[PATCH v1 net-next 00/13] mpls: Remove RTNL dependency.

[PATCH v1 net-next 00/13] mpls: Remove RTNL dependency.

[Kuniyuki Iwashima]

MPLS uses RTNL

  1) to guarantee the lifetime of struct mpls_nh.nh_dev
  2) to protect net->mpls.platform_label

, but neither actually requires RTNL.

If struct mpls_nh holds a refcnt for nh_dev, we do not need RTNL,
and it can be replaced with a dedicated mutex.

The series removes RTNL from net/mpls/.

Overview:

  Patch 1 is misc cleanup.

  Patch 2 - 9 are prep to drop RTNL for RTM_{NEW,DEL,GET}ROUTE
  handlers.

  Patch 10 & 11 converts mpls_dump_routes() and RTM_GETNETCONF to RCU.

  Patch 12 replaces RTNL with a new per-netns mutex.

  Patch 13 drops RTNL from RTM_{NEW,DEL,GET}ROUTE.


Kuniyuki Iwashima (13):
  mpls: Return early in mpls_label_ok().
  mpls: Hold dev refcnt for mpls_nh.
  mpls: Unify return paths in mpls_dev_notify().
  ipv6: Add in6_dev_rcu().
  mpls: Use in6_dev_rcu() and dev_net_rcu() in mpls_forward() and
    mpls_xmit().
  mpls: Add mpls_dev_rcu().
  mpls: Pass net to mpls_dev_get().
  mpls: Add mpls_route_input().
  mpls: Use mpls_route_input() where appropriate.
  mpls: Convert mpls_dump_routes() to RCU.
  mpls: Convert RTM_GETNETCONF to RCU.
  mpls: Protect net->mpls.platform_label with a per-netns mutex.
  mpls: Drop RTNL for RTM_NEWROUTE, RTM_DELROUTE, and RTM_GETROUTE.

 include/net/addrconf.h   |   5 +
 include/net/netns/mpls.h |   1 +
 net/mpls/af_mpls.c       | 320 ++++++++++++++++++++++++---------------
 net/mpls/internal.h      |  19 ++-
 net/mpls/mpls_iptunnel.c |   6 +-
 5 files changed, 224 insertions(+), 127 deletions(-)

-- 
2.51.1.838.g19442a804e-goog

[PATCH v1 net-next 01/13] mpls: Return early in mpls_label_ok().

[Kuniyuki Iwashima]

When mpls_label_ok() returns false, it does not need to update *index.

Let's remove is_ok and return early.

Signed-off-by: Kuniyuki Iwashima <kuniyu@google.com>
---
 net/mpls/af_mpls.c | 11 +++++------
 1 file changed, 5 insertions(+), 6 deletions(-)

diff --git a/net/mpls/af_mpls.c b/net/mpls/af_mpls.c
index 25c88cba5c48..e3533d85d372 100644
--- a/net/mpls/af_mpls.c
+++ b/net/mpls/af_mpls.c
@@ -940,24 +940,23 @@ static int mpls_nh_build_multi(struct mpls_route_config *cfg,
 static bool mpls_label_ok(struct net *net, unsigned int *index,
 			  struct netlink_ext_ack *extack)
 {
-	bool is_ok = true;
-
 	/* Reserved labels may not be set */
 	if (*index < MPLS_LABEL_FIRST_UNRESERVED) {
 		NL_SET_ERR_MSG(extack,
 			       "Invalid label - must be MPLS_LABEL_FIRST_UNRESERVED or higher");
-		is_ok = false;
+		return false;
 	}
 
 	/* The full 20 bit range may not be supported. */
-	if (is_ok && *index >= net->mpls.platform_labels) {
+	if (*index >= net->mpls.platform_labels) {
 		NL_SET_ERR_MSG(extack,
 			       "Label >= configured maximum in platform_labels");
-		is_ok = false;
+		return false;
 	}
 
 	*index = array_index_nospec(*index, net->mpls.platform_labels);
-	return is_ok;
+
+	return true;
 }
 
 static int mpls_route_add(struct mpls_route_config *cfg,
-- 
2.51.1.838.g19442a804e-goog

[PATCH v1 net-next 02/13] mpls: Hold dev refcnt for mpls_nh.

[Kuniyuki Iwashima]

MPLS uses RTNL

  1) to guarantee the lifetime of struct mpls_nh.nh_dev
  2) to protect net->mpls.platform_label

, but neither actually requires RTNL.

If we do not call dev_put() in find_outdev() and call it
just before freeing struct mpls_route, we can drop RTNL for 1).

Let's hold the refcnt of mpls_nh.nh_dev and track it with
netdevice_tracker.

Two notable changes:

If mpls_nh_build_multi() fails to set up a neighbour, we need
to call netdev_put() for successfully created neighbours in
mpls_rt_free_rcu(), so the number of neighbours (rt->rt_nhn)
is now updated in each iteration.

When a dev is unregistered, mpls_ifdown() clones mpls_route
and replaces it with the clone, so the clone requires extra
netdev_hold().

Signed-off-by: Kuniyuki Iwashima <kuniyu@google.com>
---
 net/mpls/af_mpls.c  | 63 +++++++++++++++++++++++++++++++--------------
 net/mpls/internal.h |  1 +
 2 files changed, 45 insertions(+), 19 deletions(-)

diff --git a/net/mpls/af_mpls.c b/net/mpls/af_mpls.c
index e3533d85d372..e7be87466809 100644
--- a/net/mpls/af_mpls.c
+++ b/net/mpls/af_mpls.c
@@ -530,10 +530,23 @@ static struct mpls_route *mpls_rt_alloc(u8 num_nh, u8 max_alen, u8 max_labels)
 	return rt;
 }
 
+static void mpls_rt_free_rcu(struct rcu_head *head)
+{
+	struct mpls_route *rt;
+
+	rt = container_of(head, struct mpls_route, rt_rcu);
+
+	change_nexthops(rt) {
+		netdev_put(nh->nh_dev, &nh->nh_dev_tracker);
+	} endfor_nexthops(rt);
+
+	kfree(rt);
+}
+
 static void mpls_rt_free(struct mpls_route *rt)
 {
 	if (rt)
-		kfree_rcu(rt, rt_rcu);
+		call_rcu(&rt->rt_rcu, mpls_rt_free_rcu);
 }
 
 static void mpls_notify_route(struct net *net, unsigned index,
@@ -587,6 +600,7 @@ static unsigned find_free_label(struct net *net)
 
 #if IS_ENABLED(CONFIG_INET)
 static struct net_device *inet_fib_lookup_dev(struct net *net,
+					      struct mpls_nh *nh,
 					      const void *addr)
 {
 	struct net_device *dev;
@@ -599,14 +613,14 @@ static struct net_device *inet_fib_lookup_dev(struct net *net,
 		return ERR_CAST(rt);
 
 	dev = rt->dst.dev;
-	dev_hold(dev);
-
+	netdev_hold(dev, &nh->nh_dev_tracker, GFP_KERNEL);
 	ip_rt_put(rt);
 
 	return dev;
 }
 #else
 static struct net_device *inet_fib_lookup_dev(struct net *net,
+					      struct mpls_nh *nh,
 					      const void *addr)
 {
 	return ERR_PTR(-EAFNOSUPPORT);
@@ -615,6 +629,7 @@ static struct net_device *inet_fib_lookup_dev(struct net *net,
 
