Merge branch 'ipmr-no-rtnl-for-rtnl_family_ipmr-rtnetlink'

Kuniyuki Iwashima says:

====================
ipmr: No RTNL for RTNL_FAMILY_IPMR rtnetlink.

This series removes RTNL from ipmr rtnetlink handlers.

After this series, there are a few RTNL left in net/ipv4/ipmr.c
and such users will be converted to per-netns RTNL in another
series.

Patch 1 adds a selftest to exercise most? of the RTNL paths
 in net/ipv4/ipmr.c

Patch 2 - 6 converts RTM_GETLINK / RTM_GETROUTE handlers
 to RCU.

Patch 7 - 9 converts ->exit_batch() to ->exit_rtnl() to
 save one RTNL in cleanup_net().

Patch 10 - 11 removes unnecessary RTNL during setup_net()
 failure.

Patch 12 is a random cleanup.

Patch 13 - 15 drops RTNL for RTM_NEWROUTE and RTM_DELROUTE.
====================

Link: https://patch.msgid.link/20260228221800.1082070-1-kuniyu@google.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

include/linux/mroute_base.h

@@ -76,7 +76,7 @@ static inline int mr_call_vif_notifiers(struct net *net,
 					struct vif_device *vif,
 					struct net_device *vif_dev,
 					unsigned short vif_index, u32 tb_id,
-					unsigned int *ipmr_seq)
+					atomic_t *ipmr_seq)
 {
 	struct vif_entry_notifier_info info = {
 		.info = {
@@ -89,7 +89,7 @@ static inline int mr_call_vif_notifiers(struct net *net,
 	};
 
 	ASSERT_RTNL();
-	(*ipmr_seq)++;
+	atomic_inc(ipmr_seq);
 	return call_fib_notifiers(net, event_type, &info.info);
 }
 
@@ -198,7 +198,7 @@ static inline int mr_call_mfc_notifiers(struct net *net,
 					unsigned short family,
 					enum fib_event_type event_type,
 					struct mr_mfc *mfc, u32 tb_id,
-					unsigned int *ipmr_seq)
+					atomic_t *ipmr_seq)
 {
 	struct mfc_entry_notifier_info info = {
 		.info = {
@@ -208,8 +208,7 @@ static inline int mr_call_mfc_notifiers(struct net *net,
 		.tb_id = tb_id
 	};
 
-	ASSERT_RTNL();
-	(*ipmr_seq)++;
+	atomic_inc(ipmr_seq);
 	return call_fib_notifiers(net, event_type, &info.info);
 }
 

include/net/netns/ipv4.h

@@ -279,6 +279,9 @@ struct netns_ipv4 {
 	struct list_head	mr_tables;
 	struct fib_rules_ops	*mr_rules_ops;
 #endif
+	struct fib_notifier_ops	*ipmr_notifier_ops;
+	atomic_t		ipmr_seq;
+	struct mutex		mfc_mutex;
 #endif
 #ifdef CONFIG_IP_ROUTE_MULTIPATH
 	struct sysctl_fib_multipath_hash_seed sysctl_fib_multipath_hash_seed;
@@ -290,9 +293,6 @@ struct netns_ipv4 {
 	struct fib_notifier_ops	*notifier_ops;
 	unsigned int	fib_seq;	/* writes protected by rtnl_mutex */
 
-	struct fib_notifier_ops	*ipmr_notifier_ops;
-	unsigned int	ipmr_seq;	/* protected by rtnl_mutex */
-
 	atomic_t	rt_genid;
 	siphash_key_t	ip_id_key;
 	struct hlist_head	*inet_addr_lst;

include/net/netns/ipv6.h

@@ -118,7 +118,7 @@ struct netns_ipv6 {
 	struct seg6_pernet_data *seg6_data;
 	struct fib_notifier_ops	*notifier_ops;
 	struct fib_notifier_ops	*ip6mr_notifier_ops;
-	unsigned int ipmr_seq; /* protected by rtnl_mutex */
+	atomic_t		ipmr_seq;
 	struct {
 		struct hlist_head head;
 		spinlock_t	lock;

net/ipv4/ipmr_base.c

@@ -223,7 +223,7 @@ int mr_fill_mroute(struct mr_table *mrt, struct sk_buff *skb,
 
