[2/6] i915/gvt/kvm: a NULL ->mm does not mean a thread is a kthread
diff mbox series

Message ID 20200404094101.672954-3-hch@lst.de
State New
Headers show
Series
  • [1/6] amdgpu: a NULL ->mm does not mean a thread is a kthread
Related show

Commit Message

Christoph Hellwig April 4, 2020, 9:40 a.m. UTC
Use the proper API instead.

Fixes: f440c8a572d7 ("drm/i915/gvt/kvmgt: read/write GPA via KVM API")
Signed-off-by: Christoph Hellwig <hch@lst.de>
---
 drivers/gpu/drm/i915/gvt/kvmgt.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

Comments

Sergei Shtylyov April 4, 2020, 10:05 a.m. UTC | #1
Hello!

On 04.04.2020 12:40, Christoph Hellwig wrote:

> Use the proper API instead.
> 
> Fixes: f440c8a572d7 ("drm/i915/gvt/kvmgt: read/write GPA via KVM API")
> Signed-off-by: Christoph Hellwig <hch@lst.de>
> ---
>   drivers/gpu/drm/i915/gvt/kvmgt.c | 2 +-
>   1 file changed, 1 insertion(+), 1 deletion(-)
> 
> diff --git a/drivers/gpu/drm/i915/gvt/kvmgt.c b/drivers/gpu/drm/i915/gvt/kvmgt.c
> index 074c4efb58eb..5848400620b4 100644
> --- a/drivers/gpu/drm/i915/gvt/kvmgt.c
> +++ b/drivers/gpu/drm/i915/gvt/kvmgt.c
> @@ -2037,7 +2037,7 @@ static int kvmgt_rw_gpa(unsigned long handle, unsigned long gpa,
>   	struct kvmgt_guest_info *info;
>   	struct kvm *kvm;
>   	int idx, ret;
> -	bool kthread = current->mm == NULL;
> +	bool kthread = (current->flags & PF_KTHREAD);

    Don't need the parens.

[...]

MBR, Sergei
Yan Zhao April 7, 2020, 3:08 a.m. UTC | #2
On Sat, Apr 04, 2020 at 11:40:57AM +0200, Christoph Hellwig wrote:
> Use the proper API instead.
> 
> Fixes: f440c8a572d7 ("drm/i915/gvt/kvmgt: read/write GPA via KVM API")
> Signed-off-by: Christoph Hellwig <hch@lst.de>
> ---
>  drivers/gpu/drm/i915/gvt/kvmgt.c | 2 +-
>  1 file changed, 1 insertion(+), 1 deletion(-)
> 
> diff --git a/drivers/gpu/drm/i915/gvt/kvmgt.c b/drivers/gpu/drm/i915/gvt/kvmgt.c
> index 074c4efb58eb..5848400620b4 100644
> --- a/drivers/gpu/drm/i915/gvt/kvmgt.c
> +++ b/drivers/gpu/drm/i915/gvt/kvmgt.c
> @@ -2037,7 +2037,7 @@ static int kvmgt_rw_gpa(unsigned long handle, unsigned long gpa,
>  	struct kvmgt_guest_info *info;
>  	struct kvm *kvm;
>  	int idx, ret;
> -	bool kthread = current->mm == NULL;
> +	bool kthread = (current->flags & PF_KTHREAD);
>  
>  	if (!handle_valid(handle))
>  		return -ESRCH;
> -- 
> 2.25.1
>
hi
we were removing this code. see
https://lore.kernel.org/kvm/20200313031109.7989-1-yan.y.zhao@intel.com/

The implementation of vfio_dma_rw() has been in vfio next tree.
https://github.com/awilliam/linux-vfio/commit/8d46c0cca5f4dc0538173d62cd36b1119b5105bc

in vfio_dma_rw(),  we still use
bool kthread = current->mm == NULL.
because if current->mm != NULL and current->flags & PF_KTHREAD, instead
of calling use_mm(), we first check if (current->mm == mm) and allow copy_to_user() if it's true.

Do you think it's all right?

Thanks
Yan



> _______________________________________________
> intel-gvt-dev mailing list
> intel-gvt-dev@lists.freedesktop.org
> https://lists.freedesktop.org/mailman/listinfo/intel-gvt-dev
Christoph Hellwig April 13, 2020, 1:27 p.m. UTC | #3
On Mon, Apr 06, 2020 at 11:08:46PM -0400, Yan Zhao wrote:
> hi
> we were removing this code. see
> https://lore.kernel.org/kvm/20200313031109.7989-1-yan.y.zhao@intel.com/

This didn't make 5.7-rc1.

