Merge branch 'wireguard-siphash-patches-for-5-16-rc6'

Jason A. Donenfeld says:

====================
wireguard/siphash patches for 5.16-rc

Here's quite a largeish set of stable patches I've had queued up and
testing for a number of months now:

  - Patch (1) squelches a sparse warning by fixing an annotation.
  - Patches (2), (3), and (5) are minor improvements and fixes to the
    test suite.
  - Patch (4) is part of a tree-wide cleanup to have module-specific
    init and exit functions.
  - Patch (6) fixes a an issue with dangling dst references, by having a
    function to release references immediately rather than deferring,
    and adds an associated test case to prevent this from regressing.
  - Patches (7) and (8) help mitigate somewhat a potential DoS on the
    ingress path due to the use of skb_list's locking hitting contention
    on multiple cores by switching to using a ring buffer and dropping
    packets on contention rather than locking up another core spinning.
  - Patch (9) switches kvzalloc to kvcalloc for better form.
  - Patch (10) fixes alignment traps in siphash with clang-13 (and maybe
    other compilers) on armv6, by switching to using the unaligned
    functions by default instead of the aligned functions by default.
====================

Link: https://lore.kernel.org/r/20211129153929.3457-1-Jason@zx2c4.comSigned-off-by: Jakub Kicinski <kuba@kernel.org>

drivers/net/wireguard/allowedips.c

@@ -163,7 +163,7 @@ static bool node_placement(struct allowedips_node __rcu *trie, const u8 *key,
 	return exact;
 }
 
-static inline void connect_node(struct allowedips_node **parent, u8 bit, struct allowedips_node *node)
+static inline void connect_node(struct allowedips_node __rcu **parent, u8 bit, struct allowedips_node *node)
 {
 	node->parent_bit_packed = (unsigned long)parent | bit;
 	rcu_assign_pointer(*parent, node);

drivers/net/wireguard/device.c

@@ -98,6 +98,7 @@ static int wg_stop(struct net_device *dev)
 {
 	struct wg_device *wg = netdev_priv(dev);
 	struct wg_peer *peer;
+	struct sk_buff *skb;
 
 	mutex_lock(&wg->device_update_lock);
 	list_for_each_entry(peer, &wg->peer_list, peer_list) {
@@ -108,7 +109,9 @@ static int wg_stop(struct net_device *dev)
 		wg_noise_reset_last_sent_handshake(&peer->last_sent_handshake);
 	}
 	mutex_unlock(&wg->device_update_lock);
-	skb_queue_purge(&wg->incoming_handshakes);
+	while ((skb = ptr_ring_consume(&wg->handshake_queue.ring)) != NULL)
+		kfree_skb(skb);
+	atomic_set(&wg->handshake_queue_len, 0);
 	wg_socket_reinit(wg, NULL, NULL);
 	return 0;
 }
@@ -235,14 +238,13 @@ static void wg_destruct(struct net_device *dev)
 	destroy_workqueue(wg->handshake_receive_wq);
 	destroy_workqueue(wg->handshake_send_wq);
 	destroy_workqueue(wg->packet_crypt_wq);
-	wg_packet_queue_free(&wg->decrypt_queue);
-	wg_packet_queue_free(&wg->encrypt_queue);
+	wg_packet_queue_free(&wg->handshake_queue, true);
+	wg_packet_queue_free(&wg->decrypt_queue, false);
+	wg_packet_queue_free(&wg->encrypt_queue, false);
 	rcu_barrier(); /* Wait for all the peers to be actually freed. */
 	wg_ratelimiter_uninit();
 	memzero_explicit(&wg->static_identity, sizeof(wg->static_identity));
-	skb_queue_purge(&wg->incoming_handshakes);
 	free_percpu(dev->tstats);
-	free_percpu(wg->incoming_handshakes_worker);
 	kvfree(wg->index_hashtable);
 	kvfree(wg->peer_hashtable);
 	mutex_unlock(&wg->device_update_lock);
@@ -298,7 +300,6 @@ static int wg_newlink(struct net *src_net, struct net_device *dev,
 	init_rwsem(&wg->static_identity.lock);
 	mutex_init(&wg->socket_update_lock);
 	mutex_init(&wg->device_update_lock);
-	skb_queue_head_init(&wg->incoming_handshakes);
 	wg_allowedips_init(&wg->peer_allowedips);
 	wg_cookie_checker_init(&wg->cookie_checker, wg);
 	INIT_LIST_HEAD(&wg->peer_list);
@@ -316,16 +317,10 @@ static int wg_newlink(struct net *src_net, struct net_device *dev,
 	if (!dev->tstats)
 		goto err_free_index_hashtable;
 