 #if IS_ENABLED(CONFIG_IPV6)
 static struct net_device *inet6_fib_lookup_dev(struct net *net,
+					       struct mpls_nh *nh,
 					       const void *addr)
 {
 	struct net_device *dev;
@@ -631,13 +646,14 @@ static struct net_device *inet6_fib_lookup_dev(struct net *net,
 		return ERR_CAST(dst);
 
 	dev = dst->dev;
-	dev_hold(dev);
+	netdev_hold(dev, &nh->nh_dev_tracker, GFP_KERNEL);
 	dst_release(dst);
 
 	return dev;
 }
 #else
 static struct net_device *inet6_fib_lookup_dev(struct net *net,
+					       struct mpls_nh *nh,
 					       const void *addr)
 {
 	return ERR_PTR(-EAFNOSUPPORT);
@@ -653,16 +669,17 @@ static struct net_device *find_outdev(struct net *net,
 	if (!oif) {
 		switch (nh->nh_via_table) {
 		case NEIGH_ARP_TABLE:
-			dev = inet_fib_lookup_dev(net, mpls_nh_via(rt, nh));
+			dev = inet_fib_lookup_dev(net, nh, mpls_nh_via(rt, nh));
 			break;
 		case NEIGH_ND_TABLE:
-			dev = inet6_fib_lookup_dev(net, mpls_nh_via(rt, nh));
+			dev = inet6_fib_lookup_dev(net, nh, mpls_nh_via(rt, nh));
 			break;
 		case NEIGH_LINK_TABLE:
 			break;
 		}
 	} else {
-		dev = dev_get_by_index(net, oif);
+		dev = netdev_get_by_index(net, oif,
+					  &nh->nh_dev_tracker, GFP_KERNEL);
 	}
 
 	if (!dev)
@@ -671,8 +688,7 @@ static struct net_device *find_outdev(struct net *net,
 	if (IS_ERR(dev))
 		return dev;
 
-	/* The caller is holding rtnl anyways, so release the dev reference */
-	dev_put(dev);
+	nh->nh_dev = dev;
 
 	return dev;
 }
@@ -686,20 +702,17 @@ static int mpls_nh_assign_dev(struct net *net, struct mpls_route *rt,
 	dev = find_outdev(net, rt, nh, oif);
 	if (IS_ERR(dev)) {
 		err = PTR_ERR(dev);
-		dev = NULL;
 		goto errout;
 	}
 
 	/* Ensure this is a supported device */
 	err = -EINVAL;
 	if (!mpls_dev_get(dev))
-		goto errout;
+		goto errout_put;
 
 	if ((nh->nh_via_table == NEIGH_LINK_TABLE) &&
 	    (dev->addr_len != nh->nh_via_alen))
-		goto errout;
-
-	nh->nh_dev = dev;
+		goto errout_put;
 
 	if (!(dev->flags & IFF_UP)) {
 		nh->nh_flags |= RTNH_F_DEAD;
@@ -713,6 +726,9 @@ static int mpls_nh_assign_dev(struct net *net, struct mpls_route *rt,
 
 	return 0;
 
+errout_put:
+	netdev_put(nh->nh_dev, &nh->nh_dev_tracker);
+	nh->nh_dev = NULL;
 errout:
 	return err;
 }
@@ -890,7 +906,8 @@ static int mpls_nh_build_multi(struct mpls_route_config *cfg,
 	struct nlattr *nla_via, *nla_newdst;
 	int remaining = cfg->rc_mp_len;
 	int err = 0;
-	u8 nhs = 0;
+
+	rt->rt_nhn = 0;
 
 	change_nexthops(rt) {
 		int attrlen;
@@ -926,11 +943,9 @@ static int mpls_nh_build_multi(struct mpls_route_config *cfg,
 			rt->rt_nhn_alive--;
 
 		rtnh = rtnh_next(rtnh, &remaining);
-		nhs++;
+		rt->rt_nhn++;
 	} endfor_nexthops(rt);
 
-	rt->rt_nhn = nhs;
-
 	return 0;
 
 errout:
@@ -1523,8 +1538,12 @@ static int mpls_ifdown(struct net_device *dev, int event)
 		change_nexthops(rt) {
 			unsigned int nh_flags = nh->nh_flags;
 
-			if (nh->nh_dev != dev)
+			if (nh->nh_dev != dev) {
+				if (nh_del)
+					netdev_hold(nh->nh_dev, &nh->nh_dev_tracker,
+						    GFP_KERNEL);
 				goto next;
+			}
 
 			switch (event) {
 			case NETDEV_DOWN:
@@ -2518,10 +2537,13 @@ static int resize_platform_label_table(struct net *net, size_t limit)
 	/* In case the predefined labels need to be populated */
 	if (limit > MPLS_LABEL_IPV4NULL) {
 		struct net_device *lo = net->loopback_dev;
+
 		rt0 = mpls_rt_alloc(1, lo->addr_len, 0);
 		if (IS_ERR(rt0))
 			goto nort0;
+
 		rt0->rt_nh->nh_dev = lo;
+		netdev_hold(lo, &rt0->rt_nh->nh_dev_tracker, GFP_KERNEL);
 		rt0->rt_protocol = RTPROT_KERNEL;
 		rt0->rt_payload_type = MPT_IPV4;
 		rt0->rt_ttl_propagate = MPLS_TTL_PROP_DEFAULT;
@@ -2532,10 +2554,13 @@ static int resize_platform_label_table(struct net *net, size_t limit)
 	}
 	if (limit > MPLS_LABEL_IPV6NULL) {
 		struct net_device *lo = net->loopback_dev;
+
 		rt2 = mpls_rt_alloc(1, lo->addr_len, 0);
 		if (IS_ERR(rt2))
 			goto nort2;
+
 		rt2->rt_nh->nh_dev = lo;
+		netdev_hold(lo, &rt2->rt_nh->nh_dev_tracker, GFP_KERNEL);
 		rt2->rt_protocol = RTPROT_KERNEL;
 		rt2->rt_payload_type = MPT_IPV6;
 		rt2->rt_ttl_propagate = MPLS_TTL_PROP_DEFAULT;
diff --git a/net/mpls/internal.h b/net/mpls/internal.h
index 83c629529b57..3a5feca27d6a 100644
--- a/net/mpls/internal.h
+++ b/net/mpls/internal.h
@@ -88,6 +88,7 @@ enum mpls_payload_type {
 
 struct mpls_nh { /* next hop label forwarding entry */
 	struct net_device	*nh_dev;
+	netdevice_tracker	nh_dev_tracker;
 
 	/* nh_flags is accessed under RCU in the packet path; it is
 	 * modified handling netdev events with rtnl lock held
-- 
2.51.1.838.g19442a804e-goog

[PATCH v1 net-next 03/13] mpls: Unify return paths in mpls_dev_notify().

[Kuniyuki Iwashima]

We will protect net->mpls.platform_label by a dedicated mutex.

Then, we need to wrap functions called from mpls_dev_notify()
with the mutex.

As a prep, let's unify the return paths.

Signed-off-by: Kuniyuki Iwashima <kuniyu@google.com>
---
 net/mpls/af_mpls.c | 24 ++++++++++++++++--------
 1 file changed, 16 insertions(+), 8 deletions(-)

diff --git a/net/mpls/af_mpls.c b/net/mpls/af_mpls.c
index e7be87466809..c5bbf712f8be 100644
--- a/net/mpls/af_mpls.c
+++ b/net/mpls/af_mpls.c
@@ -1616,22 +1616,24 @@ static int mpls_dev_notify(struct notifier_block *this, unsigned long event,
 
 	if (event == NETDEV_REGISTER) {
 		mdev = mpls_add_dev(dev);
-		if (IS_ERR(mdev))
-			return notifier_from_errno(PTR_ERR(mdev));
+		if (IS_ERR(mdev)) {
+			err = PTR_ERR(mdev);
+			goto err;
+		}
 
-		return NOTIFY_OK;
+		goto out;
 	}
 
 	mdev = mpls_dev_get(dev);
 	if (!mdev)
-		return NOTIFY_OK;
+		goto out;
 
 	switch (event) {
 
 	case NETDEV_DOWN:
 		err = mpls_ifdown(dev, event);
 		if (err)
-			return notifier_from_errno(err);
+			goto err;
 		break;
 	case NETDEV_UP:
 		flags = netif_get_flags(dev);
@@ -1647,13 +1649,14 @@ static int mpls_dev_notify(struct notifier_block *this, unsigned long event,
 		} else {
 			err = mpls_ifdown(dev, event);
 			if (err)
-				return notifier_from_errno(err);
+				goto err;
 		}
 		break;
 	case NETDEV_UNREGISTER:
 		err = mpls_ifdown(dev, event);
 		if (err)
-			return notifier_from_errno(err);
+			goto err;
+
 		mdev = mpls_dev_get(dev);
 		if (mdev) {
 			mpls_dev_sysctl_unregister(dev, mdev);
@@ -1667,11 +1670,16 @@ static int mpls_dev_notify(struct notifier_block *this, unsigned long event,
 			mpls_dev_sysctl_unregister(dev, mdev);
 			err = mpls_dev_sysctl_register(dev, mdev);
 			if (err)
-				return notifier_from_errno(err);
+				goto err;
 		}
 		break;
 	}
+
+out:
 	return NOTIFY_OK;
+
+err:
+	return notifier_from_errno(err);
 }
 
 static struct notifier_block mpls_dev_notifier = {
-- 
2.51.1.838.g19442a804e-goog

[PATCH v1 net-next 04/13] ipv6: Add in6_dev_rcu().

