Re: [Qemu-devel] [PATCH RFC v6 06/20] virtio: endianness checks for virtio 1.0 devices

[Greg Kurz <gkurz@linux.vnet.ibm.com>]

[-- Attachment #1: Type: text/plain, Size: 4257 bytes --]

On Thu, 22 Jan 2015 12:54:09 +1100
David Gibson <david@gibson.dropbear.id.au> wrote:

> On Thu, Dec 11, 2014 at 02:25:08PM +0100, Cornelia Huck wrote:
> > Add code that checks for the VERSION_1 feature bit in order to make
> > decisions about the device's endianness. This allows us to support
> > transitional devices.
> > 
> > Signed-off-by: Cornelia Huck <cornelia.huck@de.ibm.com>
> > ---
> >  hw/virtio/virtio.c                |    6 +++++-
> >  include/hw/virtio/virtio-access.h |    4 ++++
> >  include/hw/virtio/virtio.h        |    8 ++++++--
> >  3 files changed, 15 insertions(+), 3 deletions(-)
> > 
> > diff --git a/hw/virtio/virtio.c b/hw/virtio/virtio.c
> > index 7f74ae5..8f69ffa 100644
> > --- a/hw/virtio/virtio.c
> > +++ b/hw/virtio/virtio.c
> > @@ -881,7 +881,11 @@ static bool virtio_device_endian_needed(void *opaque)
> >      VirtIODevice *vdev = opaque;
> >  
> >      assert(vdev->device_endian != VIRTIO_DEVICE_ENDIAN_UNKNOWN);
> > -    return vdev->device_endian != virtio_default_endian();
> > +    if (!virtio_has_feature(vdev, VIRTIO_F_VERSION_1)) {
> > +        return vdev->device_endian != virtio_default_endian();
> > +    }
> > +    /* Devices conforming to VIRTIO 1.0 or later are always LE. */
> > +    return vdev->device_endian != VIRTIO_DEVICE_ENDIAN_LITTLE;
> 
> This doesn't seem quite right.  Since virtio 1.0 is always LE, this
> should just assert that device_endian == LE and return false,
> right?
> 

The device_endian field is ONLY used by devices when the software is
legacy.

It is set at device reset time (see virtio_reset()) since we can reasonably
assume that when the software changes endianness, it always reset the device
before using it again (aka. reboot or kexec).

In the case we would have a BE guest, device_endian would be BE, even if the
software is virtio 1.0. So, no, we shouldn't assert.


I had questioned Cornelia about why we care to migrate device_endian when in
virtio 1.0 mode. I got these answers:

https://lists.nongnu.org/archive/html/qemu-devel/2014-10/msg03979.html
https://lists.nongnu.org/archive/html/qemu-devel/2014-11/msg03888.html

My understanding is that a transitional device will necessarily be reset
if the software changes from legacy to 1.0 or vice-versa. So, yes, I still
think virtio_device_endian_needed() should return false.

> >  }
> >  
> >  static const VMStateDescription vmstate_virtio_device_endian = {
> > diff --git a/include/hw/virtio/virtio-access.h b/include/hw/virtio/virtio-access.h
> > index 46456fd..ee28c21 100644
> > --- a/include/hw/virtio/virtio-access.h
> > +++ b/include/hw/virtio/virtio-access.h
> > @@ -19,6 +19,10 @@
> >  
> >  static inline bool virtio_access_is_big_endian(VirtIODevice *vdev)
> >  {
> > +    if (virtio_has_feature(vdev, VIRTIO_F_VERSION_1)) {
> > +        /* Devices conforming to VIRTIO 1.0 or later are always LE. */
> > +        return false;
> > +    }
> >  #if defined(TARGET_IS_BIENDIAN)
> >      return virtio_is_big_endian(vdev);
> >  #elif defined(TARGET_WORDS_BIGENDIAN)
> > diff --git a/include/hw/virtio/virtio.h b/include/hw/virtio/virtio.h
> > index 08141c7..68c40db 100644
> > --- a/include/hw/virtio/virtio.h
> > +++ b/include/hw/virtio/virtio.h
> > @@ -297,7 +297,11 @@ static inline bool virtio_has_feature(VirtIODevice *vdev, unsigned int fbit)
> >  
> >  static inline bool virtio_is_big_endian(VirtIODevice *vdev)
> >  {
> > -    assert(vdev->device_endian != VIRTIO_DEVICE_ENDIAN_UNKNOWN);
> > -    return vdev->device_endian == VIRTIO_DEVICE_ENDIAN_BIG;
> > +    if (!virtio_has_feature(vdev, VIRTIO_F_VERSION_1)) {
> > +        assert(vdev->device_endian != VIRTIO_DEVICE_ENDIAN_UNKNOWN);
> > +        return vdev->device_endian == VIRTIO_DEVICE_ENDIAN_BIG;
> > +    }
> > +    /* Devices conforming to VIRTIO 1.0 or later are always LE. */
> > +    return false;
> >  }
> >  #endif
> 
> AFAICT, the only real difference between virtio_is_big_endian() and
> virtio_access_is_big_endian() is that the latter will become
> compile-time constant on targets that don't do bi-endian.
> 
> With virtio 1.0 support, that's no longer true, so those two macros
> should just be merged, I think.
> 


[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 181 bytes --]