All of lore.kernel.org
 help / color / mirror / Atom feed
* [PATCH] xen/gntdev: Grant maps should not be subject to NUMA balancing
@ 2015-11-10 19:22 Boris Ostrovsky
  2015-11-10 19:36 ` Konrad Rzeszutek Wilk
  0 siblings, 1 reply; 5+ messages in thread
From: Boris Ostrovsky @ 2015-11-10 19:22 UTC (permalink / raw)
  To: konrad.wilk, david.vrabel; +Cc: xen-devel, linux-kernel, boris.ostrovsky

Doing so will cause the grant to be unmapped and then, during
fault handling, the fault to be mistakenly treated as NUMA hint
fault.

In addition, even if we those maps could partcipate in NUMA
balancing it wouldn't provide any benefit since we wouldn't be
able to determine physical page's node (even if/when VNUMA is
implemented).

Marking grant maps' VMAs as VM_IO will exclude them from being
part of NUMA balancing.

Signed-off-by: Boris Ostrovsky <boris.ostrovsky@oracle.com>
---
 drivers/xen/gntdev.c |    2 +-
 1 files changed, 1 insertions(+), 1 deletions(-)

diff --git a/drivers/xen/gntdev.c b/drivers/xen/gntdev.c
index 67b9163..bf312df 100644
--- a/drivers/xen/gntdev.c
+++ b/drivers/xen/gntdev.c
@@ -802,7 +802,7 @@ static int gntdev_mmap(struct file *flip, struct vm_area_struct *vma)
 
 	vma->vm_ops = &gntdev_vmops;
 
-	vma->vm_flags |= VM_DONTEXPAND | VM_DONTDUMP;
+	vma->vm_flags |= VM_DONTEXPAND | VM_DONTDUMP | VM_IO;
 
 	if (use_ptemod)
 		vma->vm_flags |= VM_DONTCOPY;
-- 
1.7.1


^ permalink raw reply related	[flat|nested] 5+ messages in thread

* [PATCH] xen/gntdev: Grant maps should not be subject to NUMA balancing
@ 2015-11-10 19:22 Boris Ostrovsky
  0 siblings, 0 replies; 5+ messages in thread
From: Boris Ostrovsky @ 2015-11-10 19:22 UTC (permalink / raw)
  To: konrad.wilk, david.vrabel; +Cc: boris.ostrovsky, linux-kernel, xen-devel

Doing so will cause the grant to be unmapped and then, during
fault handling, the fault to be mistakenly treated as NUMA hint
fault.

In addition, even if we those maps could partcipate in NUMA
balancing it wouldn't provide any benefit since we wouldn't be
able to determine physical page's node (even if/when VNUMA is
implemented).

Marking grant maps' VMAs as VM_IO will exclude them from being
part of NUMA balancing.

Signed-off-by: Boris Ostrovsky <boris.ostrovsky@oracle.com>
---
 drivers/xen/gntdev.c |    2 +-
 1 files changed, 1 insertions(+), 1 deletions(-)

diff --git a/drivers/xen/gntdev.c b/drivers/xen/gntdev.c
index 67b9163..bf312df 100644
--- a/drivers/xen/gntdev.c
+++ b/drivers/xen/gntdev.c
@@ -802,7 +802,7 @@ static int gntdev_mmap(struct file *flip, struct vm_area_struct *vma)
 
 	vma->vm_ops = &gntdev_vmops;
 
-	vma->vm_flags |= VM_DONTEXPAND | VM_DONTDUMP;
+	vma->vm_flags |= VM_DONTEXPAND | VM_DONTDUMP | VM_IO;
 
 	if (use_ptemod)
 		vma->vm_flags |= VM_DONTCOPY;
-- 
1.7.1

^ permalink raw reply related	[flat|nested] 5+ messages in thread

* Re: [PATCH] xen/gntdev: Grant maps should not be subject to NUMA balancing
  2015-11-10 19:22 [PATCH] xen/gntdev: Grant maps should not be subject to NUMA balancing Boris Ostrovsky
@ 2015-11-10 19:36 ` Konrad Rzeszutek Wilk
  2015-11-10 19:54   ` Boris Ostrovsky
  2015-11-10 19:54   ` Boris Ostrovsky
  0 siblings, 2 replies; 5+ messages in thread
From: Konrad Rzeszutek Wilk @ 2015-11-10 19:36 UTC (permalink / raw)
  To: Boris Ostrovsky; +Cc: linux-kernel, david.vrabel, xen-devel

