On Wed, Jul 19, 2023 at 07:46:05AM +0300, Arseniy Krasnov wrote:

On 18.07.2023 23:27, Michael S. Tsirkin wrote:

quoted

On Tue, Jul 18, 2023 at 09:02:35PM +0300, Arseniy Krasnov wrote:

quoted

For non-linear skb use its pages from fragment array as buffers in
virtio tx queue. These pages are already pinned by 'get_user_pages()'
during such skb creation.

Signed-off-by: Arseniy Krasnov <redacted>
Reviewed-by: Stefano Garzarella <sgarzare@redhat.com>
---
 net/vmw_vsock/virtio_transport.c | 40 +++++++++++++++++++++++++++-----
 1 file changed, 34 insertions(+), 6 deletions(-)

diff --git a/net/vmw_vsock/virtio_transport.c b/net/vmw_vsock/virtio_transport.c
index e95df847176b..6cbb45bb12d2 100644
--- a/net/vmw_vsock/virtio_transport.c
+++ b/net/vmw_vsock/virtio_transport.c

@@ -100,7 +100,9 @@ virtio_transport_send_pkt_work(struct work_struct *work)
 	vq = vsock->vqs[VSOCK_VQ_TX];

 	for (;;) {
-		struct scatterlist hdr, buf, *sgs[2];
+		/* +1 is for packet header. */
+		struct scatterlist *sgs[MAX_SKB_FRAGS + 1];
+		struct scatterlist bufs[MAX_SKB_FRAGS + 1];
 		int ret, in_sg = 0, out_sg = 0;
 		struct sk_buff *skb;
 		bool reply;

@@ -111,12 +113,38 @@ virtio_transport_send_pkt_work(struct work_struct *work)

 		virtio_transport_deliver_tap_pkt(skb);
 		reply = virtio_vsock_skb_reply(skb);
+		sg_init_one(&bufs[out_sg], virtio_vsock_hdr(skb),
+			    sizeof(*virtio_vsock_hdr(skb)));
+		sgs[out_sg] = &bufs[out_sg];
+		out_sg++;
+
+		if (!skb_is_nonlinear(skb)) {
+			if (skb->len > 0) {
+				sg_init_one(&bufs[out_sg], skb->data, skb->len);
+				sgs[out_sg] = &bufs[out_sg];
+				out_sg++;
+			}
+		} else {
+			struct skb_shared_info *si;
+			int i;
+
+			si = skb_shinfo(skb);
+
+			for (i = 0; i < si->nr_frags; i++) {
+				skb_frag_t *skb_frag = &si->frags[i];
+				void *va = page_to_virt(skb_frag->bv_page);

-		sg_init_one(&hdr, virtio_vsock_hdr(skb), sizeof(*virtio_vsock_hdr(skb)));
-		sgs[out_sg++] = &hdr;
-		if (skb->len > 0) {
-			sg_init_one(&buf, skb->data, skb->len);
-			sgs[out_sg++] = &buf;
+				/* We will use 'page_to_virt()' for userspace page here,

don't put comments after code they refer to, please?

quoted

+				 * because virtio layer will call 'virt_to_phys()' later

it will but not always. sometimes it's the dma mapping layer.

quoted

+				 * to fill buffer descriptor. We don't touch memory at
+				 * "virtual" address of this page.

you need to stick "the" in a bunch of places above.

Ok, I'll fix this comment!

quoted

quoted

+				 */
+				sg_init_one(&bufs[out_sg],
+					    va + skb_frag->bv_offset,
+					    skb_frag->bv_len);
+				sgs[out_sg] = &bufs[out_sg];
+				out_sg++;
+			}
 		}

 		ret = virtqueue_add_sgs(vq, sgs, out_sg, in_sg, skb, GFP_KERNEL);

There's a problem here: if there vq is small this will fail.
So you really should check free vq s/gs and switch to non-zcopy
if too small.

Ok, so idea is that:

if (out_sg > vq->num_free)
   reorganise current skb for copy mode (e.g. 2 out_sg - header and data)
   and try to add it to vq again.

?

@Stefano, I'll remove net-next tag (guess RFC is not required again, but not net-next
anyway) as this change will require review. R-b I think should be also removed. All
other patches in this set still unchanged.

It's still a new feature so we have net-next tree as the target, right?

I think we should keep net-next. Even if patches require to be
re-reviewed, net-next indicates the tree where we want these to be merge
and for new features is the right one.

Ack for not putting RFC again and for R-b removal for this patch.

Thanks,
Stefano