[PATCH] net: allow unbound socket for packets in VRF when tcp_l3mdev_accept set

[PATCH] net: allow unbound socket for packets in VRF when tcp_l3mdev_accept set

[Mike Manning]

The commit 3c82a21f4320 ("net: allow binding socket in a VRF when
there's an unbound socket") changed the inet socket lookup to avoid
packets in a VRF from matching an unbound socket. This is to ensure the
necessary isolation between the default and other VRFs for routing and
forwarding. VRF-unaware processes running in the default VRF cannot
access another VRF and have to be run with 'ip vrf exec <vrf>'. This is
to be expected with tcp_l3mdev_accept disabled, but could be reallowed
when this sysctl option is enabled. So instead of directly checking dif
and sdif in inet[6]_match, here call inet_sk_bound_dev_eq(). This
allows a match on unbound socket for non-zero sdif i.e. for packets in
a VRF, if tcp_l3mdev_accept is enabled.

Fixes: 3c82a21f4320 ("net: allow binding socket in a VRF when there's an unbound socket")
Signed-off-by: Mike Manning <mvrmanning@gmail.com>
Link: https://lore.kernel.org/netdev/a54c149aed38fded2d3b5fdb1a6c89e36a083b74.camel@lasnet.de/
---

Nettest results for VRF testing remain unchanged:

$ diff nettest-baseline-502c6f8cedcc.txt nettest-fix.txt
$ tail -3 nettest-fix.txt
Tests passed: 869
Tests failed:   5

Disclaimer: While I do not think that there should be any noticeable
socket throughput degradation due to these changes, I am unable to
carry out any performance tests for this, nor provide any follow-up
support if there is any such degradation.

---
 include/net/inet6_hashtables.h |  7 +++----
 include/net/inet_hashtables.h  | 19 +++----------------
 include/net/inet_sock.h        | 11 +++++++++++
 3 files changed, 17 insertions(+), 20 deletions(-)

diff --git a/include/net/inet6_hashtables.h b/include/net/inet6_hashtables.h
index f259e1ae14ba..56f1286583d3 100644
--- a/include/net/inet6_hashtables.h
+++ b/include/net/inet6_hashtables.h
@@ -110,8 +110,6 @@ static inline bool inet6_match(struct net *net, const struct sock *sk,
 			       const __portpair ports,
 			       const int dif, const int sdif)
 {
-	int bound_dev_if;
-
 	if (!net_eq(sock_net(sk), net) ||
 	    sk->sk_family != AF_INET6 ||
 	    sk->sk_portpair != ports ||
@@ -119,8 +117,9 @@ static inline bool inet6_match(struct net *net, const struct sock *sk,
 	    !ipv6_addr_equal(&sk->sk_v6_rcv_saddr, daddr))
 		return false;
 
-	bound_dev_if = READ_ONCE(sk->sk_bound_dev_if);
-	return bound_dev_if == dif || bound_dev_if == sdif;
+	/* READ_ONCE() paired with WRITE_ONCE() in sock_bindtoindex_locked() */
+	return inet_sk_bound_dev_eq(net, READ_ONCE(sk->sk_bound_dev_if), dif,
+				    sdif);
 }
 #endif /* IS_ENABLED(CONFIG_IPV6) */
 
diff --git a/include/net/inet_hashtables.h b/include/net/inet_hashtables.h
index fd6b510d114b..e9cf2157ed8a 100644
--- a/include/net/inet_hashtables.h
+++ b/include/net/inet_hashtables.h
@@ -175,17 +175,6 @@ static inline void inet_ehash_locks_free(struct inet_hashinfo *hashinfo)
 	hashinfo->ehash_locks = NULL;
 }
 