 	rcu_read_lock();
 	vif_dev = rcu_dereference(mrt->vif_table[c->mfc_parent].dev);
-	if (vif_dev && nla_put_u32(skb, RTA_IIF, vif_dev->ifindex) < 0) {
+	if (vif_dev && nla_put_u32(skb, RTA_IIF, READ_ONCE(vif_dev->ifindex)) < 0) {
 		rcu_read_unlock();
 		return -EMSGSIZE;
 	}
@@ -252,7 +252,7 @@ int mr_fill_mroute(struct mr_table *mrt, struct sk_buff *skb,
 
 			nhp->rtnh_flags = 0;
 			nhp->rtnh_hops = c->mfc_un.res.ttls[ct];
-			nhp->rtnh_ifindex = vif_dev->ifindex;
+			nhp->rtnh_ifindex = READ_ONCE(vif_dev->ifindex);
 			nhp->rtnh_len = sizeof(*nhp);
 		}
 	}

net/ipv6/ip6mr.c

@@ -1280,7 +1280,7 @@ static int ip6mr_device_event(struct notifier_block *this,
 
 static unsigned int ip6mr_seq_read(const struct net *net)
 {
-	return READ_ONCE(net->ipv6.ipmr_seq) + ip6mr_rules_seq_read(net);
+	return atomic_read(&net->ipv6.ipmr_seq) + ip6mr_rules_seq_read(net);
 }
 
 static int ip6mr_dump(struct net *net, struct notifier_block *nb,
@@ -1305,7 +1305,7 @@ static int __net_init ip6mr_notifier_init(struct net *net)
 {
 	struct fib_notifier_ops *ops;
 
-	net->ipv6.ipmr_seq = 0;
+	atomic_set(&net->ipv6.ipmr_seq, 0);
 
 	ops = fib_notifier_ops_register(&ip6mr_notifier_ops_template, net);
 	if (IS_ERR(ops))

tools/testing/selftests/net/forwarding/.gitignore

@@ -1,2 +1,3 @@
 # SPDX-License-Identifier: GPL-2.0-only
 forwarding.config
+ipmr

tools/testing/selftests/net/forwarding/Makefile

@@ -133,6 +133,10 @@ TEST_FILES := \
 	tc_common.sh \
 # end of TEST_FILES
 
+TEST_GEN_PROGS := \
+	ipmr
+# end of TEST_GEN_PROGS
+
 TEST_INCLUDES := \
 	$(wildcard ../lib/sh/*.sh) \
 	../lib.sh \

