Re: [PATCH net-next v4 3/3] net: sched: make skip_sw actually skip software

[Marcelo Ricardo Leitner]

On Mon, Mar 25, 2024 at 08:47:36PM +0000, Asbjørn Sloth Tønnesen wrote:
..
>  +----------------------------+--------+--------+--------+
>  | tests with only skip_sw rules below:                  |
>  +----------------------------+--------+--------+--------+
>  | 1 non-matching rule        | 2694.7 | 3058.7 |  1.14x |
>  | 1 n-m rule, match trap     | 2611.2 | 3323.1 |  1.27x |
>  | 1 n-m rule, goto non-chain | 2886.8 | 2945.9 |  1.02x |
>  | 5 non-matching rules       | 1958.2 | 3061.3 |  1.56x |
>  | 5 n-m rules, match trap    | 1911.9 | 3327.0 |  1.74x |
>  | 5 n-m rules, goto non-chain| 2883.1 | 2947.5 |  1.02x |
>  | 10 non-matching rules      | 1466.3 | 3062.8 |  2.09x |
>  | 10 n-m rules, match trap   | 1444.3 | 3317.9 |  2.30x |
>  | 10 n-m rules,goto non-chain| 2883.1 | 2939.5 |  1.02x |
>  | 25 non-matching rules      |  838.5 | 3058.9 |  3.65x |
>  | 25 n-m rules, match trap   |  824.5 | 3323.0 |  4.03x |
>  | 25 n-m rules,goto non-chain| 2875.8 | 2944.7 |  1.02x |
>  | 50 non-matching rules      |  488.1 | 3054.7 |  6.26x |
                                            [A]

>  | 50 n-m rules, match trap   |  484.9 | 3318.5 |  6.84x |

Interesting. I can't explain why it consistently got 10% better than
[A] after the patch. If you check tcf_classify(), even though it
resumes to action, it still searches for the right chain. Maybe
something works differently in the driver.

In on the logs,
https://files.fiberby.net/ast/2024/tc_skip_sw/v2_tests/test_runs/netnext/tests/non_matching_and_trap_007/tc.txt

filter protocol 802.1Q pref 8 flower chain 0
filter protocol 802.1Q pref 8 flower chain 0 handle 0x1
  vlan_ethtype ip
  eth_type ipv4
  dst_ip 10.53.22.3
  skip_sw
  in_hw in_hw_count 1
	action order 1: gact action trap
	 random type none pass val 0
	 index 8 ref 1 bind 1 installed 20 sec used 0 sec
	Action statistics:
	Sent 29894330340 bytes 439622505 pkt (dropped 0, overlimits 0 requeues 0)
	Sent software 0 bytes 0 pkt
	Sent hardware 29894330340 bytes 439622505 pkt
	backlog 0b 0p requeues 0
	used_hw_stats delayed

It matched nicely.

>  | 50 n-m rules,goto non-chain| 2884.1 | 2939.7 |  1.02x |
                                   [B]

If we compare [A] and [B], there's still a 5.9% increase, plus
not requiring somewhat hacky rules.

Reviewed-by: Marcelo Ricardo Leitner <marcelo.leitner@xxxxxxxxx>