On Thu, Jun 04, 2020 at 05:18:00PM +0800, Jason Wang wrote:

On 2020/6/4 下午5:03, Michael S. Tsirkin wrote:

quoted

quoted

quoted

   static bool vhost_notify(struct vhost_dev *dev, struct vhost_virtqueue *vq)
   {
   	__u16 old, new;

diff --git a/drivers/vhost/vhost.h b/drivers/vhost/vhost.h
index a67bda9792ec..6c10e99ff334 100644
--- a/drivers/vhost/vhost.h
+++ b/drivers/vhost/vhost.h

@@ -67,6 +67,13 @@ struct vhost_desc {
   	u16 id;
   };
+struct vhost_buf {
+	u32 out_len;
+	u32 in_len;
+	u16 descs;
+	u16 id;
+};

So it looks to me the struct vhost_buf can work for both split ring and
packed ring.

If this is true, we'd better make struct vhost_desc work for both.

Thanks

Both vhost_desc and vhost_buf can work for split and packed.

Do you mean we should add packed ring support based on this?
For sure, this is one of the motivators for the patchset.

Somehow. But the reason I ask is that I see "split" suffix is used in patch
1 as:

peek_split_desc()
pop_split_desc()
push_split_desc()

But that suffix is not used for the new used ring API invented in this
patch.

Thanks

And that is intentional: split is *not* part of API. The whole idea is
that ring APIs are format agnostic using "buffer" terminology from spec.
The split things are all static within vhost.c

OK so where I had to add a bunch of new format specific code, that was
tagged as "split" to make it easier to spot that they only support a
specific format.  At the same time, I did not rename existing code
adding "split" in the name. I agree it's a useful additional step for
packed ring format support, and it's fairly easy. I just didn't want
to do it automatically.



-- 
MST