-	wg->incoming_handshakes_worker =
-		wg_packet_percpu_multicore_worker_alloc(
-				wg_packet_handshake_receive_worker, wg);
-	if (!wg->incoming_handshakes_worker)
-		goto err_free_tstats;
-
 	wg->handshake_receive_wq = alloc_workqueue("wg-kex-%s",
 			WQ_CPU_INTENSIVE | WQ_FREEZABLE, 0, dev->name);
 	if (!wg->handshake_receive_wq)
-		goto err_free_incoming_handshakes;
+		goto err_free_tstats;
 
 	wg->handshake_send_wq = alloc_workqueue("wg-kex-%s",
 			WQ_UNBOUND | WQ_FREEZABLE, 0, dev->name);
@@ -347,10 +342,15 @@ static int wg_newlink(struct net *src_net, struct net_device *dev,
 	if (ret < 0)
 		goto err_free_encrypt_queue;
 
-	ret = wg_ratelimiter_init();
+	ret = wg_packet_queue_init(&wg->handshake_queue, wg_packet_handshake_receive_worker,
+				   MAX_QUEUED_INCOMING_HANDSHAKES);
 	if (ret < 0)
 		goto err_free_decrypt_queue;
 
+	ret = wg_ratelimiter_init();
+	if (ret < 0)
+		goto err_free_handshake_queue;
+
 	ret = register_netdevice(dev);
 	if (ret < 0)
 		goto err_uninit_ratelimiter;
@@ -367,18 +367,18 @@ static int wg_newlink(struct net *src_net, struct net_device *dev,
 
 err_uninit_ratelimiter:
 	wg_ratelimiter_uninit();
+err_free_handshake_queue:
+	wg_packet_queue_free(&wg->handshake_queue, false);
 err_free_decrypt_queue:
-	wg_packet_queue_free(&wg->decrypt_queue);
+	wg_packet_queue_free(&wg->decrypt_queue, false);
 err_free_encrypt_queue:
-	wg_packet_queue_free(&wg->encrypt_queue);
+	wg_packet_queue_free(&wg->encrypt_queue, false);
 err_destroy_packet_crypt:
 	destroy_workqueue(wg->packet_crypt_wq);
 err_destroy_handshake_send:
 	destroy_workqueue(wg->handshake_send_wq);
 err_destroy_handshake_receive:
 	destroy_workqueue(wg->handshake_receive_wq);
-err_free_incoming_handshakes:
-	free_percpu(wg->incoming_handshakes_worker);
 err_free_tstats:
 	free_percpu(dev->tstats);
 err_free_index_hashtable:
@@ -398,6 +398,7 @@ static struct rtnl_link_ops link_ops __read_mostly = {
 static void wg_netns_pre_exit(struct net *net)
 {
 	struct wg_device *wg;
+	struct wg_peer *peer;
 
 	rtnl_lock();
 	list_for_each_entry(wg, &device_list, device_list) {
@@ -407,6 +408,8 @@ static void wg_netns_pre_exit(struct net *net)
 			mutex_lock(&wg->device_update_lock);
 			rcu_assign_pointer(wg->creating_net, NULL);
 			wg_socket_reinit(wg, NULL, NULL);
+			list_for_each_entry(peer, &wg->peer_list, peer_list)
+				wg_socket_clear_peer_endpoint_src(peer);
 			mutex_unlock(&wg->device_update_lock);
 		}
 	}

