Re: [RFC PATCH v1 2/8] vfio-ccw: Don't inject an I/O interrupt if the subchannel is not enabled

[Cornelia Huck <cohuck@redhat.com>]

On Fri, 15 Nov 2019 04:34:31 +0100
Eric Farman <farman@linux.ibm.com> wrote:

> From: Farhan Ali <alifm@linux.ibm.com>
> 
> According to PoPs, when the SCHIB.PMCW bit 8 is 0 status presented by
> the device is not made available to the program. So don't inject an
> interrupt in the guest if the guest OS has not enabled the
> subchannel.

Have you managed to trigger this state in real life?

> 
> Signed-off-by: Farhan Ali <alifm@linux.ibm.com>
> Signed-off-by: Eric Farman <farman@linux.ibm.com>
> ---
> 
> Notes:
>     v0->v1: [EF]
>      - Update commit message, as the bit in question is bit 8 not 15
> 
>  hw/vfio/ccw.c | 5 +++++
>  1 file changed, 5 insertions(+)
> 
> diff --git a/hw/vfio/ccw.c b/hw/vfio/ccw.c
> index 0919ddbeb8..0590a6f512 100644
> --- a/hw/vfio/ccw.c
> +++ b/hw/vfio/ccw.c
> @@ -230,6 +230,11 @@ static void vfio_ccw_io_notifier_handler(void *opaque)
>          return;
>      }
>  
> +    /* Virtual subchannel is not enabled */
> +    if (!(schib->pmcw.flags & PMCW_FLAGS_MASK_ENA)) {

How can that happen? We should not be able to disable the subchannel
while it is in active use, should we? I fear I'm missing something
here...

> +        return;
> +    }
> +
>      size = pread(vcdev->vdev.fd, region, vcdev->io_region_size,
>                   vcdev->io_region_offset);
>      if (size == -1) {