Re: [PATCH V4 5/5] arch/powerpc/mm/hugetlb: NestMMU workaround for hugetlb mprotect RW upgrade
On Thu, Dec 20, 2018 at 11:30:12AM +1100, Benjamin Herrenschmidt wrote:
> On Wed, 2018-12-19 at 08:50 +0530, Aneesh Kumar K.V wrote:
> > Christoph Hellwig writes:
> >
> > > On Tue, Dec 18, 2018 at 03:11:37PM +0530, Aneesh Kumar K.V wrote:
> > > > +EXPORT_SYMBOL(huge_ptep_modify_prot_start);
> > >
> > > The only user of this function is the one you added in the last patch
> > > in mm/hugetlb.c, so there is no need to export this function.
> > >
> > > > +
> > > > +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);
> > >
> > > Same here.
> >
> > That was done considering that ptep_modify_prot_start/commit was defined
> > in asm-generic/pgtable.h. I was trying to make sure I didn't break
> > anything with the patch. Also s390 do have that EXPORT_SYMBOL() for the
> > same. hugetlb just inherited that.
> >
> > If you feel strongly about it, I can drop the EXPORT_SYMBOL().
>
> At the very least it should be _GPL
In general yes, but in this case it just needs to go. The s390
maintainers already removed a patch to remove their
ptep_modify_prot_start / ptep_modify_prot_commit exports.
Re: [PATCH V4 5/5] arch/powerpc/mm/hugetlb: NestMMU workaround for hugetlb mprotect RW upgrade
On Wed, 2018-12-19 at 08:50 +0530, Aneesh Kumar K.V wrote:
> Christoph Hellwig writes:
>
> > On Tue, Dec 18, 2018 at 03:11:37PM +0530, Aneesh Kumar K.V wrote:
> > > +EXPORT_SYMBOL(huge_ptep_modify_prot_start);
> >
> > The only user of this function is the one you added in the last patch
> > in mm/hugetlb.c, so there is no need to export this function.
> >
> > > +
> > > +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);
> >
> > Same here.
>
> That was done considering that ptep_modify_prot_start/commit was defined
> in asm-generic/pgtable.h. I was trying to make sure I didn't break
> anything with the patch. Also s390 do have that EXPORT_SYMBOL() for the
> same. hugetlb just inherited that.
>
> If you feel strongly about it, I can drop the EXPORT_SYMBOL().
At the very least it should be _GPL
Cheers,
Ben.
Re: [PATCH V4 5/5] arch/powerpc/mm/hugetlb: NestMMU workaround for hugetlb mprotect RW upgrade
On Wed, Dec 19, 2018 at 08:50:57AM +0530, Aneesh Kumar K.V wrote: > That was done considering that ptep_modify_prot_start/commit was defined > in asm-generic/pgtable.h. I was trying to make sure I didn't break > anything with the patch. Also s390 do have that EXPORT_SYMBOL() for the > same. hugetlb just inherited that. > > If you feel strongly about it, I can drop the EXPORT_SYMBOL(). Yes. And we should probably remove the s390 as well as it isn't used either.
Re: [PATCH V4 5/5] arch/powerpc/mm/hugetlb: NestMMU workaround for hugetlb mprotect RW upgrade
Christoph Hellwig writes:
> On Tue, Dec 18, 2018 at 03:11:37PM +0530, Aneesh Kumar K.V wrote:
>> +EXPORT_SYMBOL(huge_ptep_modify_prot_start);
>
> The only user of this function is the one you added in the last patch
> in mm/hugetlb.c, so there is no need to export this function.
>
>> +
>> +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);
>
> Same here.
That was done considering that ptep_modify_prot_start/commit was defined
in asm-generic/pgtable.h. I was trying to make sure I didn't break
anything with the patch. Also s390 do have that EXPORT_SYMBOL() for the
same. hugetlb just inherited that.
If you feel strongly about it, I can drop the EXPORT_SYMBOL().
-aneesh
Re: [PATCH V4 5/5] arch/powerpc/mm/hugetlb: NestMMU workaround for hugetlb mprotect RW upgrade
On Tue, Dec 18, 2018 at 03:11:37PM +0530, Aneesh Kumar K.V wrote:
> +EXPORT_SYMBOL(huge_ptep_modify_prot_start);
The only user of this function is the one you added in the last patch
in mm/hugetlb.c, so there is no need to export this function.
> +
> +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);
Same here.
[PATCH V4 5/5] arch/powerpc/mm/hugetlb: NestMMU workaround for hugetlb mprotect RW upgrade
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
---
arch/powerpc/include/asm/book3s/64/hugetlb.h | 12 +
arch/powerpc/mm/hugetlbpage-hash64.c | 27
arch/powerpc/mm/hugetlbpage-radix.c | 17
3 files changed, 56 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(0x8000UL)
+#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-hash64.c
b/arch/powerpc/mm/hugetlbpage-hash64.c
index 2e6a8f9345d3..48fe74bfeab1 100644
--- a/arch/powerpc/mm/hugetlbpage-hash64.c
+++ b/arch/powerpc/mm/hugetlbpage-hash64.c
@@ -121,3 +121,30 @@ int __hash_page_huge(unsigned long ea, unsigned long
access, unsigned long vsid,
*ptep = __pte(new_pte & ~H_PAGE_BUSY);
return 0;
}
+
+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);
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();
}
+
+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(>context.copros) > 0))
+ radix__flush_hugetlb_page(vma, addr);
+
+ set_huge_pte_at(vma->vm_mm, addr, ptep, pte);
+}
--
2.19.2
