[PATCH V2] mm: mempolicy: skip non-migratable VMAs when setting MPOL_MF_LAZY

[PATCH V2] mm: mempolicy: skip non-migratable VMAs when setting MPOL_MF_LAZY

[Liang Chen]

MPOL_MF_LAZY is not visible from userspace since 'commit a720094ded8c
("mm: mempolicy: Hide MPOL_NOOP and MPOL_MF_LAZY from userspace for now")'
, but it should still skip non-migratable VMAs such as VM_IO, VM_PFNMAP,
and VM_HUGETLB VMAs, and avoid useless overhead of minor faults.

Signed-off-by: Liang Chen <liangchen.linux@gmail.com>
Signed-off-by: Gavin Guo <gavin.guo@canonical.com>
---
Changes since v2:
- Add more description into the changelog

We have been evaluating the enablement of MPOL_MF_LAZY again, and found
this issue. And we decided to push this patch upstream no matter if we
finally determine to propose re-enablement of MPOL_MF_LAZY or not. Since
it can be a potential problem even if MPOL_MF_LAZY is not enabled this
time.
---
 mm/mempolicy.c | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)

diff --git a/mm/mempolicy.c b/mm/mempolicy.c
index 87a1779..436ff411 100644
--- a/mm/mempolicy.c
+++ b/mm/mempolicy.c
@@ -610,7 +610,8 @@ static int queue_pages_test_walk(unsigned long start, unsigned long end,
 
 	if (flags & MPOL_MF_LAZY) {
 		/* Similar to task_numa_work, skip inaccessible VMAs */
-		if (vma->vm_flags & (VM_READ | VM_EXEC | VM_WRITE))
+		if (vma_migratable(vma) &&
+			vma->vm_flags & (VM_READ | VM_EXEC | VM_WRITE))
 			change_prot_numa(vma, start, endvma);
 		return 1;
 	}
-- 
1.9.1

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

Re: [PATCH V2] mm: mempolicy: skip non-migratable VMAs when setting MPOL_MF_LAZY

[Naoya Horiguchi]

Hello Liang,

On Thu, Jan 07, 2016 at 11:52:38AM +0800, Liang Chen wrote:
> MPOL_MF_LAZY is not visible from userspace since 'commit a720094ded8c
> ("mm: mempolicy: Hide MPOL_NOOP and MPOL_MF_LAZY from userspace for now")'
> , but it should still skip non-migratable VMAs such as VM_IO, VM_PFNMAP,
> and VM_HUGETLB VMAs, and avoid useless overhead of minor faults.
> 
> Signed-off-by: Liang Chen <liangchen.linux@gmail.com>
> Signed-off-by: Gavin Guo <gavin.guo@canonical.com>
> ---
> Changes since v2:
> - Add more description into the changelog
> 
> We have been evaluating the enablement of MPOL_MF_LAZY again, and found
> this issue. And we decided to push this patch upstream no matter if we
> finally determine to propose re-enablement of MPOL_MF_LAZY or not. Since
> it can be a potential problem even if MPOL_MF_LAZY is not enabled this
> time.
> ---
>  mm/mempolicy.c | 3 ++-
>  1 file changed, 2 insertions(+), 1 deletion(-)
> 
> diff --git a/mm/mempolicy.c b/mm/mempolicy.c
> index 87a1779..436ff411 100644
> --- a/mm/mempolicy.c
> +++ b/mm/mempolicy.c
> @@ -610,7 +610,8 @@ static int queue_pages_test_walk(unsigned long start, unsigned long end,
>  
>  	if (flags & MPOL_MF_LAZY) {
>  		/* Similar to task_numa_work, skip inaccessible VMAs */
> -		if (vma->vm_flags & (VM_READ | VM_EXEC | VM_WRITE))
> +		if (vma_migratable(vma) &&
> +			vma->vm_flags & (VM_READ | VM_EXEC | VM_WRITE))
>  			change_prot_numa(vma, start, endvma);
>  		return 1;
>  	}

task_numa_work() does more vma checks before entering change_prot_numa() like
vma_policy_mof(), is_vm_hugetlb_page(), and (vma->vm_flags & VM_MIXEDMAP).
So is it better to use the same check set to limit the target vmas to auto-numa
enabled ones?

Thanks,
Naoya Horiguchi
--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>

Re: [PATCH V2] mm: mempolicy: skip non-migratable VMAs when setting MPOL_MF_LAZY

[Liang Chen]

Hi Naoya,

Yeah. Thanks for the reminding！

vma_policy_mof doesn't need to be checked because with MPOL_MF_LAZY
do_mbind always sets the MPOL_F_MOF flag.
VM_HUGETLB and VM_MIXEDMAP vma should be excluded to avoid compound
pages being marked for migration and unexpected COWs when handling
hugetlb fault.

I will send a patch to add these check soon.

Thanks,
Liang

On Wed, Jan 13, 2016 at 4:16 PM, Naoya Horiguchi
<n-horiguchi@ah.jp.nec.com> wrote:
> Hello Liang,
>
> On Thu, Jan 07, 2016 at 11:52:38AM +0800, Liang Chen wrote:
>> MPOL_MF_LAZY is not visible from userspace since 'commit a720094ded8c
>> ("mm: mempolicy: Hide MPOL_NOOP and MPOL_MF_LAZY from userspace for now")'
>> , but it should still skip non-migratable VMAs such as VM_IO, VM_PFNMAP,
>> and VM_HUGETLB VMAs, and avoid useless overhead of minor faults.
>>
>> Signed-off-by: Liang Chen <liangchen.linux@gmail.com>
>> Signed-off-by: Gavin Guo <gavin.guo@canonical.com>
>> ---
>> Changes since v2:
>> - Add more description into the changelog
>>
>> We have been evaluating the enablement of MPOL_MF_LAZY again, and found
>> this issue. And we decided to push this patch upstream no matter if we
>> finally determine to propose re-enablement of MPOL_MF_LAZY or not. Since
>> it can be a potential problem even if MPOL_MF_LAZY is not enabled this
>> time.
>> ---
>>  mm/mempolicy.c | 3 ++-
>>  1 file changed, 2 insertions(+), 1 deletion(-)
>>
>> diff --git a/mm/mempolicy.c b/mm/mempolicy.c
>> index 87a1779..436ff411 100644
>> --- a/mm/mempolicy.c
>> +++ b/mm/mempolicy.c
>> @@ -610,7 +610,8 @@ static int queue_pages_test_walk(unsigned long start, unsigned long end,
>>
>>       if (flags & MPOL_MF_LAZY) {
>>               /* Similar to task_numa_work, skip inaccessible VMAs */
>> -             if (vma->vm_flags & (VM_READ | VM_EXEC | VM_WRITE))
>> +             if (vma_migratable(vma) &&
>> +                     vma->vm_flags & (VM_READ | VM_EXEC | VM_WRITE))
>>                       change_prot_numa(vma, start, endvma);
>>               return 1;
>>       }
>
> task_numa_work() does more vma checks before entering change_prot_numa() like
> vma_policy_mof(), is_vm_hugetlb_page(), and (vma->vm_flags & VM_MIXEDMAP).
> So is it better to use the same check set to limit the target vmas to auto-numa
> enabled ones?
>
> Thanks,
> Naoya Horiguchi

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