Re: [PATCH 5/6] netfilter: use sequence number synchronization for counters

[Eric Dumazet <dada1@cosmosbay.com>]

Stephen Hemminger a écrit :
> Change how synchronization is done on the iptables counters. Use seqcount
> wrapper instead of depending on reader/writer lock.
> 
> Signed-off-by: Stephen Hemminger <shemminger@vyatta.com>
> 

> --- a/net/netfilter/x_tables.c	2009-01-29 11:08:38.747070716 -0800
> +++ b/net/netfilter/x_tables.c	2009-01-29 11:10:03.595571234 -0800
> @@ -577,6 +577,34 @@ int xt_compat_target_to_user(struct xt_e
>  EXPORT_SYMBOL_GPL(xt_compat_target_to_user);
>  #endif
>  
> +static DEFINE_PER_CPU(seqcount_t, xt_counter_sequence);
> +
> +void xt_fetch_counter(struct xt_counters *v, int cpu,
> +		      const struct xt_counters *c)
> +{
> +	seqcount_t *seq = &per_cpu(xt_counter_sequence, cpu);
> +	unsigned start;
> +
> +	do {
> +		start = read_seqcount_begin(seq);
> +		*v = *c;
> +	} while (read_seqcount_retry(seq, start));
> +}
> +EXPORT_SYMBOL_GPL(xt_fetch_counter);
> +
> +void xt_incr_counter(struct xt_counters *c, unsigned b, unsigned p)

You really want an inline xt_incr_counter() function here to speedup ipt_do_table()

I agree xt_fetch_counter() is not time critical and can be outlined.



> +{
> +	seqcount_t *seq = &__get_cpu_var(xt_counter_sequence);
> +
> +	write_seqcount_begin(seq);
> +	c->pcnt += p;
> +	c->bcnt += b;
> +	write_seqcount_end(seq);
> +
> +}
> +EXPORT_SYMBOL_GPL(xt_incr_counter);
> +
> +