Merge git://git.kernel.org/pub/scm/linux/kernel/git/netfilter/nf

Pablo Neira Ayuso says:

====================
Netfilter fixes for net

1) Fix broken listing of set elements when table has an owner.

2) Fix conntrack refcount leak in ctnetlink with related conntrack
entries, from Hangyu Hua.

3) Fix use-after-free/double-free in ctnetlink conntrack insert path,
from Florian Westphal.

4) Fix ip6t_rpfilter with VRF, from Phil Sutter.

5) Fix use-after-free in ebtables reported by syzbot, also from Florian.

6) Use skb->len in xt_length to deal with IPv6 jumbo packets,
from Xin Long.

7) Fix NETLINK_LISTEN_ALL_NSID with ctnetlink, from Florian Westphal.

8) Fix memleak in {ip_,ip6_,arp_}tables in ENOMEM error case,
from Pavel Tikhomirov.

* git://git.kernel.org/pub/scm/linux/kernel/git/netfilter/nf:
netfilter: x_tables: fix percpu counter block leak on error path when creating new netns
netfilter: ctnetlink: make event listener tracking global
netfilter: xt_length: use skb len to match in length_mt6
netfilter: ebtables: fix table blob use-after-free
netfilter: ip6t_rpfilter: Fix regression with VRF interfaces
netfilter: conntrack: fix rmmod double-free race
netfilter: ctnetlink: fix possible refcount leak in ctnetlink_create_conntrack()
netfilter: nf_tables: allow to fetch set elements when table has an owner
====================

Link: https://lore.kernel.org/r/[e-mail]
Signed-off-by: Jakub Kicinski <[e-mail]>

include/linux/netfilter.h

···
  */
 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

···
 
 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/bridge/netfilter/ebtables.c

···
 
 	audit_log_nfcfg(repl->name, AF_BRIDGE, repl->nentries,
 			AUDIT_XT_OP_REPLACE, GFP_KERNEL);
-	return ret;
+	return 0;
 
 free_unlock:
 	mutex_unlock(&ebt_mutex);

net/ipv4/netfilter/arp_tables.c

···
 
 	new_table = xt_register_table(net, table, &bootstrap, newinfo);
 	if (IS_ERR(new_table)) {
+		struct arpt_entry *iter;
+
+		xt_entry_foreach(iter, loc_cpu_entry, newinfo->size)
+			cleanup_entry(iter, net);
 		xt_free_table_info(newinfo);
 		return PTR_ERR(new_table);
 	}

net/ipv4/netfilter/ip_tables.c

···
 	struct xt_counters *counters;
 	struct ipt_entry *iter;
 
-	ret = 0;
 	counters = xt_counters_alloc(num_counters);
 	if (!counters) {
 		ret = -ENOMEM;
···
 		net_warn_ratelimited("iptables: counters copy to user failed while replacing table\n");
 	}
 	vfree(counters);
-	return ret;
+	return 0;
 
  put_module:
 	module_put(t->me);
···
 
 	new_table = xt_register_table(net, table, &bootstrap, newinfo);
 	if (IS_ERR(new_table)) {
+		struct ipt_entry *iter;
+
+		xt_entry_foreach(iter, loc_cpu_entry, newinfo->size)
+			cleanup_entry(iter, net);
 		xt_free_table_info(newinfo);
 		return PTR_ERR(new_table);
 	}

net/ipv6/netfilter/ip6_tables.c

···
 	struct xt_counters *counters;
 	struct ip6t_entry *iter;
 
-	ret = 0;
 	counters = xt_counters_alloc(num_counters);
 	if (!counters) {
 		ret = -ENOMEM;
···
 		net_warn_ratelimited("ip6tables: counters copy to user failed while replacing table\n");
 	}
 	vfree(counters);
-	return ret;
+	return 0;
 
  put_module:
 	module_put(t->me);
···
 
 	new_table = xt_register_table(net, table, &bootstrap, newinfo);
 	if (IS_ERR(new_table)) {
+		struct ip6t_entry *iter;
+
+		xt_entry_foreach(iter, loc_cpu_entry, newinfo->size)
+			cleanup_entry(iter, net);
 		xt_free_table_info(newinfo);
 		return PTR_ERR(new_table);
 	}

