ipv6 udp early demux breaks udp_l3mdev_accept=0

ipv6 udp early demux breaks udp_l3mdev_accept=0

[David Ahern]

Subash:

My understanding of early demux is that it should only match connected
sockets (src/sport + dst/dport are all set). The ipv6 udp early demux
code added in 5425077d73e0c is doing a generic socket lookup which does
not require an exact match and accordingly breaks udp_l3mdev_accept=0.
Please fix.

David

Re: ipv6 udp early demux breaks udp_l3mdev_accept=0

[Subash Abhinov Kasiviswanathan]

On 2017-04-14 16:23, David Ahern wrote:
> Subash:
> 
> My understanding of early demux is that it should only match connected
> sockets (src/sport + dst/dport are all set). The ipv6 udp early demux
> code added in 5425077d73e0c is doing a generic socket lookup which does
> not require an exact match and accordingly breaks udp_l3mdev_accept=0.
> Please fix.
> 
> David

Hi David

Would it be possible to disable udp_early_demux for your case 
temporarily.
I'll look into this.

Re: ipv6 udp early demux breaks udp_l3mdev_accept=0

[David Ahern]

On 4/14/17 6:38 PM, Subash Abhinov Kasiviswanathan wrote:
> On 2017-04-14 16:23, David Ahern wrote:
>> Subash:
>>
>> My understanding of early demux is that it should only match connected
>> sockets (src/sport + dst/dport are all set). The ipv6 udp early demux
>> code added in 5425077d73e0c is doing a generic socket lookup which does
>> not require an exact match and accordingly breaks udp_l3mdev_accept=0.
>> Please fix.
>>
>> David
> 
> Hi David
> 
> Would it be possible to disable udp_early_demux for your case temporarily.
> I'll look into this.
> 

It should be fixed for 4.12. Basically, __udp6_lib_demux_lookup needs to
be more like __udp4_lib_demux_lookup

Re: ipv6 udp early demux breaks udp_l3mdev_accept=0

[David Miller]

From: David Ahern <dsa@cumulusnetworks.com>
Date: Fri, 14 Apr 2017 18:42:14 -0600

> On 4/14/17 6:38 PM, Subash Abhinov Kasiviswanathan wrote:
>> On 2017-04-14 16:23, David Ahern wrote:
>>> Subash:
>>>
>>> My understanding of early demux is that it should only match connected
>>> sockets (src/sport + dst/dport are all set). The ipv6 udp early demux
>>> code added in 5425077d73e0c is doing a generic socket lookup which does
>>> not require an exact match and accordingly breaks udp_l3mdev_accept=0.
>>> Please fix.
>>>
>>> David
>> 
>> Hi David
>> 
>> Would it be possible to disable udp_early_demux for your case temporarily.
>> I'll look into this.
>> 
> 
> It should be fixed for 4.12. Basically, __udp6_lib_demux_lookup needs to
> be more like __udp4_lib_demux_lookup

Agreed.

Re: ipv6 udp early demux breaks udp_l3mdev_accept=0

[Subash Abhinov Kasiviswanathan]

>> It should be fixed for 4.12. Basically, __udp6_lib_demux_lookup needs 
>> to
>> be more like __udp4_lib_demux_lookup

Hi David

Would it be possible for you to test this with udp_l3mdev_accept=0

diff --git a/net/ipv6/udp.c b/net/ipv6/udp.c
index b793ed1..0e307e5 100644
--- a/net/ipv6/udp.c
+++ b/net/ipv6/udp.c
@@ -46,6 +46,7 @@
  #include <net/tcp_states.h>
  #include <net/ip6_checksum.h>
  #include <net/xfrm.h>
+#include <net/inet_hashtables.h>
  #include <net/inet6_hashtables.h>
  #include <net/busy_poll.h>
  #include <net/sock_reuseport.h>
@@ -864,21 +865,25 @@ int __udp6_lib_rcv(struct sk_buff *skb, struct 
udp_table *udptable,
         return 0;
  }

+
  static struct sock *__udp6_lib_demux_lookup(struct net *net,
                         __be16 loc_port, const struct in6_addr 
*loc_addr,
                         __be16 rmt_port, const struct in6_addr 
*rmt_addr,
                         int dif)
  {
+       unsigned short hnum = ntohs(loc_port);
+       unsigned int hash2 = udp6_portaddr_hash(net, loc_addr, hnum);
+       unsigned int slot2 = hash2 & udp_table.mask;
+       struct udp_hslot *hslot2 = &udp_table.hash2[slot2];
+       const __portpair ports = INET_COMBINED_PORTS(rmt_port, hnum);
         struct sock *sk;

-       rcu_read_lock();
-       sk = __udp6_lib_lookup(net, rmt_addr, rmt_port, loc_addr, 
loc_port,
-                              dif, &udp_table, NULL);
-       if (sk && !atomic_inc_not_zero(&sk->sk_refcnt))
-               sk = NULL;
-       rcu_read_unlock();
-
-       return sk;
+       udp_portaddr_for_each_entry_rcu(sk, &hslot2->head) {
+               if (INET6_MATCH(sk, net, rmt_addr, loc_addr, ports, 
dif))
+                       return sk;
+               break;
+       }
+       return NULL;
  }

  static void udp_v6_early_demux(struct sk_buff *skb)

