Re: [PATCH] conf, pasta: Add --no-tap option

[Stefano Brivio <sbrivio@redhat.com>]

On Tue, 13 Jan 2026 19:20:47 +0800
Yumei Huang <yuhuang@redhat.com> wrote:

> On Sun, Jan 11, 2026 at 2:12 AM Stefano Brivio <sbrivio@redhat.com> wrote:
> >
> > On Mon, 29 Dec 2025 17:55:58 +0800
> > Yumei Huang <yuhuang@redhat.com> wrote:
> >  
> > > This patch introduces a mode where we only forward loopback connections
> > > and traffic between two namespaces (via the loopback interface, 'lo'),
> > > without a tap device.
> > >
> > > With this, podman can support forwarding ::1 in custom networks when using
> > > rootlesskit for forwarding ports.
> > >
> > > In --no-tap mode, --host-lo-to-ns-lo, --no-icmp and --no-ra is automatically
> > > enabled. Options requiring a tap device (--ns-ifname, --ns-mac-addr,
> > > --config-net, --outbound-if4/6) are rejected.
> > >
> > > Link: https://bugs.passt.top/show_bug.cgi?id=149
> > > Signed-off-by: Yumei Huang <yuhuang@redhat.com>
> > > ---
> > >  conf.c  | 56 +++++++++++++++++++++++++++++++++++++++++---------------
> > >  fwd.c   |  3 +++
> > >  passt.1 |  5 +++++
> > >  passt.h |  2 ++
> > >  pasta.c |  3 +++
> > >  tap.c   | 11 +++++++----
> > >  6 files changed, 61 insertions(+), 19 deletions(-)
> > >
> > > diff --git a/conf.c b/conf.c
> > > index 84ae12b..353d0a5 100644
> > > --- a/conf.c
> > > +++ b/conf.c
> > > @@ -1049,7 +1049,8 @@ pasta_opts:
> > >               "  --no-copy-addrs      DEPRECATED:\n"
> > >               "                       Don't copy all addresses to namespace\n"
> > >               "  --ns-mac-addr ADDR   Set MAC address on tap interface\n"
> > > -             "  --no-splice          Disable inbound socket splicing\n");
> > > +             "  --no-splice          Disable inbound socket splicing\n"
> > > +             "  --no-tap             Don't create tap device\n");
> > >
> > >       passt_exit(status);
> > >  }
> > > @@ -1451,6 +1452,7 @@ void conf(struct ctx *c, int argc, char **argv)
> > >               {"no-ndp",      no_argument,            &c->no_ndp,     1 },
> > >               {"no-ra",       no_argument,            &c->no_ra,      1 },
> > >               {"no-splice",   no_argument,            &c->no_splice,  1 },
> > > +             {"no-tap",      no_argument,            &c->no_tap,     1 },
> > >               {"freebind",    no_argument,            &c->freebind,   1 },
> > >               {"no-map-gw",   no_argument,            &no_map_gw,     1 },
> > >               {"ipv4-only",   no_argument,            NULL,           '4' },
> > > @@ -1947,8 +1949,11 @@ void conf(struct ctx *c, int argc, char **argv)
> > >               }
> > >       } while (name != -1);
> > >
> > > -     if (c->mode != MODE_PASTA)
> > > +     if (c->mode != MODE_PASTA) {
> > >               c->no_splice = 1;
> > > +             if (c->no_tap)
> > > +                     die("--no-tap is for pasta mode only");
> > > +     }
> > >
> > >       if (c->mode == MODE_PASTA && !c->pasta_conf_ns) {
> > >               if (copy_routes_opt)
> > > @@ -1957,6 +1962,25 @@ void conf(struct ctx *c, int argc, char **argv)
> > >                       die("--no-copy-addrs needs --config-net");
> > >       }
> > >
> > > +     if (c->mode == MODE_PASTA && c->no_tap) {
> > > +             if (c->no_splice)
> > > +                     die("--no-tap is incompatible with --no-splice");  
> >
> > I'm not sure if you need this for other reasons, but as long as it's
> > called --no-tap, it's not really incompatible with --no-splice.  
> 
> I will update it to --splice-only
> 
> >
> > Maybe users just want to get a disconnected namespace for whatever
> > reason ('pasta' is shorter to type than 'unshare -rUn').
> >  
> > > +             if (*c->ip4.ifname_out || *c->ip6.ifname_out)
> > > +                     die("--no-tap is incompatible with --outbound-if4/6");
> > > +             if (*c->pasta_ifn)
> > > +                     die("--no-tap is incompatible with --ns-ifname");
> > > +             if (*c->guest_mac)
> > > +                     die("--no-tap is incompatible with --ns-mac-addr");
> > > +             if (c->pasta_conf_ns)
> > > +                     die("--no-tap is incompatible with --config-net");  
> >
> > I guess all these checks are to save some checks later, which looks like
> > a good reason to have them here.
> >
> > If not, though, I don't think we *really* need to tell the user that
> > --ns-ifname will be ignored with --no-tap.
> >
> > One thing that might confuse users, though, is this:
> >
> > $ ./pasta --no-tap --mtu 1500 -- ip l
> > 1: lo: <LOOPBACK,UP,LOWER_UP> mtu 65536 qdisc noqueue state UNKNOWN mode DEFAULT group default qlen 1000
> >     link/loopback 00:00:00:00:00:00 brd 00:00:00:00:00:00
> >
> > or even this:
> >
> > $ ./pasta --no-tap -a 192.0.2.1 -- ip a
> > 1: lo: <LOOPBACK,UP,LOWER_UP> mtu 65536 qdisc noqueue state UNKNOWN group default qlen 1000
> >     link/loopback 00:00:00:00:00:00 brd 00:00:00:00:00:00
> >     inet 127.0.0.1/8 scope host lo
> >        valid_lft forever preferred_lft forever
> >     inet6 ::1/128 scope host proto kernel_lo
> >        valid_lft forever preferred_lft forever
> >
> > but I would rather *not* add conditions and checks for those even if
> > there's a *slight* potential for confusion, otherwise this becomes
> > really long. And it's really not worth it, I think.  
> 
> Then I guess we only need the c->no_splice check, right?