net/ipv6/netfilter/ip6t_rpfilter.c

···
 		goto out;
 	}
 
-	if (rt->rt6i_idev->dev == dev || (flags & XT_RPFILTER_LOOSE))
+	if (rt->rt6i_idev->dev == dev ||
+	    l3mdev_master_ifindex_rcu(rt->rt6i_idev->dev) == dev->ifindex ||
+	    (flags & XT_RPFILTER_LOOSE))
 		ret = true;
  out:
 	ip6_rt_put(rt);

net/netfilter/core.c

···
 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_bpf.c

···
 	struct nf_conn *nfct = (struct nf_conn *)nfct_i;
 	int err;
 
-	nfct->status |= IPS_CONFIRMED;
 	err = nf_conntrack_hash_check_insert(nfct);
 	if (err < 0) {
 		nf_conntrack_free(nfct);

net/netfilter/nf_conntrack_core.c

···
 
 	zone = nf_ct_zone(ct);
 
-	if (!nf_ct_ext_valid_pre(ct->ext)) {
-		NF_CT_STAT_INC_ATOMIC(net, insert_failed);
-		return -ETIMEDOUT;
-	}
+	if (!nf_ct_ext_valid_pre(ct->ext))
+		return -EAGAIN;
 
 	local_bh_disable();
 	do {
···
 			goto chaintoolong;
 	}
 
+	/* If genid has changed, we can't insert anymore because ct
+	 * extensions could have stale pointers and nf_ct_iterate_destroy
+	 * might have completed its table scan already.
+	 *
+	 * Increment of the ext genid right after this check is fine:
+	 * nf_ct_iterate_destroy blocks until locks are released.
+	 */
+	if (!nf_ct_ext_valid_post(ct->ext)) {
+		err = -EAGAIN;
+		goto out;
+	}
+
+	ct->status |= IPS_CONFIRMED;
 	smp_wmb();
 	/* The caller holds a reference to this object */
 	refcount_set(&ct->ct_general.use, 2);
···
 	nf_conntrack_double_unlock(hash, reply_hash);
 	NF_CT_STAT_INC(net, insert);
 	local_bh_enable();
-
-	if (!nf_ct_ext_valid_post(ct->ext)) {
-		nf_ct_kill(ct);
-		NF_CT_STAT_INC_ATOMIC(net, drop);
-		return -ETIMEDOUT;
-	}
 
 	return 0;
 chaintoolong:

net/netfilter/nf_conntrack_ecache.c

···
 			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/nf_conntrack_netlink.c

···
 	nfct_seqadj_ext_add(ct);
 	nfct_synproxy_ext_add(ct);
 
-	/* we must add conntrack extensions before confirmation. */
-	ct->status |= IPS_CONFIRMED;
-
 	if (cda[CTA_STATUS]) {
 		err = ctnetlink_change_status(ct, cda);
 		if (err < 0)
···
 
 	err = nf_conntrack_hash_check_insert(ct);
 	if (err < 0)
-		goto err2;
+		goto err3;
 
 	rcu_read_unlock();
 
 	return ct;
 
+err3:
+	if (ct->master)
+		nf_ct_put(ct->master);
 err2:
 	rcu_read_unlock();
 err1:

net/netfilter/nf_tables_api.c

···
 	int rem, err = 0;
 
 	table = nft_table_lookup(net, nla[NFTA_SET_ELEM_LIST_TABLE], family,
-				 genmask, NETLINK_CB(skb).portid);
+				 genmask, 0);
 	if (IS_ERR(table)) {
 		NL_SET_BAD_ATTR(extack, nla[NFTA_SET_ELEM_LIST_TABLE]);
 		return PTR_ERR(table);

net/netfilter/nfnetlink.c

···
 
 #include <net/netlink.h>
 #include <net/netns/generic.h>
+#include <linux/netfilter.h>
 #include <linux/netfilter/nfnetlink.h>
 
 MODULE_LICENSE("GPL");
···
 	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);
···
 
