netfilter: ctnetlink: make event listener tracking global

Bugzilla: https://bugzilla.redhat.com/show_bug.cgi?id=2189550
Upstream Status: commit fdf6491

commit fdf6491
Author: Florian Westphal <fw@strlen.de>
Date:   Mon Feb 20 17:24:00 2023 +0100

    netfilter: ctnetlink: make event listener tracking global

    pernet tracking doesn't work correctly because other netns might have
    set NETLINK_LISTEN_ALL_NSID on its event socket.

    In this case its expected that events originating in other net
    namespaces are also received.

    Making pernet-tracking work while also honoring NETLINK_LISTEN_ALL_NSID
    requires much more intrusive changes both in netlink and nfnetlink,
    f.e. adding a 'setsockopt' callback that lets nfnetlink know that the
    event socket entered (or left) ALL_NSID mode.

    Move to global tracking instead: if there is an event socket anywhere
    on the system, all net namespaces which have conntrack enabled and
    use autobind mode will allocate the ecache extension.

    netlink_has_listeners() returns false only if the given group has no
    subscribers in any net namespace, the 'net' argument passed to
    nfnetlink_has_listeners is only used to derive the protocol (nfnetlink),
    it has no other effect.

    For proper NETLINK_LISTEN_ALL_NSID-aware pernet tracking of event
    listeners a new netlink_has_net_listeners() is also needed.

    Fixes: 90d1daa ("netfilter: conntrack: add nf_conntrack_events autodetect mode")
    Reported-by: Bryce Kahle <bryce.kahle@datadoghq.com>
    Signed-off-by: Florian Westphal <fw@strlen.de>
    Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>

Signed-off-by: Florian Westphal <fwestpha@redhat.com>

Author: fw-strlen

include/linux/netfilter.h

@@ -493,4 +493,9 @@ extern const struct nfnl_ct_hook __rcu *nfnl_ct_hook;
  */
 DECLARE_PER_CPU(bool, nf_skb_duplicated);
 
+/**
+ * Contains bitmask of ctnetlink event subscribers, if any.
+ * Can't be pernet due to NETLINK_LISTEN_ALL_NSID setsockopt flag.
+ */
+extern u8 nf_ctnetlink_has_listener;
 #endif /*__LINUX_NETFILTER_H*/

include/net/netns/conntrack.h

@@ -95,7 +95,6 @@ struct nf_ip_net {
 
 struct netns_ct {
 #ifdef CONFIG_NF_CONNTRACK_EVENTS
-	u8 ctnetlink_has_listener;
 	bool ecache_dwork_pending;
 #endif
 	u8			sysctl_log_invalid; /* Log invalid packets */

net/netfilter/core.c

@@ -675,6 +675,9 @@ const struct nf_ct_hook __rcu *nf_ct_hook __read_mostly;
 EXPORT_SYMBOL_GPL(nf_ct_hook);
 
 #if IS_ENABLED(CONFIG_NF_CONNTRACK)
+u8 nf_ctnetlink_has_listener;
+EXPORT_SYMBOL_GPL(nf_ctnetlink_has_listener);
+
 const struct nf_nat_hook __rcu *nf_nat_hook __read_mostly;
 EXPORT_SYMBOL_GPL(nf_nat_hook);
 

net/netfilter/nf_conntrack_ecache.c

@@ -309,7 +309,7 @@ bool nf_ct_ecache_ext_add(struct nf_conn *ct, u16 ctmask, u16 expmask, gfp_t gfp
 			break;
 		return true;
 	case 2: /* autodetect: no event listener, don't allocate extension. */
-		if (!READ_ONCE(net->ct.ctnetlink_has_listener))
+		if (!READ_ONCE(nf_ctnetlink_has_listener))
 			return true;
 		fallthrough;
 	case 1:

net/netfilter/nfnetlink.c

@@ -29,6 +29,7 @@
 
 #include <net/netlink.h>
 #include <net/netns/generic.h>
+#include <linux/netfilter.h>
 #include <linux/netfilter/nfnetlink.h>
 
 MODULE_LICENSE("GPL");
@@ -685,12 +686,12 @@ static void nfnetlink_bind_event(struct net *net, unsigned int group)
 	group_bit = (1 << group);
 
 	spin_lock(&nfnl_grp_active_lock);
-	v = READ_ONCE(net->ct.ctnetlink_has_listener);
+	v = READ_ONCE(nf_ctnetlink_has_listener);
 	if ((v & group_bit) == 0) {
 		v |= group_bit;
 
 		/* read concurrently without nfnl_grp_active_lock held. */
-		WRITE_ONCE(net->ct.ctnetlink_has_listener, v);
+		WRITE_ONCE(nf_ctnetlink_has_listener, v);
 	}
 
 	spin_unlock(&nfnl_grp_active_lock);
@@ -744,12 +745,12 @@ static void nfnetlink_unbind(struct net *net, int group)
 
 	spin_lock(&nfnl_grp_active_lock);
 	if (!nfnetlink_has_listeners(net, group)) {
-		u8 v = READ_ONCE(net->ct.ctnetlink_has_listener);
+		u8 v = READ_ONCE(nf_ctnetlink_has_listener);
 
 		v &= ~group_bit;
 
 		/* read concurrently without nfnl_grp_active_lock held. */
-		WRITE_ONCE(net->ct.ctnetlink_has_listener, v);
+		WRITE_ONCE(nf_ctnetlink_has_listener, v);
 	}
 	spin_unlock(&nfnl_grp_active_lock);
 #endif