tools/testing/selftests/net/forwarding/ipmr.c

@@ -0,0 +1,455 @@
+// SPDX-License-Identifier: GPL-2.0
+/* Copyright 2026 Google LLC */
+
+#include <linux/if.h>
+#include <linux/mroute.h>
+#include <linux/netlink.h>
+#include <linux/rtnetlink.h>
+#include <linux/socket.h>
+#include <sched.h>
+#include <sys/ioctl.h>
+#include <sys/socket.h>
+
+#include "kselftest_harness.h"
+
+FIXTURE(ipmr)
+{
+	int netlink_sk;
+	int raw_sk;
+	int veth_ifindex;
+};
+
+FIXTURE_VARIANT(ipmr)
+{
+	int family;
+	int protocol;
+	int level;
+	int opts[MRT_MAX - MRT_BASE + 1];
+};
+
+FIXTURE_VARIANT_ADD(ipmr, ipv4)
+{
+	.family = AF_INET,
+	.protocol = IPPROTO_IGMP,
+	.level = IPPROTO_IP,
+	.opts = {
+		MRT_INIT,
+		MRT_DONE,
+		MRT_ADD_VIF,
+		MRT_DEL_VIF,
+		MRT_ADD_MFC,
+		MRT_DEL_MFC,
+		MRT_VERSION,
+		MRT_ASSERT,
+		MRT_PIM,
+		MRT_TABLE,
+		MRT_ADD_MFC_PROXY,
+		MRT_DEL_MFC_PROXY,
+		MRT_FLUSH,
+	},
+};
+
+struct mfc_attr {
+	int table;
+	__u32 origin;
+	__u32 group;
+	int ifindex;
+	bool proxy;
+};
+
+static struct rtattr *nl_add_rtattr(struct nlmsghdr *nlmsg, struct rtattr *rta,
+				    int type, const void *data, int len)
+{
+	int unused = 0;
+
+	rta->rta_type = type;
+	rta->rta_len = RTA_LENGTH(len);
+	memcpy(RTA_DATA(rta), data, len);
+
+	nlmsg->nlmsg_len += NLMSG_ALIGN(rta->rta_len);
+
+	return RTA_NEXT(rta, unused);
+}
+
+static int nl_sendmsg_mfc(struct __test_metadata *_metadata, FIXTURE_DATA(ipmr) *self,
+			  __u16 nlmsg_type, struct mfc_attr *mfc_attr)
+{
+	struct {
+		struct nlmsghdr nlmsg;
+		struct rtmsg rtm;
+		char buf[4096];
+	} req = {
+		.nlmsg = {
+			.nlmsg_len = NLMSG_LENGTH(sizeof(req.rtm)),
+			/* ipmr does not care about NLM_F_CREATE and NLM_F_EXCL ... */
+			.nlmsg_flags = NLM_F_REQUEST | NLM_F_ACK,
+			.nlmsg_type = nlmsg_type,
+		},
+		.rtm = {
+			/* hard requirements in rtm_to_ipmr_mfcc() */
+			.rtm_family = RTNL_FAMILY_IPMR,
+			.rtm_dst_len = 32,
+			.rtm_type = RTN_MULTICAST,
+			.rtm_scope = RT_SCOPE_UNIVERSE,
+			.rtm_protocol = RTPROT_MROUTED,
+		},
+	};
+	struct nlmsghdr *nlmsg = &req.nlmsg;
+	struct nlmsgerr *errmsg;
+	struct rtattr *rta;
+	int err;
+
+	rta = (struct rtattr *)&req.buf;
+	rta = nl_add_rtattr(nlmsg, rta, RTA_TABLE, &mfc_attr->table, sizeof(mfc_attr->table));
+	rta = nl_add_rtattr(nlmsg, rta, RTA_SRC, &mfc_attr->origin, sizeof(mfc_attr->origin));
+	rta = nl_add_rtattr(nlmsg, rta, RTA_DST, &mfc_attr->group, sizeof(mfc_attr->group));
+	if (mfc_attr->ifindex)
+		rta = nl_add_rtattr(nlmsg, rta, RTA_IIF, &mfc_attr->ifindex, sizeof(mfc_attr->ifindex));
+	if (mfc_attr->proxy)
+		rta = nl_add_rtattr(nlmsg, rta, RTA_PREFSRC, NULL, 0);
+
+	err = send(self->netlink_sk, &req, req.nlmsg.nlmsg_len, 0);
+	ASSERT_EQ(err, req.nlmsg.nlmsg_len);
+
+	memset(&req, 0, sizeof(req));
+
+	err = recv(self->netlink_sk, &req, sizeof(req), 0);
+	ASSERT_TRUE(NLMSG_OK(nlmsg, err));
