Merge branch 'ipv4-namespacify-ipv4-address-hash-table'

Kuniyuki Iwashima says:

====================
ipv4: Namespacify IPv4 address hash table.

This is a prep of per-net RTNL conversion for RTM_(NEW|DEL|SET)ADDR.

Currently, each IPv4 address is linked to the global hash table, and
this needs to be protected by another global lock or namespacified to
support per-net RTNL.

Adding a global lock will cause deadlock in the rtnetlink path and GC,

  rtnetlink                      check_lifetime
  |- rtnl_net_lock(net)          |- acquire the global lock
  |- acquire the global lock     |- check ifa's netns
  `- put ifa into hash table     `- rtnl_net_lock(net)

so we need to namespacify the hash table.

The IPv6 one is already namespacified, let's follow that.

v2: https://lore.kernel.org/netdev/20241004195958.64396-1-kuniyu@amazon.com/
v1: https://lore.kernel.org/netdev/20241001024837.96425-1-kuniyu@amazon.com/
====================

Link: https://patch.msgid.link/20241008172906.1326-1-kuniyu@amazon.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

Author: kuba-moo

include/linux/inetdevice.h

@@ -141,7 +141,7 @@ static inline void ipv4_devconf_setall(struct in_device *in_dev)
 							  ARP_EVICT_NOCARRIER)
 
 struct in_ifaddr {
-	struct hlist_node	hash;
+	struct hlist_node	addr_lst;
 	struct in_ifaddr	__rcu *ifa_next;
 	struct in_device	*ifa_dev;
 	struct rcu_head		rcu_head;

include/net/netns/ipv4.h

@@ -270,5 +270,7 @@ struct netns_ipv4 {
 
 	atomic_t	rt_genid;
 	siphash_key_t	ip_id_key;
+	struct hlist_head	*inet_addr_lst;
+	struct delayed_work	addr_chk_work;
 };
 #endif

net/ipv4/devinet.c

@@ -119,8 +119,6 @@ struct inet_fill_args {
 #define IN4_ADDR_HSIZE_SHIFT	8
 #define IN4_ADDR_HSIZE		(1U << IN4_ADDR_HSIZE_SHIFT)
 
-static struct hlist_head inet_addr_lst[IN4_ADDR_HSIZE];
-
 static u32 inet_addr_hash(const struct net *net, __be32 addr)
 {
 	u32 val = (__force u32) addr ^ net_hash_mix(net);
@@ -133,13 +131,13 @@ static void inet_hash_insert(struct net *net, struct in_ifaddr *ifa)
 	u32 hash = inet_addr_hash(net, ifa->ifa_local);
 
 	ASSERT_RTNL();
-	hlist_add_head_rcu(&ifa->hash, &inet_addr_lst[hash]);
+	hlist_add_head_rcu(&ifa->addr_lst, &net->ipv4.inet_addr_lst[hash]);
 }
 
 static void inet_hash_remove(struct in_ifaddr *ifa)
 {
 	ASSERT_RTNL();
-	hlist_del_init_rcu(&ifa->hash);
+	hlist_del_init_rcu(&ifa->addr_lst);
 }
 
 /**
@@ -186,9 +184,8 @@ struct in_ifaddr *inet_lookup_ifaddr_rcu(struct net *net, __be32 addr)
 	u32 hash = inet_addr_hash(net, addr);
 	struct in_ifaddr *ifa;
 
-	hlist_for_each_entry_rcu(ifa, &inet_addr_lst[hash], hash)
-		if (ifa->ifa_local == addr &&
-		    net_eq(dev_net(ifa->ifa_dev->dev), net))
+	hlist_for_each_entry_rcu(ifa, &net->ipv4.inet_addr_lst[hash], addr_lst)
+		if (ifa->ifa_local == addr)
 			return ifa;
 
 	return NULL;
@@ -227,7 +224,7 @@ static struct in_ifaddr *inet_alloc_ifa(struct in_device *in_dev)
 	in_dev_hold(in_dev);
 	ifa->ifa_dev = in_dev;
 
