Skip to contentSkip to conversation

net: udp: increase UDP_MIB_RCVBUFERRORS when ENOBUFS

PATCH In review Menglong Dong 2 participants 3 messages Oct 26, 2020

Subsystems (3)
lore
Share & tools
lore
#0 lore raw

PATCH 1/1 net: udp: increase UDP_MIB_RCVBUFERRORS when ENOBUFS

The error returned from __udp_enqueue_schedule_skb is ENOMEM or ENOBUFS. For now, only ENOMEM is counted into UDP_MIB_RCVBUFERRORS in __udp_queue_rcv_skb. UDP_MIB_RCVBUFERRORS should count all of the failed skb because of memory errors during udp receiving, not just those because of the limit of sock receive queue. We can see this in __udp4_lib_mcast_deliver:

		nskb = skb_clone(skb, GFP_ATOMIC);

		if (unlikely(!nskb)) {
			atomic_inc(&sk->sk_drops);
			__UDP_INC_STATS(net, UDP_MIB_RCVBUFERRORS,
					IS_UDPLITE(sk));
			__UDP_INC_STATS(net, UDP_MIB_INERRORS,
					IS_UDPLITE(sk));
			continue;
		}

See, UDP_MIB_RCVBUFERRORS is increased when skb clone failed. From this point, ENOBUFS from __udp_enqueue_schedule_skb should be counted, too. It means that the buffer used by all of the UDP sock is to the limit, and it ought to be counted.

---
 net/ipv4/udp.c | 4 +---
 net/ipv6/udp.c | 4 +---
 2 files changed, 2 insertions(+), 6 deletions(-)
net/ipv4/udp.c +1 -3
index 09f0a23d1a01..49a69d8d55b3 100644
@@ -2035,9 +2035,7 @@ static int __udp_queue_rcv_skb(struct sock *sk, struct sk_buff *skb)
 		int is_udplite = IS_UDPLITE(sk);
 
 		/* Note that an ENOMEM error is charged twice */
-		if (rc == -ENOMEM)
-			UDP_INC_STATS(sock_net(sk), UDP_MIB_RCVBUFERRORS,
-					is_udplite);
+		UDP_INC_STATS(sock_net(sk), UDP_MIB_RCVBUFERRORS, is_udplite);
 		UDP_INC_STATS(sock_net(sk), UDP_MIB_INERRORS, is_udplite);
 		kfree_skb(skb);
 		trace_udp_fail_queue_rcv_skb(rc, sk);
net/ipv6/udp.c +1 -3
index 29d9691359b9..d5e23b150fd9 100644
@@ -634,9 +634,7 @@ static int __udpv6_queue_rcv_skb(struct sock *sk, struct sk_buff *skb)
 		int is_udplite = IS_UDPLITE(sk);
 
 		/* Note that an ENOMEM error is charged twice */
-		if (rc == -ENOMEM)
-			UDP6_INC_STATS(sock_net(sk),
-					 UDP_MIB_RCVBUFERRORS, is_udplite);
+		UDP6_INC_STATS(sock_net(sk), UDP_MIB_RCVBUFERRORS, is_udplite);
 		UDP6_INC_STATS(sock_net(sk), UDP_MIB_INERRORS, is_udplite);
 		kfree_skb(skb);
 		return -1;
Signature · 1 line
2.28.0
maintainer
#1 lore raw

Hello,

Quoted 48 lines · Menglong Dong

The error returned from __udp_enqueue_schedule_skb is ENOMEM or ENOBUFS.
For now, only ENOMEM is counted into UDP_MIB_RCVBUFERRORS in
__udp_queue_rcv_skb. UDP_MIB_RCVBUFERRORS should count all of the
failed skb because of memory errors during udp receiving, not just because of the limit of sock receive queue. We can see this in __udp4_lib_mcast_deliver:

		nskb = skb_clone(skb, GFP_ATOMIC);

		if (unlikely(!nskb)) {
			atomic_inc(&sk->sk_drops);
			__UDP_INC_STATS(net, UDP_MIB_RCVBUFERRORS,
					IS_UDPLITE(sk));
			__UDP_INC_STATS(net, UDP_MIB_INERRORS,
					IS_UDPLITE(sk));
			continue;
		}

See, UDP_MIB_RCVBUFERRORS is increased when skb clone failed. From this point, ENOBUFS from __udp_enqueue_schedule_skb should be counted, too. It means that the buffer used by all of the UDP sock is to the limit, and it ought to be counted.

---
 net/ipv4/udp.c | 4 +---
 net/ipv6/udp.c | 4 +---
 2 files changed, 2 insertions(+), 6 deletions(-)
net/ipv4/udp.c +1 -3
index 09f0a23d1a01..49a69d8d55b3 100644
@@ -2035,9 +2035,7 @@ static int __udp_queue_rcv_skb(struct sock *sk, struct sk_buff *skb)
 		int is_udplite = IS_UDPLITE(sk);
 
