Re: connbytes & 64bit counters

[Patrick McHardy <kaber@trash.net>]

Krzysztof Oledzki wrote:
> [NETFILTER] Reimplementation of 64bit counters for bytes/packets accounting
> 
> Initially netfilter has had 64bit counters for conntrack-based
> accounting, but
> it was changed in 2.6.14 to save memory. Unfortunately in-kernel 64bit
> counters are
> still required, for example for "connbytes" extension. Add possibility
> to choose
> between 32 and 64bits but keep default 32bit counters.
> 
> Signed-off-by: Krzysztof Piotr Oledzki <ole@ans.pl>
> 
> +++ linux-2.6.19-rc3/net/netfilter/nf_conntrack_netlink.c    2006-10-28
> 18:24:12.000000000 +0200
> @@ -194,13 +194,23 @@
>  {
>      enum ctattr_type type = dir ? CTA_COUNTERS_REPLY: CTA_COUNTERS_ORIG;
>      struct nfattr *nest_count = NFA_NEST(skb, type);
> -    u_int32_t tmp;
> +#ifdef CONFIG_NF_CT_ACCT_64BIT_COUNTERS
> +    __be64 tmp;
> +
> +    tmp = cpu_to_be64(ct->counters[dir].packets);
> +    NFA_PUT(skb, CTA_COUNTERS_PACKETS, sizeof(__be64), &tmp);
> +
> +    tmp = cpu_to_be64(ct->counters[dir].bytes);
> +    NFA_PUT(skb, CTA_COUNTERS_BYTES, sizeof(__be64), &tmp);
> +#else


We can't make the API depend on config options (and I don't
like all those ifdefs). For now I would suggest to keep the
ctnetlink interface as it is and use 64 bit counters either
unconditionally or only when the connbytes match is enabled.