drivers/net/wireguard/device.h

@@ -39,21 +39,18 @@ struct prev_queue {
 
 struct wg_device {
 	struct net_device *dev;
-	struct crypt_queue encrypt_queue, decrypt_queue;
+	struct crypt_queue encrypt_queue, decrypt_queue, handshake_queue;
 	struct sock __rcu *sock4, *sock6;
 	struct net __rcu *creating_net;
 	struct noise_static_identity static_identity;
-	struct workqueue_struct *handshake_receive_wq, *handshake_send_wq;
-	struct workqueue_struct *packet_crypt_wq;
-	struct sk_buff_head incoming_handshakes;
-	int incoming_handshake_cpu;
-	struct multicore_worker __percpu *incoming_handshakes_worker;
+	struct workqueue_struct *packet_crypt_wq,*handshake_receive_wq, *handshake_send_wq;
 	struct cookie_checker cookie_checker;
 	struct pubkey_hashtable *peer_hashtable;
 	struct index_hashtable *index_hashtable;
 	struct allowedips peer_allowedips;
 	struct mutex device_update_lock, socket_update_lock;
 	struct list_head device_list, peer_list;
+	atomic_t handshake_queue_len;
 	unsigned int num_peers, device_update_gen;
 	u32 fwmark;
 	u16 incoming_port;

drivers/net/wireguard/main.c

@@ -17,7 +17,7 @@
 #include <linux/genetlink.h>
 #include <net/rtnetlink.h>
 
-static int __init mod_init(void)
+static int __init wg_mod_init(void)
 {
 	int ret;
 
@@ -60,7 +60,7 @@ static int __init mod_init(void)
 	return ret;
 }
 
-static void __exit mod_exit(void)
+static void __exit wg_mod_exit(void)
 {
 	wg_genetlink_uninit();
 	wg_device_uninit();
@@ -68,8 +68,8 @@ static void __exit mod_exit(void)
 	wg_allowedips_slab_uninit();
 }
 
-module_init(mod_init);
-module_exit(mod_exit);
+module_init(wg_mod_init);
+module_exit(wg_mod_exit);
 MODULE_LICENSE("GPL v2");
 MODULE_DESCRIPTION("WireGuard secure network tunnel");
 MODULE_AUTHOR("Jason A. Donenfeld <Jason@zx2c4.com>");

drivers/net/wireguard/queueing.c

@@ -38,11 +38,11 @@ int wg_packet_queue_init(struct crypt_queue *queue, work_func_t function,
 	return 0;
 }
 
-void wg_packet_queue_free(struct crypt_queue *queue)
+void wg_packet_queue_free(struct crypt_queue *queue, bool purge)
 {
 	free_percpu(queue->worker);
-	WARN_ON(!__ptr_ring_empty(&queue->ring));
-	ptr_ring_cleanup(&queue->ring, NULL);
+	WARN_ON(!purge && !__ptr_ring_empty(&queue->ring));
+	ptr_ring_cleanup(&queue->ring, purge ? (void(*)(void*))kfree_skb : NULL);
 }
 
 #define NEXT(skb) ((skb)->prev)

drivers/net/wireguard/queueing.h

@@ -23,7 +23,7 @@ struct sk_buff;
 /* queueing.c APIs: */
 int wg_packet_queue_init(struct crypt_queue *queue, work_func_t function,
 			 unsigned int len);
-void wg_packet_queue_free(struct crypt_queue *queue);
+void wg_packet_queue_free(struct crypt_queue *queue, bool purge);
 struct multicore_worker __percpu *
 wg_packet_percpu_multicore_worker_alloc(work_func_t function, void *ptr);
 

drivers/net/wireguard/ratelimiter.c

@@ -176,12 +176,12 @@ int wg_ratelimiter_init(void)
 			(1U << 14) / sizeof(struct hlist_head)));
 	max_entries = table_size * 8;
 
