Re: [PATCH net-next 2/2] net/sched: act_police: don't use spinlock in the data path

[Eric Dumazet]

On 09/13/2018 10:29 AM, Davide Caratti wrote:
> use RCU instead of spinlocks, to protect concurrent read/write on
> act_police configuration. This reduces the effects of contention in the
> data path, in case multiple readers are present.
> 
> Signed-off-by: Davide Caratti <dcara...@redhat.com>
> ---
>  net/sched/act_police.c | 156 ++++++++++++++++++++++++-----------------
>  1 file changed, 92 insertions(+), 64 deletions(-)
> 

I must be missing something obvious with this patch.

How can the following piece of code in tcf_police_act() can possibly be run
without a spinlock or something preventing multiple cpus messing badly with the 
state variables  ?


                now = ktime_get_ns();
                toks = min_t(s64, now - p->tcfp_t_c, p->tcfp_burst);
                if (p->peak_present) {
                        ptoks = toks + p->tcfp_ptoks;
                        if (ptoks > p->tcfp_mtu_ptoks)
                                ptoks = p->tcfp_mtu_ptoks;
                        ptoks -= (s64)psched_l2t_ns(&p->peak,
                                                    qdisc_pkt_len(skb));
                }
                toks += p->tcfp_toks;
                if (toks > p->tcfp_burst)
                        toks = p->tcfp_burst;
                toks -= (s64)psched_l2t_ns(&p->rate, qdisc_pkt_len(skb));
                if ((toks|ptoks) >= 0) {
                        p->tcfp_t_c = now;
                        p->tcfp_toks = toks;
                        p->tcfp_ptoks = ptoks;
                        ret = p->tcfp_result;
                        goto inc_drops;
                }