[PATCH v2] vsock/virtio: rework MSG_ZEROCOPY flag handling

[PATCH v2] vsock/virtio: rework MSG_ZEROCOPY flag handling

[Arseniy Krasnov]

Logically it was based on TCP implementation, so make further support
easier, rewrite it in the TCP way.

Signed-off-by: Arseniy Krasnov <avkrasnov@rulkc.org>
---
 Changelog v1->v2:
 * Rebase on last 'net-next'. Don't need 'skb_zcopy_set()' now - it was
   already added.

 net/vmw_vsock/virtio_transport_common.c | 48 ++++++++++++-------------
 1 file changed, 23 insertions(+), 25 deletions(-)

diff --git a/net/vmw_vsock/virtio_transport_common.c b/net/vmw_vsock/virtio_transport_common.c
index 09475007165b..787524b8cb44 100644
--- a/net/vmw_vsock/virtio_transport_common.c
+++ b/net/vmw_vsock/virtio_transport_common.c
@@ -328,38 +328,36 @@ static int virtio_transport_send_pkt_info(struct vsock_sock *vsk,
 	if (pkt_len == 0 && info->op == VIRTIO_VSOCK_OP_RW)
 		return pkt_len;
 
-	if (info->msg) {
-		/* If zerocopy is not enabled by 'setsockopt()', we behave as
-		 * there is no MSG_ZEROCOPY flag set.
+	if (info->msg && (info->msg->msg_flags & MSG_ZEROCOPY)) {
+		/* If 'info->msg' is not NULL, this is only VIRTIO_VSOCK_OP_RW.
+		 * 'MSG_ZEROCOPY' flag handling here is based on the same flag
+		 * handling from 'tcp_sendmsg_locked()'.
 		 */
-		if (!sock_flag(sk_vsock(vsk), SOCK_ZEROCOPY))
-			info->msg->msg_flags &= ~MSG_ZEROCOPY;
+		if (info->msg->msg_ubuf) {
+			uarg = info->msg->msg_ubuf;
+			can_zcopy = virtio_transport_can_zcopy(t_ops, info, pkt_len);
+		} else if (sock_flag(sk_vsock(vsk), SOCK_ZEROCOPY)) {
+			uarg = msg_zerocopy_realloc(sk_vsock(vsk), pkt_len,
+						    NULL, false);
+			if (!uarg) {
+				virtio_transport_put_credit(vvs, pkt_len);
+				return -ENOMEM;
+			}
 
-		if (info->msg->msg_flags & MSG_ZEROCOPY)
 			can_zcopy = virtio_transport_can_zcopy(t_ops, info, pkt_len);
 
+			if (!can_zcopy)
+				uarg_to_msgzc(uarg)->zerocopy = 0;
+
+			have_uref = true;
+		}
+
+		/* 'can_zcopy' means that this transmission will be
+		 * in zerocopy way (e.g. using 'frags' array).
+		 */
 		if (can_zcopy)
 			max_skb_len = min_t(u32, VIRTIO_VSOCK_MAX_PKT_BUF_SIZE,
 					    (MAX_SKB_FRAGS * PAGE_SIZE));
-
-		if (info->msg->msg_flags & MSG_ZEROCOPY &&
-		    info->op == VIRTIO_VSOCK_OP_RW) {
-			uarg = info->msg->msg_ubuf;
-
-			if (!uarg) {
-				uarg = msg_zerocopy_realloc(sk_vsock(vsk),
-							    pkt_len, NULL, false);
-				if (!uarg) {
-					virtio_transport_put_credit(vvs, pkt_len);
-					return -ENOMEM;
-				}
-
-				if (!can_zcopy)
-					uarg_to_msgzc(uarg)->zerocopy = 0;
-
-				have_uref = true;
-			}
-		}
 	}
 
 	rest_len = pkt_len;
-- 
2.25.1

Re: [PATCH v2] vsock/virtio: rework MSG_ZEROCOPY flag handling

[Stefano Garzarella]

On Sun, Jun 14, 2026 at 08:47:56PM +0300, Arseniy Krasnov wrote:
>Logically it was based on TCP implementation, so make further support
>easier, rewrite it in the TCP way.

Hi Arseniy, and thank you so much for the patch!

I’d like to ask you to expand on the message a bit, especially to 
explain why we’re making this change.

In particular, I’d like to better understand whether this is just a 
cosmetic change or if we’re fixing any issues (and if so, which ones), 
so we can determine whether this patch should be backported to the 
stable branches.

>
>Signed-off-by: Arseniy Krasnov <avkrasnov@rulkc.org>
>---
> Changelog v1->v2:
> * Rebase on last 'net-next'. Don't need 'skb_zcopy_set()' now - it was
>   already added.

Ah, okay is net-next material, please use the net-next tag (ie. [PATCH 
net-next v2]).

