Re: [PATCH 32/32] tcp: Use dual stack sockets for port forwarding when possible

[David Gibson <david@gibson.dropbear.id.au>]

[-- Attachment #1: Type: text/plain, Size: 3121 bytes --]

On Thu, Nov 17, 2022 at 01:15:30AM +0100, Stefano Brivio wrote:
> On Wed, 16 Nov 2022 15:42:12 +1100
> David Gibson <david@gibson.dropbear.id.au> wrote:
> 
> > Platforms like Linux allow IPv6 sockets to listen for IPv4 connections as
> > well as native IPv6 connections.  By doing this we halve the number of
> > listening sockets we need for TCP (assuming passt/pasta is listening on the
> > same ports for IPv4 and IPv6).  When forwarding many ports (e.g. -t all)
> > this can significantly reduce the amount of kernel memory that passt
> > consumes.
> > 
> > When forwarding all TCP and UDP ports for both IPv4 and IPv6 (-t all
> > -u all), this reduces kernel memory usage from ~677MiB to ~487MiB
> > (kernel version 6.0.8 on Fedora 37, x86_64).
> 
> Oh, nice, that's quite significant.
> 
> > Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
> > ---
> >  tcp.c | 14 ++++++++++++--
> >  1 file changed, 12 insertions(+), 2 deletions(-)
> > 
> > diff --git a/tcp.c b/tcp.c
> > index 616b9d0..5860c9f 100644
> > --- a/tcp.c
> > +++ b/tcp.c
> > @@ -2991,8 +2991,12 @@ static int tcp_sock_init_af(const struct ctx *c, int af, in_port_t port,
> >  
> >  	s = sock_l4(c, af, IPPROTO_TCP, addr, ifname, port, tref.u32);
> >  
> > -	if (c->tcp.fwd_in.mode == FWD_AUTO)
> > -		tcp_sock_init_ext[port][(af == AF_INET) ? V4 : V6] = s;
> > +	if (c->tcp.fwd_in.mode == FWD_AUTO) {
> > +		if (af == AF_INET || af == AF_UNSPEC)
> > +			tcp_sock_init_ext[port][V4] = s;
> > +		if (af == AF_INET6 || af == AF_UNSPEC)
> 
> Nit: you could align the || af == AF_UNSPEC above with an extra
> whitespace (as it's done in the context below).

Done.

> > +			tcp_sock_init_ext[port][V6] = s;
> > +	}
> >  
> >  	if (s < 0)
> >  		return -1;
> > @@ -3012,6 +3016,12 @@ static int tcp_sock_init_af(const struct ctx *c, int af, in_port_t port,
> >  void tcp_sock_init(const struct ctx *c, sa_family_t af, const void *addr,
> >  		   const char *ifname, in_port_t port)
> >  {
> > +	if (af == AF_UNSPEC && c->ifi4 && c->ifi6)
> > +		/* Attempt to get a dual stack socket */
> > +		if (tcp_sock_init_af(c, AF_UNSPEC, port, addr, ifname) >= 0)
> > +			return;
> > +
> > +	/* Otherwise create a socket per IP version */
> 
> ...this looks surprisingly clean by the way, at least much cleaner than
> I expected.

Right.  The trick is in realizing that the properties (spliced, IP
version) of an established connection don't need to be tied to the
properties of the listening socket which created it in the first
place.

> >  	if ((af == AF_INET  || af == AF_UNSPEC) && c->ifi4)
> >  		tcp_sock_init_af(c, AF_INET, port, addr, ifname);
> >  	if ((af == AF_INET6 || af == AF_UNSPEC) && c->ifi6)
> 
> I just finished reviewing this series, in general it looks great to me,
> I would have another look (and test!) on Thursday -- either using this
> version or a re-spin.
> 

-- 
David Gibson			| I'll have my music baroque, and my code
david AT gibson.dropbear.id.au	| minimalist, thank you.  NOT _the_ _other_
				| _way_ _around_!
http://www.ozlabs.org/~dgibson

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 833 bytes --]