[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <20210126183812.180d0d61@kicinski-fedora-pc1c0hjn.dhcp.thefacebook.com>
Date: Tue, 26 Jan 2021 18:38:12 -0800
From: Jakub Kicinski <kuba@...nel.org>
To: Simon Horman <simon.horman@...ronome.com>
Cc: Jamal Hadi Salim <jhs@...atatu.com>,
Cong Wang <xiyou.wangcong@...il.com>,
Jiri Pirko <jiri@...lanox.com>, netdev@...r.kernel.org,
oss-drivers@...ronome.com,
Baowen Zheng <baowen.zheng@...igine.com>,
Louis Peens <louis.peens@...ronome.com>
Subject: Re: [PATCH RFC net-next] net/sched: act_police: add support for
packet-per-second policing
On Mon, 25 Jan 2021 16:18:19 +0100 Simon Horman wrote:
> From: Baowen Zheng <baowen.zheng@...igine.com>
>
> Allow a policer action to enforce a rate-limit based on packets-per-second,
> configurable using a packet-per-second rate and burst parameters. This may
> be used in conjunction with existing byte-per-second rate limiting in the
> same policer action.
>
> e.g.
> tc filter add dev tap1 parent ffff: u32 match \
> u32 0 0 police pkts_rate 3000 pkts_burst 1000
>
> Testing was unable to uncover a performance impact of this change on
> existing features.
>
> Signed-off-by: Baowen Zheng <baowen.zheng@...igine.com>
> Signed-off-by: Simon Horman <simon.horman@...ronome.com>
> Signed-off-by: Louis Peens <louis.peens@...ronome.com>
> diff --git a/net/sched/act_police.c b/net/sched/act_police.c
> index 8d8452b1cdd4..d700b2105535 100644
> --- a/net/sched/act_police.c
> +++ b/net/sched/act_police.c
> @@ -42,6 +42,8 @@ static const struct nla_policy police_policy[TCA_POLICE_MAX + 1] = {
> [TCA_POLICE_RESULT] = { .type = NLA_U32 },
> [TCA_POLICE_RATE64] = { .type = NLA_U64 },
> [TCA_POLICE_PEAKRATE64] = { .type = NLA_U64 },
> + [TCA_POLICE_PKTRATE64] = { .type = NLA_U64 },
> + [TCA_POLICE_PKTBURST64] = { .type = NLA_U64 },
Should we set the policy so that .min = 1?
> };
>
> static int tcf_police_init(struct net *net, struct nlattr *nla,
> @@ -61,6 +63,7 @@ static int tcf_police_init(struct net *net, struct nlattr *nla,
> bool exists = false;
> u32 index;
> u64 rate64, prate64;
> + u64 pps, ppsburst;
>
> if (nla == NULL)
> return -EINVAL;
> @@ -183,6 +186,16 @@ static int tcf_police_init(struct net *net, struct nlattr *nla,
> if (tb[TCA_POLICE_AVRATE])
> new->tcfp_ewma_rate = nla_get_u32(tb[TCA_POLICE_AVRATE]);
>
> + if (tb[TCA_POLICE_PKTRATE64] && tb[TCA_POLICE_PKTBURST64]) {
Should we reject if only one is present?
> + pps = nla_get_u64(tb[TCA_POLICE_PKTRATE64]);
> + ppsburst = nla_get_u64(tb[TCA_POLICE_PKTBURST64]);
> + if (pps) {
> + new->pps_present = true;
> + new->tcfp_pkt_burst = PSCHED_TICKS2NS(ppsburst);
> + psched_ppscfg_precompute(&new->ppsrate, pps);
> + }
> + }
> +
> spin_lock_bh(&police->tcf_lock);
> spin_lock_bh(&police->tcfp_lock);
> police->tcfp_t_c = ktime_get_ns();
> +void psched_ppscfg_precompute(struct psched_pktrate *r,
> + u64 pktrate64)
> +{
> + memset(r, 0, sizeof(*r));
> + r->rate_pkts_ps = pktrate64;
> + r->mult = 1;
> + /* The deal here is to replace a divide by a reciprocal one
> + * in fast path (a reciprocal divide is a multiply and a shift)
> + *
> + * Normal formula would be :
> + * time_in_ns = (NSEC_PER_SEC * pkt_num) / pktrate64
> + *
> + * We compute mult/shift to use instead :
> + * time_in_ns = (len * mult) >> shift;
> + *
> + * We try to get the highest possible mult value for accuracy,
> + * but have to make sure no overflows will ever happen.
> + */
> + if (r->rate_pkts_ps > 0) {
> + u64 factor = NSEC_PER_SEC;
> +
> + for (;;) {
> + r->mult = div64_u64(factor, r->rate_pkts_ps);
> + if (r->mult & (1U << 31) || factor & (1ULL << 63))
> + break;
> + factor <<= 1;
> + r->shift++;
Aren't there helpers somewhere for the reciprocal divide
pre-calculation?
> + }
> + }
> +}
> +EXPORT_SYMBOL(psched_ppscfg_precompute);
Powered by blists - more mailing lists