[PATCH net 0/2] udp: Fix reuseport selection with connected sockets.

[PATCH net 0/2] udp: Fix reuseport selection with connected sockets.

[Kuniyuki Iwashima]

From: kuniyu <kuniyu@amazon.co.jp>

This patch set addresses two issues which happen when both connected and
unconnected sockets are in the same UDP reuseport group.

Kuniyuki Iwashima (2):
  udp: Copy has_conns in reuseport_grow().
  udp: Improve load balancing for SO_REUSEPORT.

 net/core/sock_reuseport.c |  1 +
 net/ipv4/udp.c            | 15 +++++++++------
 net/ipv6/udp.c            | 15 +++++++++------
 3 files changed, 19 insertions(+), 12 deletions(-)

-- 
2.17.2 (Apple Git-113)

[PATCH net 1/2] udp: Copy has_conns in reuseport_grow().

[Kuniyuki Iwashima]

If an unconnected socket in a UDP reuseport group connect()s, has_conns is
set to 1. Then, when a packet is received, udp[46]_lib_lookup2() scans all
sockets in udp_hslot looking for the connected socket with the highest
score.

However, when the number of sockets bound to the port exceeds max_socks,
reuseport_grow() resets has_conns to 0. It can cause udp[46]_lib_lookup2()
to return without scanning all sockets, resulting in that packets sent to
connected sockets may be distributed to unconnected sockets.

Therefore, reuseport_grow() should copy has_conns.

Fixes: acdcecc61285 ("udp: correct reuseport selection with connected sockets")
CC: Willem de Bruijn <willemb@google.com>
Reviewed-by: Benjamin Herrenschmidt <benh@amazon.com>
Signed-off-by: Kuniyuki Iwashima <kuniyu@amazon.co.jp>
---
 net/core/sock_reuseport.c | 1 +
 1 file changed, 1 insertion(+)

diff --git a/net/core/sock_reuseport.c b/net/core/sock_reuseport.c
index adcb3aea576d..bbdd3c7b6cb5 100644
--- a/net/core/sock_reuseport.c
+++ b/net/core/sock_reuseport.c
@@ -101,6 +101,7 @@ static struct sock_reuseport *reuseport_grow(struct sock_reuseport *reuse)
 	more_reuse->prog = reuse->prog;
 	more_reuse->reuseport_id = reuse->reuseport_id;
 	more_reuse->bind_inany = reuse->bind_inany;
+	more_reuse->has_conns = reuse->has_conns;
 
 	memcpy(more_reuse->socks, reuse->socks,
 	       reuse->num_socks * sizeof(struct sock *));
-- 
2.17.2 (Apple Git-113)

Re: [PATCH net 1/2] udp: Copy has_conns in reuseport_grow().

[Willem de Bruijn]

On Tue, Jul 21, 2020 at 2:16 AM Kuniyuki Iwashima <kuniyu@amazon.co.jp> wrote:
>
> If an unconnected socket in a UDP reuseport group connect()s, has_conns is
> set to 1. Then, when a packet is received, udp[46]_lib_lookup2() scans all
> sockets in udp_hslot looking for the connected socket with the highest
> score.
>
> However, when the number of sockets bound to the port exceeds max_socks,
> reuseport_grow() resets has_conns to 0. It can cause udp[46]_lib_lookup2()
> to return without scanning all sockets, resulting in that packets sent to
> connected sockets may be distributed to unconnected sockets.
>
> Therefore, reuseport_grow() should copy has_conns.
>
> Fixes: acdcecc61285 ("udp: correct reuseport selection with connected sockets")
> CC: Willem de Bruijn <willemb@google.com>
> Reviewed-by: Benjamin Herrenschmidt <benh@amazon.com>
> Signed-off-by: Kuniyuki Iwashima <kuniyu@amazon.co.jp>

Acked-by: Willem de Bruijn <willemb@google.com>

Thanks. Yes, I missed this resize operation when adding the field.