+	ASSERT_EQ(NLMSG_ERROR, nlmsg->nlmsg_type);
+
+	errmsg = (struct nlmsgerr *)NLMSG_DATA(nlmsg);
+	return errmsg->error;
+}
+
+FIXTURE_SETUP(ipmr)
+{
+	struct ifreq ifr = {
+		.ifr_name = "veth0",
+	};
+	int err;
+
+	err = unshare(CLONE_NEWNET);
+	ASSERT_EQ(0, err);
+
+	self->netlink_sk = socket(AF_NETLINK, SOCK_RAW, NETLINK_ROUTE);
+	ASSERT_LE(0, self->netlink_sk);
+
+	self->raw_sk = socket(variant->family, SOCK_RAW, variant->protocol);
+	ASSERT_LT(0, self->raw_sk);
+
+	err = system("ip link add veth0 type veth peer veth1");
+	ASSERT_EQ(0, err);
+
+	err = ioctl(self->raw_sk, SIOCGIFINDEX, &ifr);
+	ASSERT_EQ(0, err);
+
+	self->veth_ifindex = ifr.ifr_ifindex;
+}
+
+FIXTURE_TEARDOWN(ipmr)
+{
+	close(self->raw_sk);
+	close(self->netlink_sk);
+}
+
+TEST_F(ipmr, mrt_init)
+{
+	int err, val = 0;  /* any value is ok, but size must be int for MRT_INIT. */
+
+	err = setsockopt(self->raw_sk,
+			 variant->level, variant->opts[MRT_INIT - MRT_BASE],
+			 &val,  sizeof(val));
+	ASSERT_EQ(0, err);
+
+	err = setsockopt(self->raw_sk,
+			 variant->level, variant->opts[MRT_DONE - MRT_BASE],
+			 &val,  sizeof(val));
+	ASSERT_EQ(0, err);
+}
+
+TEST_F(ipmr, mrt_add_vif_register)
+{
+	struct vifctl vif = {
+		.vifc_vifi = 0,
+		.vifc_flags = VIFF_REGISTER,
+	};
+	int err;
+
+	err = setsockopt(self->raw_sk,
+			 variant->level, variant->opts[MRT_ADD_VIF - MRT_BASE],
+			 &vif,  sizeof(vif));
+	ASSERT_EQ(0, err);
+
+	err = system("cat /proc/net/ip_mr_vif | grep -q pimreg");
+	ASSERT_EQ(0, err);
+
+	err = setsockopt(self->raw_sk,
+			 variant->level, variant->opts[MRT_DEL_VIF - MRT_BASE],
+			 &vif,  sizeof(vif));
+	ASSERT_EQ(0, err);
+}
+
+TEST_F(ipmr, mrt_del_vif_unreg)
+{
+	struct vifctl vif = {
+		.vifc_vifi = 0,
+		.vifc_flags = VIFF_USE_IFINDEX,
+		.vifc_lcl_ifindex = self->veth_ifindex,
+	};
+	int err;
+
+	err = setsockopt(self->raw_sk,
+			 variant->level, variant->opts[MRT_ADD_VIF - MRT_BASE],
+			 &vif,  sizeof(vif));
+	ASSERT_EQ(0, err);
+
+	err = system("cat /proc/net/ip_mr_vif | grep -q veth0");
+	ASSERT_EQ(0, err);
+
+	/* VIF is removed along with its device. */
+	err = system("ip link del veth0");
+	ASSERT_EQ(0, err);
+
+	/* mrt->vif_table[veth_ifindex]->dev is NULL. */
+	err = setsockopt(self->raw_sk,
+			 variant->level, variant->opts[MRT_DEL_VIF - MRT_BASE],
+			 &vif,  sizeof(vif));
+	ASSERT_EQ(-1, err);
+	ASSERT_EQ(EADDRNOTAVAIL, errno);
+}
+
+TEST_F(ipmr, mrt_del_vif_netns_dismantle)
+{
+	struct vifctl vif = {
+		.vifc_vifi = 0,
+		.vifc_flags = VIFF_USE_IFINDEX,
+		.vifc_lcl_ifindex = self->veth_ifindex,
+	};
+	int err;
+
+	err = setsockopt(self->raw_sk,
+			 variant->level, variant->opts[MRT_ADD_VIF - MRT_BASE],
+			 &vif,  sizeof(vif));
+	ASSERT_EQ(0, err);
+
+	/* Let cleanup_net() remove veth0 and VIF. */
+}
+
+TEST_F(ipmr, mrt_add_mfc)
+{
+	struct mfcctl mfc = {};