 		/* Note that an ENOMEM error is charged twice */
-		if (rc == -ENOMEM)
-			UDP_INC_STATS(sock_net(sk), UDP_MIB_RCVBUFERRORS,
-					is_udplite);
+		UDP_INC_STATS(sock_net(sk), UDP_MIB_RCVBUFERRORS, is_udplite);
 		UDP_INC_STATS(sock_net(sk), UDP_MIB_INERRORS, is_udplite);
 		kfree_skb(skb);
 		trace_udp_fail_queue_rcv_skb(rc, sk);
net/ipv6/udp.c +0 -0
index 29d9691359b9..d5e23b150fd9 100644

net/ipv6/udp.c static int __udpv6_queue_rcv_skb(struct sock *sk, struct sk_buff *skb) · line 640

Quoted 10 lines
net/ipv6/udp.c +1 -3
@@ -634,9 +634,7 @@ static int __udpv6_queue_rcv_skb(struct sock *sk, struct sk_buff *skb)
 		int is_udplite = IS_UDPLITE(sk);
 
 		/* Note that an ENOMEM error is charged twice */
-		if (rc == -ENOMEM)
-			UDP6_INC_STATS(sock_net(sk),
-					 UDP_MIB_RCVBUFERRORS, is_udplite);
+		UDP6_INC_STATS(sock_net(sk), UDP_MIB_RCVBUFERRORS, is_udplite);
 		UDP6_INC_STATS(sock_net(sk), UDP_MIB_INERRORS, is_udplite);
 		kfree_skb(skb);
 		return -1;

The diffstat is nice, but I'm unsure we can do this kind of change (well, I really think we should not do it): it will fool any kind of existing users (application, scripts, admin) currently reading the above counters and expecting UDP_MIB_RCVBUFERRORS being increased with the existing schema.

Cheers,

Paolo

#2 lore raw

Hello~

Quoted 69 lines · Paolo Abeni

Hello,

Quoted 48 lines · Menglong Dong

The error returned from __udp_enqueue_schedule_skb is ENOMEM or ENOBUFS.
For now, only ENOMEM is counted into UDP_MIB_RCVBUFERRORS in
__udp_queue_rcv_skb. UDP_MIB_RCVBUFERRORS should count all of the
failed skb because of memory errors during udp receiving, not just because of the limit of sock receive queue. We can see this in __udp4_lib_mcast_deliver:

              nskb = skb_clone(skb, GFP_ATOMIC);

              if (unlikely(!nskb)) {
                      atomic_inc(&sk->sk_drops);
                      __UDP_INC_STATS(net, UDP_MIB_RCVBUFERRORS,
                                      IS_UDPLITE(sk));
                      __UDP_INC_STATS(net, UDP_MIB_INERRORS,
                                      IS_UDPLITE(sk));
                      continue;
              }

See, UDP_MIB_RCVBUFERRORS is increased when skb clone failed. From this point, ENOBUFS from __udp_enqueue_schedule_skb should be counted, too. It means that the buffer used by all of the UDP sock is to the limit, and it ought to be counted.

---
 net/ipv4/udp.c | 4 +---
 net/ipv6/udp.c | 4 +---
 2 files changed, 2 insertions(+), 6 deletions(-)
net/ipv4/udp.c +1 -3
index 09f0a23d1a01..49a69d8d55b3 100644
@@ -2035,9 +2035,7 @@ static int __udp_queue_rcv_skb(struct sock *sk, struct sk_buff *skb)
              int is_udplite = IS_UDPLITE(sk);
 
              /* Note that an ENOMEM error is charged twice */
-             if (rc == -ENOMEM)
-                     UDP_INC_STATS(sock_net(sk), UDP_MIB_RCVBUFERRORS,
-                                     is_udplite);
+             UDP_INC_STATS(sock_net(sk), UDP_MIB_RCVBUFERRORS, is_udplite);
              UDP_INC_STATS(sock_net(sk), UDP_MIB_INERRORS, is_udplite);
              kfree_skb(skb);
              trace_udp_fail_queue_rcv_skb(rc, sk);
net/ipv6/udp.c +0 -0
index 29d9691359b9..d5e23b150fd9 100644

net/ipv6/udp.c static int __udpv6_queue_rcv_skb(struct sock *sk, struct sk_buff *skb) · line 640

Quoted 10 lines
net/ipv6/udp.c +1 -3
@@ -634,9 +634,7 @@ static int __udpv6_queue_rcv_skb(struct sock *sk, struct sk_buff *skb)
              int is_udplite = IS_UDPLITE(sk);
 
              /* Note that an ENOMEM error is charged twice */
-             if (rc == -ENOMEM)
-                     UDP6_INC_STATS(sock_net(sk),
-                                      UDP_MIB_RCVBUFERRORS, is_udplite);
+             UDP6_INC_STATS(sock_net(sk), UDP_MIB_RCVBUFERRORS, is_udplite);
              UDP6_INC_STATS(sock_net(sk), UDP_MIB_INERRORS, is_udplite);
              kfree_skb(skb);
              return -1;

The diffstat is nice, but I'm unsure we can do this kind of change (well, I really think we should not do it): it will fool any kind of existing users (application, scripts, admin) currently reading the above counters and expecting UDP_MIB_RCVBUFERRORS being increased with the existing schema.

Quoted 4 lines

Cheers,

Paolo

Well, your words make sense, this change isn't friendly for the existing users. It really puzzled me when this ENOBUFS happened, no counters were done and I hardly figured out what happened.

So, is it a good idea to introduce a 'UDP_MIB_MEMERRORS'?

Cheers,

Menglong Dong