-	table_v4 = kvzalloc(table_size * sizeof(*table_v4), GFP_KERNEL);
+	table_v4 = kvcalloc(table_size, sizeof(*table_v4), GFP_KERNEL);
 	if (unlikely(!table_v4))
 		goto err_kmemcache;
 
 #if IS_ENABLED(CONFIG_IPV6)
-	table_v6 = kvzalloc(table_size * sizeof(*table_v6), GFP_KERNEL);
+	table_v6 = kvcalloc(table_size, sizeof(*table_v6), GFP_KERNEL);
 	if (unlikely(!table_v6)) {
 		kvfree(table_v4);
 		goto err_kmemcache;

drivers/net/wireguard/receive.c

@@ -116,8 +116,8 @@ static void wg_receive_handshake_packet(struct wg_device *wg,
 		return;
 	}
 
-	under_load = skb_queue_len(&wg->incoming_handshakes) >=
-		     MAX_QUEUED_INCOMING_HANDSHAKES / 8;
+	under_load = atomic_read(&wg->handshake_queue_len) >=
+			MAX_QUEUED_INCOMING_HANDSHAKES / 8;
 	if (under_load) {
 		last_under_load = ktime_get_coarse_boottime_ns();
 	} else if (last_under_load) {
@@ -212,13 +212,14 @@ static void wg_receive_handshake_packet(struct wg_device *wg,
 
 void wg_packet_handshake_receive_worker(struct work_struct *work)
 {
-	struct wg_device *wg = container_of(work, struct multicore_worker,
-					    work)->ptr;
+	struct crypt_queue *queue = container_of(work, struct multicore_worker, work)->ptr;
+	struct wg_device *wg = container_of(queue, struct wg_device, handshake_queue);
 	struct sk_buff *skb;
 
-	while ((skb = skb_dequeue(&wg->incoming_handshakes)) != NULL) {
+	while ((skb = ptr_ring_consume_bh(&queue->ring)) != NULL) {
 		wg_receive_handshake_packet(wg, skb);
 		dev_kfree_skb(skb);
+		atomic_dec(&wg->handshake_queue_len);
 		cond_resched();
 	}
 }
@@ -553,22 +554,28 @@ void wg_packet_receive(struct wg_device *wg, struct sk_buff *skb)
 	case cpu_to_le32(MESSAGE_HANDSHAKE_INITIATION):
 	case cpu_to_le32(MESSAGE_HANDSHAKE_RESPONSE):
 	case cpu_to_le32(MESSAGE_HANDSHAKE_COOKIE): {
-		int cpu;
-
-		if (skb_queue_len(&wg->incoming_handshakes) >
-			    MAX_QUEUED_INCOMING_HANDSHAKES ||
-		    unlikely(!rng_is_initialized())) {
+		int cpu, ret = -EBUSY;
+
+		if (unlikely(!rng_is_initialized()))
+			goto drop;
+		if (atomic_read(&wg->handshake_queue_len) > MAX_QUEUED_INCOMING_HANDSHAKES / 2) {
+			if (spin_trylock_bh(&wg->handshake_queue.ring.producer_lock)) {
+				ret = __ptr_ring_produce(&wg->handshake_queue.ring, skb);
+				spin_unlock_bh(&wg->handshake_queue.ring.producer_lock);
+			}
+		} else
+			ret = ptr_ring_produce_bh(&wg->handshake_queue.ring, skb);
+		if (ret) {
+	drop:
 			net_dbg_skb_ratelimited("%s: Dropping handshake packet from %pISpfsc\n",
 						wg->dev->name, skb);
 			goto err;
 		}
-		skb_queue_tail(&wg->incoming_handshakes, skb);
-		/* Queues up a call to packet_process_queued_handshake_
-		 * packets(skb):
-		 */
-		cpu = wg_cpumask_next_online(&wg->incoming_handshake_cpu);
+		atomic_inc(&wg->handshake_queue_len);
+		cpu = wg_cpumask_next_online(&wg->handshake_queue.last_cpu);
+		/* Queues up a call to packet_process_queued_handshake_packets(skb): */
 		queue_work_on(cpu, wg->handshake_receive_wq,
-			&per_cpu_ptr(wg->incoming_handshakes_worker, cpu)->work);
+			      &per_cpu_ptr(wg->handshake_queue.worker, cpu)->work);
 		break;
 	}
 	case cpu_to_le32(MESSAGE_DATA):

