[PATCH v2 net-next 0/3] inet: ping: misc changes

[PATCH v2 net-next 0/3] inet: ping: misc changes

[Eric Dumazet]

First and third patches improve security a bit.

Second patch (ping_hash removal) is a cleanup.

Eric Dumazet (3):
  inet: ping: check sock_net() in ping_get_port() and ping_lookup()
  inet: ping: remove ping_hash()
  inet: ping: make ping_port_rover per netns

 include/net/netns/ipv4.h |  1 +
 net/ipv4/ping.c          | 34 +++++++++++++++-------------------
 net/ipv6/ping.c          |  1 -
 3 files changed, 16 insertions(+), 20 deletions(-)

-- 
2.51.0.268.g9569e192d0-goog

[PATCH v2 net-next 1/3] inet: ping: check sock_net() in ping_get_port() and ping_lookup()

[Eric Dumazet]

We need to check socket netns before considering them in ping_get_port().
Otherwise, one malicious netns could 'consume' all ports.

Add corresponding check in ping_lookup().

Fixes: c319b4d76b9e ("net: ipv4: add IPPROTO_ICMP socket kind")
Signed-off-by: Eric Dumazet <edumazet@google.com>
Reviewed-by: David Ahern <dsahern@kernel.org>
---
v2: added change to ping_lookup().
v1: https://lore.kernel.org/netdev/CANn89iKF+DFQQyNJoYA2U-wf4QDPOUG2yNOd8fSu45hQ+TxJ5Q@mail.gmail.com/T/#u

 net/ipv4/ping.c | 14 ++++++++++----
 1 file changed, 10 insertions(+), 4 deletions(-)

diff --git a/net/ipv4/ping.c b/net/ipv4/ping.c
index f119da68fc301be00719213ad33615b6754e6272..74a0beddfcc41d8ba17792a11a9d027c9d590bac 100644
--- a/net/ipv4/ping.c
+++ b/net/ipv4/ping.c
@@ -77,6 +77,7 @@ static inline struct hlist_head *ping_hashslot(struct ping_table *table,
 
 int ping_get_port(struct sock *sk, unsigned short ident)
 {
+	struct net *net = sock_net(sk);
 	struct inet_sock *isk, *isk2;
 	struct hlist_head *hlist;
 	struct sock *sk2 = NULL;
@@ -90,9 +91,10 @@ int ping_get_port(struct sock *sk, unsigned short ident)
 		for (i = 0; i < (1L << 16); i++, result++) {
 			if (!result)
 				result++; /* avoid zero */
-			hlist = ping_hashslot(&ping_table, sock_net(sk),
-					    result);
+			hlist = ping_hashslot(&ping_table, net, result);
 			sk_for_each(sk2, hlist) {
+				if (!net_eq(sock_net(sk2), net))
+					continue;
 				isk2 = inet_sk(sk2);
 
 				if (isk2->inet_num == result)
@@ -108,8 +110,10 @@ int ping_get_port(struct sock *sk, unsigned short ident)
 		if (i >= (1L << 16))
 			goto fail;
 	} else {
-		hlist = ping_hashslot(&ping_table, sock_net(sk), ident);
+		hlist = ping_hashslot(&ping_table, net, ident);
 		sk_for_each(sk2, hlist) {
+			if (!net_eq(sock_net(sk2), net))
+				continue;
 			isk2 = inet_sk(sk2);
 
 			/* BUG? Why is this reuse and not reuseaddr? ping.c
@@ -129,7 +133,7 @@ int ping_get_port(struct sock *sk, unsigned short ident)
 		pr_debug("was not hashed\n");
 		sk_add_node_rcu(sk, hlist);
 		sock_set_flag(sk, SOCK_RCU_FREE);
-		sock_prot_inuse_add(sock_net(sk), sk->sk_prot, 1);
+		sock_prot_inuse_add(net, sk->sk_prot, 1);
 	}
 	spin_unlock(&ping_table.lock);
 	return 0;
@@ -188,6 +192,8 @@ static struct sock *ping_lookup(struct net *net, struct sk_buff *skb, u16 ident)
 	}
 
 	sk_for_each_rcu(sk, hslot) {
+		if (!net_eq(sock_net(sk), net))
+			continue;
 		isk = inet_sk(sk);
 
 		pr_debug("iterate\n");
-- 
2.51.0.268.g9569e192d0-goog

[PATCH v2 net-next 2/3] inet: ping: remove ping_hash()

[Eric Dumazet]

There is no point in keeping ping_hash().

