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,
--
1.8.3.1
