On Wed, Feb 12, 2025 at 7:11 PM Philippe Mathieu-Daudé
<phi...@linaro.org> wrote:
>
> On 12/2/25 17:49, Konstantin Shkolnyy wrote:
> > VDPA didn't work on a big-endian machine due to missing/incorrect
> > CPU<->LE data format conversions.
> >
> > Signed-off-by: Konstantin Shkolnyy <k...@linux.ibm.com>
> > ---
> > Changes in v2: Change desc_next[] from LE format to "CPU".
> >
> > hw/virtio/vhost-shadow-virtqueue.c | 18 ++++++++++--------
> > 1 file changed, 10 insertions(+), 8 deletions(-)
>
>
> > @@ -228,10 +228,12 @@ static void vhost_svq_kick(VhostShadowVirtqueue *svq)
> > smp_mb();
> >
> > if (virtio_vdev_has_feature(svq->vdev, VIRTIO_RING_F_EVENT_IDX)) {
> > - uint16_t avail_event = *(uint16_t
> > *)(&svq->vring.used->ring[svq->vring.num]);
> > + uint16_t avail_event = le16_to_cpu(
> > + *(uint16_t *)(&svq->vring.used->ring[svq->vring.num]));
>
> Nitpicking, sometimes using the ld/st API is cleaner (here lduw_le_p).
>
I'm not sure if it is right in SVQ, as it is not accessing guest
memory but QEMU memory that has been mapped to a device. But if you
think it is still a valid use case for ld* and st* family I'd be
totally ok with that usage.
> > needs_kick = vring_need_event(avail_event, svq->shadow_avail_idx,
> > svq->shadow_avail_idx - 1);
> > } else {
> > - needs_kick = !(svq->vring.used->flags & VRING_USED_F_NO_NOTIFY);
> > + needs_kick =
> > + !(svq->vring.used->flags &
> > cpu_to_le16(VRING_USED_F_NO_NOTIFY));
> > }
>