[Kuniyuki Iwashima]

rcu_dereference_rtnl() does not clearly tell whether the caller
is under RCU or RTNL.

Let's add in6_dev_rcu() to make it easy to remove __in6_dev_get()
in the future.

Signed-off-by: Kuniyuki Iwashima <kuniyu@google.com>
---
 include/net/addrconf.h | 5 +++++
 1 file changed, 5 insertions(+)

diff --git a/include/net/addrconf.h b/include/net/addrconf.h
index 9e5e95988b9e..78e8b877fb25 100644
--- a/include/net/addrconf.h
+++ b/include/net/addrconf.h
@@ -347,6 +347,11 @@ static inline struct inet6_dev *__in6_dev_get(const struct net_device *dev)
 	return rcu_dereference_rtnl(dev->ip6_ptr);
 }
 
+static inline struct inet6_dev *in6_dev_rcu(const struct net_device *dev)
+{
+	return rcu_dereference(dev->ip6_ptr);
+}
+
 static inline struct inet6_dev *__in6_dev_get_rtnl_net(const struct net_device *dev)
 {
 	return rtnl_net_dereference(dev_net(dev), dev->ip6_ptr);
-- 
2.51.1.838.g19442a804e-goog

[PATCH v1 net-next 05/13] mpls: Use in6_dev_rcu() and dev_net_rcu() in mpls_forward() and mpls_xmit().

[Kuniyuki Iwashima]

mpls_forward() and mpls_xmit() are called under RCU.

Let's use in6_dev_rcu() and dev_net_rcu() there to annotate
as such.

Now we pass net to mpls_stats_inc_outucastpkts() not to read
dev_net_rcu() twice.

Signed-off-by: Kuniyuki Iwashima <kuniyu@google.com>
---
 net/mpls/af_mpls.c       | 15 ++++++++-------
 net/mpls/internal.h      |  3 ++-
 net/mpls/mpls_iptunnel.c |  4 ++--
 3 files changed, 12 insertions(+), 10 deletions(-)

diff --git a/net/mpls/af_mpls.c b/net/mpls/af_mpls.c
index c5bbf712f8be..efc6c7da5766 100644
--- a/net/mpls/af_mpls.c
+++ b/net/mpls/af_mpls.c
@@ -129,7 +129,8 @@ bool mpls_pkt_too_big(const struct sk_buff *skb, unsigned int mtu)
 }
 EXPORT_SYMBOL_GPL(mpls_pkt_too_big);
 
-void mpls_stats_inc_outucastpkts(struct net_device *dev,
+void mpls_stats_inc_outucastpkts(struct net *net,
+				 struct net_device *dev,
 				 const struct sk_buff *skb)
 {
 	struct mpls_dev *mdev;
@@ -141,13 +142,13 @@ void mpls_stats_inc_outucastpkts(struct net_device *dev,
 					   tx_packets,
 					   tx_bytes);
 	} else if (skb->protocol == htons(ETH_P_IP)) {
-		IP_UPD_PO_STATS(dev_net(dev), IPSTATS_MIB_OUT, skb->len);
+		IP_UPD_PO_STATS(net, IPSTATS_MIB_OUT, skb->len);
 #if IS_ENABLED(CONFIG_IPV6)
 	} else if (skb->protocol == htons(ETH_P_IPV6)) {
-		struct inet6_dev *in6dev = __in6_dev_get(dev);
+		struct inet6_dev *in6dev = in6_dev_rcu(dev);
 
 		if (in6dev)
-			IP6_UPD_PO_STATS(dev_net(dev), in6dev,
+			IP6_UPD_PO_STATS(net, in6dev,
 					 IPSTATS_MIB_OUT, skb->len);
 #endif
 	}
@@ -342,7 +343,7 @@ static bool mpls_egress(struct net *net, struct mpls_route *rt,
 static int mpls_forward(struct sk_buff *skb, struct net_device *dev,
 			struct packet_type *pt, struct net_device *orig_dev)
 {
-	struct net *net = dev_net(dev);
+	struct net *net = dev_net_rcu(dev);
 	struct mpls_shim_hdr *hdr;
 	const struct mpls_nh *nh;
 	struct mpls_route *rt;
@@ -434,7 +435,7 @@ static int mpls_forward(struct sk_buff *skb, struct net_device *dev,
 	dec.ttl -= 1;
 	if (unlikely(!new_header_size && dec.bos)) {
 		/* Penultimate hop popping */
-		if (!mpls_egress(dev_net(out_dev), rt, skb, dec))
+		if (!mpls_egress(net, rt, skb, dec))
 			goto err;
 	} else {
 		bool bos;
@@ -451,7 +452,7 @@ static int mpls_forward(struct sk_buff *skb, struct net_device *dev,
 		}
 	}
 
-	mpls_stats_inc_outucastpkts(out_dev, skb);
+	mpls_stats_inc_outucastpkts(net, out_dev, skb);
 
 	/* If via wasn't specified then send out using device address */
 	if (nh->nh_via_table == MPLS_NEIGH_TABLE_UNSPEC)
diff --git a/net/mpls/internal.h b/net/mpls/internal.h
index 3a5feca27d6a..e491427ea08a 100644
--- a/net/mpls/internal.h
+++ b/net/mpls/internal.h
@@ -197,7 +197,8 @@ int nla_get_labels(const struct nlattr *nla, u8 max_labels, u8 *labels,
 bool mpls_output_possible(const struct net_device *dev);
 unsigned int mpls_dev_mtu(const struct net_device *dev);
 bool mpls_pkt_too_big(const struct sk_buff *skb, unsigned int mtu);
-void mpls_stats_inc_outucastpkts(struct net_device *dev,
+void mpls_stats_inc_outucastpkts(struct net *net,
+				 struct net_device *dev,
 				 const struct sk_buff *skb);
 
 #endif /* MPLS_INTERNAL_H */
diff --git a/net/mpls/mpls_iptunnel.c b/net/mpls/mpls_iptunnel.c
index 6e73da94af7f..cfbab7b2fec7 100644
--- a/net/mpls/mpls_iptunnel.c
+++ b/net/mpls/mpls_iptunnel.c
@@ -53,7 +53,7 @@ static int mpls_xmit(struct sk_buff *skb)
 
 	/* Find the output device */
 	out_dev = dst->dev;
-	net = dev_net(out_dev);
+	net = dev_net_rcu(out_dev);
 
 	if (!mpls_output_possible(out_dev) ||
 	    !dst->lwtstate || skb_warn_if_lro(skb))
@@ -128,7 +128,7 @@ static int mpls_xmit(struct sk_buff *skb)
 		bos = false;
 	}
 
-	mpls_stats_inc_outucastpkts(out_dev, skb);
+	mpls_stats_inc_outucastpkts(net, out_dev, skb);
 
 	if (rt) {
 		if (rt->rt_gw_family == AF_INET6)
-- 
2.51.1.838.g19442a804e-goog

[PATCH v1 net-next 06/13] mpls: Add mpls_dev_rcu().

[Kuniyuki Iwashima]

mpls_dev_get() uses rcu_dereference_rtnl() to fetch dev->mpls_ptr.

We will replace RTNL with a dedicated mutex to protect the field.

Then, we will use rcu_dereference_protected() for clarity.

Let's add mpls_dev_rcu() for the RCU reader.