drivers/net/wireguard/socket.c

@@ -308,7 +308,7 @@ void wg_socket_clear_peer_endpoint_src(struct wg_peer *peer)
 {
 	write_lock_bh(&peer->endpoint_lock);
 	memset(&peer->endpoint.src6, 0, sizeof(peer->endpoint.src6));
-	dst_cache_reset(&peer->endpoint_cache);
+	dst_cache_reset_now(&peer->endpoint_cache);
 	write_unlock_bh(&peer->endpoint_lock);
 }
 

include/linux/siphash.h

@@ -27,9 +27,7 @@ static inline bool siphash_key_is_zero(const siphash_key_t *key)
 }
 
 u64 __siphash_aligned(const void *data, size_t len, const siphash_key_t *key);
-#ifndef CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS
 u64 __siphash_unaligned(const void *data, size_t len, const siphash_key_t *key);
-#endif
 
 u64 siphash_1u64(const u64 a, const siphash_key_t *key);
 u64 siphash_2u64(const u64 a, const u64 b, const siphash_key_t *key);
@@ -82,10 +80,9 @@ static inline u64 ___siphash_aligned(const __le64 *data, size_t len,
 static inline u64 siphash(const void *data, size_t len,
 			  const siphash_key_t *key)
 {
-#ifndef CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS
-	if (!IS_ALIGNED((unsigned long)data, SIPHASH_ALIGNMENT))
+	if (IS_ENABLED(CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS) ||
+	    !IS_ALIGNED((unsigned long)data, SIPHASH_ALIGNMENT))
 		return __siphash_unaligned(data, len, key);
-#endif
 	return ___siphash_aligned(data, len, key);
 }
 
@@ -96,10 +93,8 @@ typedef struct {
 
 u32 __hsiphash_aligned(const void *data, size_t len,
 		       const hsiphash_key_t *key);
-#ifndef CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS
 u32 __hsiphash_unaligned(const void *data, size_t len,
 			 const hsiphash_key_t *key);
-#endif
 
 u32 hsiphash_1u32(const u32 a, const hsiphash_key_t *key);
 u32 hsiphash_2u32(const u32 a, const u32 b, const hsiphash_key_t *key);
@@ -135,10 +130,9 @@ static inline u32 ___hsiphash_aligned(const __le32 *data, size_t len,
 static inline u32 hsiphash(const void *data, size_t len,
 			   const hsiphash_key_t *key)
 {
-#ifndef CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS
-	if (!IS_ALIGNED((unsigned long)data, HSIPHASH_ALIGNMENT))
+	if (IS_ENABLED(CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS) ||
+	    !IS_ALIGNED((unsigned long)data, HSIPHASH_ALIGNMENT))
 		return __hsiphash_unaligned(data, len, key);
-#endif
 	return ___hsiphash_aligned(data, len, key);
 }
 

include/net/dst_cache.h

@@ -79,6 +79,17 @@ static inline void dst_cache_reset(struct dst_cache *dst_cache)
 	dst_cache->reset_ts = jiffies;
 }
 
+/**
+ *	dst_cache_reset_now - invalidate the cache contents immediately
+ *	@dst_cache: the cache
+ *
+ *	The caller must be sure there are no concurrent users, as this frees
+ *	all dst_cache users immediately, rather than waiting for the next
+ *	per-cpu usage like dst_cache_reset does. Most callers should use the
+ *	higher speed lazily-freed dst_cache_reset function instead.
+ */
+void dst_cache_reset_now(struct dst_cache *dst_cache);
+
 /**
  *	dst_cache_init - initialize the cache, allocating the required storage
  *	@dst_cache: the cache

lib/siphash.c

@@ -49,6 +49,7 @@
 	SIPROUND; \
 	return (v0 ^ v1) ^ (v2 ^ v3);
 
