[PATCH 2.4 NETLINK]: Unhash sockets correctly

[PATCH 2.4 NETLINK]: Unhash sockets correctly

[Patrick McHardy]

[-- Attachment #1: Type: text/plain, Size: 286 bytes --]

Hi Dave,

this patch fixes a nasty bug introduced by the pid-hash backport.
Sockets are only unhashed if they are contained in the first hash
bucket. This leads to leaking sockets and, over time, to bind
conflicts which confuse iproute and result in sporadic failure.

Regards
Patrick


[-- Attachment #2: x --]
[-- Type: text/plain, Size: 1347 bytes --]

# This is a BitKeeper generated diff -Nru style patch.
#
# ChangeSet
#   2005/02/16 19:32:38+01:00 kaber@coreworks.de 
#   [NETLINK]: Unhash sockets correctly
#   
#   netlink_remove() only unhashes sockets contained in the
#   first hash bucket. This leads to leaking sockets and,
#   over time, to bind conflicts which confuse iproute.
#   
#   Signed-off-by: Patrick McHardy <kaber@trash.net>
# 
# net/netlink/af_netlink.c
#   2005/02/16 19:32:36+01:00 kaber@coreworks.de +2 -1
#   [NETLINK]: Unhash sockets correctly
#   
#   netlink_remove() only unhashes sockets contained in the
#   first hash bucket. This leads to leaking sockets and,
#   over time, to bind conflicts which confuse iproute.
#   
#   Signed-off-by: Patrick McHardy <kaber@trash.net>
# 
diff -Nru a/net/netlink/af_netlink.c b/net/netlink/af_netlink.c
--- a/net/netlink/af_netlink.c	2005-02-16 19:32:51 +01:00
+++ b/net/netlink/af_netlink.c	2005-02-16 19:32:51 +01:00
@@ -327,10 +327,11 @@
 	struct sock **skp;
 	struct netlink_table *table = &nl_table[sk->protocol];
 	struct nl_pid_hash *hash = &table->hash;
+	u32 pid = nlk_sk(sk)->pid;
 
 	netlink_table_grab();
 	hash->entries--;
-	for (skp = hash->table; *skp; skp = &((*skp)->next)) {
+	for (skp = nl_pid_hashfn(hash, pid); *skp; skp = &((*skp)->next)) {
 		if (*skp == sk) {
 			*skp = sk->next;
 			__sock_put(sk);

Re: [PATCH 2.4 NETLINK]: Unhash sockets correctly

[David S. Miller]

On Wed, 16 Feb 2005 19:54:17 +0100
Patrick McHardy <kaber@trash.net> wrote:

> this patch fixes a nasty bug introduced by the pid-hash backport.
> Sockets are only unhashed if they are contained in the first hash
> bucket. This leads to leaking sockets and, over time, to bind
> conflicts which confuse iproute and result in sporadic failure.

Good catch Patrick.  Applied, and I'll push upstream momentarily.