On Tue, Nov 10, 2015 at 02:22:44PM -0500, Boris Ostrovsky wrote:
> Doing so will cause the grant to be unmapped and then, during
> fault handling, the fault to be mistakenly treated as NUMA hint
> fault.
> 
> In addition, even if we those maps could partcipate in NUMA
> balancing it wouldn't provide any benefit since we wouldn't be
> able to determine physical page's node (even if/when VNUMA is
> implemented).
> 
> Marking grant maps' VMAs as VM_IO will exclude them from being
> part of NUMA balancing.
> 

Should this be 'cc:stable@vger.kernel.org' ?


> Signed-off-by: Boris Ostrovsky <boris.ostrovsky@oracle.com>
> ---
>  drivers/xen/gntdev.c |    2 +-
>  1 files changed, 1 insertions(+), 1 deletions(-)
> 
> diff --git a/drivers/xen/gntdev.c b/drivers/xen/gntdev.c
> index 67b9163..bf312df 100644
> --- a/drivers/xen/gntdev.c
> +++ b/drivers/xen/gntdev.c
> @@ -802,7 +802,7 @@ static int gntdev_mmap(struct file *flip, struct vm_area_struct *vma)
>  
>  	vma->vm_ops = &gntdev_vmops;
>  
> -	vma->vm_flags |= VM_DONTEXPAND | VM_DONTDUMP;
> +	vma->vm_flags |= VM_DONTEXPAND | VM_DONTDUMP | VM_IO;
>  
>  	if (use_ptemod)
>  		vma->vm_flags |= VM_DONTCOPY;
> -- 
> 1.7.1
> 

^ permalink raw reply	[flat|nested] 5+ messages in thread

* Re: [PATCH] xen/gntdev: Grant maps should not be subject to NUMA balancing
  2015-11-10 19:36 ` Konrad Rzeszutek Wilk
@ 2015-11-10 19:54   ` Boris Ostrovsky
  2015-11-10 19:54   ` Boris Ostrovsky
  1 sibling, 0 replies; 5+ messages in thread
From: Boris Ostrovsky @ 2015-11-10 19:54 UTC (permalink / raw)
  To: Konrad Rzeszutek Wilk; +Cc: david.vrabel, xen-devel, linux-kernel

On 11/10/2015 02:36 PM, Konrad Rzeszutek Wilk wrote:
> On Tue, Nov 10, 2015 at 02:22:44PM -0500, Boris Ostrovsky wrote:
>> Doing so will cause the grant to be unmapped and then, during
>> fault handling, the fault to be mistakenly treated as NUMA hint
>> fault.
>>
>> In addition, even if we those maps could partcipate in NUMA
>> balancing it wouldn't provide any benefit since we wouldn't be
>> able to determine physical page's node (even if/when VNUMA is
>> implemented).
>>
>> Marking grant maps' VMAs as VM_IO will exclude them from being
>> part of NUMA balancing.
>>
> Should this be 'cc:stable@vger.kernel.org' ?

This bug becomes a real problem with 4.4, now that NUMA balancing is on 
by default.

But we should probably include it in stable too. I'll resubmit it then 
with CC tag.

