[patch net-next RFC] net: sched: cls_api: make reclassify return all the way back to the original tp

[patch net-next RFC] net: sched: cls_api: make reclassify return all the way back to the original tp

[Jiri Pirko]

From: Jiri Pirko <jiri@mellanox.com>

With the introduction of chain goto action, the reclassification would
cause the re-iteration of the actual chain. But it perhaps makes more
sense to restart the whole thing. Thoughts?

Signed-off-by: Jiri Pirko <jiri@mellanox.com>
---
 net/sched/cls_api.c | 8 +++++---
 1 file changed, 5 insertions(+), 3 deletions(-)

diff --git a/net/sched/cls_api.c b/net/sched/cls_api.c
index 01a8b8b..89fbb35 100644
--- a/net/sched/cls_api.c
+++ b/net/sched/cls_api.c
@@ -300,7 +300,8 @@ int tcf_classify(struct sk_buff *skb, const struct tcf_proto *tp,
 	__be16 protocol = tc_skb_protocol(skb);
 #ifdef CONFIG_NET_CLS_ACT
 	const int max_reclassify_loop = 4;
-	const struct tcf_proto *old_tp = tp;
+	const struct tcf_proto *orig_tp = tp;
+	const struct tcf_proto *first_tp;
 	int limit = 0;
 
 reclassify:
@@ -315,9 +316,10 @@ int tcf_classify(struct sk_buff *skb, const struct tcf_proto *tp,
 		err = tp->classify(skb, tp, res);
 #ifdef CONFIG_NET_CLS_ACT
 		if (unlikely(err == TC_ACT_RECLASSIFY && !compat_mode)) {
+			first_tp = orig_tp;
 			goto reset;
 		} else if (unlikely(TC_ACT_EXT_CMP(err, TC_ACT_GOTO_CHAIN))) {
-			old_tp = res->goto_tp;
+			first_tp = res->goto_tp;
 			goto reset;
 		}
 #endif
@@ -335,7 +337,7 @@ int tcf_classify(struct sk_buff *skb, const struct tcf_proto *tp,
 		return TC_ACT_SHOT;
 	}
 
-	tp = old_tp;
+	tp = first_tp;
 	protocol = tc_skb_protocol(skb);
 	goto reclassify;
 #endif
-- 
2.9.3

Re: [patch net-next RFC] net: sched: cls_api: make reclassify return all the way back to the original tp

[Cong Wang]

On Mon, May 22, 2017 at 8:09 AM, Jiri Pirko <jiri@resnulli.us> wrote:
> From: Jiri Pirko <jiri@mellanox.com>
>
> With the introduction of chain goto action, the reclassification would
> cause the re-iteration of the actual chain. But it perhaps makes more
> sense to restart the whole thing. Thoughts?

I think reclassification is meant to restart the whole logic rather
than just one chain. So your patch makes sense to me, but not
sure if there is any corner case I miss.

Re: [patch net-next RFC] net: sched: cls_api: make reclassify return all the way back to the original tp

[Jiri Pirko]

Tue, May 23, 2017 at 01:57:47AM CEST, xiyou.wangcong@gmail.com wrote:
>On Mon, May 22, 2017 at 8:09 AM, Jiri Pirko <jiri@resnulli.us> wrote:
>> From: Jiri Pirko <jiri@mellanox.com>
>>
>> With the introduction of chain goto action, the reclassification would
>> cause the re-iteration of the actual chain. But it perhaps makes more
>> sense to restart the whole thing. Thoughts?
>
>I think reclassification is meant to restart the whole logic rather
>than just one chain. So your patch makes sense to me, but not
>sure if there is any corner case I miss.

Agreed. Will send v1. Thanks