[PATCH net 2/2] udp: Improve load balancing for SO_REUSEPORT.

[Kuniyuki Iwashima]

Currently, SO_REUSEPORT does not work well if connected sockets are in a
UDP reuseport group.

Then reuseport_has_conns() returns true and the result of
reuseport_select_sock() is discarded. Also, unconnected sockets have the
same score, hence only does the first unconnected socket in udp_hslot
always receive all packets sent to unconnected sockets.

So, the result of reuseport_select_sock() should be used for load
balancing.

The noteworthy point is that the unconnected sockets placed after
connected sockets in sock_reuseport.socks will receive more packets than
others because of the algorithm in reuseport_select_sock().

    index | connected | reciprocal_scale | result
    ---------------------------------------------
    0     | no        | 20%              | 40%
    1     | no        | 20%              | 20%
    2     | yes       | 20%              | 0%
    3     | no        | 20%              | 40%
    4     | yes       | 20%              | 0%

If most of the sockets are connected, this can be a problem, but it still
works better than now.

Fixes: acdcecc61285 ("udp: correct reuseport selection with connected sockets")
CC: Willem de Bruijn <willemb@google.com>
Reviewed-by: Benjamin Herrenschmidt <benh@amazon.com>
Signed-off-by: Kuniyuki Iwashima <kuniyu@amazon.co.jp>
---
 net/ipv4/udp.c | 15 +++++++++------
 net/ipv6/udp.c | 15 +++++++++------
 2 files changed, 18 insertions(+), 12 deletions(-)

diff --git a/net/ipv4/udp.c b/net/ipv4/udp.c
index 1b7ebbcae497..99251d3c70d0 100644
--- a/net/ipv4/udp.c
+++ b/net/ipv4/udp.c
@@ -416,7 +416,7 @@ static struct sock *udp4_lib_lookup2(struct net *net,
 				     struct udp_hslot *hslot2,
 				     struct sk_buff *skb)
 {
-	struct sock *sk, *result;
+	struct sock *sk, *result, *reuseport_result;
 	int score, badness;
 	u32 hash = 0;
 
@@ -426,17 +426,20 @@ static struct sock *udp4_lib_lookup2(struct net *net,
 		score = compute_score(sk, net, saddr, sport,
 				      daddr, hnum, dif, sdif);
 		if (score > badness) {
+			reuseport_result = NULL;
+
 			if (sk->sk_reuseport &&
 			    sk->sk_state != TCP_ESTABLISHED) {
 				hash = udp_ehashfn(net, daddr, hnum,
 						   saddr, sport);
-				result = reuseport_select_sock(sk, hash, skb,
-							sizeof(struct udphdr));
-				if (result && !reuseport_has_conns(sk, false))
-					return result;
+				reuseport_result = reuseport_select_sock(sk, hash, skb,
+									 sizeof(struct udphdr));
+				if (reuseport_result && !reuseport_has_conns(sk, false))
+					return reuseport_result;
 			}
+
+			result = reuseport_result ? : sk;
 			badness = score;
-			result = sk;
 		}
 	}
 	return result;
diff --git a/net/ipv6/udp.c b/net/ipv6/udp.c
index 7d4151747340..9503c87ac0b3 100644
--- a/net/ipv6/udp.c
+++ b/net/ipv6/udp.c
@@ -148,7 +148,7 @@ static struct sock *udp6_lib_lookup2(struct net *net,
 		int dif, int sdif, struct udp_hslot *hslot2,
 		struct sk_buff *skb)
 {
-	struct sock *sk, *result;
+	struct sock *sk, *result, *reuseport_result;
 	int score, badness;
 	u32 hash = 0;
 
