[PATCH 0/1] vfio-ccw: be less trusting

[PATCH 0/1] vfio-ccw: be less trusting

[Cornelia Huck]

I'm currently prototyping hsch()/csch() support for vfio-ccw, and
stumbled over this: We currently expect any I/O to be in command-mode,
but don't take care to actively reject transport-mode. This is not an
issue with QEMU (which rejects transport-mode ORBs before even the css
backend gets to it), but better play safe.

Should we want to support transport-mode in the future, we'll probably
need to introduce a capability for it that needs to be enabled by user
space.

[I also still have the doc patch from some time ago queued. Let's see
if we'll collect more patches.]

Cornelia Huck (1):
  vfio-ccw: fence off transport mode

 drivers/s390/cio/vfio_ccw_fsm.c | 5 +++++
 1 file changed, 5 insertions(+)

-- 
2.13.6

[PATCH 1/1] vfio-ccw: fence off transport mode

[Cornelia Huck]

vfio-ccw only supports command mode for channel programs, not transport
mode. User space is supposed to already take care of that and pass us
command-mode ORBs only, but better make sure and return an error to
the caller instead of trying to process tcws as ccws.

Signed-off-by: Cornelia Huck <cohuck@redhat.com>
---
 drivers/s390/cio/vfio_ccw_fsm.c | 5 +++++
 1 file changed, 5 insertions(+)

diff --git a/drivers/s390/cio/vfio_ccw_fsm.c b/drivers/s390/cio/vfio_ccw_fsm.c
index c30420c517b1..ff6963ad6e39 100644
--- a/drivers/s390/cio/vfio_ccw_fsm.c
+++ b/drivers/s390/cio/vfio_ccw_fsm.c
@@ -124,6 +124,11 @@ static void fsm_io_request(struct vfio_ccw_private *private,
 	if (scsw->cmd.fctl & SCSW_FCTL_START_FUNC) {
 		orb = (union orb *)io_region->orb_area;
 
+		/* Don't try to build a cp if transport mode is specified. */
+		if (orb->tm.b) {
+			io_region->ret_code = -EOPNOTSUPP;
+			goto err_out;
+		}
 		io_region->ret_code = cp_init(&private->cp, mdev_dev(mdev),
 					      orb);
 		if (io_region->ret_code)
-- 
2.13.6

Re: [PATCH 1/1] vfio-ccw: fence off transport mode

[Halil Pasic]

On 02/22/2018 04:39 PM, Cornelia Huck wrote:
> vfio-ccw only supports command mode for channel programs, not transport
> mode. User space is supposed to already take care of that and pass us
> command-mode ORBs only, but better make sure and return an error to
> the caller instead of trying to process tcws as ccws.
> 
> Signed-off-by: Cornelia Huck <cohuck@redhat.com>

Acked-by: Halil Pasic <pasic@linux.vnet.ibm.com>

> ---
>  drivers/s390/cio/vfio_ccw_fsm.c | 5 +++++
>  1 file changed, 5 insertions(+)
> 
> diff --git a/drivers/s390/cio/vfio_ccw_fsm.c b/drivers/s390/cio/vfio_ccw_fsm.c
> index c30420c517b1..ff6963ad6e39 100644
> --- a/drivers/s390/cio/vfio_ccw_fsm.c
> +++ b/drivers/s390/cio/vfio_ccw_fsm.c
> @@ -124,6 +124,11 @@ static void fsm_io_request(struct vfio_ccw_private *private,
>  	if (scsw->cmd.fctl & SCSW_FCTL_START_FUNC) {
>  		orb = (union orb *)io_region->orb_area;
> 
> +		/* Don't try to build a cp if transport mode is specified. */
> +		if (orb->tm.b) {
> +			io_region->ret_code = -EOPNOTSUPP;

I guess now we communicate this as appropriately as possible.

