Patchwork virtio_net: introduce VIRTIO_NET_HDR_F_DATA_VALID

login
register
mail settings
Submitter Jason Wang
Date June 10, 2011, 10:56 a.m.
Message ID <1307703377-3798-1-git-send-email-jasowang@redhat.com>
Download mbox | patch
Permalink /patch/99880/
State Accepted
Delegated to: David Miller
Headers show

Comments

Jason Wang - June 10, 2011, 10:56 a.m.
There's no need for the guest to validate the checksum if it have been
validated by host nics. So this patch introduces a new flag -
VIRTIO_NET_HDR_F_DATA_VALID which is used to bypass the checksum
examing in guest. The backend (tap/macvtap) may set this flag when
met skbs with CHECKSUM_UNNECESSARY to save cpu utilization.

No feature negotiation is needed as old driver just ignore this flag.

Iperf shows 12%-30% performance improvement for UDP traffic. For TCP,
when gro is on no difference as it produces skb with partial
checksum. But when gro is disabled, 20% or even higher improvement
could be measured by netperf.

Signed-off-by: Jason Wang <jasowang@redhat.com>
---
 drivers/net/macvtap.c      |    2 ++
 drivers/net/tun.c          |    2 ++
 drivers/net/virtio_net.c   |    2 ++
 include/linux/virtio_net.h |    1 +
 net/packet/af_packet.c     |    2 ++
 5 files changed, 9 insertions(+), 0 deletions(-)
Michael S. Tsirkin - June 10, 2011, 11:28 a.m.
On Fri, Jun 10, 2011 at 06:56:17PM +0800, Jason Wang wrote:
> There's no need for the guest to validate the checksum if it have been
> validated by host nics. So this patch introduces a new flag -
> VIRTIO_NET_HDR_F_DATA_VALID which is used to bypass the checksum
> examing in guest. The backend (tap/macvtap) may set this flag when
> met skbs with CHECKSUM_UNNECESSARY to save cpu utilization.
> 
> No feature negotiation is needed as old driver just ignore this flag.

This wasn't required by the spec, but maybe it should be.

> Iperf shows 12%-30% performance improvement for UDP traffic. For TCP,
> when gro is on no difference as it produces skb with partial
> checksum. But when gro is disabled, 20% or even higher improvement
> could be measured by netperf.
> 
> Signed-off-by: Jason Wang <jasowang@redhat.com>


