Re: net/netfilter/nf_conntrack_proto_tcp.c:1606:9: error: ‘struct nf_proto_net’ has no member named ‘user’

[Pablo Neira Ayuso <pablo@netfilter.org>]

On Wed, Jun 13, 2012 at 04:51:03PM +0800, Gao feng wrote:
> 于 2012年06月13日 00:03, Pablo Neira Ayuso 写道:
> > Still, we have that pn->users thing:
> > 
> >         if (!pn->users++) {
> >                 for (i = 0; i < TCP_CONNTRACK_TIMEOUT_MAX; i++)
> >                         tn->timeouts[i] = tcp_timeouts[i];
> > 
> >                 tn->tcp_loose = nf_ct_tcp_loose;
> >                 tn->tcp_be_liberal = nf_ct_tcp_be_liberal;
> >                 tn->tcp_max_retrans = nf_ct_tcp_max_retrans;
> >         }
> > 
> > Define some pn->initialized boolean. Set it to true at the end of
> > the new tcp_init_net.
> 
> It should not be boolean,because We don't know what's the right time to set it false.
> Just like nf_conntrack_l4proto_tcp(4,6), they both use nf_tcp_net,we can set it false
> only when tcp4 and tcp6 are unregistered, so we need a referenct count here to point
> out if one proto has initialied the nf_tcp_net, or if there are no protos use nf_tcp_net.
>
> Though it maybe confusing,we can regard it as the reference count of nf_proto_net(how
> many protos are using nf_proto_net), and this reference count is increased when
> l4proto being registered success(no matter sysctl is enabled or disabled).

Yes, it makes sense indeed to regard pn->users as the refcount for the
per-net nf_proto_net. But in that case, I think we have to clean up
the code to increment pn->users everytime nf_conntrack_l[3|4]proto_register
is called, and decrement it for nf_conntrack_l[3|4]_unregister case.
Currently, by looking at the code, it seems that pn->users is some
sysctl specific refcount.