Re: ipv6 udp early demux breaks udp_l3mdev_accept=0

[David Ahern]

On 4/15/17 8:23 PM, Subash Abhinov Kasiviswanathan wrote:
>>> It should be fixed for 4.12. Basically, __udp6_lib_demux_lookup needs to
>>> be more like __udp4_lib_demux_lookup
> 
> Hi David
> 
> Would it be possible for you to test this with udp_l3mdev_accept=0

formatting was botched so it does not apply cleanly. please send a patch
file

Re: ipv6 udp early demux breaks udp_l3mdev_accept=0

[Subash Abhinov Kasiviswanathan]

Can you try this

---
 net/ipv6/udp.c | 21 +++++++++++++--------
 1 file changed, 13 insertions(+), 8 deletions(-)

diff --git a/net/ipv6/udp.c b/net/ipv6/udp.c
index b793ed1..0e307e5 100644
--- a/net/ipv6/udp.c
+++ b/net/ipv6/udp.c
@@ -46,6 +46,7 @@
 #include <net/tcp_states.h>
 #include <net/ip6_checksum.h>
 #include <net/xfrm.h>
+#include <net/inet_hashtables.h>
 #include <net/inet6_hashtables.h>
 #include <net/busy_poll.h>
 #include <net/sock_reuseport.h>
@@ -864,21 +865,25 @@ int __udp6_lib_rcv(struct sk_buff *skb, struct udp_table *udptable,
 	return 0;
 }
 
+
 static struct sock *__udp6_lib_demux_lookup(struct net *net,
 			__be16 loc_port, const struct in6_addr *loc_addr,
 			__be16 rmt_port, const struct in6_addr *rmt_addr,
 			int dif)
 {
+	unsigned short hnum = ntohs(loc_port);
+	unsigned int hash2 = udp6_portaddr_hash(net, loc_addr, hnum);
+	unsigned int slot2 = hash2 & udp_table.mask;
+	struct udp_hslot *hslot2 = &udp_table.hash2[slot2];
+	const __portpair ports = INET_COMBINED_PORTS(rmt_port, hnum);
 	struct sock *sk;
 
-	rcu_read_lock();
-	sk = __udp6_lib_lookup(net, rmt_addr, rmt_port, loc_addr, loc_port,
-			       dif, &udp_table, NULL);
-	if (sk && !atomic_inc_not_zero(&sk->sk_refcnt))
-		sk = NULL;
-	rcu_read_unlock();
-
-	return sk;
+	udp_portaddr_for_each_entry_rcu(sk, &hslot2->head) {
+		if (INET6_MATCH(sk, net, rmt_addr, loc_addr, ports, dif))
+			return sk;
+		break;
+	}
+	return NULL;
 }
 
 static void udp_v6_early_demux(struct sk_buff *skb)
-- 
1.9.1

Re: ipv6 udp early demux breaks udp_l3mdev_accept=0

[David Ahern]

On 4/16/17 6:44 PM, Subash Abhinov Kasiviswanathan wrote:
> Can you try this
> 
> ---
>  net/ipv6/udp.c | 21 +++++++++++++--------
>  1 file changed, 13 insertions(+), 8 deletions(-)
> 
> diff --git a/net/ipv6/udp.c b/net/ipv6/udp.c
> index b793ed1..0e307e5 100644
> --- a/net/ipv6/udp.c
> +++ b/net/ipv6/udp.c
> @@ -46,6 +46,7 @@
>  #include <net/tcp_states.h>
>  #include <net/ip6_checksum.h>
>  #include <net/xfrm.h>
> +#include <net/inet_hashtables.h>
>  #include <net/inet6_hashtables.h>
>  #include <net/busy_poll.h>
>  #include <net/sock_reuseport.h>
> @@ -864,21 +865,25 @@ int __udp6_lib_rcv(struct sk_buff *skb, struct udp_table *udptable,
>  	return 0;
>  }
>  
> +
>  static struct sock *__udp6_lib_demux_lookup(struct net *net,
>  			__be16 loc_port, const struct in6_addr *loc_addr,
>  			__be16 rmt_port, const struct in6_addr *rmt_addr,
>  			int dif)
>  {
> +	unsigned short hnum = ntohs(loc_port);
> +	unsigned int hash2 = udp6_portaddr_hash(net, loc_addr, hnum);
> +	unsigned int slot2 = hash2 & udp_table.mask;
> +	struct udp_hslot *hslot2 = &udp_table.hash2[slot2];
> +	const __portpair ports = INET_COMBINED_PORTS(rmt_port, hnum);
>  	struct sock *sk;
>  
> -	rcu_read_lock();
> -	sk = __udp6_lib_lookup(net, rmt_addr, rmt_port, loc_addr, loc_port,
> -			       dif, &udp_table, NULL);
> -	if (sk && !atomic_inc_not_zero(&sk->sk_refcnt))
> -		sk = NULL;
> -	rcu_read_unlock();
> -
> -	return sk;
> +	udp_portaddr_for_each_entry_rcu(sk, &hslot2->head) {
> +		if (INET6_MATCH(sk, net, rmt_addr, loc_addr, ports, dif))
> +			return sk;
> +		break;
> +	}
> +	return NULL;
>  }
>  
>  static void udp_v6_early_demux(struct sk_buff *skb)
> 

That works for me. Thanks for the fast turn around.