Acked-by: Michael S. Tsirkin <mst@redhat.com>

> ---
>  drivers/net/macvtap.c      |    2 ++
>  drivers/net/tun.c          |    2 ++
>  drivers/net/virtio_net.c   |    2 ++
>  include/linux/virtio_net.h |    1 +
>  net/packet/af_packet.c     |    2 ++
>  5 files changed, 9 insertions(+), 0 deletions(-)
> 
> diff --git a/drivers/net/macvtap.c b/drivers/net/macvtap.c
> index 6696e56..ecee0fe 100644
> --- a/drivers/net/macvtap.c
> +++ b/drivers/net/macvtap.c
> @@ -508,6 +508,8 @@ static int macvtap_skb_to_vnet_hdr(const struct sk_buff *skb,
>  		vnet_hdr->flags = VIRTIO_NET_HDR_F_NEEDS_CSUM;
>  		vnet_hdr->csum_start = skb_checksum_start_offset(skb);
>  		vnet_hdr->csum_offset = skb->csum_offset;
> +	} else if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
> +		vnet_hdr->flags = VIRTIO_NET_HDR_F_DATA_VALID;
>  	} /* else everything is zero */
>  
>  	return 0;
> diff --git a/drivers/net/tun.c b/drivers/net/tun.c
> index 74e9405..f43fa45 100644
> --- a/drivers/net/tun.c
> +++ b/drivers/net/tun.c
> @@ -788,6 +788,8 @@ static __inline__ ssize_t tun_put_user(struct tun_struct *tun,
>  			gso.flags = VIRTIO_NET_HDR_F_NEEDS_CSUM;
>  			gso.csum_start = skb_checksum_start_offset(skb);
>  			gso.csum_offset = skb->csum_offset;
> +		} else if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
> +			gso.flags = VIRTIO_NET_HDR_F_DATA_VALID;
>  		} /* else everything is zero */
>  
>  		if (unlikely(memcpy_toiovecend(iv, (void *)&gso, total,
> diff --git a/drivers/net/virtio_net.c b/drivers/net/virtio_net.c
> index f685324..be3686a 100644
> --- a/drivers/net/virtio_net.c
> +++ b/drivers/net/virtio_net.c
> @@ -274,6 +274,8 @@ static void receive_buf(struct net_device *dev, void *buf, unsigned int len)
>  					  hdr->hdr.csum_start,
>  					  hdr->hdr.csum_offset))
>  			goto frame_err;
> +	} else if (hdr->hdr.flags & VIRTIO_NET_HDR_F_DATA_VALID) {
> +		skb->ip_summed = CHECKSUM_UNNECESSARY;
>  	}
>  
>  	skb->protocol = eth_type_trans(skb, dev);
> diff --git a/include/linux/virtio_net.h b/include/linux/virtio_net.h
> index 136040b..970d5a2 100644
> --- a/include/linux/virtio_net.h
> +++ b/include/linux/virtio_net.h
> @@ -63,6 +63,7 @@ struct virtio_net_config {
>   * specify GSO or CSUM features, you can simply ignore the header. */
>  struct virtio_net_hdr {
>  #define VIRTIO_NET_HDR_F_NEEDS_CSUM	1	// Use csum_start, csum_offset
> +#define VIRTIO_NET_HDR_F_DATA_VALID	2	// Csum is valid
>  	__u8 flags;
>  #define VIRTIO_NET_HDR_GSO_NONE		0	// Not a GSO frame
>  #define VIRTIO_NET_HDR_GSO_TCPV4	1	// GSO frame, IPv4 TCP (TSO)
> diff --git a/net/packet/af_packet.c b/net/packet/af_packet.c
> index c0c3cda..99ca471 100644
> --- a/net/packet/af_packet.c
> +++ b/net/packet/af_packet.c
> @@ -1681,6 +1681,8 @@ static int packet_recvmsg(struct kiocb *iocb, struct socket *sock,
>  			vnet_hdr.flags = VIRTIO_NET_HDR_F_NEEDS_CSUM;
>  			vnet_hdr.csum_start = skb_checksum_start_offset(skb);
>  			vnet_hdr.csum_offset = skb->csum_offset;
> +		} else if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
> +			vnet_hdr.flags = VIRTIO_NET_HDR_F_DATA_VALID;
>  		} /* else everything is zero */
>  
>  		err = memcpy_toiovec(msg->msg_iov, (void *)&vnet_hdr,
> -- 
> 1.7.1
--
To unsubscribe from this list: send the line "unsubscribe netdev" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
David Miller - June 11, 2011, 10:58 p.m.
From: "Michael S. Tsirkin" <mst@redhat.com>
Date: Fri, 10 Jun 2011 14:28:28 +0300

> On Fri, Jun 10, 2011 at 06:56:17PM +0800, Jason Wang wrote:
>> There's no need for the guest to validate the checksum if it have been
>> validated by host nics. So this patch introduces a new flag -
>> VIRTIO_NET_HDR_F_DATA_VALID which is used to bypass the checksum
>> examing in guest. The backend (tap/macvtap) may set this flag when
>> met skbs with CHECKSUM_UNNECESSARY to save cpu utilization.
>> 
>> No feature negotiation is needed as old driver just ignore this flag.
> 
> This wasn't required by the spec, but maybe it should be.
> 
>> Iperf shows 12%-30% performance improvement for UDP traffic. For TCP,
>> when gro is on no difference as it produces skb with partial
>> checksum. But when gro is disabled, 20% or even higher improvement
>> could be measured by netperf.
>> 
>> Signed-off-by: Jason Wang <jasowang@redhat.com>
> 
> 
> Acked-by: Michael S. Tsirkin <mst@redhat.com>

