Re: [PATCH net-next v2 2/4] vsock/virtio: support to send non-linear skb

From: "Michael S. Tsirkin" <mst@redhat.com>
To: Arseniy Krasnov <AVKrasnov@sberdevices.ru>
Cc: Bobby Eshleman <bobby.eshleman@bytedance.com>,
	kvm@vger.kernel.org, netdev@vger.kernel.org,
	linux-kernel@vger.kernel.org,
	virtualization@lists.linux-foundation.org, oxffffaa@gmail.com,
	Eric Dumazet <edumazet@google.com>,
	Stefan Hajnoczi <stefanha@redhat.com>,
	kernel@sberdevices.ru, Jakub Kicinski <kuba@kernel.org>,
	Paolo Abeni <pabeni@redhat.com>,
	"David S. Miller" <davem@davemloft.net>
Subject: Re: [PATCH net-next v2 2/4] vsock/virtio: support to send non-linear skb
Date: Tue, 18 Jul 2023 16:27:32 -0400	[thread overview]
Message-ID: <20230718162202-mutt-send-email-mst@kernel.org> (raw)
In-Reply-To: <20230718180237.3248179-3-AVKrasnov@sberdevices.ru>

On Tue, Jul 18, 2023 at 09:02:35PM +0300, Arseniy Krasnov wrote:
> 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 <AVKrasnov@sberdevices.ru>
> 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?

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

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

> +				 * 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.

> +				 */
> +				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.

> -- 
> 2.25.1

_______________________________________________
Virtualization mailing list
Virtualization@lists.linux-foundation.org
https://lists.linuxfoundation.org/mailman/listinfo/virtualization