-boris

>
>
>> Signed-off-by: Boris Ostrovsky <boris.ostrovsky@oracle.com>
>> ---
>>   drivers/xen/gntdev.c |    2 +-
>>   1 files changed, 1 insertions(+), 1 deletions(-)
>>
>> diff --git a/drivers/xen/gntdev.c b/drivers/xen/gntdev.c
>> index 67b9163..bf312df 100644
>> --- a/drivers/xen/gntdev.c
>> +++ b/drivers/xen/gntdev.c
>> @@ -802,7 +802,7 @@ static int gntdev_mmap(struct file *flip, struct vm_area_struct *vma)
>>   
>>   	vma->vm_ops = &gntdev_vmops;
>>   
>> -	vma->vm_flags |= VM_DONTEXPAND | VM_DONTDUMP;
>> +	vma->vm_flags |= VM_DONTEXPAND | VM_DONTDUMP | VM_IO;
>>   
>>   	if (use_ptemod)
>>   		vma->vm_flags |= VM_DONTCOPY;
>> -- 
>> 1.7.1
>>


^ permalink raw reply	[flat|nested] 5+ messages in thread

* Re: [PATCH] xen/gntdev: Grant maps should not be subject to NUMA balancing
  2015-11-10 19:36 ` Konrad Rzeszutek Wilk
  2015-11-10 19:54   ` Boris Ostrovsky
@ 2015-11-10 19:54   ` Boris Ostrovsky
  1 sibling, 0 replies; 5+ messages in thread
From: Boris Ostrovsky @ 2015-11-10 19:54 UTC (permalink / raw)
  To: Konrad Rzeszutek Wilk; +Cc: linux-kernel, david.vrabel, xen-devel

On 11/10/2015 02:36 PM, Konrad Rzeszutek Wilk wrote:
> On Tue, Nov 10, 2015 at 02:22:44PM -0500, Boris Ostrovsky wrote:
>> Doing so will cause the grant to be unmapped and then, during
>> fault handling, the fault to be mistakenly treated as NUMA hint
>> fault.
>>
>> In addition, even if we those maps could partcipate in NUMA
>> balancing it wouldn't provide any benefit since we wouldn't be
>> able to determine physical page's node (even if/when VNUMA is
>> implemented).
>>
>> Marking grant maps' VMAs as VM_IO will exclude them from being
>> part of NUMA balancing.
>>
> Should this be 'cc:stable@vger.kernel.org' ?

This bug becomes a real problem with 4.4, now that NUMA balancing is on 
by default.

But we should probably include it in stable too. I'll resubmit it then 
with CC tag.

-boris

>
>
>> Signed-off-by: Boris Ostrovsky <boris.ostrovsky@oracle.com>
>> ---
>>   drivers/xen/gntdev.c |    2 +-
>>   1 files changed, 1 insertions(+), 1 deletions(-)
>>
>> diff --git a/drivers/xen/gntdev.c b/drivers/xen/gntdev.c
>> index 67b9163..bf312df 100644
>> --- a/drivers/xen/gntdev.c
>> +++ b/drivers/xen/gntdev.c
>> @@ -802,7 +802,7 @@ static int gntdev_mmap(struct file *flip, struct vm_area_struct *vma)
>>   
>>   	vma->vm_ops = &gntdev_vmops;
>>   
>> -	vma->vm_flags |= VM_DONTEXPAND | VM_DONTDUMP;
>> +	vma->vm_flags |= VM_DONTEXPAND | VM_DONTDUMP | VM_IO;
>>   
>>   	if (use_ptemod)
>>   		vma->vm_flags |= VM_DONTCOPY;
>> -- 
>> 1.7.1
>>

^ permalink raw reply	[flat|nested] 5+ messages in thread

end of thread, other threads:[~2015-11-10 19:55 UTC | newest]

Thread overview: 5+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2015-11-10 19:22 [PATCH] xen/gntdev: Grant maps should not be subject to NUMA balancing Boris Ostrovsky
2015-11-10 19:36 ` Konrad Rzeszutek Wilk
2015-11-10 19:54   ` Boris Ostrovsky
2015-11-10 19:54   ` Boris Ostrovsky
  -- strict thread matches above, loose matches on Subject: below --
2015-11-10 19:22 Boris Ostrovsky

This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.