Re: [PATCH 2/4] vhost-vdpa: Introduce RESUME backend feature bit

["Michael S. Tsirkin" <mst@redhat.com>]

On Fri, Oct 14, 2022 at 09:40:46AM +0000, Boeuf, Sebastien wrote:
> On Fri, 2022-10-14 at 05:37 -0400, Michael S. Tsirkin wrote:
> > On Fri, Oct 14, 2022 at 08:07:08AM +0000, Boeuf, Sebastien wrote:
> > > On Fri, 2022-10-14 at 02:11 -0400, Michael S. Tsirkin wrote:
> > > > On Fri, Oct 14, 2022 at 02:09:02PM +0800, Jason Wang wrote:
> > > > > On Fri, Oct 14, 2022 at 2:05 PM Michael S. Tsirkin
> > > > > <mst@redhat.com>
> > > > > wrote:
> > > > > > 
> > > > > > On Fri, Oct 14, 2022 at 01:58:38PM +0800, Jason Wang wrote:
> > > > > > > On Thu, Oct 13, 2022 at 11:35 PM
> > > > > > > <sebastien.boeuf@intel.com>
> > > > > > > wrote:
> > > > > > > > 
> > > > > > > > From: Sebastien Boeuf <sebastien.boeuf@intel.com>
> > > > > > > > 
> > > > > > > > Userspace knows if the device can be resumed or not by
> > > > > > > > checking this
> > > > > > > > feature bit.
> > > > > > > > 
> > > > > > > > It's only exposed if the vdpa driver backend implements
> > > > > > > > the
> > > > > > > > resume()
> > > > > > > > operation callback. Userspace trying to negotiate this
> > > > > > > > feature when it
> > > > > > > > hasn't been exposed will result in an error.
> > > > > > > > 
> > > > > > > > Signed-off-by: Sebastien Boeuf
> > > > > > > > <sebastien.boeuf@intel.com>
> > > > > > > > ---
> > > > > > > >  drivers/vhost/vdpa.c             | 19
> > > > > > > > ++++++++++++++++++-
> > > > > > > >  include/uapi/linux/vhost_types.h |  2 ++
> > > > > > > >  2 files changed, 20 insertions(+), 1 deletion(-)
> > > > > > > > 
> > > > > > > > diff --git a/drivers/vhost/vdpa.c b/drivers/vhost/vdpa.c
> > > > > > > > index 166044642fd5..161727e1a9a5 100644
> > > > > > > > --- a/drivers/vhost/vdpa.c
> > > > > > > > +++ b/drivers/vhost/vdpa.c
> > > > > > > > @@ -355,6 +355,14 @@ static bool
> > > > > > > > vhost_vdpa_can_suspend(const
> > > > > > > > struct vhost_vdpa *v)
> > > > > > > >         return ops->suspend;
> > > > > > > >  }
> > > > > > > > 
> > > > > > > > +static bool vhost_vdpa_can_resume(const struct
> > > > > > > > vhost_vdpa
> > > > > > > > *v)
> > > > > > > > +{
> > > > > > > > +       struct vdpa_device *vdpa = v->vdpa;
> > > > > > > > +       const struct vdpa_config_ops *ops = vdpa->config;
> > > > > > > > +
> > > > > > > > +       return ops->resume;
> > > > > > > > +}
> > > > > > > > +
> > > > > > > >  static long vhost_vdpa_get_features(struct vhost_vdpa
> > > > > > > > *v,
> > > > > > > > u64 __user *featurep)
> > > > > > > >  {
> > > > > > > >         struct vdpa_device *vdpa = v->vdpa;
> > > > > > > > @@ -602,11 +610,18 @@ static long
> > > > > > > > vhost_vdpa_unlocked_ioctl(struct file *filep,
> > > > > > > >                 if (copy_from_user(&features, featurep,
> > > > > > > > sizeof(features)))
> > > > > > > >                         return -EFAULT;
> > > > > > > >                 if (features &
> > > > > > > > ~(VHOST_VDPA_BACKEND_FEATURES
> > > > > > > > > 
> > > > > > > > -                               
> > > > > > > > BIT_ULL(VHOST_BACKEND_F_SUSPEND)))
> > > > > > > > +                               
> > > > > > > > BIT_ULL(VHOST_BACKEND_F_SUSPEND) |
> > > > > > > > +                               
> > > > > > > > BIT_ULL(VHOST_BACKEND_F_RESUME)))
> > > > > > > >                         return -EOPNOTSUPP;