-	INIT_HLIST_NODE(&ifa->hash);
+	INIT_HLIST_NODE(&ifa->addr_lst);
 
 	return ifa;
 }
@@ -484,15 +481,12 @@ static void inet_del_ifa(struct in_device *in_dev,
 	__inet_del_ifa(in_dev, ifap, destroy, NULL, 0);
 }
 
-static void check_lifetime(struct work_struct *work);
-
-static DECLARE_DELAYED_WORK(check_lifetime_work, check_lifetime);
-
 static int __inet_insert_ifa(struct in_ifaddr *ifa, struct nlmsghdr *nlh,
 			     u32 portid, struct netlink_ext_ack *extack)
 {
 	struct in_ifaddr __rcu **last_primary, **ifap;
 	struct in_device *in_dev = ifa->ifa_dev;
+	struct net *net = dev_net(in_dev->dev);
 	struct in_validator_info ivi;
 	struct in_ifaddr *ifa1;
 	int ret;
@@ -561,8 +555,8 @@ static int __inet_insert_ifa(struct in_ifaddr *ifa, struct nlmsghdr *nlh,
 
 	inet_hash_insert(dev_net(in_dev->dev), ifa);
 
-	cancel_delayed_work(&check_lifetime_work);
-	queue_delayed_work(system_power_efficient_wq, &check_lifetime_work, 0);
+	cancel_delayed_work(&net->ipv4.addr_chk_work);
+	queue_delayed_work(system_power_efficient_wq, &net->ipv4.addr_chk_work, 0);
 
 	/* Send message first, then call notifier.
 	   Notifier will trigger FIB update, so that
@@ -708,16 +702,19 @@ static void check_lifetime(struct work_struct *work)
 	unsigned long now, next, next_sec, next_sched;
 	struct in_ifaddr *ifa;
 	struct hlist_node *n;
+	struct net *net;
 	int i;
 
+	net = container_of(to_delayed_work(work), struct net, ipv4.addr_chk_work);
 	now = jiffies;
 	next = round_jiffies_up(now + ADDR_CHECK_FREQUENCY);
 
 	for (i = 0; i < IN4_ADDR_HSIZE; i++) {
+		struct hlist_head *head = &net->ipv4.inet_addr_lst[i];
 		bool change_needed = false;
 
 		rcu_read_lock();
-		hlist_for_each_entry_rcu(ifa, &inet_addr_lst[i], hash) {
+		hlist_for_each_entry_rcu(ifa, head, addr_lst) {
 			unsigned long age, tstamp;
 			u32 preferred_lft;
 			u32 valid_lft;
@@ -755,7 +752,7 @@ static void check_lifetime(struct work_struct *work)
 		if (!change_needed)
 			continue;
 		rtnl_lock();
-		hlist_for_each_entry_safe(ifa, n, &inet_addr_lst[i], hash) {
+		hlist_for_each_entry_safe(ifa, n, head, addr_lst) {
 			unsigned long age;
 
 			if (ifa->ifa_flags & IFA_F_PERMANENT)
@@ -804,8 +801,8 @@ static void check_lifetime(struct work_struct *work)
 	if (time_before(next_sched, now + ADDRCONF_TIMER_FUZZ_MAX))
 		next_sched = now + ADDRCONF_TIMER_FUZZ_MAX;
 
