Re: [PATCH v2 2/2] drm/i915/gvt: subsitute kvm_read/write_guest with vfio_dma_rw

[Alex Williamson <alex.williamson@redhat.com>]

On Thu, 16 Jan 2020 00:49:41 -0500
Yan Zhao <yan.y.zhao@intel.com> wrote:

> On Thu, Jan 16, 2020 at 04:06:51AM +0800, Alex Williamson wrote:
> > On Tue, 14 Jan 2020 22:54:55 -0500
> > Yan Zhao <yan.y.zhao@intel.com> wrote:
> >   
> > > As a device model, it is better to read/write guest memory using vfio
> > > interface, so that vfio is able to maintain dirty info of device IOVAs.
> > > 
> > > Compared to kvm interfaces kvm_read/write_guest(), vfio_dma_rw() has ~600
> > > cycles more overhead on average.
> > > 
> > > -------------------------------------
> > > |    interface     | avg cpu cycles |
> > > |-----------------------------------|
> > > | kvm_write_guest  |     1554       |
> > > | ----------------------------------|
> > > | kvm_read_guest   |     707        |
> > > |-----------------------------------|
> > > | vfio_dma_rw(w)   |     2274       |
> > > |-----------------------------------|
> > > | vfio_dma_rw(r)   |     1378       |
> > > -------------------------------------  
> > 
> > In v1 you had:
> > 
> > -------------------------------------
> > |    interface     | avg cpu cycles |
> > |-----------------------------------|
> > | kvm_write_guest  |     1546       |
> > | ----------------------------------|
> > | kvm_read_guest   |     686        |
> > |-----------------------------------|
> > | vfio_iova_rw(w)  |     2233       |
> > |-----------------------------------|
> > | vfio_iova_rw(r)  |     1262       |
> > -------------------------------------
> > 
> > So the kvm numbers remained within +0.5-3% while the vfio numbers are
> > now +1.8-9.2%.  I would have expected the algorithm change to at least
> > not be worse for small accesses and be better for accesses crossing
> > page boundaries.  Do you know what happened?
> >  
> I only tested the 4 interfaces in GVT's environment, where most of the
> guest memory accesses are less than one page.
> And the different fluctuations should be caused by the locks.
> vfio_dma_rw contends locks with other vfio accesses which are assumed to
> be abundant in the case of GVT.

Hmm, so maybe it's time to convert vfio_iommu.lock from a mutex to a
rwsem?  Thanks,

Alex

> > > Comparison of benchmarks scores are as blow:
> > > ------------------------------------------------------
> > > |  avg score  | kvm_read/write_guest  | vfio_dma_rw  |
> > > |----------------------------------------------------|
> > > |   Glmark2   |         1284          |    1296      |
> > > |----------------------------------------------------|
> > > |  Lightsmark |         61.24         |    61.27     |
> > > |----------------------------------------------------|
> > > |  OpenArena  |         140.9         |    137.4     |
> > > |----------------------------------------------------|
> > > |   Heaven    |          671          |     670      |
> > > ------------------------------------------------------
> > > No obvious performance downgrade found.
> > > 
> > > Cc: Kevin Tian <kevin.tian@intel.com>
> > > Signed-off-by: Yan Zhao <yan.y.zhao@intel.com>
> > > ---
> > >  drivers/gpu/drm/i915/gvt/kvmgt.c | 26 +++++++-------------------
> > >  1 file changed, 7 insertions(+), 19 deletions(-)
> > > 
> > > diff --git a/drivers/gpu/drm/i915/gvt/kvmgt.c b/drivers/gpu/drm/i915/gvt/kvmgt.c
> > > index bd79a9718cc7..17edc9a7ff05 100644
> > > --- a/drivers/gpu/drm/i915/gvt/kvmgt.c
> > > +++ b/drivers/gpu/drm/i915/gvt/kvmgt.c
> > > @@ -1966,31 +1966,19 @@ static int kvmgt_rw_gpa(unsigned long handle, unsigned long gpa,
> > >  			void *buf, unsigned long len, bool write)
> > >  {
> > >  	struct kvmgt_guest_info *info;
> > > -	struct kvm *kvm;
> > > -	int idx, ret;
> > > -	bool kthread = current->mm == NULL;
> > > +	int ret;
> > > +	struct intel_vgpu *vgpu;
> > > +	struct device *dev;
> > >  
> > >  	if (!handle_valid(handle))
> > >  		return -ESRCH;
> > >  
> > >  	info = (struct kvmgt_guest_info *)handle;
> > > -	kvm = info->kvm;
> > > -
> > > -	if (kthread) {
> > > -		if (!mmget_not_zero(kvm->mm))
> > > -			return -EFAULT;
> > > -		use_mm(kvm->mm);
> > > -	}
> > > -
> > > -	idx = srcu_read_lock(&kvm->srcu);
> > > -	ret = write ? kvm_write_guest(kvm, gpa, buf, len) :
> > > -		      kvm_read_guest(kvm, gpa, buf, len);
> > > -	srcu_read_unlock(&kvm->srcu, idx);
> > > +	vgpu = info->vgpu;
> > > +	dev = mdev_dev(vgpu->vdev.mdev);
> > >  
> > > -	if (kthread) {
> > > -		unuse_mm(kvm->mm);
> > > -		mmput(kvm->mm);
> > > -	}
> > > +	ret = write ? vfio_dma_rw(dev, gpa, buf, len, true) :
> > > +			vfio_dma_rw(dev, gpa, buf, len, false);  
> > 
> > As Paolo suggested previously, this can be simplified:
> > 
> > ret = vfio_dma_rw(dev, gpa, buf, len, write);
> >  
> > >  
> > >  	return ret;  
> > 
> > Or even more simple, remove the ret variable:
> > 
> > return vfio_dma_rw(dev, gpa, buf, len, write);
> >   
> oh, it seems that I missed Paolo's mail. will change it. thank you!
> 
> Thanks
> Yan
> >   
> > >  }  
> >   
>