Signed-off-by: Kuniyuki Iwashima <kuniyu@google.com>
---
 net/mpls/af_mpls.c       | 12 ++++++------
 net/mpls/internal.h      |  5 +++++
 net/mpls/mpls_iptunnel.c |  2 +-
 3 files changed, 12 insertions(+), 7 deletions(-)

diff --git a/net/mpls/af_mpls.c b/net/mpls/af_mpls.c
index efc6c7da5766..10130b90c439 100644
--- a/net/mpls/af_mpls.c
+++ b/net/mpls/af_mpls.c
@@ -136,7 +136,7 @@ void mpls_stats_inc_outucastpkts(struct net *net,
 	struct mpls_dev *mdev;
 
 	if (skb->protocol == htons(ETH_P_MPLS_UC)) {
-		mdev = mpls_dev_get(dev);
+		mdev = mpls_dev_rcu(dev);
 		if (mdev)
 			MPLS_INC_STATS_LEN(mdev, skb->len,
 					   tx_packets,
@@ -358,7 +358,7 @@ static int mpls_forward(struct sk_buff *skb, struct net_device *dev,
 
 	/* Careful this entire function runs inside of an rcu critical section */
 
-	mdev = mpls_dev_get(dev);
+	mdev = mpls_dev_rcu(dev);
 	if (!mdev)
 		goto drop;
 
@@ -467,7 +467,7 @@ static int mpls_forward(struct sk_buff *skb, struct net_device *dev,
 	return 0;
 
 tx_err:
-	out_mdev = out_dev ? mpls_dev_get(out_dev) : NULL;
+	out_mdev = out_dev ? mpls_dev_rcu(out_dev) : NULL;
 	if (out_mdev)
 		MPLS_INC_STATS(out_mdev, tx_errors);
 	goto drop;
@@ -1118,7 +1118,7 @@ static int mpls_fill_stats_af(struct sk_buff *skb,
 	struct mpls_dev *mdev;
 	struct nlattr *nla;
 
-	mdev = mpls_dev_get(dev);
+	mdev = mpls_dev_rcu(dev);
 	if (!mdev)
 		return -ENODATA;
 
@@ -1138,7 +1138,7 @@ static size_t mpls_get_stats_af_size(const struct net_device *dev)
 {
 	struct mpls_dev *mdev;
 
-	mdev = mpls_dev_get(dev);
+	mdev = mpls_dev_rcu(dev);
 	if (!mdev)
 		return 0;
 
@@ -1341,7 +1341,7 @@ static int mpls_netconf_dump_devconf(struct sk_buff *skb,
 
 	rcu_read_lock();
 	for_each_netdev_dump(net, dev, ctx->ifindex) {
-		mdev = mpls_dev_get(dev);
+		mdev = mpls_dev_rcu(dev);
 		if (!mdev)
 			continue;
 		err = mpls_netconf_fill_devconf(skb, mdev,
diff --git a/net/mpls/internal.h b/net/mpls/internal.h
index e491427ea08a..080e82010022 100644
--- a/net/mpls/internal.h
+++ b/net/mpls/internal.h
@@ -185,6 +185,11 @@ static inline struct mpls_entry_decoded mpls_entry_decode(struct mpls_shim_hdr *
 	return result;
 }
 
+static inline struct mpls_dev *mpls_dev_rcu(const struct net_device *dev)
+{
+	return rcu_dereference(dev->mpls_ptr);
+}
+
 static inline struct mpls_dev *mpls_dev_get(const struct net_device *dev)
 {
 	return rcu_dereference_rtnl(dev->mpls_ptr);
diff --git a/net/mpls/mpls_iptunnel.c b/net/mpls/mpls_iptunnel.c
index cfbab7b2fec7..1a1a0eb5b787 100644
--- a/net/mpls/mpls_iptunnel.c
+++ b/net/mpls/mpls_iptunnel.c
@@ -153,7 +153,7 @@ static int mpls_xmit(struct sk_buff *skb)
 	return LWTUNNEL_XMIT_DONE;
 
 drop:
-	out_mdev = out_dev ? mpls_dev_get(out_dev) : NULL;
+	out_mdev = out_dev ? mpls_dev_rcu(out_dev) : NULL;
 	if (out_mdev)
 		MPLS_INC_STATS(out_mdev, tx_errors);
 	kfree_skb(skb);
-- 
2.51.1.838.g19442a804e-goog

[PATCH v1 net-next 07/13] mpls: Pass net to mpls_dev_get().

[Kuniyuki Iwashima]

We will replace RTNL with a per-netns mutex to protect dev->mpls_ptr.

Then, we will use rcu_dereference_protected() with the lockdep_is_held()
annotation, which requires net to access the per-netns mutex.

However, dev_net(dev) is not safe without RTNL.

Let's pass net to mpls_dev_get().

Signed-off-by: Kuniyuki Iwashima <kuniyu@google.com>
---
 net/mpls/af_mpls.c  | 11 ++++++-----
 net/mpls/internal.h |  3 ++-
 2 files changed, 8 insertions(+), 6 deletions(-)

diff --git a/net/mpls/af_mpls.c b/net/mpls/af_mpls.c
index 10130b90c439..a715b12860e9 100644
--- a/net/mpls/af_mpls.c
+++ b/net/mpls/af_mpls.c
@@ -708,7 +708,7 @@ static int mpls_nh_assign_dev(struct net *net, struct mpls_route *rt,
 
 	/* Ensure this is a supported device */
 	err = -EINVAL;
-	if (!mpls_dev_get(dev))
+	if (!mpls_dev_get(net, dev))
 		goto errout_put;
 
 	if ((nh->nh_via_table == NEIGH_LINK_TABLE) &&
@@ -1288,7 +1288,7 @@ static int mpls_netconf_get_devconf(struct sk_buff *in_skb,
 	if (!dev)
 		goto errout;
 
-	mdev = mpls_dev_get(dev);
+	mdev = mpls_dev_get(net, dev);
 	if (!mdev)
 		goto errout;
 
@@ -1611,6 +1611,7 @@ static int mpls_dev_notify(struct notifier_block *this, unsigned long event,
 			   void *ptr)
 {
 	struct net_device *dev = netdev_notifier_info_to_dev(ptr);
+	struct net *net = dev_net(dev);
 	struct mpls_dev *mdev;
 	unsigned int flags;
 	int err;
@@ -1625,7 +1626,7 @@ static int mpls_dev_notify(struct notifier_block *this, unsigned long event,
 		goto out;
 	}
 
-	mdev = mpls_dev_get(dev);
+	mdev = mpls_dev_get(net, dev);
 	if (!mdev)
 		goto out;
 
@@ -1658,7 +1659,7 @@ static int mpls_dev_notify(struct notifier_block *this, unsigned long event,
 		if (err)
 			goto err;
 
-		mdev = mpls_dev_get(dev);
+		mdev = mpls_dev_get(net, dev);
 		if (mdev) {
 			mpls_dev_sysctl_unregister(dev, mdev);
 			RCU_INIT_POINTER(dev->mpls_ptr, NULL);
@@ -1666,7 +1667,7 @@ static int mpls_dev_notify(struct notifier_block *this, unsigned long event,
 		}
 		break;
 	case NETDEV_CHANGENAME:
-		mdev = mpls_dev_get(dev);
+		mdev = mpls_dev_get(net, dev);
 		if (mdev) {
 			mpls_dev_sysctl_unregister(dev, mdev);
 			err = mpls_dev_sysctl_register(dev, mdev);
diff --git a/net/mpls/internal.h b/net/mpls/internal.h
index 080e82010022..0df01a5395ee 100644
--- a/net/mpls/internal.h
+++ b/net/mpls/internal.h
@@ -190,7 +190,8 @@ static inline struct mpls_dev *mpls_dev_rcu(const struct net_device *dev)
 	return rcu_dereference(dev->mpls_ptr);
 }
 
-static inline struct mpls_dev *mpls_dev_get(const struct net_device *dev)
+static inline struct mpls_dev *mpls_dev_get(const struct net *net,
+					    const struct net_device *dev)
 {
 	return rcu_dereference_rtnl(dev->mpls_ptr);
 }
-- 
2.51.1.838.g19442a804e-goog

[PATCH v1 net-next 08/13] mpls: Add mpls_route_input().