-	queue_delayed_work(system_power_efficient_wq, &check_lifetime_work,
-			next_sched - now);
+	queue_delayed_work(system_power_efficient_wq, &net->ipv4.addr_chk_work,
+			   next_sched - now);
 }
 
 static void set_ifa_lifetime(struct in_ifaddr *ifa, __u32 valid_lft,
@@ -1002,9 +999,9 @@ static int inet_rtm_newaddr(struct sk_buff *skb, struct nlmsghdr *nlh,
 		ifa->ifa_proto = new_proto;
 
 		set_ifa_lifetime(ifa, valid_lft, prefered_lft);
-		cancel_delayed_work(&check_lifetime_work);
+		cancel_delayed_work(&net->ipv4.addr_chk_work);
 		queue_delayed_work(system_power_efficient_wq,
-				&check_lifetime_work, 0);
+				   &net->ipv4.addr_chk_work, 0);
 		rtmsg_ifa(RTM_NEWADDR, ifa, nlh, NETLINK_CB(skb).portid);
 	}
 	return 0;
@@ -2663,14 +2660,21 @@ static struct ctl_table ctl_forward_entry[] = {
 
 static __net_init int devinet_init_net(struct net *net)
 {
-	int err;
-	struct ipv4_devconf *all, *dflt;
 #ifdef CONFIG_SYSCTL
-	struct ctl_table *tbl;
 	struct ctl_table_header *forw_hdr;
+	struct ctl_table *tbl;
 #endif
+	struct ipv4_devconf *all, *dflt;
+	int err;
+	int i;
 
 	err = -ENOMEM;
+	net->ipv4.inet_addr_lst = kmalloc_array(IN4_ADDR_HSIZE,
+						sizeof(struct hlist_head),
+						GFP_KERNEL);
+	if (!net->ipv4.inet_addr_lst)
+		goto err_alloc_hash;
+
 	all = kmemdup(&ipv4_devconf, sizeof(ipv4_devconf), GFP_KERNEL);
 	if (!all)
 		goto err_alloc_all;
@@ -2731,6 +2735,11 @@ static __net_init int devinet_init_net(struct net *net)
 	net->ipv4.forw_hdr = forw_hdr;
 #endif
 
+	for (i = 0; i < IN4_ADDR_HSIZE; i++)
+		INIT_HLIST_HEAD(&net->ipv4.inet_addr_lst[i]);
+
+	INIT_DEFERRABLE_WORK(&net->ipv4.addr_chk_work, check_lifetime);
+
 	net->ipv4.devconf_all = all;
 	net->ipv4.devconf_dflt = dflt;
 	return 0;
@@ -2748,14 +2757,20 @@ static __net_init int devinet_init_net(struct net *net)
 err_alloc_dflt:
 	kfree(all);
 err_alloc_all:
+	kfree(net->ipv4.inet_addr_lst);
+err_alloc_hash:
 	return err;
 }
 
 static __net_exit void devinet_exit_net(struct net *net)
 {
 #ifdef CONFIG_SYSCTL
 	const struct ctl_table *tbl;
+#endif
+
+	cancel_delayed_work_sync(&net->ipv4.addr_chk_work);
 
+#ifdef CONFIG_SYSCTL
 	tbl = net->ipv4.forw_hdr->ctl_table_arg;
 	unregister_net_sysctl_table(net->ipv4.forw_hdr);
 	__devinet_sysctl_unregister(net, net->ipv4.devconf_dflt,
@@ -2766,6 +2781,7 @@ static __net_exit void devinet_exit_net(struct net *net)
 #endif
 	kfree(net->ipv4.devconf_dflt);
 	kfree(net->ipv4.devconf_all);
+	kfree(net->ipv4.inet_addr_lst);
 }
 
 static __net_initdata struct pernet_operations devinet_ops = {
@@ -2783,16 +2799,9 @@ static struct rtnl_af_ops inet_af_ops __read_mostly = {
 
 void __init devinet_init(void)
 {
-	int i;
-
-	for (i = 0; i < IN4_ADDR_HSIZE; i++)
-		INIT_HLIST_HEAD(&inet_addr_lst[i]);
-
 	register_pernet_subsys(&devinet_ops);
 	register_netdevice_notifier(&ip_netdev_notifier);
 
-	queue_delayed_work(system_power_efficient_wq, &check_lifetime_work, 0);
-
 	rtnl_af_register(&inet_af_ops);
 
 	rtnl_register(PF_INET, RTM_NEWADDR, inet_rtm_newaddr, NULL, 0);