...maybe? About *needing*, yes, I guess so, but if other checks save
more checks later, I would keep them.

> > > +
> > > +             c->host_lo_to_ns_lo = 1;
> > > +             c->no_icmp = 1;
> > > +             c->no_ra = 1;
> > > +             c->no_dns = 1;
> > > +             c->no_dns_search = 1;
> > > +     }
> > > +
> > >       if (!ifi4 && *c->ip4.ifname_out)
> > >               ifi4 = if_nametoindex(c->ip4.ifname_out);
> > >
> > > @@ -1980,9 +2004,9 @@ void conf(struct ctx *c, int argc, char **argv)
> > >       log_conf_parsed = true;         /* Stop printing everything */
> > >
> > >       nl_sock_init(c, false);
> > > -     if (!v6_only)
> > > +     if (!v6_only && !c->no_tap)
> > >               c->ifi4 = conf_ip4(ifi4, &c->ip4);
> > > -     if (!v4_only)
> > > +     if (!v4_only && !c->no_tap)
> > >               c->ifi6 = conf_ip6(ifi6, &c->ip6);
> > >
> > >       if (c->ifi4 && c->mtu < IPV4_MIN_MTU) {
> > > @@ -1998,30 +2022,32 @@ void conf(struct ctx *c, int argc, char **argv)
> > >           (*c->ip6.ifname_out && !c->ifi6))
> > >               die("External interface not usable");
> > >
> > > -     if (!c->ifi4 && !c->ifi6 && !*c->pasta_ifn) {
> > > +     if (!c->ifi4 && !c->ifi6 && !*c->pasta_ifn && !c->no_tap) {  
> >
> > You already checked that !*c->pasta_ifn above.  
> 
> I guess the check above (aka. if (*c->pasta_ifn && c->no_tap)) doesn't
> affect this one? If c->pasta_ifn is assigned, we won't come to the
> check  !c->no_tap here. Otherwise, we do need to check !c->no_tap.

Right, but you don't care about resetting c->pasta_ifn to the default
value if !c->no_tap, because in that case you know that c->pasta_ifn
wasn't set, so you can happily override it.

I guess, at least, I haven't thoroughly checked what might happen later
with it.