[Kuniyuki Iwashima]

mpls_route_input_rcu() is called from mpls_forward() and
mpls_getroute().

The former is under RCU, and the latter is under RTNL, so
mpls_route_input_rcu() uses rcu_dereference_rtnl().

Let's use rcu_dereference() in mpls_route_input_rcu() and
add an RTNL variant for mpls_getroute().

Later, we will remove rtnl_dereference() there.

Signed-off-by: Kuniyuki Iwashima <kuniyu@google.com>
---
 net/mpls/af_mpls.c | 28 ++++++++++++++++++----------
 1 file changed, 18 insertions(+), 10 deletions(-)

diff --git a/net/mpls/af_mpls.c b/net/mpls/af_mpls.c
index a715b12860e9..530f7e6f7b3c 100644
--- a/net/mpls/af_mpls.c
+++ b/net/mpls/af_mpls.c
@@ -75,16 +75,23 @@ static void rtmsg_lfib(int event, u32 label, struct mpls_route *rt,
 		       struct nlmsghdr *nlh, struct net *net, u32 portid,
 		       unsigned int nlm_flags);
 
-static struct mpls_route *mpls_route_input_rcu(struct net *net, unsigned index)
+static struct mpls_route *mpls_route_input(struct net *net, unsigned int index)
 {
-	struct mpls_route *rt = NULL;
+	struct mpls_route __rcu **platform_label;
 
-	if (index < net->mpls.platform_labels) {
-		struct mpls_route __rcu **platform_label =
-			rcu_dereference_rtnl(net->mpls.platform_label);
-		rt = rcu_dereference_rtnl(platform_label[index]);
-	}
-	return rt;
+	platform_label = rtnl_dereference(net->mpls.platform_label);
+	return rtnl_dereference(platform_label[index]);
+}
+
+static struct mpls_route *mpls_route_input_rcu(struct net *net, unsigned int index)
+{
+	struct mpls_route __rcu **platform_label;
+
+	if (index >= net->mpls.platform_labels)
+		return NULL;
+
+	platform_label = rcu_dereference(net->mpls.platform_label);
+	return rcu_dereference(platform_label[index]);
 }
 
 bool mpls_output_possible(const struct net_device *dev)
@@ -2373,12 +2380,12 @@ static int mpls_getroute(struct sk_buff *in_skb, struct nlmsghdr *in_nlh,
 	u32 portid = NETLINK_CB(in_skb).portid;
 	u32 in_label = LABEL_NOT_SPECIFIED;
 	struct nlattr *tb[RTA_MAX + 1];
+	struct mpls_route *rt = NULL;
 	u32 labels[MAX_NEW_LABELS];
 	struct mpls_shim_hdr *hdr;
 	unsigned int hdr_size = 0;
 	const struct mpls_nh *nh;
 	struct net_device *dev;
-	struct mpls_route *rt;
 	struct rtmsg *rtm, *r;
 	struct nlmsghdr *nlh;
 	struct sk_buff *skb;
@@ -2406,7 +2413,8 @@ static int mpls_getroute(struct sk_buff *in_skb, struct nlmsghdr *in_nlh,
 		}
 	}
 
-	rt = mpls_route_input_rcu(net, in_label);
+	if (in_label < net->mpls.platform_labels)
+		rt = mpls_route_input(net, in_label);
 	if (!rt) {
 		err = -ENETUNREACH;
 		goto errout;
-- 
2.51.1.838.g19442a804e-goog

[PATCH v1 net-next 09/13] mpls: Use mpls_route_input() where appropriate.

[Kuniyuki Iwashima]

In many places, we uses rtnl_dereference() twice for
net->mpls.platform_label and net->mpls.platform_label[index].

Let's replace the code with mpls_route_input().

We do not use mpls_route_input() in mpls_dump_routes() since
we will rely on RCU there.

Signed-off-by: Kuniyuki Iwashima <kuniyu@google.com>
---
 net/mpls/af_mpls.c | 32 +++++++++++++-------------------
 1 file changed, 13 insertions(+), 19 deletions(-)

diff --git a/net/mpls/af_mpls.c b/net/mpls/af_mpls.c
index 530f7e6f7b3c..35ae3dbd7bdc 100644
--- a/net/mpls/af_mpls.c
+++ b/net/mpls/af_mpls.c
@@ -590,19 +590,17 @@ static void mpls_route_update(struct net *net, unsigned index,
 	mpls_rt_free(rt);
 }
 
-static unsigned find_free_label(struct net *net)
+static unsigned int find_free_label(struct net *net)
 {
-	struct mpls_route __rcu **platform_label;
-	size_t platform_labels;
-	unsigned index;
+	unsigned int index;
 
-	platform_label = rtnl_dereference(net->mpls.platform_label);
-	platform_labels = net->mpls.platform_labels;
-	for (index = MPLS_LABEL_FIRST_UNRESERVED; index < platform_labels;
+	for (index = MPLS_LABEL_FIRST_UNRESERVED;
+	     index < net->mpls.platform_labels;
 	     index++) {
-		if (!rtnl_dereference(platform_label[index]))
+		if (!mpls_route_input(net, index))
 			return index;
 	}
+
 	return LABEL_NOT_SPECIFIED;
 }
 
@@ -985,7 +983,6 @@ static bool mpls_label_ok(struct net *net, unsigned int *index,
 static int mpls_route_add(struct mpls_route_config *cfg,
 			  struct netlink_ext_ack *extack)
 {
-	struct mpls_route __rcu **platform_label;
 	struct net *net = cfg->rc_nlinfo.nl_net;
 	struct mpls_route *rt, *old;
 	int err = -EINVAL;
@@ -1013,8 +1010,7 @@ static int mpls_route_add(struct mpls_route_config *cfg,
 	}
 
 	err = -EEXIST;
-	platform_label = rtnl_dereference(net->mpls.platform_label);
-	old = rtnl_dereference(platform_label[index]);
+	old = mpls_route_input(net, index);
 	if ((cfg->rc_nlflags & NLM_F_EXCL) && old)
 		goto errout;
 
@@ -1503,16 +1499,15 @@ static void mpls_dev_destroy_rcu(struct rcu_head *head)
 
 static int mpls_ifdown(struct net_device *dev, int event)
 {
-	struct mpls_route __rcu **platform_label;
 	struct net *net = dev_net(dev);
-	unsigned index;
+	unsigned int index;
 
-	platform_label = rtnl_dereference(net->mpls.platform_label);
 	for (index = 0; index < net->mpls.platform_labels; index++) {
-		struct mpls_route *rt = rtnl_dereference(platform_label[index]);
+		struct mpls_route *rt;
 		bool nh_del = false;
 		u8 alive = 0;
 
+		rt = mpls_route_input(net, index);
 		if (!rt)
 			continue;
 
@@ -1583,15 +1578,14 @@ static int mpls_ifdown(struct net_device *dev, int event)
 
 static void mpls_ifup(struct net_device *dev, unsigned int flags)
 {
-	struct mpls_route __rcu **platform_label;
 	struct net *net = dev_net(dev);
-	unsigned index;
+	unsigned int index;
 	u8 alive;
 
-	platform_label = rtnl_dereference(net->mpls.platform_label);
 	for (index = 0; index < net->mpls.platform_labels; index++) {
-		struct mpls_route *rt = rtnl_dereference(platform_label[index]);
+		struct mpls_route *rt;
 
+		rt = mpls_route_input(net, index);
 		if (!rt)
 			continue;
 
-- 
2.51.1.838.g19442a804e-goog

[PATCH v1 net-next 10/13] mpls: Convert mpls_dump_routes() to RCU.

[Kuniyuki Iwashima]

mpls_dump_routes() sets fib_dump_filter.rtnl_held to true and
calls __dev_get_by_index() in mpls_valid_fib_dump_req().

This is the only RTNL dependant in mpls_dump_routes().

Also, synchronize_rcu() in resize_platform_label_table()
guarantees that net->mpls.platform_label is alive under RCU.

