On Tue, 12 Apr 2022 13:53:44 +0800, Jason Wang <jasowang@xxxxxxxxxx> wrote:
>
> 在 2022/4/6 上午11:43, Xuan Zhuo 写道:
> > virtio ring split supports resize.
> >
> > Only after the new vring is successfully allocated based on the new num,
> > we will release the old vring. In any case, an error is returned,
> > indicating that the vring still points to the old vring.
> >
> > In the case of an error, the caller must
> > re-initialize(virtqueue_reinit_split()) the virtqueue to ensure that the
> > vring can be used.
> >
> > In addition, vring_align, may_reduce_num are necessary for reallocating
> > vring, so they are retained for creating vq.
> >
> > Signed-off-by: Xuan Zhuo <xuanzhuo@xxxxxxxxxxxxxxxxx>
> > ---
> > drivers/virtio/virtio_ring.c | 47 ++++++++++++++++++++++++++++++++++++
> > 1 file changed, 47 insertions(+)
> >
> > diff --git a/drivers/virtio/virtio_ring.c b/drivers/virtio/virtio_ring.c
> > index 3dc6ace2ba7a..33864134a744 100644
> > --- a/drivers/virtio/virtio_ring.c
> > +++ b/drivers/virtio/virtio_ring.c
> > @@ -139,6 +139,12 @@ struct vring_virtqueue {
> > /* DMA address and size information */
> > dma_addr_t queue_dma_addr;
> > size_t queue_size_in_bytes;
> > +
> > + /* The parameters for creating vrings are reserved for
> > + * creating new vring.
> > + */
> > + u32 vring_align;
> > + bool may_reduce_num;
> > } split;
> >
> > /* Available for packed ring */
> > @@ -199,6 +205,7 @@ struct vring_virtqueue {
> > };
> >
> > static struct vring_desc_extra *vring_alloc_desc_extra(unsigned int num);
> > +static void vring_free(struct virtqueue *_vq);
> >
> > /*
> > * Helpers.
> > @@ -1088,6 +1095,8 @@ static struct virtqueue *vring_create_virtqueue_split(
> > return NULL;
> > }
> >
> > + to_vvq(vq)->split.vring_align = vring_align;
> > + to_vvq(vq)->split.may_reduce_num = may_reduce_num;
>
>
> It looks to me the above should belong to patch 6.
patch 6 just extracts a function, no logical modification.
to_vvq(vq)->split.may_reduce_num is newly added, so I don't think it should be
merged into patch 6.
>
>
> > to_vvq(vq)->split.queue_dma_addr = dma_addr;
> > to_vvq(vq)->split.queue_size_in_bytes = queue_size_in_bytes;
> > to_vvq(vq)->we_own_ring = true;
> > @@ -1095,6 +1104,44 @@ static struct virtqueue *vring_create_virtqueue_split(
> > return vq;
> > }
> >
> > +static int virtqueue_resize_split(struct virtqueue *_vq, u32 num)
> > +{
> > + struct vring_virtqueue *vq = to_vvq(_vq);
> > + struct virtio_device *vdev = _vq->vdev;
> > + struct vring_desc_state_split *state;
> > + struct vring_desc_extra *extra;
> > + size_t queue_size_in_bytes;
> > + dma_addr_t dma_addr;
> > + struct vring vring;
> > + int err = -ENOMEM;
> > + void *queue;
> > +
> > + queue = vring_alloc_queue_split(vdev, &dma_addr, &num,
> > + vq->split.vring_align,
> > + vq->weak_barriers,
> > + vq->split.may_reduce_num);
> > + if (!queue)
> > + return -ENOMEM;
> > +
> > + queue_size_in_bytes = vring_size(num, vq->split.vring_align);
> > +
> > + err = vring_alloc_state_extra_split(num, &state, &extra);
> > + if (err) {
> > + vring_free_queue(vdev, queue_size_in_bytes, queue, dma_addr);
> > + return -ENOMEM;
> > + }
> > +
> > + vring_free(&vq->vq);
> > +
> > + vring_init(&vring, num, queue, vq->split.vring_align);
> > + vring_virtqueue_attach_split(vq, vring, state, extra);
> > + vq->split.queue_dma_addr = dma_addr;
> > + vq->split.queue_size_in_bytes = queue_size_in_bytes;
>
>
> I wonder if it's better to move the above assignments to
> vring_virtqueue_attach_split().
I also think so, the reason for not doing this is that there is no dma_addr and
queue_size_in_bytes when vring_virtqueue_attach_split is called in
__vring_new_virtqueue.
As discussed in patch 12, we can pass the struct struct vring_virtqueue_split to
vring_virtqueue_attach_split(). This is much more convenient.
Thanks.
>
> Other looks good.
>
> Thanks
>
>
> > +
> > + vring_virtqueue_init_split(vq, vdev, true);
> > + return 0;
> > +}
> > +
> >
> > /*
> > * Packed ring specific functions - *_packed().
>
