netdev - [PATCH net] tcp: fix connect() invalid -EADDRNOTAVAIL error

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-next>] [day] [month] [year] [list]

Message-Id: <1416379060-15685-1-git-send-email-jmaxwell37@gmail.com>
Date:	Wed, 19 Nov 2014 17:37:40 +1100
From:	Jon Maxwell <jmaxwell37@...il.com>
To:	davem@...emloft.net
Cc:	kuznet@....inr.ac.ru, jmorris@...ei.org, yoshfuji@...ux-ipv6.org,
	kaber@...sh.net, netdev@...r.kernel.org,
	linux-kernel@...r.kernel.org, jmaxwell@...hat.com,
	Jon Maxwell <jmaxwell37@...il.com>
Subject: [PATCH net] tcp: fix connect() invalid -EADDRNOTAVAIL error

The connect() routine returns -EADDRNOTAVAIL without doing a 4 
tuple check when the hash buckets were previously allocated by 
bind() and all local ports are used.

The bind() routine creates the local port hash buckets in 
inet_csk_get_port(). Depending on the socket options it sets 
tb->fastreuse and tb->fastreuseport to 0 or 1 in the bucket.

However the __inet_hash_connect() routine initializes the hash 
buckets differently and sets these to -1. The end result is 
that connect() calling into __inet_hash_connect() will 
subsequently ignore the check_established() routine if, here

__inet_hash_connect()
.
.
if (tb->fastreuse >= 0 ||↩
    tb->fastreuseport >= 0)↩
    goto next_port;

and cycle through all local ports until it returns -EADDRNOTAVAIL. 
The 4 tuple check is in check_established() so connect() can fail 
unnecessarily.

Prerequisites for this to happen:
1) The local tcp port range must be exhausted.
2) A process must have called bind() followed by connect() for all 
local ports.
3) A different process calls connect() only which returns -EADDRNOTAVAIL. 
4) The system more than 1 interface configured.

If a system has 2 IP Addresses and all local tcp ports are in use
for connection from IP Address (1). Connecting to the same ports 
via IP Address (2) should work based on the 4 tuple rule. But it 
fails under this condition. 

To fix this make __inet_hash_connect() honour inet_csk_get_port()'s
tb->fastreuse* variables.

Signed-off-by: Jon Maxwell <jmaxwell37@...il.com>
---
 net/ipv4/inet_hashtables.c | 6 ++----
 1 file changed, 2 insertions(+), 4 deletions(-)

diff --git a/net/ipv4/inet_hashtables.c b/net/ipv4/inet_hashtables.c
index 9111a4e..b39e89e 100644
--- a/net/ipv4/inet_hashtables.c
+++ b/net/ipv4/inet_hashtables.c
@@ -513,8 +513,8 @@ int __inet_hash_connect(struct inet_timewait_death_row *death_row,
 			inet_bind_bucket_for_each(tb, &head->chain) {
 				if (net_eq(ib_net(tb), net) &&
 				    tb->port == port) {
-					if (tb->fastreuse >= 0 ||
-					    tb->fastreuseport >= 0)
+					if (tb->fastreuse > 0 ||
+					    tb->fastreuseport > 0)
 						goto next_port;
 					WARN_ON(hlist_empty(&tb->owners));
 					if (!check_established(death_row, sk,
@@ -530,8 +530,6 @@ int __inet_hash_connect(struct inet_timewait_death_row *death_row,
 				spin_unlock(&head->lock);
 				break;
 			}
-			tb->fastreuse = -1;
-			tb->fastreuseport = -1;
 			goto ok;

 		next_port:
-- 
1.8.3.1

--
To unsubscribe from this list: send the line "unsubscribe netdev" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html