>
> net/vmw_vsock/virtio_transport_common.c | 48 ++++++++++++-------------
> 1 file changed, 23 insertions(+), 25 deletions(-)
>
>diff --git a/net/vmw_vsock/virtio_transport_common.c b/net/vmw_vsock/virtio_transport_common.c
>index 09475007165b..787524b8cb44 100644
>--- a/net/vmw_vsock/virtio_transport_common.c
>+++ b/net/vmw_vsock/virtio_transport_common.c
>@@ -328,38 +328,36 @@ static int virtio_transport_send_pkt_info(struct vsock_sock *vsk,
> 	if (pkt_len == 0 && info->op == VIRTIO_VSOCK_OP_RW)
> 		return pkt_len;
>
>-	if (info->msg) {
>-		/* If zerocopy is not enabled by 'setsockopt()', we behave as
>-		 * there is no MSG_ZEROCOPY flag set.
>+	if (info->msg && (info->msg->msg_flags & MSG_ZEROCOPY)) {
>+		/* If 'info->msg' is not NULL, this is only VIRTIO_VSOCK_OP_RW.
>+		 * 'MSG_ZEROCOPY' flag handling here is based on the same flag
>+		 * handling from 'tcp_sendmsg_locked()'.
> 		 */
>-		if (!sock_flag(sk_vsock(vsk), SOCK_ZEROCOPY))
>-			info->msg->msg_flags &= ~MSG_ZEROCOPY;
>+		if (info->msg->msg_ubuf) {
>+			uarg = info->msg->msg_ubuf;
>+			can_zcopy = virtio_transport_can_zcopy(t_ops, info, pkt_len);
>+		} else if (sock_flag(sk_vsock(vsk), SOCK_ZEROCOPY)) {
>+			uarg = msg_zerocopy_realloc(sk_vsock(vsk), pkt_len,
>+						    NULL, false);
>+			if (!uarg) {
>+				virtio_transport_put_credit(vvs, pkt_len);
>+				return -ENOMEM;
>+			}
>
>-		if (info->msg->msg_flags & MSG_ZEROCOPY)
> 			can_zcopy = virtio_transport_can_zcopy(t_ops, info, pkt_len);
>

nit: we can remove this extra blank line.

For the rest I can't see anything wrong, but a bit more context in the 
commit would help me in the review.

Thanks,
Stefano

Re: [PATCH v2] vsock/virtio: rework MSG_ZEROCOPY flag handling

[Arseniy Krasnov]

On 16/06/2026 16:09, Stefano Garzarella wrote:
> On Sun, Jun 14, 2026 at 08:47:56PM +0300, Arseniy Krasnov wrote:
>> Logically it was based on TCP implementation, so make further support
>> easier, rewrite it in the TCP way.
>
> Hi Arseniy, and thank you so much for the patch!
>
> I’d like to ask you to expand on the message a bit, especially to explain why we’re making this change.
>
> In particular, I’d like to better understand whether this is just a cosmetic change or if we’re fixing any issues (and if so, which ones), so we can determine whether this patch should be backported to the stable branches.

This is cosmetic change. I'll update commit message in v3.

>
>>
>> Signed-off-by: Arseniy Krasnov <avkrasnov@rulkc.org>
>> ---
>> Changelog v1->v2:
>> * Rebase on last 'net-next'. Don't need 'skb_zcopy_set()' now - it was
>>   already added.
>
> Ah, okay is net-next material, please use the net-next tag (ie. [PATCH net-next v2]). 

Sure!

>
>>
>> net/vmw_vsock/virtio_transport_common.c | 48 ++++++++++++-------------
>> 1 file changed, 23 insertions(+), 25 deletions(-)
>>
>> diff --git a/net/vmw_vsock/virtio_transport_common.c b/net/vmw_vsock/virtio_transport_common.c
>> index 09475007165b..787524b8cb44 100644
>> --- a/net/vmw_vsock/virtio_transport_common.c
>> +++ b/net/vmw_vsock/virtio_transport_common.c
>> @@ -328,38 +328,36 @@ static int virtio_transport_send_pkt_info(struct vsock_sock *vsk,
>>     if (pkt_len == 0 && info->op == VIRTIO_VSOCK_OP_RW)
>>         return pkt_len;
>>
>> -    if (info->msg) {
>> -        /* If zerocopy is not enabled by 'setsockopt()', we behave as
>> -         * there is no MSG_ZEROCOPY flag set.
>> +    if (info->msg && (info->msg->msg_flags & MSG_ZEROCOPY)) {
>> +        /* If 'info->msg' is not NULL, this is only VIRTIO_VSOCK_OP_RW.
>> +         * 'MSG_ZEROCOPY' flag handling here is based on the same flag
>> +         * handling from 'tcp_sendmsg_locked()'.
>>          */
>> -        if (!sock_flag(sk_vsock(vsk), SOCK_ZEROCOPY))
>> -            info->msg->msg_flags &= ~MSG_ZEROCOPY;
>> +        if (info->msg->msg_ubuf) {
>> +            uarg = info->msg->msg_ubuf;
>> +            can_zcopy = virtio_transport_can_zcopy(t_ops, info, pkt_len);
>> +        } else if (sock_flag(sk_vsock(vsk), SOCK_ZEROCOPY)) {
>> +            uarg = msg_zerocopy_realloc(sk_vsock(vsk), pkt_len,
>> +                            NULL, false);
>> +            if (!uarg) {
>> +                virtio_transport_put_credit(vvs, pkt_len);
>> +                return -ENOMEM;
>> +            }
>>
>> -        if (info->msg->msg_flags & MSG_ZEROCOPY)
>>             can_zcopy = virtio_transport_can_zcopy(t_ops, info, pkt_len);
>>
>
> nit: we can remove this extra blank line.
>
> For the rest I can't see anything wrong, but a bit more context in the commit would help me in the review.

Ack, I'll update commit message in v3. Also need to check some reports about pre-existing issues from sashiko, triggered by this patch.

Thanks!

>
> Thanks,
> Stefano
>