Commit e5f80fcf authored by Eric Dumazet's avatar Eric Dumazet Committed by David S. Miller
Browse files

ipv6: give an IPv6 dev to blackhole_netdev 



IPv6 addrconf notifiers wants the loopback device to
be the last device being dismantled at netns deletion.

This caused many limitations and work arounds.

Back in linux-5.3, Mahesh added a per host blackhole_netdev
that can be used whenever we need to make sure objects no longer
refer to a disappearing device.

If we attach to blackhole_netdev an ip6_ptr (allocate an idev),
then we can use this special device (which is never freed)
in place of the loopback_dev (which can be freed).

This will permit improvements in netdev_run_todo() and other parts
of the stack where had steps to make sure loopback_dev was
the last device to disappear.

Signed-off-by: default avatarEric Dumazet <edumazet@google.com>
Cc: Mahesh Bandewar <maheshb@google.com>
Signed-off-by: default avatarDavid S. Miller <davem@davemloft.net>
parent 2d4feb2c

net/ipv6/addrconf.c

+32 −46
Original line number Diff line number Diff line
@@ -372,7 +372,7 @@ static struct inet6_dev *ipv6_add_dev(struct net_device *dev) 


	ASSERT_RTNL();



	if (dev->mtu < IPV6_MIN_MTU)

	if (dev->mtu < IPV6_MIN_MTU && dev != blackhole_netdev)

		return ERR_PTR(-EINVAL);



	ndev = kzalloc(sizeof(struct inet6_dev), GFP_KERNEL);
@@ -400,6 +400,7 @@ static struct inet6_dev *ipv6_add_dev(struct net_device *dev) 
	/* We refer to the device */

	dev_hold_track(dev, &ndev->dev_tracker, GFP_KERNEL);



	if (dev != blackhole_netdev) {

		if (snmp6_alloc_dev(ndev) < 0) {

			netdev_dbg(dev, "%s: cannot allocate memory for statistics\n",

				   __func__);
@@ -414,7 +415,7 @@ static struct inet6_dev *ipv6_add_dev(struct net_device *dev) 
				   __func__, dev->name);

			goto err_release;

		}



	}

	/* One reference from device. */

	refcount_set(&ndev->refcnt, 1);
@@ -445,15 +446,18 @@ static struct inet6_dev *ipv6_add_dev(struct net_device *dev) 


	ipv6_mc_init_dev(ndev);

	ndev->tstamp = jiffies;

	if (dev != blackhole_netdev) {

		err = addrconf_sysctl_register(ndev);

		if (err) {

			ipv6_mc_destroy_dev(ndev);

			snmp6_unregister_dev(ndev);

			goto err_release;

		}

	}

	/* protected by rtnl_lock */

	rcu_assign_pointer(dev->ip6_ptr, ndev);



	if (dev != blackhole_netdev) {

		/* Join interface-local all-node multicast group */

		ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allnodes);
@@ -463,7 +467,7 @@ static struct inet6_dev *ipv6_add_dev(struct net_device *dev) 
		/* Join all-router multicast group if forwarding is set */

		if (ndev->cnf.forwarding && (dev->flags & IFF_MULTICAST))

			ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);



	}

	return ndev;



err_release:
@@ -7233,26 +7237,8 @@ int __init addrconf_init(void) 
		goto out_nowq;

	}



	/* The addrconf netdev notifier requires that loopback_dev

	 * has it's ipv6 private information allocated and setup

	 * before it can bring up and give link-local addresses

	 * to other devices which are up.

	 *

	 * Unfortunately, loopback_dev is not necessarily the first

	 * entry in the global dev_base list of net devices.  In fact,

	 * it is likely to be the very last entry on that list.

	 * So this causes the notifier registry below to try and

	 * give link-local addresses to all devices besides loopback_dev

	 * first, then loopback_dev, which cases all the non-loopback_dev

	 * devices to fail to get a link-local address.

	 *

	 * So, as a temporary fix, allocate the ipv6 structure for

	 * loopback_dev first by hand.

	 * Longer term, all of the dependencies ipv6 has upon the loopback

	 * device and it being up should be removed.

	 */

	rtnl_lock();

	idev = ipv6_add_dev(init_net.loopback_dev);

	idev = ipv6_add_dev(blackhole_netdev);

	rtnl_unlock();

	if (IS_ERR(idev)) {

		err = PTR_ERR(idev);

net/ipv6/route.c

+8 −13
Original line number Diff line number Diff line
@@ -156,14 +156,10 @@ void rt6_uncached_list_del(struct rt6_info *rt) 
	}

}



static void rt6_uncached_list_flush_dev(struct net *net, struct net_device *dev)

static void rt6_uncached_list_flush_dev(struct net_device *dev)

{

	struct net_device *loopback_dev = net->loopback_dev;

	int cpu;



	if (dev == loopback_dev)

		return;



	for_each_possible_cpu(cpu) {

		struct uncached_list *ul = per_cpu_ptr(&rt6_uncached_list, cpu);

		struct rt6_info *rt;
@@ -174,7 +170,7 @@ static void rt6_uncached_list_flush_dev(struct net *net, struct net_device *dev) 
			struct net_device *rt_dev = rt->dst.dev;



			if (rt_idev->dev == dev) {

				rt->rt6i_idev = in6_dev_get(loopback_dev);

				rt->rt6i_idev = in6_dev_get(blackhole_netdev);

				in6_dev_put(rt_idev);

			}
@@ -371,13 +367,12 @@ static void ip6_dst_ifdown(struct dst_entry *dst, struct net_device *dev, 
{

	struct rt6_info *rt = (struct rt6_info *)dst;

	struct inet6_dev *idev = rt->rt6i_idev;

	struct net_device *loopback_dev =

		dev_net(dev)->loopback_dev;



	if (idev && idev->dev != loopback_dev) {

		struct inet6_dev *loopback_idev = in6_dev_get(loopback_dev);

		if (loopback_idev) {

			rt->rt6i_idev = loopback_idev;

	if (idev && idev->dev != blackhole_netdev) {

		struct inet6_dev *blackhole_idev = in6_dev_get(blackhole_netdev);



		if (blackhole_idev) {

			rt->rt6i_idev = blackhole_idev;

			in6_dev_put(idev);

		}

	}
@@ -4892,7 +4887,7 @@ void rt6_sync_down_dev(struct net_device *dev, unsigned long event) 
void rt6_disable_ip(struct net_device *dev, unsigned long event)

{

	rt6_sync_down_dev(dev, event);

	rt6_uncached_list_flush_dev(dev_net(dev), dev);

	rt6_uncached_list_flush_dev(dev);

	neigh_ifdown(&nd_tbl, dev);

}