Let's convert mpls_dump_routes() to RCU and use dev_get_by_index_rcu().

Signed-off-by: Kuniyuki Iwashima <kuniyu@google.com>
---
 net/mpls/af_mpls.c | 25 ++++++++++++++++---------
 1 file changed, 16 insertions(+), 9 deletions(-)

diff --git a/net/mpls/af_mpls.c b/net/mpls/af_mpls.c
index 35ae3dbd7bdc..f6cf38673742 100644
--- a/net/mpls/af_mpls.c
+++ b/net/mpls/af_mpls.c
@@ -2153,7 +2153,7 @@ static int mpls_valid_fib_dump_req(struct net *net, const struct nlmsghdr *nlh,
 
 		if (i == RTA_OIF) {
 			ifindex = nla_get_u32(tb[i]);
-			filter->dev = __dev_get_by_index(net, ifindex);
+			filter->dev = dev_get_by_index_rcu(net, ifindex);
 			if (!filter->dev)
 				return -ENODEV;
 			filter->filter_set = 1;
@@ -2191,20 +2191,19 @@ static int mpls_dump_routes(struct sk_buff *skb, struct netlink_callback *cb)
 	struct net *net = sock_net(skb->sk);
 	struct mpls_route __rcu **platform_label;
 	struct fib_dump_filter filter = {
-		.rtnl_held = true,
+		.rtnl_held = false,
 	};
 	unsigned int flags = NLM_F_MULTI;
 	size_t platform_labels;
 	unsigned int index;
+	int err;
 
-	ASSERT_RTNL();
+	rcu_read_lock();
 
 	if (cb->strict_check) {
-		int err;
-
 		err = mpls_valid_fib_dump_req(net, nlh, &filter, cb);
 		if (err < 0)
-			return err;
+			goto err;
 
 		/* for MPLS, there is only 1 table with fixed type and flags.
 		 * If either are set in the filter then return nothing.
@@ -2212,14 +2211,14 @@ static int mpls_dump_routes(struct sk_buff *skb, struct netlink_callback *cb)
 		if ((filter.table_id && filter.table_id != RT_TABLE_MAIN) ||
 		    (filter.rt_type && filter.rt_type != RTN_UNICAST) ||
 		     filter.flags)
-			return skb->len;
+			goto unlock;
 	}
 
 	index = cb->args[0];
 	if (index < MPLS_LABEL_FIRST_UNRESERVED)
 		index = MPLS_LABEL_FIRST_UNRESERVED;
 
-	platform_label = rtnl_dereference(net->mpls.platform_label);
+	platform_label = rcu_dereference(net->mpls.platform_label);
 	platform_labels = net->mpls.platform_labels;
 
 	if (filter.filter_set)
@@ -2228,7 +2227,7 @@ static int mpls_dump_routes(struct sk_buff *skb, struct netlink_callback *cb)
 	for (; index < platform_labels; index++) {
 		struct mpls_route *rt;
 
-		rt = rtnl_dereference(platform_label[index]);
+		rt = rcu_dereference(platform_label[index]);
 		if (!rt)
 			continue;
 
@@ -2243,7 +2242,13 @@ static int mpls_dump_routes(struct sk_buff *skb, struct netlink_callback *cb)
 	}
 	cb->args[0] = index;
 
+unlock:
+	rcu_read_unlock();
 	return skb->len;
+
+err:
+	rcu_read_unlock();
+	return err;
 }
 
 static inline size_t lfib_nlmsg_size(struct mpls_route *rt)
@@ -2768,6 +2773,8 @@ static const struct rtnl_msg_handler mpls_rtnl_msg_handlers[] __initdata_or_modu
 	{THIS_MODULE, PF_MPLS, RTM_NEWROUTE, mpls_rtm_newroute, NULL, 0},
 	{THIS_MODULE, PF_MPLS, RTM_DELROUTE, mpls_rtm_delroute, NULL, 0},
 	{THIS_MODULE, PF_MPLS, RTM_GETROUTE, mpls_getroute, mpls_dump_routes, 0},
+	{THIS_MODULE, PF_MPLS, RTM_GETROUTE, mpls_getroute, mpls_dump_routes,
+	 RTNL_FLAG_DUMP_UNLOCKED},
 	{THIS_MODULE, PF_MPLS, RTM_GETNETCONF,
 	 mpls_netconf_get_devconf, mpls_netconf_dump_devconf,
 	 RTNL_FLAG_DUMP_UNLOCKED},
-- 
2.51.1.838.g19442a804e-goog

Re: [PATCH v1 net-next 10/13] mpls: Convert mpls_dump_routes() to RCU.

[Guillaume Nault]

On Tue, Oct 28, 2025 at 03:37:05AM +0000, Kuniyuki Iwashima wrote:
> @@ -2768,6 +2773,8 @@ static const struct rtnl_msg_handler mpls_rtnl_msg_handlers[] __initdata_or_modu
>  	{THIS_MODULE, PF_MPLS, RTM_NEWROUTE, mpls_rtm_newroute, NULL, 0},
>  	{THIS_MODULE, PF_MPLS, RTM_DELROUTE, mpls_rtm_delroute, NULL, 0},
>  	{THIS_MODULE, PF_MPLS, RTM_GETROUTE, mpls_getroute, mpls_dump_routes, 0},
> +	{THIS_MODULE, PF_MPLS, RTM_GETROUTE, mpls_getroute, mpls_dump_routes,
> +	 RTNL_FLAG_DUMP_UNLOCKED},

I can't see any reason to keep the old RTM_GETROUTE declaration here.
It's going to be overridden by the one with RTNL_FLAG_DUMP_UNLOCKED.

>  	{THIS_MODULE, PF_MPLS, RTM_GETNETCONF,
>  	 mpls_netconf_get_devconf, mpls_netconf_dump_devconf,
>  	 RTNL_FLAG_DUMP_UNLOCKED},
> -- 
> 2.51.1.838.g19442a804e-goog
> 
> 

Re: [PATCH v1 net-next 10/13] mpls: Convert mpls_dump_routes() to RCU.

[Kuniyuki Iwashima]

On Tue, Oct 28, 2025 at 10:41 AM Guillaume Nault <gnault@redhat.com> wrote:
>
> On Tue, Oct 28, 2025 at 03:37:05AM +0000, Kuniyuki Iwashima wrote:
> > @@ -2768,6 +2773,8 @@ static const struct rtnl_msg_handler mpls_rtnl_msg_handlers[] __initdata_or_modu
> >       {THIS_MODULE, PF_MPLS, RTM_NEWROUTE, mpls_rtm_newroute, NULL, 0},
> >       {THIS_MODULE, PF_MPLS, RTM_DELROUTE, mpls_rtm_delroute, NULL, 0},
> >       {THIS_MODULE, PF_MPLS, RTM_GETROUTE, mpls_getroute, mpls_dump_routes, 0},
> > +     {THIS_MODULE, PF_MPLS, RTM_GETROUTE, mpls_getroute, mpls_dump_routes,
> > +      RTNL_FLAG_DUMP_UNLOCKED},
>
> I can't see any reason to keep the old RTM_GETROUTE declaration here.
> It's going to be overridden by the one with RTNL_FLAG_DUMP_UNLOCKED.

Oh I think I made a mistake while rebasing and reordering patches :)

Will remove it in v2.

Thanks!

---
pw-bot: cr

[PATCH v1 net-next 11/13] mpls: Convert RTM_GETNETCONF to RCU.

[Kuniyuki Iwashima]

mpls_netconf_get_devconf() calls __dev_get_by_index(),
and this only depends on RTNL.

Let's convert mpls_netconf_get_devconf() to RCU and use
dev_get_by_index_rcu().

Note that nlmsg_new() is moved ahead to use GFP_KERNEL.