-static inline bool inet_sk_bound_dev_eq(struct net *net, int bound_dev_if,
-					int dif, int sdif)
-{
-#if IS_ENABLED(CONFIG_NET_L3_MASTER_DEV)
-	return inet_bound_dev_eq(!!READ_ONCE(net->ipv4.sysctl_tcp_l3mdev_accept),
-				 bound_dev_if, dif, sdif);
-#else
-	return inet_bound_dev_eq(true, bound_dev_if, dif, sdif);
-#endif
-}
-
 struct inet_bind_bucket *
 inet_bind_bucket_create(struct kmem_cache *cachep, struct net *net,
 			struct inet_bind_hashbucket *head,
@@ -271,16 +260,14 @@ static inline bool inet_match(struct net *net, const struct sock *sk,
 			      const __addrpair cookie, const __portpair ports,
 			      int dif, int sdif)
 {
-	int bound_dev_if;
-
 	if (!net_eq(sock_net(sk), net) ||
 	    sk->sk_portpair != ports ||
 	    sk->sk_addrpair != cookie)
 	        return false;
 
-	/* Paired with WRITE_ONCE() from sock_bindtoindex_locked() */
-	bound_dev_if = READ_ONCE(sk->sk_bound_dev_if);
-	return bound_dev_if == dif || bound_dev_if == sdif;
+	/* READ_ONCE() paired with WRITE_ONCE() in sock_bindtoindex_locked() */
+	return inet_sk_bound_dev_eq(net, READ_ONCE(sk->sk_bound_dev_if), dif,
+				    sdif);
 }
 
 /* Sockets in TCP_CLOSE state are _always_ taken out of the hash, so we need
diff --git a/include/net/inet_sock.h b/include/net/inet_sock.h
index 6395f6b9a5d2..bf5654ce711e 100644
--- a/include/net/inet_sock.h
+++ b/include/net/inet_sock.h
@@ -149,6 +149,17 @@ static inline bool inet_bound_dev_eq(bool l3mdev_accept, int bound_dev_if,
 	return bound_dev_if == dif || bound_dev_if == sdif;
 }
 
+static inline bool inet_sk_bound_dev_eq(struct net *net, int bound_dev_if,
+					int dif, int sdif)
+{
+#if IS_ENABLED(CONFIG_NET_L3_MASTER_DEV)
+	return inet_bound_dev_eq(!!READ_ONCE(net->ipv4.sysctl_tcp_l3mdev_accept),
+				 bound_dev_if, dif, sdif);
+#else
+	return inet_bound_dev_eq(true, bound_dev_if, dif, sdif);
+#endif
+}
+
 struct inet_cork {
 	unsigned int		flags;
 	__be32			addr;
-- 
2.20.1

Re: [PATCH] net: allow unbound socket for packets in VRF when tcp_l3mdev_accept set

[David Ahern]

[ cc Andy who reported a similar issue ]