 	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

net/netfilter/xt_length.c

···
 length_mt6(const struct sk_buff *skb, struct xt_action_param *par)
 {
 	const struct xt_length_info *info = par->matchinfo;
-	const u_int16_t pktlen = ntohs(ipv6_hdr(skb)->payload_len) +
-				 sizeof(struct ipv6hdr);
+	u32 pktlen = skb->len;
 
 	return (pktlen >= info->min && pktlen <= info->max) ^ info->invert;
 }

tools/testing/selftests/netfilter/rpath.sh

···
 ip -net "$ns2" a a fec0:42::1/64 dev d0 nodad
 
 # firewall matches to test
-[ -n "$iptables" ] && ip netns exec "$ns2" \
-	"$iptables" -t raw -A PREROUTING -s 192.168.0.0/16 -m rpfilter
-[ -n "$ip6tables" ] && ip netns exec "$ns2" \
-	"$ip6tables" -t raw -A PREROUTING -s fec0::/16 -m rpfilter
+[ -n "$iptables" ] && {
+	common='-t raw -A PREROUTING -s 192.168.0.0/16'
+	ip netns exec "$ns2" "$iptables" $common -m rpfilter
+	ip netns exec "$ns2" "$iptables" $common -m rpfilter --invert
+}
+[ -n "$ip6tables" ] && {
+	common='-t raw -A PREROUTING -s fec0::/16'
+	ip netns exec "$ns2" "$ip6tables" $common -m rpfilter
+	ip netns exec "$ns2" "$ip6tables" $common -m rpfilter --invert
+}
 [ -n "$nft" ] && ip netns exec "$ns2" $nft -f - <<EOF
 table inet t {
 	chain c {
···
 	[ -n "$1" ] || return 0
 	ip netns exec "$ns2" "$1" -t raw -vS | grep -q -- "-m rpfilter -c 0 0"
 }
+ipt_zero_reverse_rule() { # (command)
+	[ -n "$1" ] || return 0
+	ip netns exec "$ns2" "$1" -t raw -vS | \
+		grep -q -- "-m rpfilter --invert -c 0 0"
+}
 nft_zero_rule() { # (family)
 	[ -n "$nft" ] || return 0
 	ip netns exec "$ns2" "$nft" list chain inet t c | \
···
 	ip netns exec "$netns" ping -q -c 1 -W 1 "$@" >/dev/null
 }
 
-testrun() {
-	# clear counters first
+clear_counters() {
 	[ -n "$iptables" ] && ip netns exec "$ns2" "$iptables" -t raw -Z
 	[ -n "$ip6tables" ] && ip netns exec "$ns2" "$ip6tables" -t raw -Z
 	if [ -n "$nft" ]; then
···
 			ip netns exec "$ns2" $nft -s list table inet t;
 		) | ip netns exec "$ns2" $nft -f -
 	fi
+}
+
+testrun() {
+	clear_counters
 
 	# test 1: martian traffic should fail rpfilter matches
 	netns_ping "$ns1" -I v0 192.168.42.1 && \
···
 
 	ipt_zero_rule "$iptables" || die "iptables matched martian"
 	ipt_zero_rule "$ip6tables" || die "ip6tables matched martian"
+	ipt_zero_reverse_rule "$iptables" && die "iptables not matched martian"
+	ipt_zero_reverse_rule "$ip6tables" && die "ip6tables not matched martian"
 	nft_zero_rule ip || die "nft IPv4 matched martian"
 	nft_zero_rule ip6 || die "nft IPv6 matched martian"
+
+	clear_counters
 
 	# test 2: rpfilter match should pass for regular traffic
 	netns_ping "$ns1" 192.168.23.1 || \
···
 
 	ipt_zero_rule "$iptables" && die "iptables match not effective"
 	ipt_zero_rule "$ip6tables" && die "ip6tables match not effective"
+	ipt_zero_reverse_rule "$iptables" || die "iptables match over-effective"
+	ipt_zero_reverse_rule "$ip6tables" || die "ip6tables match over-effective"
 	nft_zero_rule ip && die "nft IPv4 match not effective"
 	nft_zero_rule ip6 && die "nft IPv6 match not effective"