+#ifndef CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS
 u64 __siphash_aligned(const void *data, size_t len, const siphash_key_t *key)
 {
 	const u8 *end = data + len - (len % sizeof(u64));
@@ -80,8 +81,8 @@ u64 __siphash_aligned(const void *data, size_t len, const siphash_key_t *key)
 	POSTAMBLE
 }
 EXPORT_SYMBOL(__siphash_aligned);
+#endif
 
-#ifndef CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS
 u64 __siphash_unaligned(const void *data, size_t len, const siphash_key_t *key)
 {
 	const u8 *end = data + len - (len % sizeof(u64));
@@ -113,7 +114,6 @@ u64 __siphash_unaligned(const void *data, size_t len, const siphash_key_t *key)
 	POSTAMBLE
 }
 EXPORT_SYMBOL(__siphash_unaligned);
-#endif
 
 /**
  * siphash_1u64 - compute 64-bit siphash PRF value of a u64
@@ -250,6 +250,7 @@ EXPORT_SYMBOL(siphash_3u32);
 	HSIPROUND; \
 	return (v0 ^ v1) ^ (v2 ^ v3);
 
+#ifndef CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS
 u32 __hsiphash_aligned(const void *data, size_t len, const hsiphash_key_t *key)
 {
 	const u8 *end = data + len - (len % sizeof(u64));
@@ -280,8 +281,8 @@ u32 __hsiphash_aligned(const void *data, size_t len, const hsiphash_key_t *key)
 	HPOSTAMBLE
 }
 EXPORT_SYMBOL(__hsiphash_aligned);
+#endif
 
-#ifndef CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS
 u32 __hsiphash_unaligned(const void *data, size_t len,
 			 const hsiphash_key_t *key)
 {
@@ -313,7 +314,6 @@ u32 __hsiphash_unaligned(const void *data, size_t len,
 	HPOSTAMBLE
 }
 EXPORT_SYMBOL(__hsiphash_unaligned);
-#endif
 
 /**
  * hsiphash_1u32 - compute 64-bit hsiphash PRF value of a u32
@@ -418,6 +418,7 @@ EXPORT_SYMBOL(hsiphash_4u32);
 	HSIPROUND; \
 	return v1 ^ v3;
 
+#ifndef CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS
 u32 __hsiphash_aligned(const void *data, size_t len, const hsiphash_key_t *key)
 {
 	const u8 *end = data + len - (len % sizeof(u32));
@@ -438,8 +439,8 @@ u32 __hsiphash_aligned(const void *data, size_t len, const hsiphash_key_t *key)
 	HPOSTAMBLE
 }
 EXPORT_SYMBOL(__hsiphash_aligned);
+#endif
 
-#ifndef CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS
 u32 __hsiphash_unaligned(const void *data, size_t len,
 			 const hsiphash_key_t *key)
 {
@@ -461,7 +462,6 @@ u32 __hsiphash_unaligned(const void *data, size_t len,
 	HPOSTAMBLE
 }
 EXPORT_SYMBOL(__hsiphash_unaligned);
-#endif
 
 /**
  * hsiphash_1u32 - compute 32-bit hsiphash PRF value of a u32

net/core/dst_cache.c

@@ -162,3 +162,22 @@ void dst_cache_destroy(struct dst_cache *dst_cache)
 	free_percpu(dst_cache->cache);
 }
 EXPORT_SYMBOL_GPL(dst_cache_destroy);
+
+void dst_cache_reset_now(struct dst_cache *dst_cache)
+{
+	int i;
+
+	if (!dst_cache->cache)
+		return;
+
+	dst_cache->reset_ts = jiffies;
+	for_each_possible_cpu(i) {
+		struct dst_cache_pcpu *idst = per_cpu_ptr(dst_cache->cache, i);
+		struct dst_entry *dst = idst->dst;
+
+		idst->cookie = 0;
+		idst->dst = NULL;
+		dst_release(dst);
+	}
+}
+EXPORT_SYMBOL_GPL(dst_cache_reset_now);