Signed-off-by: Kuniyuki Iwashima <kuniyu@google.com>
---
 net/mpls/af_mpls.c | 44 ++++++++++++++++++++++++++++++--------------
 1 file changed, 30 insertions(+), 14 deletions(-)

diff --git a/net/mpls/af_mpls.c b/net/mpls/af_mpls.c
index f6cf38673742..ffd8bc96be55 100644
--- a/net/mpls/af_mpls.c
+++ b/net/mpls/af_mpls.c
@@ -1282,23 +1282,32 @@ static int mpls_netconf_get_devconf(struct sk_buff *in_skb,
 	if (err < 0)
 		goto errout;
 
-	err = -EINVAL;
-	if (!tb[NETCONFA_IFINDEX])
+	if (!tb[NETCONFA_IFINDEX]) {
+		err = -EINVAL;
 		goto errout;
+	}
 
 	ifindex = nla_get_s32(tb[NETCONFA_IFINDEX]);
-	dev = __dev_get_by_index(net, ifindex);
-	if (!dev)
-		goto errout;
-
-	mdev = mpls_dev_get(net, dev);
-	if (!mdev)
-		goto errout;
 
-	err = -ENOBUFS;
 	skb = nlmsg_new(mpls_netconf_msgsize_devconf(NETCONFA_ALL), GFP_KERNEL);
-	if (!skb)
+	if (!skb) {
+		err = -ENOBUFS;
 		goto errout;
+	}
+
+	rcu_read_lock();
+
+	dev = dev_get_by_index_rcu(net, ifindex);
+	if (!dev) {
+		err = -EINVAL;
+		goto errout_unlock;
+	}
+
+	mdev = mpls_dev_rcu(dev);
+	if (!mdev) {
+		err = -EINVAL;
+		goto errout_unlock;
+	}
 
 	err = mpls_netconf_fill_devconf(skb, mdev,
 					NETLINK_CB(in_skb).portid,
@@ -1307,12 +1316,19 @@ static int mpls_netconf_get_devconf(struct sk_buff *in_skb,
 	if (err < 0) {
 		/* -EMSGSIZE implies BUG in mpls_netconf_msgsize_devconf() */
 		WARN_ON(err == -EMSGSIZE);
-		kfree_skb(skb);
-		goto errout;
+		goto errout_unlock;
 	}
+
 	err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
+
+	rcu_read_unlock();
 errout:
 	return err;
+
+errout_unlock:
+	rcu_read_unlock();
+	kfree_skb(skb);
+	goto errout;
 }
 
 static int mpls_netconf_dump_devconf(struct sk_buff *skb,
@@ -2777,7 +2793,7 @@ static const struct rtnl_msg_handler mpls_rtnl_msg_handlers[] __initdata_or_modu
 	 RTNL_FLAG_DUMP_UNLOCKED},
 	{THIS_MODULE, PF_MPLS, RTM_GETNETCONF,
 	 mpls_netconf_get_devconf, mpls_netconf_dump_devconf,
-	 RTNL_FLAG_DUMP_UNLOCKED},
+	 RTNL_FLAG_DOIT_UNLOCKED | RTNL_FLAG_DUMP_UNLOCKED},
 };
 
 static int __init mpls_init(void)
-- 
2.51.1.838.g19442a804e-goog

[PATCH v1 net-next 12/13] mpls: Protect net->mpls.platform_label with a per-netns mutex.

[Kuniyuki Iwashima]

MPLS (re)uses RTNL to protect net->mpls.platform_label,
but the lock does not need to be RTNL at all.

Let's protect net->mpls.platform_label with a dedicated
per-netns mutex.

Signed-off-by: Kuniyuki Iwashima <kuniyu@google.com>
---
 include/net/netns/mpls.h |  1 +
 net/mpls/af_mpls.c       | 55 ++++++++++++++++++++++++++--------------
 net/mpls/internal.h      |  7 ++++-
 3 files changed, 43 insertions(+), 20 deletions(-)

diff --git a/include/net/netns/mpls.h b/include/net/netns/mpls.h
index 19ad2574b267..6682e51513ef 100644
--- a/include/net/netns/mpls.h
+++ b/include/net/netns/mpls.h
@@ -16,6 +16,7 @@ struct netns_mpls {
 	int default_ttl;
 	size_t platform_labels;
 	struct mpls_route __rcu * __rcu *platform_label;
+	struct mutex platform_mutex;
 
 	struct ctl_table_header *ctl;
 };
diff --git a/net/mpls/af_mpls.c b/net/mpls/af_mpls.c
index ffd8bc96be55..afe5b0b70b23 100644
--- a/net/mpls/af_mpls.c
+++ b/net/mpls/af_mpls.c
@@ -79,8 +79,8 @@ static struct mpls_route *mpls_route_input(struct net *net, unsigned int index)
 {
 	struct mpls_route __rcu **platform_label;
 
-	platform_label = rtnl_dereference(net->mpls.platform_label);
-	return rtnl_dereference(platform_label[index]);
+	platform_label = mpls_dereference(net, net->mpls.platform_label);
+	return mpls_dereference(net, platform_label[index]);
 }
 
 static struct mpls_route *mpls_route_input_rcu(struct net *net, unsigned int index)
@@ -578,10 +578,8 @@ static void mpls_route_update(struct net *net, unsigned index,
 	struct mpls_route __rcu **platform_label;
 	struct mpls_route *rt;
 
-	ASSERT_RTNL();
-
-	platform_label = rtnl_dereference(net->mpls.platform_label);
-	rt = rtnl_dereference(platform_label[index]);
+	platform_label = mpls_dereference(net, net->mpls.platform_label);
+	rt = mpls_dereference(net, platform_label[index]);
 	rcu_assign_pointer(platform_label[index], new);
 
 	mpls_notify_route(net, index, rt, new, info);
@@ -1472,8 +1470,6 @@ static struct mpls_dev *mpls_add_dev(struct net_device *dev)
 	int err = -ENOMEM;
 	int i;
 
-	ASSERT_RTNL();
-
 	mdev = kzalloc(sizeof(*mdev), GFP_KERNEL);
 	if (!mdev)
 		return ERR_PTR(err);
@@ -1633,6 +1629,8 @@ static int mpls_dev_notify(struct notifier_block *this, unsigned long event,
 	unsigned int flags;
 	int err;
 
+	mutex_lock(&net->mpls.platform_mutex);
+
 	if (event == NETDEV_REGISTER) {
 		mdev = mpls_add_dev(dev);
 		if (IS_ERR(mdev)) {
@@ -1695,9 +1693,11 @@ static int mpls_dev_notify(struct notifier_block *this, unsigned long event,
 	}
 
 out:
+	mutex_unlock(&net->mpls.platform_mutex);
 	return NOTIFY_OK;
 
 err:
+	mutex_unlock(&net->mpls.platform_mutex);
 	return notifier_from_errno(err);
 }
 
@@ -1973,6 +1973,7 @@ static int rtm_to_route_config(struct sk_buff *skb,
 static int mpls_rtm_delroute(struct sk_buff *skb, struct nlmsghdr *nlh,
 			     struct netlink_ext_ack *extack)
 {
+	struct net *net = sock_net(skb->sk);
 	struct mpls_route_config *cfg;
 	int err;
 
@@ -1984,7 +1985,9 @@ static int mpls_rtm_delroute(struct sk_buff *skb, struct nlmsghdr *nlh,
 	if (err < 0)
 		goto out;
 
+	mutex_lock(&net->mpls.platform_mutex);
 	err = mpls_route_del(cfg, extack);
+	mutex_unlock(&net->mpls.platform_mutex);
 out:
 	kfree(cfg);
 
@@ -1995,6 +1998,7 @@ static int mpls_rtm_delroute(struct sk_buff *skb, struct nlmsghdr *nlh,
 static int mpls_rtm_newroute(struct sk_buff *skb, struct nlmsghdr *nlh,
 			     struct netlink_ext_ack *extack)
 {
+	struct net *net = sock_net(skb->sk);
 	struct mpls_route_config *cfg;
 	int err;
 
@@ -2006,7 +2010,9 @@ static int mpls_rtm_newroute(struct sk_buff *skb, struct nlmsghdr *nlh,
 	if (err < 0)
 		goto out;
 
+	mutex_lock(&net->mpls.platform_mutex);
 	err = mpls_route_add(cfg, extack);
+	mutex_unlock(&net->mpls.platform_mutex);
 out:
 	kfree(cfg);
 
