On 04/26/2016 09:55 AM, Jason Wang wrote:
> There's no need to calculate rps hash if it was not enabled. So this
> patch export rps_needed and check it before trying to get rps
> hash. Tests (using pktgen to inject packets to guest) shows this can
> improve pps about 13% (when rps is disabled).
>
> Before:
> ~1150000 pps
> After:
> ~1300000 pps
>
> Cc: Michael S. Tsirkin <m...@redhat.com>
> Signed-off-by: Jason Wang <jasow...@redhat.com>
> ---
> drivers/net/tun.c | 2 +-
> net/core/dev.c | 1 +
> 2 files changed, 2 insertions(+), 1 deletion(-)
>
> diff --git a/drivers/net/tun.c b/drivers/net/tun.c
> index afdf950..746877f 100644
> --- a/drivers/net/tun.c
> +++ b/drivers/net/tun.c
> @@ -819,7 +819,7 @@ static netdev_tx_t tun_net_xmit(struct sk_buff *skb,
> struct net_device *dev)
> if (txq >= numqueues)
> goto drop;
>
> - if (numqueues == 1) {
> + if (numqueues == 1 && static_key_false(&rps_needed)) {
> /* Select queue was not called for the skbuff, so we extract the
> * RPS hash and save it into the flow_table here.
> */
> diff --git a/net/core/dev.c b/net/core/dev.c
> index b9bcbe7..d4ba936 100644
> --- a/net/core/dev.c
> +++ b/net/core/dev.c
> @@ -3428,6 +3428,7 @@ u32 rps_cpu_mask __read_mostly;
> EXPORT_SYMBOL(rps_cpu_mask);
>
> struct static_key rps_needed __read_mostly;
> +EXPORT_SYMBOL(rps_needed);
>
> static struct rps_dev_flow *
> set_rps_cpu(struct net_device *dev, struct sk_buff *skb,
Kbuild bot reports an error when !CONFIG_RPS. Will send V2 to fix this.
