diff mbox series

[V3,5/5] arch/powerpc/mm/hugetlb: NestMMU workaround for hugetlb mprotect RW upgrade

Message ID 20181205030931.12037-6-aneesh.kumar@linux.ibm.com (mailing list archive)
State New, archived
Headers show
Series NestMMU pte upgrade workaround for mprotect | expand

Commit Message

Aneesh Kumar K.V Dec. 5, 2018, 3:09 a.m. UTC
NestMMU requires us to mark the pte invalid and flush the tlb when we do a
RW upgrade of pte. We fixed a variant of this in the fault path in commit
Fixes: bd5050e38aec ("powerpc/mm/radix: Change pte relax sequence to handle nest MMU hang")

Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.ibm.com>
---
 arch/powerpc/include/asm/book3s/64/hugetlb.h | 12 ++++++++
 arch/powerpc/mm/hugetlbpage-radix.c          | 17 ++++++++++++
 arch/powerpc/mm/hugetlbpage.c                | 29 ++++++++++++++++++++
 3 files changed, 58 insertions(+)

Comments

Christophe Leroy Dec. 5, 2018, 3:57 a.m. UTC | #1
Le 05/12/2018 à 04:09, Aneesh Kumar K.V a écrit :
> NestMMU requires us to mark the pte invalid and flush the tlb when we do a
> RW upgrade of pte. We fixed a variant of this in the fault path in commit
> Fixes: bd5050e38aec ("powerpc/mm/radix: Change pte relax sequence to handle nest MMU hang")
> 
> Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.ibm.com>
> ---
>   arch/powerpc/include/asm/book3s/64/hugetlb.h | 12 ++++++++
>   arch/powerpc/mm/hugetlbpage-radix.c          | 17 ++++++++++++
>   arch/powerpc/mm/hugetlbpage.c                | 29 ++++++++++++++++++++
>   3 files changed, 58 insertions(+)
> 
> diff --git a/arch/powerpc/include/asm/book3s/64/hugetlb.h b/arch/powerpc/include/asm/book3s/64/hugetlb.h
> index 5b0177733994..66c1e4f88d65 100644
> --- a/arch/powerpc/include/asm/book3s/64/hugetlb.h
> +++ b/arch/powerpc/include/asm/book3s/64/hugetlb.h
> @@ -13,6 +13,10 @@ radix__hugetlb_get_unmapped_area(struct file *file, unsigned long addr,
>   				unsigned long len, unsigned long pgoff,
>   				unsigned long flags);
>   
> +extern void radix__huge_ptep_modify_prot_commit(struct vm_area_struct *vma,
> +						unsigned long addr, pte_t *ptep,
> +						pte_t old_pte, pte_t pte);
> +
>   static inline int hstate_get_psize(struct hstate *hstate)
>   {
>   	unsigned long shift;
> @@ -42,4 +46,12 @@ static inline bool gigantic_page_supported(void)
>   /* hugepd entry valid bit */
>   #define HUGEPD_VAL_BITS		(0x8000000000000000UL)
>   
> +#define huge_ptep_modify_prot_start huge_ptep_modify_prot_start
> +extern pte_t huge_ptep_modify_prot_start(struct vm_area_struct *vma,
> +					 unsigned long addr, pte_t *ptep);
> +
> +#define huge_ptep_modify_prot_commit huge_ptep_modify_prot_commit
> +extern void huge_ptep_modify_prot_commit(struct vm_area_struct *vma,
> +					 unsigned long addr, pte_t *ptep,
> +					 pte_t old_pte, pte_t new_pte);
>   #endif
> diff --git a/arch/powerpc/mm/hugetlbpage-radix.c b/arch/powerpc/mm/hugetlbpage-radix.c
> index 2486bee0f93e..11d9ea28a816 100644
> --- a/arch/powerpc/mm/hugetlbpage-radix.c
> +++ b/arch/powerpc/mm/hugetlbpage-radix.c
> @@ -90,3 +90,20 @@ radix__hugetlb_get_unmapped_area(struct file *file, unsigned long addr,
>   
>   	return vm_unmapped_area(&info);
>   }
> +
> +void radix__huge_ptep_modify_prot_commit(struct vm_area_struct *vma,
> +					 unsigned long addr, pte_t *ptep,
> +					 pte_t old_pte, pte_t pte)
> +{
> +	struct mm_struct *mm = vma->vm_mm;
> +
> +	/*
> +	 * To avoid NMMU hang while relaxing access we need to flush the tlb before
> +	 * we set the new value.
> +	 */
> +	if (is_pte_rw_upgrade(pte_val(old_pte), pte_val(pte)) &&
> +	    (atomic_read(&mm->context.copros) > 0))
> +		radix__flush_hugetlb_page(vma, addr);
> +
> +	set_huge_pte_at(vma->vm_mm, addr, ptep, pte);
> +}
> diff --git a/arch/powerpc/mm/hugetlbpage.c b/arch/powerpc/mm/hugetlbpage.c
> index 8cf035e68378..39d33a3d0dc6 100644
> --- a/arch/powerpc/mm/hugetlbpage.c
> +++ b/arch/powerpc/mm/hugetlbpage.c
> @@ -912,3 +912,32 @@ int gup_hugepte(pte_t *ptep, unsigned long sz, unsigned long addr,
>   
>   	return 1;
>   }
> +
> +#ifdef CONFIG_PPC_BOOK3S_64