@@ -158,17 +158,20 @@ static struct sock *udp6_lib_lookup2(struct net *net,
 		score = compute_score(sk, net, saddr, sport,
 				      daddr, hnum, dif, sdif);
 		if (score > badness) {
+			reuseport_result = NULL;
+
 			if (sk->sk_reuseport &&
 			    sk->sk_state != TCP_ESTABLISHED) {
 				hash = udp6_ehashfn(net, daddr, hnum,
 						    saddr, sport);
 
-				result = reuseport_select_sock(sk, hash, skb,
-							sizeof(struct udphdr));
-				if (result && !reuseport_has_conns(sk, false))
-					return result;
+				reuseport_result = reuseport_select_sock(sk, hash, skb,
+									 sizeof(struct udphdr));
+				if (reuseport_result && !reuseport_has_conns(sk, false))
+					return reuseport_result;
 			}
-			result = sk;
+
+			result = reuseport_result ? : sk;
 			badness = score;
 		}
 	}
-- 
2.17.2 (Apple Git-113)

Re: [PATCH net 2/2] udp: Improve load balancing for SO_REUSEPORT.

[Willem de Bruijn]

On Tue, Jul 21, 2020 at 2:17 AM Kuniyuki Iwashima <kuniyu@amazon.co.jp> wrote:
>
> Currently, SO_REUSEPORT does not work well if connected sockets are in a
> UDP reuseport group.
>
> Then reuseport_has_conns() returns true and the result of
> reuseport_select_sock() is discarded. Also, unconnected sockets have the
> same score, hence only does the first unconnected socket in udp_hslot
> always receive all packets sent to unconnected sockets.
>
> So, the result of reuseport_select_sock() should be used for load
> balancing.
>
> The noteworthy point is that the unconnected sockets placed after
> connected sockets in sock_reuseport.socks will receive more packets than
> others because of the algorithm in reuseport_select_sock().
>
>     index | connected | reciprocal_scale | result
>     ---------------------------------------------
>     0     | no        | 20%              | 40%
>     1     | no        | 20%              | 20%
>     2     | yes       | 20%              | 0%
>     3     | no        | 20%              | 40%
>     4     | yes       | 20%              | 0%
>
> If most of the sockets are connected, this can be a problem, but it still
> works better than now.
>
> Fixes: acdcecc61285 ("udp: correct reuseport selection with connected sockets")
> CC: Willem de Bruijn <willemb@google.com>
> Reviewed-by: Benjamin Herrenschmidt <benh@amazon.com>
> Signed-off-by: Kuniyuki Iwashima <kuniyu@amazon.co.jp>

Acked-by: Willem de Bruijn <willemb@google.com>

Re: [PATCH net 0/2] udp: Fix reuseport selection with connected sockets.

[David Miller]

From: Kuniyuki Iwashima <kuniyu@amazon.co.jp>
Date: Tue, 21 Jul 2020 15:15:29 +0900

> From: kuniyu <kuniyu@amazon.co.jp>

Please fix your configuration to show your full name in this
"From: " field, I had to edit it out and use the one from your
email headers.

> This patch set addresses two issues which happen when both connected and
> unconnected sockets are in the same UDP reuseport group.

Series applied and queued up for -stable, th anks.

Re: [PATCH net 0/2] udp: Fix reuseport selection with connected sockets.

[Kuniyuki Iwashima]

From:   David Miller <davem@davemloft.net>
Date:   Tue, 21 Jul 2020 15:32:39 -0700 (PDT)
> From: Kuniyuki Iwashima <kuniyu@amazon.co.jp>
> Date: Tue, 21 Jul 2020 15:15:29 +0900
> 
> > From: kuniyu <kuniyu@amazon.co.jp>
> 
> Please fix your configuration to show your full name in this
> "From: " field, I had to edit it out and use the one from your
> email headers.
> 
> > This patch set addresses two issues which happen when both connected and
> > unconnected sockets are in the same UDP reuseport group.
> 
> Series applied and queued up for -stable, th anks.

I am sorry for bothering you... and grateful for your kind advice.
I have fixed my configuration for net repository as with net-next.

Thank you.