.ibm.com> wrote:
>
>> Hello,
>> 
>> This patch series fix random application crashes observed on ppc64 with numa
>> balancing enabled. Without the patch we see crashes like
>> 
>> anacron[14551]: unhandled signal 11 at 0000000000000041 nip 000000003cfd54b4 lr 000000003cfd5464 code 30001
>> anacron[14599]: unhandled signal 11 at 0000000000000041 nip 000000003efc54b4 lr 000000003efc5464 code 30001
>> 
>
> Random application crashes are bad.  Which kernel version(s) do you think
> need fixing here?
>
> I grabbed the patches but would like to hear from Ben (or something
> approximating him) before doing anything with them, please.
>

Considering this impact only ppc64 and also only when numa balancing is enabled, we
only need to send this upstream. (no need to backport to any other
kernel versions)

We merged numa balancing support for ppc64
(c34a51ce49b40b9667cd7f5cc2e40475af8b4c3d) only in this merge window.

$git describe --contains c34a51ce49b40b9667cd7f5cc2e40475af8b4c3d
v3.14-rc1~80^2~35


-aneesh

Re: [PATCH V2 0/3] powerpc: Fix random application crashes with NUMA_BALANCING enabled

[Benjamin Herrenschmidt]

On Thu, 2014-02-13 at 15:06 -0800, Andrew Morton wrote:
> On Wed, 12 Feb 2014 09:13:35 +0530 "Aneesh Kumar K.V" <aneesh.kumar@linux.vnet.ibm.com> wrote:
> 
> > Hello,
> > 
> > This patch series fix random application crashes observed on ppc64 with numa
> > balancing enabled. Without the patch we see crashes like
> > 
> > anacron[14551]: unhandled signal 11 at 0000000000000041 nip 000000003cfd54b4 lr 000000003cfd5464 code 30001
> > anacron[14599]: unhandled signal 11 at 0000000000000041 nip 000000003efc54b4 lr 000000003efc5464 code 30001
> > 
> 
> Random application crashes are bad.  Which kernel version(s) do you think
> need fixing here?
> 
> I grabbed the patches but would like to hear from Ben (or something
> approximating him) before doing anything with them, please.

Ah good. Did you grab v2 ? v1 had a compile breakage. I was about to
send them to Linus today as well but then got distracted by a sick
child, so I'm happy for you to pick them up and send them to the
boss :-)

Cheers,
Ben.