+	int err;
+
+	/* MRT_ADD_MFC / MRT_ADD_MFC_PROXY does not need vif to exist (unlike netlink). */
+	err = setsockopt(self->raw_sk,
+			 variant->level, variant->opts[MRT_ADD_MFC - MRT_BASE],
+			 &mfc,  sizeof(mfc));
+	ASSERT_EQ(0, err);
+
+	/* (0.0.0.0 -> 0.0.0.0) */
+	err = system("cat /proc/net/ip_mr_cache | grep -q '00000000 00000000' ");
+	ASSERT_EQ(0, err);
+
+	err = setsockopt(self->raw_sk,
+			 variant->level, variant->opts[MRT_DEL_MFC - MRT_BASE],
+			 &mfc,  sizeof(mfc));
+}
+
+TEST_F(ipmr, mrt_add_mfc_proxy)
+{
+	struct mfcctl mfc = {};
+	int err;
+
+	err = setsockopt(self->raw_sk,
+			 variant->level, variant->opts[MRT_ADD_MFC_PROXY - MRT_BASE],
+			 &mfc,  sizeof(mfc));
+	ASSERT_EQ(0, err);
+
+	err = system("cat /proc/net/ip_mr_cache | grep -q '00000000 00000000' ");
+	ASSERT_EQ(0, err);
+
+	err = setsockopt(self->raw_sk,
+			 variant->level, variant->opts[MRT_DEL_MFC_PROXY - MRT_BASE],
+			 &mfc,  sizeof(mfc));
+}
+
+TEST_F(ipmr, mrt_add_mfc_netlink)
+{
+	struct vifctl vif = {
+		.vifc_vifi = 0,
+		.vifc_flags = VIFF_USE_IFINDEX,
+		.vifc_lcl_ifindex = self->veth_ifindex,
+	};
+	struct mfc_attr mfc_attr = {
+		.table = RT_TABLE_DEFAULT,
+		.origin = 0,
+		.group = 0,
+		.ifindex = self->veth_ifindex,
+		.proxy = false,
+	};
+	int err;
+
+	err = setsockopt(self->raw_sk,
+			 variant->level, variant->opts[MRT_ADD_VIF - MRT_BASE],
+			 &vif,  sizeof(vif));
+	ASSERT_EQ(0, err);
+
+	err = nl_sendmsg_mfc(_metadata, self, RTM_NEWROUTE, &mfc_attr);
+	ASSERT_EQ(0, err);
+
+	err = system("cat /proc/net/ip_mr_cache | grep -q '00000000 00000000' ");
+	ASSERT_EQ(0, err);
+
+	err = nl_sendmsg_mfc(_metadata, self, RTM_DELROUTE, &mfc_attr);
+	ASSERT_EQ(0, err);
+}
+
+TEST_F(ipmr, mrt_add_mfc_netlink_proxy)
+{
+	struct vifctl vif = {
+		.vifc_vifi = 0,
+		.vifc_flags = VIFF_USE_IFINDEX,
+		.vifc_lcl_ifindex = self->veth_ifindex,
+	};
+	struct mfc_attr mfc_attr = {
+		.table = RT_TABLE_DEFAULT,
+		.origin = 0,
+		.group = 0,
+		.ifindex = self->veth_ifindex,
+		.proxy = true,
+	};
+	int err;
+
+	err = setsockopt(self->raw_sk,
+			 variant->level, variant->opts[MRT_ADD_VIF - MRT_BASE],
+			 &vif,  sizeof(vif));
+	ASSERT_EQ(0, err);
+
+	err = nl_sendmsg_mfc(_metadata, self, RTM_NEWROUTE, &mfc_attr);
+	ASSERT_EQ(0, err);
+
+	err = system("cat /proc/net/ip_mr_cache | grep -q '00000000 00000000' ");
+	ASSERT_EQ(0, err);
+
+	err = nl_sendmsg_mfc(_metadata, self, RTM_DELROUTE, &mfc_attr);
+	ASSERT_EQ(0, err);
+}
+
+TEST_F(ipmr, mrt_add_mfc_netlink_no_vif)
+{
+	struct mfc_attr mfc_attr = {
+		.table = RT_TABLE_DEFAULT,
+		.origin = 0,
+		.group = 0,
+		.proxy = false,
+	};
+	int err;
+
+	/* netlink always requires RTA_IIF of an existing vif. */
+	mfc_attr.ifindex = 0;
+	err = nl_sendmsg_mfc(_metadata, self, RTM_NEWROUTE, &mfc_attr);
+	ASSERT_EQ(-ENFILE, err);
+