> >  
> > >               strncpy(c->pasta_ifn, pasta_default_ifn,
> > >                       sizeof(c->pasta_ifn) - 1);
> > >       }
> > >
> > >       if (!c->ifi4 && !v6_only) {
> > > -             info("IPv4: no external interface as template, use local mode");
> > > -
> > > -             conf_ip4_local(&c->ip4);
> > > +             if (!c->no_tap) {
> > > +                     info("IPv4: no external interface as template, use local mode");
> > > +                     conf_ip4_local(&c->ip4);
> > > +             }
> > >               c->ifi4 = -1;
> > >       }
> > >
> > >       if (!c->ifi6 && !v4_only) {
> > > -             info("IPv6: no external interface as template, use local mode");
> > > -
> > > -             conf_ip6_local(&c->ip6);
> > > +             if (!c->no_tap) {
> > > +                     info("IPv6: no external interface as template, use local mode");
> > > +                     conf_ip6_local(&c->ip6);
> > > +             }
> > >               c->ifi6 = -1;
> > >       }
> > >
> > > -     if (c->ifi4 && !no_map_gw &&
> > > +     if (c->ifi4 > 0 && !no_map_gw &&
> > >           IN4_IS_ADDR_UNSPECIFIED(&c->ip4.map_host_loopback))
> > >               c->ip4.map_host_loopback = c->ip4.guest_gw;
> > >
> > > -     if (c->ifi6 && !no_map_gw &&
> > > +     if (c->ifi6 > 0 && !no_map_gw &&
> > >           IN6_IS_ADDR_UNSPECIFIED(&c->ip6.map_host_loopback))
> > >               c->ip6.map_host_loopback = c->ip6.guest_gw;
> > >
> > > @@ -2116,10 +2142,10 @@ void conf(struct ctx *c, int argc, char **argv)
> > >                       conf_ports(c, name, optarg, &c->udp.fwd_out);
> > >       } while (name != -1);
> > >
> > > -     if (!c->ifi4)
> > > +     if (c->ifi4 <= 0)
> > >               c->no_dhcp = 1;
> > >
> > > -     if (!c->ifi6) {
> > > +     if (c->ifi6 <= 0) {
> > >               c->no_ndp = 1;
> > >               c->no_dhcpv6 = 1;
> > >       } else if (IN6_IS_ADDR_UNSPECIFIED(&c->ip6.addr)) {
> > > diff --git a/fwd.c b/fwd.c
> > > index 44a0e10..2f4a89a 100644
> > > --- a/fwd.c
> > > +++ b/fwd.c
> > > @@ -780,6 +780,9 @@ uint8_t fwd_nat_from_host(const struct ctx *c, uint8_t proto,
> > >               return PIF_SPLICE;
> > >       }
> > >
> > > +     if (c->no_tap)
> > > +             return PIF_NONE;
> > > +
> > >       if (!nat_inbound(c, &ini->eaddr, &tgt->oaddr)) {
> > >               if (inany_v4(&ini->eaddr)) {
> > >                       if (IN4_IS_ADDR_UNSPECIFIED(&c->ip4.our_tap_addr))
> > > diff --git a/passt.1 b/passt.1
> > > index db0d662..2d643f7 100644
> > > --- a/passt.1
> > > +++ b/passt.1
> > > @@ -755,6 +755,11 @@ Default is to let the tap driver build a pseudorandom hardware address.
> > >  Disable the bypass path for inbound, local traffic. See the section \fBHandling
> > >  of local traffic in pasta\fR in the \fBNOTES\fR for more details.
> > >
> > > +.TP
> > > +.BR \-\-no-tap
> > > +Do not create a tap device in the namespace. In this mode, only local loopback
> > > +traffic between namespaces is forwarded using splice.  
> >
> > "Using splice" isn't really clear, and it's not entirely correct in the
> > case of UDP: there's no splice() system call there, even if we call
> > some UDP flows "spliced" for analogy with TCP.
> >
> > Maybe just omit it, say:
> >
> >   [...] In this mode, \fIpasta\fR only
> >   forwards loopback traffic between namespaces.  
> 
> Do you think we still need "Do not create a tap device in the
> namespace" after updating it to --splice-only?

Yes, I think so, because otherwise it might look like we create the tap
device anyway, but discard all traffic from/to it.

-- 
Stefano