> The implementation of vfio_dma_rw() has been in vfio next tree.
> https://github.com/awilliam/linux-vfio/commit/8d46c0cca5f4dc0538173d62cd36b1119b5105bc


This made 5.7-rc1, so I'll update the series to take it into account.

T
> in vfio_dma_rw(),  we still use
> bool kthread = current->mm == NULL.
> because if current->mm != NULL and current->flags & PF_KTHREAD, instead
> of calling use_mm(), we first check if (current->mm == mm) and allow copy_to_user() if it's true.
> 
> Do you think it's all right?

I can't think of another way for a kernel thread to have a mm indeed.
Yan Zhao April 14, 2020, 12:04 a.m. UTC | #4
On Mon, Apr 13, 2020 at 03:27:30PM +0200, Christoph Hellwig wrote:
> On Mon, Apr 06, 2020 at 11:08:46PM -0400, Yan Zhao wrote:
> > hi
> > we were removing this code. see
> > https://lore.kernel.org/kvm/20200313031109.7989-1-yan.y.zhao@intel.com/
> 
> This didn't make 5.7-rc1.
> 
> > The implementation of vfio_dma_rw() has been in vfio next tree.
> > https://github.com/awilliam/linux-vfio/commit/8d46c0cca5f4dc0538173d62cd36b1119b5105bc
> 
> 
> This made 5.7-rc1, so I'll update the series to take it into account.
> 
> T
> > in vfio_dma_rw(),  we still use
> > bool kthread = current->mm == NULL.
> > because if current->mm != NULL and current->flags & PF_KTHREAD, instead
> > of calling use_mm(), we first check if (current->mm == mm) and allow copy_to_user() if it's true.
> > 
> > Do you think it's all right?
> 
> I can't think of another way for a kernel thread to have a mm indeed.
for example, before calling to vfio_dma_rw(), a kernel thread has already
called use_mm(), then its current->mm is not null, and it has flag
PF_KTHREAD.
in this case, we just want to allow the copy_to_user() directly if
current->mm == mm, rather than call another use_mm() again.

do you think it makes sense?

Thanks
Yan

> _______________________________________________
> intel-gvt-dev mailing list
> intel-gvt-dev@lists.freedesktop.org
> https://lists.freedesktop.org/mailman/listinfo/intel-gvt-dev
Christoph Hellwig April 14, 2020, 7 a.m. UTC | #5
On Mon, Apr 13, 2020 at 08:04:10PM -0400, Yan Zhao wrote:
> > I can't think of another way for a kernel thread to have a mm indeed.
> for example, before calling to vfio_dma_rw(), a kernel thread has already
> called use_mm(), then its current->mm is not null, and it has flag
> PF_KTHREAD.
> in this case, we just want to allow the copy_to_user() directly if
> current->mm == mm, rather than call another use_mm() again.
> 
> do you think it makes sense?

I mean no other way than using use_mm.  That being said nesting
potentional use_mm callers sounds like a rather bad idea, and we
should avoid that.
Yan Zhao April 14, 2020, 7:03 a.m. UTC | #6
On Tue, Apr 14, 2020 at 09:00:13AM +0200, Christoph Hellwig wrote:
> On Mon, Apr 13, 2020 at 08:04:10PM -0400, Yan Zhao wrote:
> > > I can't think of another way for a kernel thread to have a mm indeed.
> > for example, before calling to vfio_dma_rw(), a kernel thread has already
> > called use_mm(), then its current->mm is not null, and it has flag
> > PF_KTHREAD.
> > in this case, we just want to allow the copy_to_user() directly if
> > current->mm == mm, rather than call another use_mm() again.
> > 
> > do you think it makes sense?
> 
> I mean no other way than using use_mm.  That being said nesting
> potentional use_mm callers sounds like a rather bad idea, and we
> should avoid that.
yes, agree.
I was explaining why we just use "current->mm == NULL"
(not "current->flag & PF_KTHREAD") as a criteria to call use_mm()
in vfio_dma_rw(), which you might ask us when you take that part into your
series. :)

Patch
diff mbox series

diff --git a/drivers/gpu/drm/i915/gvt/kvmgt.c b/drivers/gpu/drm/i915/gvt/kvmgt.c
index 074c4efb58eb..5848400620b4 100644
--- a/drivers/gpu/drm/i915/gvt/kvmgt.c
+++ b/drivers/gpu/drm/i915/gvt/kvmgt.c
@@ -2037,7 +2037,7 @@  static int kvmgt_rw_gpa(unsigned long handle, unsigned long gpa,
 	struct kvmgt_guest_info *info;
 	struct kvm *kvm;
 	int idx, ret;
-	bool kthread = current->mm == NULL;
+	bool kthread = (current->flags & PF_KTHREAD);
 
 	if (!handle_valid(handle))
 		return -ESRCH;