Could this go in hugetlbpage-hash64.c instead to avoid the #ifdef sequence ?

Christophe

> +pte_t huge_ptep_modify_prot_start(struct vm_area_struct *vma,
> +				  unsigned long addr, pte_t *ptep)
> +{
> +	unsigned long pte_val;
> +	/*
> +	 * Clear the _PAGE_PRESENT so that no hardware parallel update is
> +	 * possible. Also keep the pte_present true so that we don't take
> +	 * wrong fault.
> +	 */
> +	pte_val = pte_update(vma->vm_mm, addr, ptep,
> +			     _PAGE_PRESENT, _PAGE_INVALID, 1);
> +
> +	return __pte(pte_val);
> +}
> +EXPORT_SYMBOL(huge_ptep_modify_prot_start);
> +
> +void huge_ptep_modify_prot_commit(struct vm_area_struct *vma, unsigned long addr,
> +				  pte_t *ptep, pte_t old_pte, pte_t pte)
> +{
> +
> +	if (radix_enabled())
> +		return radix__huge_ptep_modify_prot_commit(vma, addr, ptep,
> +							   old_pte, pte);
> +	set_huge_pte_at(vma->vm_mm, addr, ptep, pte);
> +}
> +EXPORT_SYMBOL(huge_ptep_modify_prot_commit);
> +#endif
>
Aneesh Kumar K.V Dec. 5, 2018, 4:09 a.m. UTC | #2
Christophe LEROY <christophe.leroy@c-s.fr> writes:
> diff --git a/arch/powerpc/mm/hugetlbpage.c b/arch/powerpc/mm/hugetlbpage.c
>> index 8cf035e68378..39d33a3d0dc6 100644
>> --- a/arch/powerpc/mm/hugetlbpage.c
>> +++ b/arch/powerpc/mm/hugetlbpage.c
>> @@ -912,3 +912,32 @@ int gup_hugepte(pte_t *ptep, unsigned long sz, unsigned long addr,
>>   
>>   	return 1;
>>   }
>> +
>> +#ifdef CONFIG_PPC_BOOK3S_64
>
> Could this go in hugetlbpage-hash64.c instead to avoid the #ifdef sequence ?
>

yes. I will send updated patch as reply

-aneesh
diff mbox series

Patch

diff --git a/arch/powerpc/include/asm/book3s/64/hugetlb.h b/arch/powerpc/include/asm/book3s/64/hugetlb.h
index 5b0177733994..66c1e4f88d65 100644
--- a/arch/powerpc/include/asm/book3s/64/hugetlb.h
+++ b/arch/powerpc/include/asm/book3s/64/hugetlb.h
@@ -13,6 +13,10 @@  radix__hugetlb_get_unmapped_area(struct file *file, unsigned long addr,
 				unsigned long len, unsigned long pgoff,
 				unsigned long flags);
 