@@ -2407,6 +2413,8 @@ static int mpls_getroute(struct sk_buff *in_skb, struct nlmsghdr *in_nlh,
 	u8 n_labels;
 	int err;
 
+	mutex_lock(&net->mpls.platform_mutex);
+
 	err = mpls_valid_getroute_req(in_skb, in_nlh, tb, extack);
 	if (err < 0)
 		goto errout;
@@ -2450,7 +2458,8 @@ static int mpls_getroute(struct sk_buff *in_skb, struct nlmsghdr *in_nlh,
 			goto errout_free;
 		}
 
-		return rtnl_unicast(skb, net, portid);
+		err = rtnl_unicast(skb, net, portid);
+		goto errout;
 	}
 
 	if (tb[RTA_NEWDST]) {
@@ -2542,12 +2551,14 @@ static int mpls_getroute(struct sk_buff *in_skb, struct nlmsghdr *in_nlh,
 
 	err = rtnl_unicast(skb, net, portid);
 errout:
+	mutex_unlock(&net->mpls.platform_mutex);
 	return err;
 
 nla_put_failure:
 	nlmsg_cancel(skb, nlh);
 	err = -EMSGSIZE;
 errout_free:
+	mutex_unlock(&net->mpls.platform_mutex);
 	kfree_skb(skb);
 	return err;
 }
@@ -2603,9 +2614,10 @@ static int resize_platform_label_table(struct net *net, size_t limit)
 		       lo->addr_len);
 	}
 
-	rtnl_lock();
+	mutex_lock(&net->mpls.platform_mutex);
+
 	/* Remember the original table */
-	old = rtnl_dereference(net->mpls.platform_label);
+	old = mpls_dereference(net, net->mpls.platform_label);
 	old_limit = net->mpls.platform_labels;
 
 	/* Free any labels beyond the new table */
@@ -2636,7 +2648,7 @@ static int resize_platform_label_table(struct net *net, size_t limit)
 	net->mpls.platform_labels = limit;
 	rcu_assign_pointer(net->mpls.platform_label, labels);
 
-	rtnl_unlock();
+	mutex_unlock(&net->mpls.platform_mutex);
 
 	mpls_rt_free(rt2);
 	mpls_rt_free(rt0);
@@ -2709,12 +2721,13 @@ static const struct ctl_table mpls_table[] = {
 	},
 };
 
-static int mpls_net_init(struct net *net)
+static __net_init int mpls_net_init(struct net *net)
 {
 	size_t table_size = ARRAY_SIZE(mpls_table);
 	struct ctl_table *table;
 	int i;
 
+	mutex_init(&net->mpls.platform_mutex);
 	net->mpls.platform_labels = 0;
 	net->mpls.platform_label = NULL;
 	net->mpls.ip_ttl_propagate = 1;
@@ -2740,7 +2753,7 @@ static int mpls_net_init(struct net *net)
 	return 0;
 }
 
-static void mpls_net_exit(struct net *net)
+static __net_exit void mpls_net_exit(struct net *net)
 {
 	struct mpls_route __rcu **platform_label;
 	size_t platform_labels;
@@ -2760,16 +2773,20 @@ static void mpls_net_exit(struct net *net)
 	 * As such no additional rcu synchronization is necessary when
 	 * freeing the platform_label table.
 	 */
-	rtnl_lock();
-	platform_label = rtnl_dereference(net->mpls.platform_label);
+	mutex_lock(&net->mpls.platform_mutex);
+
+	platform_label = mpls_dereference(net, net->mpls.platform_label);
 	platform_labels = net->mpls.platform_labels;
+
 	for (index = 0; index < platform_labels; index++) {
-		struct mpls_route *rt = rtnl_dereference(platform_label[index]);
-		RCU_INIT_POINTER(platform_label[index], NULL);
+		struct mpls_route *rt;
+
+		rt = mpls_dereference(net, platform_label[index]);
 		mpls_notify_route(net, index, rt, NULL, NULL);
 		mpls_rt_free(rt);
 	}
-	rtnl_unlock();
+
+	mutex_unlock(&net->mpls.platform_mutex);
 
 	kvfree(platform_label);
 }
diff --git a/net/mpls/internal.h b/net/mpls/internal.h
index 0df01a5395ee..80cb5bbcd946 100644
--- a/net/mpls/internal.h
+++ b/net/mpls/internal.h
@@ -185,6 +185,11 @@ static inline struct mpls_entry_decoded mpls_entry_decode(struct mpls_shim_hdr *
 	return result;
 }
 
+#define mpls_dereference(net, p)					\
+	rcu_dereference_protected(					\
+		(p),							\
+		lockdep_is_held(&(net)->mpls.platform_mutex))
+
 static inline struct mpls_dev *mpls_dev_rcu(const struct net_device *dev)
 {
 	return rcu_dereference(dev->mpls_ptr);
@@ -193,7 +198,7 @@ static inline struct mpls_dev *mpls_dev_rcu(const struct net_device *dev)
 static inline struct mpls_dev *mpls_dev_get(const struct net *net,
 					    const struct net_device *dev)
 {
-	return rcu_dereference_rtnl(dev->mpls_ptr);
+	return mpls_dereference(net, dev->mpls_ptr);
 }
 
 int nla_put_labels(struct sk_buff *skb, int attrtype,  u8 labels,
-- 
2.51.1.838.g19442a804e-goog

[PATCH v1 net-next 13/13] mpls: Drop RTNL for RTM_NEWROUTE, RTM_DELROUTE, and RTM_GETROUTE.

[Kuniyuki Iwashima]

RTM_NEWROUTE looks up dev under RCU (ip_route_output(),
ipv6_stub->ipv6_dst_lookup_flow(), netdev_get_by_index()),
and each neighbour holds the refcnt of its dev.

Also, net->mpls.platform_label is protected by a dedicated
per-netns mutex.

Now, no MPLS code depends on RTNL.

Let's drop RTNL for RTM_NEWROUTE, RTM_DELROUTE, and RTM_GETROUTE.

Signed-off-by: Kuniyuki Iwashima <kuniyu@google.com>
---
 net/mpls/af_mpls.c | 8 +++++---
 1 file changed, 5 insertions(+), 3 deletions(-)

diff --git a/net/mpls/af_mpls.c b/net/mpls/af_mpls.c
index afe5b0b70b23..0ee161703759 100644
--- a/net/mpls/af_mpls.c
+++ b/net/mpls/af_mpls.c
@@ -2803,11 +2803,13 @@ static struct rtnl_af_ops mpls_af_ops __read_mostly = {
 };
 
 static const struct rtnl_msg_handler mpls_rtnl_msg_handlers[] __initdata_or_module = {
-	{THIS_MODULE, PF_MPLS, RTM_NEWROUTE, mpls_rtm_newroute, NULL, 0},
-	{THIS_MODULE, PF_MPLS, RTM_DELROUTE, mpls_rtm_delroute, NULL, 0},
+	{THIS_MODULE, PF_MPLS, RTM_NEWROUTE, mpls_rtm_newroute, NULL,
+	 RTNL_FLAG_DOIT_UNLOCKED},
+	{THIS_MODULE, PF_MPLS, RTM_DELROUTE, mpls_rtm_delroute, NULL,
+	 RTNL_FLAG_DOIT_UNLOCKED},
 	{THIS_MODULE, PF_MPLS, RTM_GETROUTE, mpls_getroute, mpls_dump_routes, 0},
 	{THIS_MODULE, PF_MPLS, RTM_GETROUTE, mpls_getroute, mpls_dump_routes,
-	 RTNL_FLAG_DUMP_UNLOCKED},
+	 RTNL_FLAG_DOIT_UNLOCKED | RTNL_FLAG_DUMP_UNLOCKED},
 	{THIS_MODULE, PF_MPLS, RTM_GETNETCONF,
 	 mpls_netconf_get_devconf, mpls_netconf_dump_devconf,
 	 RTNL_FLAG_DOIT_UNLOCKED | RTNL_FLAG_DUMP_UNLOCKED},
-- 
2.51.1.838.g19442a804e-goog