Signed-off-by: Eric Dumazet <edumazet@google.com>
Reviewed-by: David Ahern <dsahern@kernel.org>
---
 net/ipv4/ping.c | 10 ----------
 net/ipv6/ping.c |  1 -
 2 files changed, 11 deletions(-)

diff --git a/net/ipv4/ping.c b/net/ipv4/ping.c
index 74a0beddfcc41d8ba17792a11a9d027c9d590bac..75e1b0f5c697653e79166fde5f312f46b471344a 100644
--- a/net/ipv4/ping.c
+++ b/net/ipv4/ping.c
@@ -67,7 +67,6 @@ static inline u32 ping_hashfn(const struct net *net, u32 num, u32 mask)
 	pr_debug("hash(%u) = %u\n", num, res);
 	return res;
 }
-EXPORT_SYMBOL_GPL(ping_hash);
 
 static inline struct hlist_head *ping_hashslot(struct ping_table *table,
 					       struct net *net, unsigned int num)
@@ -144,14 +143,6 @@ int ping_get_port(struct sock *sk, unsigned short ident)
 }
 EXPORT_SYMBOL_GPL(ping_get_port);
 
-int ping_hash(struct sock *sk)
-{
-	pr_debug("ping_hash(sk->port=%u)\n", inet_sk(sk)->inet_num);
-	BUG(); /* "Please do not press this button again." */
-
-	return 0;
-}
-
 void ping_unhash(struct sock *sk)
 {
 	struct inet_sock *isk = inet_sk(sk);
@@ -1008,7 +999,6 @@ struct proto ping_prot = {
 	.bind =		ping_bind,
 	.backlog_rcv =	ping_queue_rcv_skb,
 	.release_cb =	ip4_datagram_release_cb,
-	.hash =		ping_hash,
 	.unhash =	ping_unhash,
 	.get_port =	ping_get_port,
 	.put_port =	ping_unhash,
diff --git a/net/ipv6/ping.c b/net/ipv6/ping.c
index 82b0492923d458213ac7a6f9316158af2191e30f..d7a2cdaa26312b44f1fe502d3d40f3e27f961fa8 100644
--- a/net/ipv6/ping.c
+++ b/net/ipv6/ping.c
@@ -208,7 +208,6 @@ struct proto pingv6_prot = {
 	.recvmsg =	ping_recvmsg,
 	.bind =		ping_bind,
 	.backlog_rcv =	ping_queue_rcv_skb,
-	.hash =		ping_hash,
 	.unhash =	ping_unhash,
 	.get_port =	ping_get_port,
 	.put_port =	ping_unhash,
-- 
2.51.0.268.g9569e192d0-goog

[PATCH v2 net-next 3/3] inet: ping: make ping_port_rover per netns

[Eric Dumazet]

Provide isolation between netns for ping idents.

Randomize initial ping_port_rover value at netns creation.

Signed-off-by: Eric Dumazet <edumazet@google.com>
Reviewed-by: David Ahern <dsahern@kernel.org>
---
 include/net/netns/ipv4.h |  1 +
 net/ipv4/ping.c          | 10 +++++-----
 2 files changed, 6 insertions(+), 5 deletions(-)

diff --git a/include/net/netns/ipv4.h b/include/net/netns/ipv4.h
index 6373e3f17da84ebc5c11058763932e595f0fd205..54a7d187f62a2e995076e85f1e6b2fd70f84b2c1 100644
--- a/include/net/netns/ipv4.h
+++ b/include/net/netns/ipv4.h
@@ -251,6 +251,7 @@ struct netns_ipv4 {
 	int sysctl_igmp_qrv;
 
 	struct ping_group_range ping_group_range;
+	u16			ping_port_rover;
 
 	atomic_t dev_addr_genid;
 
diff --git a/net/ipv4/ping.c b/net/ipv4/ping.c
index 75e1b0f5c697653e79166fde5f312f46b471344a..98ccd4f9ed657d2bb9c013932d0c678f2b38a746 100644
--- a/net/ipv4/ping.c
+++ b/net/ipv4/ping.c
@@ -58,8 +58,6 @@ static struct ping_table ping_table;
 struct pingv6_ops pingv6_ops;
 EXPORT_SYMBOL_GPL(pingv6_ops);
 