+extern void radix__huge_ptep_modify_prot_commit(struct vm_area_struct *vma,
+						unsigned long addr, pte_t *ptep,
+						pte_t old_pte, pte_t pte);
+
 static inline int hstate_get_psize(struct hstate *hstate)
 {
 	unsigned long shift;
@@ -42,4 +46,12 @@  static inline bool gigantic_page_supported(void)
 /* hugepd entry valid bit */
 #define HUGEPD_VAL_BITS		(0x8000000000000000UL)
 
+#define huge_ptep_modify_prot_start huge_ptep_modify_prot_start
+extern pte_t huge_ptep_modify_prot_start(struct vm_area_struct *vma,
+					 unsigned long addr, pte_t *ptep);
+
+#define huge_ptep_modify_prot_commit huge_ptep_modify_prot_commit
+extern void huge_ptep_modify_prot_commit(struct vm_area_struct *vma,
+					 unsigned long addr, pte_t *ptep,
+					 pte_t old_pte, pte_t new_pte);
 #endif
diff --git a/arch/powerpc/mm/hugetlbpage-radix.c b/arch/powerpc/mm/hugetlbpage-radix.c
index 2486bee0f93e..11d9ea28a816 100644
--- a/arch/powerpc/mm/hugetlbpage-radix.c
+++ b/arch/powerpc/mm/hugetlbpage-radix.c
@@ -90,3 +90,20 @@  radix__hugetlb_get_unmapped_area(struct file *file, unsigned long addr,
 
 	return vm_unmapped_area(&info);
 }
+
+void radix__huge_ptep_modify_prot_commit(struct vm_area_struct *vma,
+					 unsigned long addr, pte_t *ptep,
+					 pte_t old_pte, pte_t pte)
+{
+	struct mm_struct *mm = vma->vm_mm;
+
+	/*
+	 * To avoid NMMU hang while relaxing access we need to flush the tlb before
+	 * we set the new value.
+	 */
+	if (is_pte_rw_upgrade(pte_val(old_pte), pte_val(pte)) &&
+	    (atomic_read(&mm->context.copros) > 0))
+		radix__flush_hugetlb_page(vma, addr);
+
+	set_huge_pte_at(vma->vm_mm, addr, ptep, pte);
+}
diff --git a/arch/powerpc/mm/hugetlbpage.c b/arch/powerpc/mm/hugetlbpage.c
index 8cf035e68378..39d33a3d0dc6 100644
--- a/arch/powerpc/mm/hugetlbpage.c
+++ b/arch/powerpc/mm/hugetlbpage.c
@@ -912,3 +912,32 @@  int gup_hugepte(pte_t *ptep, unsigned long sz, unsigned long addr,
 
 	return 1;
 }
+
+#ifdef CONFIG_PPC_BOOK3S_64
+pte_t huge_ptep_modify_prot_start(struct vm_area_struct *vma,
+				  unsigned long addr, pte_t *ptep)
+{
+	unsigned long pte_val;
+	/*
+	 * Clear the _PAGE_PRESENT so that no hardware parallel update is
+	 * possible. Also keep the pte_present true so that we don't take
+	 * wrong fault.
+	 */
+	pte_val = pte_update(vma->vm_mm, addr, ptep,
+			     _PAGE_PRESENT, _PAGE_INVALID, 1);
+
+	return __pte(pte_val);
+}
+EXPORT_SYMBOL(huge_ptep_modify_prot_start);
+
+void huge_ptep_modify_prot_commit(struct vm_area_struct *vma, unsigned long addr,
+				  pte_t *ptep, pte_t old_pte, pte_t pte)
+{
+
+	if (radix_enabled())
+		return radix__huge_ptep_modify_prot_commit(vma, addr, ptep,
+							   old_pte, pte);
+	set_huge_pte_at(vma->vm_mm, addr, ptep, pte);
+}
+EXPORT_SYMBOL(huge_ptep_modify_prot_commit);
+#endif