Applied to net-next-2.6
--
To unsubscribe from this list: send the line "unsubscribe netdev" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Patch

diff --git a/drivers/net/macvtap.c b/drivers/net/macvtap.c
index 6696e56..ecee0fe 100644
--- a/drivers/net/macvtap.c
+++ b/drivers/net/macvtap.c
@@ -508,6 +508,8 @@  static int macvtap_skb_to_vnet_hdr(const struct sk_buff *skb,
 		vnet_hdr->flags = VIRTIO_NET_HDR_F_NEEDS_CSUM;
 		vnet_hdr->csum_start = skb_checksum_start_offset(skb);
 		vnet_hdr->csum_offset = skb->csum_offset;
+	} else if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
+		vnet_hdr->flags = VIRTIO_NET_HDR_F_DATA_VALID;
 	} /* else everything is zero */
 
 	return 0;
diff --git a/drivers/net/tun.c b/drivers/net/tun.c
index 74e9405..f43fa45 100644
--- a/drivers/net/tun.c
+++ b/drivers/net/tun.c
@@ -788,6 +788,8 @@  static __inline__ ssize_t tun_put_user(struct tun_struct *tun,
 			gso.flags = VIRTIO_NET_HDR_F_NEEDS_CSUM;
 			gso.csum_start = skb_checksum_start_offset(skb);
 			gso.csum_offset = skb->csum_offset;
+		} else if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
+			gso.flags = VIRTIO_NET_HDR_F_DATA_VALID;
 		} /* else everything is zero */
 
 		if (unlikely(memcpy_toiovecend(iv, (void *)&gso, total,
diff --git a/drivers/net/virtio_net.c b/drivers/net/virtio_net.c
index f685324..be3686a 100644
--- a/drivers/net/virtio_net.c
+++ b/drivers/net/virtio_net.c
@@ -274,6 +274,8 @@  static void receive_buf(struct net_device *dev, void *buf, unsigned int len)
 					  hdr->hdr.csum_start,
 					  hdr->hdr.csum_offset))
 			goto frame_err;
+	} else if (hdr->hdr.flags & VIRTIO_NET_HDR_F_DATA_VALID) {
+		skb->ip_summed = CHECKSUM_UNNECESSARY;
 	}
 
 	skb->protocol = eth_type_trans(skb, dev);
diff --git a/include/linux/virtio_net.h b/include/linux/virtio_net.h
index 136040b..970d5a2 100644
--- a/include/linux/virtio_net.h
+++ b/include/linux/virtio_net.h
@@ -63,6 +63,7 @@  struct virtio_net_config {
  * specify GSO or CSUM features, you can simply ignore the header. */
 struct virtio_net_hdr {
 #define VIRTIO_NET_HDR_F_NEEDS_CSUM	1	// Use csum_start, csum_offset
+#define VIRTIO_NET_HDR_F_DATA_VALID	2	// Csum is valid
 	__u8 flags;
 #define VIRTIO_NET_HDR_GSO_NONE		0	// Not a GSO frame
 #define VIRTIO_NET_HDR_GSO_TCPV4	1	// GSO frame, IPv4 TCP (TSO)
diff --git a/net/packet/af_packet.c b/net/packet/af_packet.c
index c0c3cda..99ca471 100644
--- a/net/packet/af_packet.c
+++ b/net/packet/af_packet.c
@@ -1681,6 +1681,8 @@  static int packet_recvmsg(struct kiocb *iocb, struct socket *sock,
 			vnet_hdr.flags = VIRTIO_NET_HDR_F_NEEDS_CSUM;
 			vnet_hdr.csum_start = skb_checksum_start_offset(skb);
 			vnet_hdr.csum_offset = skb->csum_offset;
+		} else if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
+			vnet_hdr.flags = VIRTIO_NET_HDR_F_DATA_VALID;
 		} /* else everything is zero */
 
 		err = memcpy_toiovec(msg->msg_iov, (void *)&vnet_hdr,