> > > > > > > >                 if ((features &
> > > > > > > > BIT_ULL(VHOST_BACKEND_F_SUSPEND)) &&
> > > > > > > >                      !vhost_vdpa_can_suspend(v))
> > > > > > > >                         return -EOPNOTSUPP;
> > > > > > > > +               if ((features &
> > > > > > > > BIT_ULL(VHOST_BACKEND_F_RESUME)) &&
> > > > > > > > +                    !vhost_vdpa_can_resume(v))
> > > > > > > > +                       return -EOPNOTSUPP;
> > > > > > > > +               if (!(features &
> > > > > > > > BIT_ULL(VHOST_BACKEND_F_SUSPEND)) &&
> > > > > > > > +                    (features &
> > > > > > > > BIT_ULL(VHOST_BACKEND_F_RESUME)))
> > > > > > > > +                       return -EINVAL;
> > > > > > > 
> > > > > > > Is it better to do the check during the probe? It should be
> > > > > > > a
> > > > > > > bug that
> > > > > > > we're having a parent that can only do resume but not
> > > > > > > suspend.
> > > > > > > 
> > > > > > > Thanks
> > > > > > 
> > > > > > well we separated them in the spec ...
> > > > > > suspend could have other uses, I won't say it's an invalid
> > > > > > config.
> > > > > 
> > > > > For suspend only, yes. But we should fail the probe with a
> > > > > resume
> > > > > only, this is somehow the above code wants to check. Or
> > > > > anything I
> > > > > missed?
> > > > > 
> > > > > Thanks
> > > > 
> > > > I am not sure but I would say failing probe is a drastic measure.
> > > > if we have no use for a given combination of features let us
> > > > clear
> > > > the
> > > > feature bit in validation.
> > > 
> > > The current patch only returns an error to the user who might be
> > > trying
> > > to set the RESUME feature bit without the SUSPEND one. But I agree
> > > if
> > > we go down this road, it might be better to also return an error
> > > during
> > > the probe of the backend driver if it provides only the resume
> > > operation.
> > > 
> > > The alternative is to never return the RESUME feature bit as
> > > available
> > > (through GET_BACKEND_FEATURES) if the device is not capable of
> > > being
> > > suspended. This way the vdpa framework would never advertise a
> > > RESUME
> > > feature bit without the SUSPEND one, and the only error that would
> > > have
> > > to be handled should be on the SET_BACKEND_FEATURES (which is what
> > > this
> > > patch does).
> > > 
> > > Please let me know which approach sounds the most appropriate.
> > > 
> > > Thanks,
> > > Sebastien
> > > 
> > > > 
> > > > > > 
> > > > > > > >                 vhost_set_backend_features(&v->vdev,
> > > > > > > > features);
> > > > > > > >                 return 0;
> > > > > > > >         }
> > > > > > > > @@ -658,6 +673,8 @@ static long
> > > > > > > > vhost_vdpa_unlocked_ioctl(struct file *filep,
> > > > > > > >                 features = VHOST_VDPA_BACKEND_FEATURES;
> > > > > > > >                 if (vhost_vdpa_can_suspend(v))
> > > > > > > >                         features |=
> > > > > > > > BIT_ULL(VHOST_BACKEND_F_SUSPEND);
> > > > > > > > +               if (vhost_vdpa_can_resume(v))
> > > > > > > > +                       features |=
> > > > > > > > BIT_ULL(VHOST_BACKEND_F_RESUME);
> > > > > > > >                 if (copy_to_user(featurep, &features,
> > > > > > > > sizeof(features)))
> > > > > > > >                         r = -EFAULT;
> > > > > > > >                 break;
> > > > > > > > diff --git a/include/uapi/linux/vhost_types.h