-static u16 ping_port_rover;
-
 static inline u32 ping_hashfn(const struct net *net, u32 num, u32 mask)
 {
 	u32 res = (num + net_hash_mix(net)) & mask;
@@ -84,12 +82,12 @@ int ping_get_port(struct sock *sk, unsigned short ident)
 	isk = inet_sk(sk);
 	spin_lock(&ping_table.lock);
 	if (ident == 0) {
+		u16 result = net->ipv4.ping_port_rover + 1;
 		u32 i;
-		u16 result = ping_port_rover + 1;
 
 		for (i = 0; i < (1L << 16); i++, result++) {
 			if (!result)
-				result++; /* avoid zero */
+				continue; /* avoid zero */
 			hlist = ping_hashslot(&ping_table, net, result);
 			sk_for_each(sk2, hlist) {
 				if (!net_eq(sock_net(sk2), net))
@@ -101,7 +99,7 @@ int ping_get_port(struct sock *sk, unsigned short ident)
 			}
 
 			/* found */
-			ping_port_rover = ident = result;
+			net->ipv4.ping_port_rover = ident = result;
 			break;
 next_port:
 			;
@@ -1146,6 +1144,8 @@ static int __net_init ping_v4_proc_init_net(struct net *net)
 	if (!proc_create_net("icmp", 0444, net->proc_net, &ping_v4_seq_ops,
 			sizeof(struct ping_iter_state)))
 		return -ENOMEM;
+
+	net->ipv4.ping_port_rover = get_random_u16();
 	return 0;
 }
 
-- 
2.51.0.268.g9569e192d0-goog

Re: [PATCH v2 net-next 2/3] inet: ping: remove ping_hash()

[Yue Haibing]

On 2025/8/29 0:41, Eric Dumazet wrote:
> There is no point in keeping ping_hash().
> 
> Signed-off-by: Eric Dumazet <edumazet@google.com>
> Reviewed-by: David Ahern <dsahern@kernel.org>
> ---
>  net/ipv4/ping.c | 10 ----------
>  net/ipv6/ping.c |  1 -
>  2 files changed, 11 deletions(-)
> 
> diff --git a/net/ipv4/ping.c b/net/ipv4/ping.c
> index 74a0beddfcc41d8ba17792a11a9d027c9d590bac..75e1b0f5c697653e79166fde5f312f46b471344a 100644
> --- a/net/ipv4/ping.c
> +++ b/net/ipv4/ping.c
> @@ -67,7 +67,6 @@ static inline u32 ping_hashfn(const struct net *net, u32 num, u32 mask)
>  	pr_debug("hash(%u) = %u\n", num, res);
>  	return res;
>  }
> -EXPORT_SYMBOL_GPL(ping_hash);

The declaration should also be removed

include/net/ping.h:58:void ping_unhash(struct sock *sk);