On 7/25/22 12:14 PM, Mike Manning wrote:
> The commit 3c82a21f4320 ("net: allow binding socket in a VRF when
> there's an unbound socket") changed the inet socket lookup to avoid
> packets in a VRF from matching an unbound socket. This is to ensure the
> necessary isolation between the default and other VRFs for routing and
> forwarding. VRF-unaware processes running in the default VRF cannot
> access another VRF and have to be run with 'ip vrf exec <vrf>'. This is
> to be expected with tcp_l3mdev_accept disabled, but could be reallowed
> when this sysctl option is enabled. So instead of directly checking dif
> and sdif in inet[6]_match, here call inet_sk_bound_dev_eq(). This
> allows a match on unbound socket for non-zero sdif i.e. for packets in
> a VRF, if tcp_l3mdev_accept is enabled.
> 
> Fixes: 3c82a21f4320 ("net: allow binding socket in a VRF when there's an unbound socket")
> Signed-off-by: Mike Manning <mvrmanning@gmail.com>
> Link: https://lore.kernel.org/netdev/a54c149aed38fded2d3b5fdb1a6c89e36a083b74.camel@lasnet.de/
> ---
> 
> Nettest results for VRF testing remain unchanged:
> 
> $ diff nettest-baseline-502c6f8cedcc.txt nettest-fix.txt
> $ tail -3 nettest-fix.txt
> Tests passed: 869
> Tests failed:   5
> 
> Disclaimer: While I do not think that there should be any noticeable
> socket throughput degradation due to these changes, I am unable to
> carry out any performance tests for this, nor provide any follow-up
> support if there is any such degradation.
> 
> ---
>  include/net/inet6_hashtables.h |  7 +++----
>  include/net/inet_hashtables.h  | 19 +++----------------
>  include/net/inet_sock.h        | 11 +++++++++++
>  3 files changed, 17 insertions(+), 20 deletions(-)
> 
> diff --git a/include/net/inet6_hashtables.h b/include/net/inet6_hashtables.h
> index f259e1ae14ba..56f1286583d3 100644
> --- a/include/net/inet6_hashtables.h
> +++ b/include/net/inet6_hashtables.h
> @@ -110,8 +110,6 @@ static inline bool inet6_match(struct net *net, const struct sock *sk,
>  			       const __portpair ports,
>  			       const int dif, const int sdif)
>  {
> -	int bound_dev_if;
> -
>  	if (!net_eq(sock_net(sk), net) ||
>  	    sk->sk_family != AF_INET6 ||
>  	    sk->sk_portpair != ports ||
> @@ -119,8 +117,9 @@ static inline bool inet6_match(struct net *net, const struct sock *sk,
>  	    !ipv6_addr_equal(&sk->sk_v6_rcv_saddr, daddr))
>  		return false;
>  
> -	bound_dev_if = READ_ONCE(sk->sk_bound_dev_if);
> -	return bound_dev_if == dif || bound_dev_if == sdif;
> +	/* READ_ONCE() paired with WRITE_ONCE() in sock_bindtoindex_locked() */
> +	return inet_sk_bound_dev_eq(net, READ_ONCE(sk->sk_bound_dev_if), dif,
> +				    sdif);
>  }
>  #endif /* IS_ENABLED(CONFIG_IPV6) */
>  
> diff --git a/include/net/inet_hashtables.h b/include/net/inet_hashtables.h
> index fd6b510d114b..e9cf2157ed8a 100644
> --- a/include/net/inet_hashtables.h
> +++ b/include/net/inet_hashtables.h
> @@ -175,17 +175,6 @@ static inline void inet_ehash_locks_free(struct inet_hashinfo *hashinfo)
>  	hashinfo->ehash_locks = NULL;
>  }
>  
> -static inline bool inet_sk_bound_dev_eq(struct net *net, int bound_dev_if,
> -					int dif, int sdif)
> -{
> -#if IS_ENABLED(CONFIG_NET_L3_MASTER_DEV)
> -	return inet_bound_dev_eq(!!READ_ONCE(net->ipv4.sysctl_tcp_l3mdev_accept),
> -				 bound_dev_if, dif, sdif);
> -#else
> -	return inet_bound_dev_eq(true, bound_dev_if, dif, sdif);
> -#endif
> -}
> -
>  struct inet_bind_bucket *
>  inet_bind_bucket_create(struct kmem_cache *cachep, struct net *net,
>  			struct inet_bind_hashbucket *head,
> @@ -271,16 +260,14 @@ static inline bool inet_match(struct net *net, const struct sock *sk,
>  			      const __addrpair cookie, const __portpair ports,
>  			      int dif, int sdif)
>  {
> -	int bound_dev_if;
> -
>  	if (!net_eq(sock_net(sk), net) ||
>  	    sk->sk_portpair != ports ||
>  	    sk->sk_addrpair != cookie)
>  	        return false;
>  
> -	/* Paired with WRITE_ONCE() from sock_bindtoindex_locked() */
> -	bound_dev_if = READ_ONCE(sk->sk_bound_dev_if);
> -	return bound_dev_if == dif || bound_dev_if == sdif;
> +	/* READ_ONCE() paired with WRITE_ONCE() in sock_bindtoindex_locked() */
> +	return inet_sk_bound_dev_eq(net, READ_ONCE(sk->sk_bound_dev_if), dif,
> +				    sdif);
>  }
>  
>  /* Sockets in TCP_CLOSE state are _always_ taken out of the hash, so we need
> diff --git a/include/net/inet_sock.h b/include/net/inet_sock.h
> index 6395f6b9a5d2..bf5654ce711e 100644
> --- a/include/net/inet_sock.h
> +++ b/include/net/inet_sock.h
> @@ -149,6 +149,17 @@ static inline bool inet_bound_dev_eq(bool l3mdev_accept, int bound_dev_if,
>  	return bound_dev_if == dif || bound_dev_if == sdif;
>  }
>  
> +static inline bool inet_sk_bound_dev_eq(struct net *net, int bound_dev_if,
> +					int dif, int sdif)
> +{
> +#if IS_ENABLED(CONFIG_NET_L3_MASTER_DEV)
> +	return inet_bound_dev_eq(!!READ_ONCE(net->ipv4.sysctl_tcp_l3mdev_accept),
> +				 bound_dev_if, dif, sdif);
> +#else
> +	return inet_bound_dev_eq(true, bound_dev_if, dif, sdif);
> +#endif
> +}
> +
>  struct inet_cork {
>  	unsigned int		flags;
>  	__be32			addr;