> > > > > > > > b/include/uapi/linux/vhost_types.h
> > > > > > > > index 53601ce2c20a..c5690a8992d8 100644
> > > > > > > > --- a/include/uapi/linux/vhost_types.h
> > > > > > > > +++ b/include/uapi/linux/vhost_types.h
> > > > > > > > @@ -163,5 +163,7 @@ struct vhost_vdpa_iova_range {
> > > > > > > >  #define VHOST_BACKEND_F_IOTLB_ASID  0x3
> > > > > > > >  /* Device can be suspended */
> > > > > > > >  #define VHOST_BACKEND_F_SUSPEND  0x4
> > > > > > > > +/* Device can be resumed */
> > > > > > > > +#define VHOST_BACKEND_F_RESUME  0x5
> > > > > > > > 
> > > > > > > >  #endif
> > > > > > > > --
> > > > > > > > 2.34.1
> > > > > > > > 
> > > > > > > > ---------------------------------------------------------
> > > > > > > > ----
> > > > > > > > --------
> > > > > > > > Intel Corporation SAS (French simplified joint stock
> > > > > > > > company)
> > > > > > > > Registered headquarters: "Les Montalets"- 2, rue de
> > > > > > > > Paris,
> > > > > > > > 92196 Meudon Cedex, France
> > > > > > > > Registration Number:  302 456 199 R.C.S. NANTERRE
> > > > > > > > Capital: 5 208 026.16 Euros
> > > > > > > > 
> > > > > > > > This e-mail and any attachments may contain confidential
> > > > > > > > material for
> > > > > > > > the sole use of the intended recipient(s). Any review or
> > > > > > > > distribution
> > > > > > > > by others is strictly prohibited. If you are not the
> > > > > > > > intended
> > > > > > > > recipient, please contact the sender and delete all
> > > > > > > > copies.
> > > > > > > > 
> > > > > > 
> > > > 
> > > 
> > > -------------------------------------------------------------------
> > > --
> > > Intel Corporation SAS (French simplified joint stock company)
> > > Registered headquarters: "Les Montalets"- 2, rue de Paris, 
> > > 92196 Meudon Cedex, France
> > > Registration Number:  302 456 199 R.C.S. NANTERRE
> > > Capital: 5 208 026.16 Euros
> > > 
> > > This e-mail and any attachments may contain confidential material
> > > for
> > > the sole use of the intended recipient(s). Any review or
> > > distribution
> > > by others is strictly prohibited. If you are not the intended
> > > recipient, please contact the sender and delete all copies.
> > 
> > 
> > I really feel it's up to the driver.
> 
> So solution number 2? I keep this code and add some logic to the
> GET_BACKEND_FEATURES ioctl so that it wouldn't return the RESUME bit if
> the backend isn't capable of being suspended?

No, what I mean is that caller of SET_BACKEND_FEATURES should not
set RESUME if SUSPEND is not set, and if it does I see no reason to
intervene.


> > 
> 
> ---------------------------------------------------------------------
> Intel Corporation SAS (French simplified joint stock company)
> Registered headquarters: "Les Montalets"- 2, rue de Paris, 
> 92196 Meudon Cedex, France
> Registration Number:  302 456 199 R.C.S. NANTERRE
> Capital: 5 208 026.16 Euros
> 
> This e-mail and any attachments may contain confidential material for
> the sole use of the intended recipient(s). Any review or distribution
> by others is strictly prohibited. If you are not the intended
> recipient, please contact the sender and delete all copies.

_______________________________________________
Virtualization mailing list
Virtualization@lists.linux-foundation.org
https://lists.linuxfoundation.org/mailman/listinfo/virtualization