+	/* netlink always requires RTA_IIF of an existing vif. */
+	mfc_attr.ifindex = self->veth_ifindex;
+	err = nl_sendmsg_mfc(_metadata, self, RTM_NEWROUTE, &mfc_attr);
+	ASSERT_EQ(-ENFILE, err);
+}
+
+TEST_F(ipmr, mrt_del_mfc_netlink_netns_dismantle)
+{
+	struct vifctl vifs[2] = {
+		{
+			.vifc_vifi = 0,
+			.vifc_flags = VIFF_USE_IFINDEX,
+			.vifc_lcl_ifindex = self->veth_ifindex,
+		},
+		{
+			.vifc_vifi = 1,
+			.vifc_flags = VIFF_REGISTER,
+		}
+	};
+	struct mfc_attr mfc_attr = {
+		.table = RT_TABLE_DEFAULT,
+		.origin = 0,
+		.group = 0,
+		.ifindex = self->veth_ifindex,
+		.proxy = false,
+	};
+	int i, err;
+
+	for (i = 0; i < 2; i++) {
+		/* Create 2 VIFs just to avoid -ENFILE later. */
+		err = setsockopt(self->raw_sk,
+				 variant->level, variant->opts[MRT_ADD_VIF - MRT_BASE],
+				 &vifs[i],  sizeof(vifs[i]));
+		ASSERT_EQ(0, err);
+	}
+
+	/* Create a MFC for mrt->vif_table[0]. */
+	err = nl_sendmsg_mfc(_metadata, self, RTM_NEWROUTE, &mfc_attr);
+	ASSERT_EQ(0, err);
+
+	err = system("cat /proc/net/ip_mr_cache | grep -q '00000000 00000000' ");
+	ASSERT_EQ(0, err);
+
+	/* Remove mrt->vif_table[0]. */
+	err = system("ip link del veth0");
+	ASSERT_EQ(0, err);
+
+	/* MFC entry is NOT removed even if the tied VIF is removed... */
+	err = system("cat /proc/net/ip_mr_cache | grep -q '00000000 00000000' ");
+	ASSERT_EQ(0, err);
+
+	/* ... and netlink is not capable of removing such an entry
+	 * because netlink always requires a valid RTA_IIF ... :/
+	 */
+	err = nl_sendmsg_mfc(_metadata, self, RTM_DELROUTE, &mfc_attr);
+	ASSERT_EQ(-ENODEV, err);
+
+	/* It can be removed by setsockopt(), but let cleanup_net() remove this time. */
+}
+
+TEST_F(ipmr, mrt_table_flush)
+{
+	struct vifctl vif = {
+		.vifc_vifi = 0,
+		.vifc_flags = VIFF_USE_IFINDEX,
+		.vifc_lcl_ifindex = self->veth_ifindex,
+	};
+	struct mfc_attr mfc_attr = {
+		.origin = 0,
+		.group = 0,
+		.ifindex = self->veth_ifindex,
+		.proxy = false,
+	};
+	int table_id = 92;
+	int err, flags;
+
+	/* Set a random table id rather than RT_TABLE_DEFAULT.
+	 * Note that /proc/net/ip_mr_{vif,cache} only supports RT_TABLE_DEFAULT.
+	 */
+	err = setsockopt(self->raw_sk,
+			 variant->level, variant->opts[MRT_TABLE - MRT_BASE],
+			 &table_id,  sizeof(table_id));
+	ASSERT_EQ(0, err);
+
+	err = setsockopt(self->raw_sk,
+			 variant->level, variant->opts[MRT_ADD_VIF - MRT_BASE],
+			 &vif,  sizeof(vif));
+	ASSERT_EQ(0, err);
+
+	mfc_attr.table = table_id;
+	err = nl_sendmsg_mfc(_metadata, self, RTM_NEWROUTE, &mfc_attr);
+	ASSERT_EQ(0, err);
+
+	/* Flush mrt->vif_table[] and all caches. */
+	flags = MRT_FLUSH_VIFS | MRT_FLUSH_VIFS_STATIC |
+		MRT_FLUSH_MFC | MRT_FLUSH_MFC_STATIC;
+	err = setsockopt(self->raw_sk,
+			 variant->level, variant->opts[MRT_FLUSH - MRT_BASE],
+			 &flags,  sizeof(flags));
+	ASSERT_EQ(0, err);
+}
+
+TEST_HARNESS_MAIN