netdev - Re: Soft lockup in inet_put

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <1481343298.4930.208.camel@edumazet-glaptop3.roam.corp.google.com>
Date:   Fri, 09 Dec 2016 20:14:58 -0800
From:   Eric Dumazet <eric.dumazet@...il.com>
To:     Josef Bacik <jbacik@...com>
Cc:     Hannes Frederic Sowa <hannes@...essinduktion.org>,
        Tom Herbert <tom@...bertland.com>,
        Linux Kernel Network Developers <netdev@...r.kernel.org>
Subject: Re: Soft lockup in inet_put_port on 4.6

On Fri, 2016-12-09 at 19:47 -0800, Eric Dumazet wrote:

> 
> Hmm... Is your ephemeral port range includes the port your load
> balancing app is using ?

I suspect that you might have processes doing bind( port = 0) that are
trapped into the bind_conflict() scan ?

With 100,000 + timewaits there, this possibly hurts.

Can you try the following loop breaker ?

diff --git a/net/ipv4/inet_connection_sock.c b/net/ipv4/inet_connection_sock.c
index d5d3ead0a6c31e42e8843d30f8c643324a91b8e9..74f0f5ee6a02c624edb0263b9ddd27813f68d0a5 100644
--- a/net/ipv4/inet_connection_sock.c
+++ b/net/ipv4/inet_connection_sock.c
@@ -51,7 +51,7 @@ int inet_csk_bind_conflict(const struct sock *sk,
 	int reuse = sk->sk_reuse;
 	int reuseport = sk->sk_reuseport;
 	kuid_t uid = sock_i_uid((struct sock *)sk);
-
+	unsigned int max_count;
 	/*
 	 * Unlike other sk lookup places we do not check
 	 * for sk_net here, since _all_ the socks listed
@@ -59,6 +59,7 @@ int inet_csk_bind_conflict(const struct sock *sk,
 	 * one this bucket belongs to.
 	 */
 
+	max_count = relax ? ~0U : 100;
 	sk_for_each_bound(sk2, &tb->owners) {
 		if (sk != sk2 &&
 		    !inet_v6_ipv6only(sk2) &&
@@ -84,6 +85,8 @@ int inet_csk_bind_conflict(const struct sock *sk,
 					break;
 			}
 		}
+		if (--max_count == 0)
+			return 1;
 	}
 	return sk2 != NULL;
 }
diff --git a/net/ipv6/inet6_connection_sock.c b/net/ipv6/inet6_connection_sock.c
index 1c86c478f578b49373e61a4c397f23f3dc7f3fc6..4f63d06e0d601da94eb3f2b35a988abd060e156c 100644
--- a/net/ipv6/inet6_connection_sock.c
+++ b/net/ipv6/inet6_connection_sock.c
@@ -35,12 +35,14 @@ int inet6_csk_bind_conflict(const struct sock *sk,
 	int reuse = sk->sk_reuse;
 	int reuseport = sk->sk_reuseport;
 	kuid_t uid = sock_i_uid((struct sock *)sk);
+	unsigned int max_count;
 
 	/* We must walk the whole port owner list in this case. -DaveM */
 	/*
 	 * See comment in inet_csk_bind_conflict about sock lookup
 	 * vs net namespaces issues.
 	 */
+	max_count = relax ? ~0U : 100;
 	sk_for_each_bound(sk2, &tb->owners) {
 		if (sk != sk2 &&
 		    (!sk->sk_bound_dev_if ||
@@ -61,6 +63,8 @@ int inet6_csk_bind_conflict(const struct sock *sk,
 			    ipv6_rcv_saddr_equal(sk, sk2, true))
 				break;
 		}
+		if (--max_count == 0)
+			return 1;
 	}
 
 	return sk2 != NULL;