On 10/13/2017 08:30 AM, Jiri Pirko wrote:
Thu, Oct 12, 2017 at 11:45:43PM CEST, dan...@iogearbox.net wrote:
On 10/12/2017 07:17 PM, Jiri Pirko wrote:
From: Jiri Pirko <j...@mellanox.com>
In sch_handle_egress and sch_handle_ingress, don't use tp->q and use
dev->ingress_queue which stores the same pointer instead.
Signed-off-by: Jiri Pirko <j...@mellanox.com>
---
net/core/dev.c | 21 +++++++++++++++------
1 file changed, 15 insertions(+), 6 deletions(-)
diff --git a/net/core/dev.c b/net/core/dev.c
index fcddccb..cb9e5e5 100644
--- a/net/core/dev.c
+++ b/net/core/dev.c
@@ -3273,14 +3273,18 @@ EXPORT_SYMBOL(dev_loopback_xmit);
static struct sk_buff *
sch_handle_egress(struct sk_buff *skb, int *ret, struct net_device *dev)
{
+ struct netdev_queue *netdev_queue =
+ rcu_dereference_bh(dev->ingress_queue);
struct tcf_proto *cl = rcu_dereference_bh(dev->egress_cl_list);
struct tcf_result cl_res;
+ struct Qdisc *q;
- if (!cl)
+ if (!cl || !netdev_queue)
return skb;
+ q = netdev_queue->qdisc;
NAK, no additional overhead in the software fast-path of
sch_handle_{ingress,egress}() like this. There are users out there
that use tc in software only, so performance is critical here.
Okay, how else do you suggest I can avoid the need to use tp->q?
I was thinking about storing q directly to net_device, which would safe
one dereference, resulting in the same amount as current cl->q.
Sorry for late reply, mostly off for few days. netdev struct has different
cachelines which are hot on rx and tx (see also the location of the two
lists, ingress_cl_list and egress_cl_list), if you add only one qdisc
pointer there, then you'd at least penalize one of the two w/ potential
cache miss. Can we leave it in cl?
