On Fri, Jun 17, 2016 at 12:52 PM, William Tu <[email protected]> wrote:
> The commit f2a4d086ed4c ("openvswitch: Add packet truncation support.")
> introduces packet truncation before sending to userspace upcall receiver.
> This patch passes up the skb->len before truncation so that the upcall
> receiver knows the original packet size. Potentially this will be used
> by sFlow, where OVS translates sFlow config header=N to a sample action,
> truncating packet to N byte in kernel datapath. Thus, only N bytes instead
> of full-packet size is copied from kernel to userspace, saving the
> kernel-to-userspace bandwidth.
>
> Signed-off-by: William Tu <[email protected]>
> Cc: Pravin Shelar <[email protected]>
> ---
> v2->v3:
> - remove platform specific name.
> - fix issue when receiving GSO packet
> - remove skblen in struct dp_upcall_info
> v1->v2:
> - pass skb->len to userspace instead of cutlen
> ---
> include/uapi/linux/openvswitch.h | 2 ++
> net/openvswitch/datapath.c | 11 ++++++++++-
> 2 files changed, 12 insertions(+), 1 deletion(-)
>
Patch looks good, I just have one minor comment.
> diff --git a/net/openvswitch/datapath.c b/net/openvswitch/datapath.c
> index 6739342..623fc04 100644
> --- a/net/openvswitch/datapath.c
> +++ b/net/openvswitch/datapath.c
> @@ -387,7 +387,8 @@ static size_t upcall_msg_size(const struct dp_upcall_info
> *upcall_info,
> {
> size_t size = NLMSG_ALIGN(sizeof(struct ovs_header))
> + nla_total_size(hdrlen) /* OVS_PACKET_ATTR_PACKET */
> - + nla_total_size(ovs_key_attr_size()); /* OVS_PACKET_ATTR_KEY
> */
> + + nla_total_size(ovs_key_attr_size()) /* OVS_PACKET_ATTR_KEY
> */
> + + nla_total_size(sizeof(unsigned int)); /*
> OVS_PACKET_ATTR_LEN */
>
> /* OVS_PACKET_ATTR_USERDATA */
> if (upcall_info->userdata)
> @@ -514,6 +515,14 @@ static int queue_userspace_packet(struct datapath *dp,
> struct sk_buff *skb,
> pad_packet(dp, user_skb);
> }
>
> + /* Add OVS_PACKET_ATTR_LEN */
> + if (nla_put_u32(user_skb, OVS_PACKET_ATTR_LEN,
> + skb->len)) {
> + err = -ENOBUFS;
> + goto out;
> + }
> + pad_packet(dp, user_skb);
> +
We should only send packet length in case of truncated packet. This
would avoid sending this extra attribute in miss flow handling case.
