Re: [PATCH net-next v4] ipv6: add `force_forwarding` sysctl to enable per-interface forwarding

[Gabriel Goller]

On 04.07.2025 10:07, Nicolas Dichtel wrote:
> Le 03/07/2025 à 18:01, Gabriel Goller a écrit :
> > It is currently impossible to enable ipv6 forwarding on a per-interface
> > basis like in ipv4. To enable forwarding on an ipv6 interface we need to
> > enable it on all interfaces and disable it on the other interfaces using
> > a netfilter rule. This is especially cumbersome if you have lots of
> > interface and only want to enable forwarding on a few. According to the
> > sysctl docs [0] the `net.ipv6.conf.all.forwarding` enables forwarding
> > for all interfaces, while the interface-specific
> > `net.ipv6.conf.<interface>.forwarding` configures the interface
> > Host/Router configuration.
> >
> > Introduce a new sysctl flag `force_forwarding`, which can be set on every
> > interface. The ip6_forwarding function will then check if the global
> > forwarding flag OR the force_forwarding flag is active and forward the
> > packet.
> >
> > To preserver backwards-compatibility reset the flag (on all interfaces)
> > to 0 if the net.ipv6.conf.all.forwarding flag is set to 0.
> >
> > Add a short selftest that checks if a packet gets forwarded with and
> > without `force_forwarding`.
> >
> > [0]: https://www.kernel.org/doc/Documentation/networking/ip-sysctl.txt
> >
> > Signed-off-by: Gabriel Goller <g.goller@xxxxxxxxxxx>
> > ---
>
> [snip]
>
> > @@ -6747,6 +6759,78 @@ static int addrconf_sysctl_disable_policy(const struct ctl_table *ctl, int write
> >  	return ret;
> >  }
> >
> > +static void addrconf_force_forward_change(struct net *net, __s32 newf)
> > +{
> > +	ASSERT_RTNL();
> > +	struct net_device *dev;
> > +	struct inet6_dev *idev;
> > +
>
> ASSERT_RTNL() is always put after variables declaration.

I removed ASSERT_RTNL completely, this is already checked by __in6_dev_get_rtnl_net.

> > +	for_each_netdev(net, dev) {
> > +		idev = __in6_dev_get_rtnl_net(dev);
> > +		if (idev) {
> > +			int changed = (!idev->cnf.force_forwarding) ^ (!newf);
> > +
> > +			WRITE_ONCE(idev->cnf.force_forwarding, newf);
> > +			if (changed) {
> > +				inet6_netconf_notify_devconf(dev_net(dev), RTM_NEWNETCONF,
> > +							     NETCONFA_FORCE_FORWARDING,
> > +							     dev->ifindex, &idev->cnf);
> > +			}
> > +		}
> > +	}
> > +}
> > +
> > +static int addrconf_sysctl_force_forwarding(const struct ctl_table *ctl, int write,
> > +					    void *buffer, size_t *lenp, loff_t *ppos)
> > +{
> > +	struct inet6_dev *idev = ctl->extra1;
> > +	struct net *net = ctl->extra2;
> > +	int *valp = ctl->data;
> > +	loff_t pos = *ppos;
> > +	int new_val = *valp;
> > +	int old_val = *valp;
> > +	int ret;
> > +
> > +	struct ctl_table tmp_ctl = *ctl;
> This declaration should be put with other declarations.

Agree.

> > +
> > +	tmp_ctl.extra1 = SYSCTL_ZERO;
> > +	tmp_ctl.extra2 = SYSCTL_ONE;
> > +	tmp_ctl.data = &new_val;
> > +
> > +	ret = proc_douintvec_minmax(&tmp_ctl, write, buffer, lenp, ppos);
> > +
> > +	if (write && old_val != new_val) {
> > +		if (!rtnl_net_trylock(net))
> > +			return restart_syscall();
> > +
> > +		if (valp == &net->ipv6.devconf_dflt->force_forwarding) {
> > +			inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
> > +						     NETCONFA_FORCE_FORWARDING,
> > +						     NETCONFA_IFINDEX_DEFAULT,
> > +						     net->ipv6.devconf_dflt);
> > +		} else if (valp == &net->ipv6.devconf_all->force_forwarding) {
> > +			inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
> > +						     NETCONFA_FORCE_FORWARDING,
> > +						     NETCONFA_IFINDEX_ALL,
> > +						     net->ipv6.devconf_all);
> > +
> > +			addrconf_force_forward_change(net, new_val);
> > +		} else {
> > +			inet6_netconf_notify_devconf(net, RTM_NEWNETCONF,
> > +						     NETCONFA_FORCE_FORWARDING,
> > +						     idev->dev->ifindex,
> > +						     &idev->cnf);
> > +		}
> > +		rtnl_net_unlock(net);
> > +	}
> > +
> > +	if (write)
> > +		WRITE_ONCE(*valp, new_val);
> Why not putting this in the above block?
> And maybe under the rtnl_lock to avoid race if two users change the value at the
> same time.

Yep, you're right.

> Nicolas

Thanks for the review!