tools/testing/selftests/wireguard/netns.sh

@@ -276,7 +276,11 @@ n0 ping -W 1 -c 1 192.168.241.2
 n1 wg set wg0 peer "$pub2" endpoint 192.168.241.2:7
 ip2 link del wg0
 ip2 link del wg1
-! n0 ping -W 1 -c 10 -f 192.168.241.2 || false # Should not crash kernel
+read _ _ tx_bytes_before < <(n0 wg show wg1 transfer)
+! n0 ping -W 1 -c 10 -f 192.168.241.2 || false
+sleep 1
+read _ _ tx_bytes_after < <(n0 wg show wg1 transfer)
+(( tx_bytes_after - tx_bytes_before < 70000 ))
 
 ip0 link del wg1
 ip1 link del wg0
@@ -609,6 +613,28 @@ ip0 link set wg0 up
 kill $ncat_pid
 ip0 link del wg0
 
+# Ensure that dst_cache references don't outlive netns lifetime
+ip1 link add dev wg0 type wireguard
+ip2 link add dev wg0 type wireguard
+configure_peers
+ip1 link add veth1 type veth peer name veth2
+ip1 link set veth2 netns $netns2
+ip1 addr add fd00:aa::1/64 dev veth1
+ip2 addr add fd00:aa::2/64 dev veth2
+ip1 link set veth1 up
+ip2 link set veth2 up
+waitiface $netns1 veth1
+waitiface $netns2 veth2
+ip1 -6 route add default dev veth1 via fd00:aa::2
+ip2 -6 route add default dev veth2 via fd00:aa::1
+n1 wg set wg0 peer "$pub2" endpoint [fd00:aa::2]:2
+n2 wg set wg0 peer "$pub1" endpoint [fd00:aa::1]:1
+n1 ping6 -c 1 fd00::2
+pp ip netns delete $netns1
+pp ip netns delete $netns2
+pp ip netns add $netns1
+pp ip netns add $netns2
+
 # Ensure there aren't circular reference loops
 ip1 link add wg1 type wireguard
 ip2 link add wg2 type wireguard
@@ -627,7 +653,7 @@ while read -t 0.1 -r line 2>/dev/null || [[ $? -ne 142 ]]; do
 done < /dev/kmsg
 alldeleted=1
 for object in "${!objects[@]}"; do
-	if [[ ${objects["$object"]} != *createddestroyed ]]; then
+	if [[ ${objects["$object"]} != *createddestroyed && ${objects["$object"]} != *createdcreateddestroyeddestroyed ]]; then
 		echo "Error: $object: merely ${objects["$object"]}" >&3
 		alldeleted=0
 	fi

tools/testing/selftests/wireguard/qemu/debug.config

@@ -47,7 +47,7 @@ CONFIG_DEBUG_ATOMIC_SLEEP=y
 CONFIG_TRACE_IRQFLAGS=y
 CONFIG_DEBUG_BUGVERBOSE=y
 CONFIG_DEBUG_LIST=y
-CONFIG_DEBUG_PI_LIST=y
+CONFIG_DEBUG_PLIST=y
 CONFIG_PROVE_RCU=y
 CONFIG_SPARSE_RCU_POINTER=y
 CONFIG_RCU_CPU_STALL_TIMEOUT=21

tools/testing/selftests/wireguard/qemu/kernel.config

@@ -66,6 +66,7 @@ CONFIG_PROC_SYSCTL=y
 CONFIG_SYSFS=y
 CONFIG_TMPFS=y
 CONFIG_CONSOLE_LOGLEVEL_DEFAULT=15
+CONFIG_LOG_BUF_SHIFT=18
 CONFIG_PRINTK_TIME=y
 CONFIG_BLK_DEV_INITRD=y
 CONFIG_LEGACY_VSYSCALL_NONE=y