Re: [PATCH] net: allow unbound socket for packets in VRF when tcp_l3mdev_accept set

[David Ahern]

On 7/25/22 12:14 PM, Mike Manning wrote:
> The commit 3c82a21f4320 ("net: allow binding socket in a VRF when
> there's an unbound socket") changed the inet socket lookup to avoid
> packets in a VRF from matching an unbound socket. This is to ensure the
> necessary isolation between the default and other VRFs for routing and
> forwarding. VRF-unaware processes running in the default VRF cannot
> access another VRF and have to be run with 'ip vrf exec <vrf>'. This is
> to be expected with tcp_l3mdev_accept disabled, but could be reallowed
> when this sysctl option is enabled. So instead of directly checking dif
> and sdif in inet[6]_match, here call inet_sk_bound_dev_eq(). This
> allows a match on unbound socket for non-zero sdif i.e. for packets in
> a VRF, if tcp_l3mdev_accept is enabled.
> 
> Fixes: 3c82a21f4320 ("net: allow binding socket in a VRF when there's an unbound socket")
> Signed-off-by: Mike Manning <mvrmanning@gmail.com>
> Link: https://lore.kernel.org/netdev/a54c149aed38fded2d3b5fdb1a6c89e36a083b74.camel@lasnet.de/
> ---
> 
> Nettest results for VRF testing remain unchanged:
> 
> $ diff nettest-baseline-502c6f8cedcc.txt nettest-fix.txt
> $ tail -3 nettest-fix.txt
> Tests passed: 869
> Tests failed:   5
> 
> Disclaimer: While I do not think that there should be any noticeable
> socket throughput degradation due to these changes, I am unable to
> carry out any performance tests for this, nor provide any follow-up
> support if there is any such degradation.
> 
> ---
>  include/net/inet6_hashtables.h |  7 +++----
>  include/net/inet_hashtables.h  | 19 +++----------------
>  include/net/inet_sock.h        | 11 +++++++++++
>  3 files changed, 17 insertions(+), 20 deletions(-)
> 

Reviewed-by: David Ahern <dsahern@kernel.org>

Re: [PATCH] net: allow unbound socket for packets in VRF when tcp_l3mdev_accept set

[patchwork-bot+netdevbpf]

Hello:

This patch was applied to netdev/net.git (master)
by David S. Miller <davem@davemloft.net>:

On Mon, 25 Jul 2022 19:14:42 +0100 you wrote:
> The commit 3c82a21f4320 ("net: allow binding socket in a VRF when
> there's an unbound socket") changed the inet socket lookup to avoid
> packets in a VRF from matching an unbound socket. This is to ensure the
> necessary isolation between the default and other VRFs for routing and
> forwarding. VRF-unaware processes running in the default VRF cannot
> access another VRF and have to be run with 'ip vrf exec <vrf>'. This is
> to be expected with tcp_l3mdev_accept disabled, but could be reallowed
> when this sysctl option is enabled. So instead of directly checking dif
> and sdif in inet[6]_match, here call inet_sk_bound_dev_eq(). This
> allows a match on unbound socket for non-zero sdif i.e. for packets in
> a VRF, if tcp_l3mdev_accept is enabled.
> 
> [...]

Here is the summary with links:
  - net: allow unbound socket for packets in VRF when tcp_l3mdev_accept set
    https://git.kernel.org/netdev/net/c/944fd1aeacb6

You are awesome, thank you!
-- 
Deet-doot-dot, I am a bot.
https://korg.docs.kernel.org/patchwork/pwbot.html