> +			goto err_out;
> +		}
>  		io_region->ret_code = cp_init(&private->cp, mdev_dev(mdev),
>  					      orb);
>  		if (io_region->ret_code)
> 

Re: [PATCH 1/1] vfio-ccw: fence off transport mode

[Cornelia Huck]

On Thu, 1 Mar 2018 15:51:05 +0100
Halil Pasic <pasic@linux.vnet.ibm.com> wrote:

> On 02/22/2018 04:39 PM, Cornelia Huck wrote:
> > vfio-ccw only supports command mode for channel programs, not transport
> > mode. User space is supposed to already take care of that and pass us
> > command-mode ORBs only, but better make sure and return an error to
> > the caller instead of trying to process tcws as ccws.
> > 
> > Signed-off-by: Cornelia Huck <cohuck@redhat.com>  
> 
> Acked-by: Halil Pasic <pasic@linux.vnet.ibm.com>

Thanks!

> 
> > ---
> >  drivers/s390/cio/vfio_ccw_fsm.c | 5 +++++
> >  1 file changed, 5 insertions(+)
> > 
> > diff --git a/drivers/s390/cio/vfio_ccw_fsm.c b/drivers/s390/cio/vfio_ccw_fsm.c
> > index c30420c517b1..ff6963ad6e39 100644
> > --- a/drivers/s390/cio/vfio_ccw_fsm.c
> > +++ b/drivers/s390/cio/vfio_ccw_fsm.c
> > @@ -124,6 +124,11 @@ static void fsm_io_request(struct vfio_ccw_private *private,
> >  	if (scsw->cmd.fctl & SCSW_FCTL_START_FUNC) {
> >  		orb = (union orb *)io_region->orb_area;
> > 
> > +		/* Don't try to build a cp if transport mode is specified. */
> > +		if (orb->tm.b) {
> > +			io_region->ret_code = -EOPNOTSUPP;  
> 
> I guess now we communicate this as appropriately as possible.

Yes. We certainly only want to return an error to user space; it is
their responsibility to correctly reflect that to the guest (and even
more their responsibility to not present us with a transport orb in the
first place...)

> 
> > +			goto err_out;
> > +		}
> >  		io_region->ret_code = cp_init(&private->cp, mdev_dev(mdev),
> >  					      orb);
> >  		if (io_region->ret_code)
> >   
> 

Re: [PATCH 1/1] vfio-ccw: fence off transport mode

[Cornelia Huck]

On Thu, 22 Feb 2018 16:39:59 +0100
Cornelia Huck <cohuck@redhat.com> wrote:

> vfio-ccw only supports command mode for channel programs, not transport
> mode. User space is supposed to already take care of that and pass us
> command-mode ORBs only, but better make sure and return an error to
> the caller instead of trying to process tcws as ccws.
> 
> Signed-off-by: Cornelia Huck <cohuck@redhat.com>
> ---
>  drivers/s390/cio/vfio_ccw_fsm.c | 5 +++++
>  1 file changed, 5 insertions(+)
> 
> diff --git a/drivers/s390/cio/vfio_ccw_fsm.c b/drivers/s390/cio/vfio_ccw_fsm.c
> index c30420c517b1..ff6963ad6e39 100644
> --- a/drivers/s390/cio/vfio_ccw_fsm.c
> +++ b/drivers/s390/cio/vfio_ccw_fsm.c
> @@ -124,6 +124,11 @@ static void fsm_io_request(struct vfio_ccw_private *private,
>  	if (scsw->cmd.fctl & SCSW_FCTL_START_FUNC) {
>  		orb = (union orb *)io_region->orb_area;
>  
> +		/* Don't try to build a cp if transport mode is specified. */
> +		if (orb->tm.b) {
> +			io_region->ret_code = -EOPNOTSUPP;
> +			goto err_out;
> +		}
>  		io_region->ret_code = cp_init(&private->cp, mdev_dev(mdev),
>  					      orb);
>  		if (io_region->ret_code)

Queued.

I'll probably send a pull request next week.