>  
>  static inline struct hlist_head *ping_hashslot(struct ping_table *table,
>  					       struct net *net, unsigned int num)
> @@ -144,14 +143,6 @@ int ping_get_port(struct sock *sk, unsigned short ident)
>  }
>  EXPORT_SYMBOL_GPL(ping_get_port);
>  
> -int ping_hash(struct sock *sk)
> -{
> -	pr_debug("ping_hash(sk->port=%u)\n", inet_sk(sk)->inet_num);
> -	BUG(); /* "Please do not press this button again." */
> -
> -	return 0;
> -}
> -
>  void ping_unhash(struct sock *sk)
>  {
>  	struct inet_sock *isk = inet_sk(sk);
> @@ -1008,7 +999,6 @@ struct proto ping_prot = {
>  	.bind =		ping_bind,
>  	.backlog_rcv =	ping_queue_rcv_skb,
>  	.release_cb =	ip4_datagram_release_cb,
> -	.hash =		ping_hash,
>  	.unhash =	ping_unhash,
>  	.get_port =	ping_get_port,
>  	.put_port =	ping_unhash,
> diff --git a/net/ipv6/ping.c b/net/ipv6/ping.c
> index 82b0492923d458213ac7a6f9316158af2191e30f..d7a2cdaa26312b44f1fe502d3d40f3e27f961fa8 100644
> --- a/net/ipv6/ping.c
> +++ b/net/ipv6/ping.c
> @@ -208,7 +208,6 @@ struct proto pingv6_prot = {
>  	.recvmsg =	ping_recvmsg,
>  	.bind =		ping_bind,
>  	.backlog_rcv =	ping_queue_rcv_skb,
> -	.hash =		ping_hash,
>  	.unhash =	ping_unhash,
>  	.get_port =	ping_get_port,
>  	.put_port =	ping_unhash,

Re: [PATCH v2 net-next 2/3] inet: ping: remove ping_hash()

[Eric Dumazet]

On Thu, Aug 28, 2025 at 11:47 PM Yue Haibing <yuehaibing@huawei.com> wrote:
>
> On 2025/8/29 0:41, Eric Dumazet wrote:
> > There is no point in keeping ping_hash().
> >
> > Signed-off-by: Eric Dumazet <edumazet@google.com>
> > Reviewed-by: David Ahern <dsahern@kernel.org>
> > ---
> >  net/ipv4/ping.c | 10 ----------
> >  net/ipv6/ping.c |  1 -
> >  2 files changed, 11 deletions(-)
> >
> > diff --git a/net/ipv4/ping.c b/net/ipv4/ping.c
> > index 74a0beddfcc41d8ba17792a11a9d027c9d590bac..75e1b0f5c697653e79166fde5f312f46b471344a 100644
> > --- a/net/ipv4/ping.c
> > +++ b/net/ipv4/ping.c
> > @@ -67,7 +67,6 @@ static inline u32 ping_hashfn(const struct net *net, u32 num, u32 mask)
> >       pr_debug("hash(%u) = %u\n", num, res);
> >       return res;
> >  }
> > -EXPORT_SYMBOL_GPL(ping_hash);
>
> The declaration should also be removed
>
> include/net/ping.h:58:void ping_unhash(struct sock *sk);

Right, but you probably meant

include/net/ping.h:57:int ping_hash(struct sock *sk);

Re: [PATCH v2 net-next 2/3] inet: ping: remove ping_hash()

[Yue Haibing]

On 2025/8/29 14:57, Eric Dumazet wrote:
> On Thu, Aug 28, 2025 at 11:47 PM Yue Haibing <yuehaibing@huawei.com> wrote:
>>
>> On 2025/8/29 0:41, Eric Dumazet wrote:
>>> There is no point in keeping ping_hash().
>>>
>>> Signed-off-by: Eric Dumazet <edumazet@google.com>
>>> Reviewed-by: David Ahern <dsahern@kernel.org>
>>> ---
>>>  net/ipv4/ping.c | 10 ----------
>>>  net/ipv6/ping.c |  1 -
>>>  2 files changed, 11 deletions(-)
>>>
>>> diff --git a/net/ipv4/ping.c b/net/ipv4/ping.c
>>> index 74a0beddfcc41d8ba17792a11a9d027c9d590bac..75e1b0f5c697653e79166fde5f312f46b471344a 100644
>>> --- a/net/ipv4/ping.c
>>> +++ b/net/ipv4/ping.c
>>> @@ -67,7 +67,6 @@ static inline u32 ping_hashfn(const struct net *net, u32 num, u32 mask)
>>>       pr_debug("hash(%u) = %u\n", num, res);
>>>       return res;
>>>  }
>>> -EXPORT_SYMBOL_GPL(ping_hash);
>>
>> The declaration should also be removed
>>
>> include/net/ping.h:58:void ping_unhash(struct sock *sk);
> 
> Right, but you probably meant

Sure，I have a copy_paste mistake.
> 
> include/net/ping.h:57:int ping_hash(struct sock *sk);