Attachment 945419 Details for Bug 1147240 – work-around

[patch] work-around

dhcp-hop-limit.patch (text/plain), 1.43 KB, created by Jiri Popelka on 2014-10-09 17:23:49 UTC

(hide)

Description:

Filename:

MIME Type:

Creator: Jiri Popelka

Created: 2014-10-09 17:23:49 UTC

Size: 1.43 KB

patch

obsolete

>From f3d697a84e90f31a86874e334240ee1360a2fbf8 Mon Sep 17 00:00:00 2001
>From: Jiri Popelka <jpopelka@redhat.com>
>Date: Thu, 9 Oct 2014 18:57:22 +0200
>Subject: [PATCH] Work-around for hop-limit set to 1
>
>---
> common/socket.c | 16 +++++++++++++++-
> 1 file changed, 15 insertions(+), 1 deletion(-)
>
>diff --git a/common/socket.c b/common/socket.c
>index b8b44a6..93406ef 100644
>--- a/common/socket.c
>+++ b/common/socket.c
>@@ -301,7 +301,21 @@ if_register_socket(struct interface_info *info, int family,
> 	}
> 
> 	if ((family == AF_INET6) &&
>-	    ((info->flags & INTERFACE_UPSTREAM) != 0)) {
>+	    /*
>+	     * rfc3315 says that: "If relay agent relays messages to
>+	     * All_DHCP_Servers multicast address or other multicast addresses,
>+	     * it sets the Hop Limit field to 32."
>+	     *
>+	     * Because we use the same socket for all (upper/lower)
>+	     * interfaces, the hop limit is the same for both/all.
>+	     * There should be INTERFACE_UPSTREAM in the below condition, but
>+	     * problem is when the interface which registers the socket is
>+	     * lower interface, in that case the hop limit was not set
>+	     * (i.e. set to 1) for both.
>+	     * Because rfc doesn't say anything about hop limit for lower,
>+	     * it'd be lesser evil to have 32 in both cases.
>+	     */
>+	    ((info->flags & INTERFACE_STREAMS) != 0)) {
> 		int hop_limit = 32;
> 		if (setsockopt(sock, IPPROTO_IPV6, IPV6_MULTICAST_HOPS,
> 			       &hop_limit, sizeof(int)) < 0) {
>-- 
>2.1.0
>

Actions: View | Diff

Attachments on bug 1147240: 943140 | 945419