Skip to contentSkip to conversation

make TLLAO option for NA packets configurable

PATCH Applied Cosmin Ratiu 4 participants 13 messages Oct 1, 2009 – Oct 7, 2009

Subsystems (4)
lore
Share & tools
lore
#0 lore raw

PATCH 1/1 make TLLAO option for NA packets configurable

Hello,

This is a patch that adds a sysctl to control the sending of the Target Link Layer Address Option (TLLAO) with Neighbor Advertisements responding to unicast NS. The patch was made for kernel 2.6.7 (yes it is ancient), but the code is similar with the current kernel and I can rework it if you want it in.

RFC 2461, page 24 suggests that this option should be included with NAs to avoid a race with the sender clearing its cache after sending an unicast NS, but before receiving a NA.

It seems there are some Juniper routers (MX series) that expect this option to be included with all NAs.

Another solution is to always send this option, as it has little overhead.

Please let me know what you think,
Cosmin.

home/z/w1/packages/linux_2.6.7/main/src/include/linux/sysctl.h renamed from /packages/linux_2.6.7/main/src/include/linux/sysctl.h +1 -0
@@ -444,6 +444,7 @@                                                
        NET_IPV6_IP6FRAG_TIME=23,                                  
        NET_IPV6_IP6FRAG_SECRET_INTERVAL=24,                       
        NET_IPV6_MLD_MAX_MSF=25,                                   
+       NET_IPV6_NDISC_FORCE_TLLAO=26,                             
 };                                                                
                                                                   
 enum {                                                            
home/z/w1/packages/linux_2.6.7/main/src/include/net/ipv6.h renamed from /packages/linux_2.6.7/main/src/include/net/ipv6.h +1 -0
@@ -479,6 +479,7 @@
 extern int sysctl_ip6frag_low_thresh;
 extern int sysctl_ip6frag_time;
 extern int sysctl_ip6frag_secret_interval;
+extern int sysctl_ndisc_force_tllao;
 
 #endif /* __KERNEL__ */
 #endif /* _NET_IPV6_H */
home/z/w1/packages/linux_2.6.7/main/src/net/ipv6/ndisc.c renamed from /packages/linux_2.6.7/main/src/net/ipv6/ndisc.c +5 -0
@@ -169,6 +169,8 @@
 
 #define NDISC_OPT_SPACE(len) (((len)+2+7)&~7)
 
+int sysctl_ndisc_force_tllao;
+
 static u8 *ndisc_fill_option(u8 *opt, int type, void *data, int data_len)
 {
        int space = NDISC_OPT_SPACE(data_len);
@@ -399,6 +401,9 @@
                return;
        }
 
+       if (sysctl_ndisc_force_tllao)
+               inc_opt = 1;
+
        if (inc_opt) {
                if (dev->addr_len)
                        len += NDISC_OPT_SPACE(dev->addr_len);
home/z/w1/packages/linux_2.6.7/main/src/net/ipv6/sysctl_net_ipv6.c renamed from /packages/linux_2.6.7/main/src/net/ipv6/sysctl_net_ipv6.c +8 -0
@@ -84,6 +84,14 @@
                .mode           = 0644,
                .proc_handler   = &proc_dointvec
        },
+       {
+               .ctl_name       = NET_IPV6_NDISC_FORCE_TLLAO,
+               .procname       = "ndisc_force_tllao",
+               .data           = &sysctl_ndisc_force_tllao,
+               .maxlen         = sizeof(int),
+               .mode           = 0644,
+               .proc_handler   = &proc_dointvec,
+       },
        { .ctl_name = 0 }
 };
 
#1 lore raw
Quoted 23 lines · On Thu, 1 Oct 2009 19:16:40 +0300 Cosmin Ratiu <[email protected]> wrote:

Hello,

This is a patch that adds a sysctl to control the sending of the Target Link Layer Address Option (TLLAO) with Neighbor Advertisements responding to unicast NS. The patch was made for kernel 2.6.7 (yes it is ancient), but the code is similar with the current kernel and I can rework it if you want it in.

RFC 2461, page 24 suggests that this option should be included with NAs to avoid a race with the sender clearing its cache after sending an unicast NS, but before receiving a NA.

It seems there are some Juniper routers (MX series) that expect this option to be included with all NAs.

Another solution is to always send this option, as it has little overhead.

Please let me know what you think,
Cosmin.

home/z/w1/packages/linux_2.6.7/main/src/include/linux/sysctl.h line 447

Quoted 4 lines
home/z/w1/packages/linux_2.6.7/main/src/include/linux/sysctl.h renamed from /packages/linux_2.6.7/main/src/include/linux/sysctl.h +1 -0
@@ -444,6 +444,7 @@                                                
        NET_IPV6_IP6FRAG_TIME=23,                                  
        NET_IPV6_IP6FRAG_SECRET_INTERVAL=24,                       
        NET_IPV6_MLD_MAX_MSF=25,                                   
+       NET_IPV6_NDISC_FORCE_TLLAO=26,                             

Since numbered sysctl values are deprecated, can you use CTL_UNNUMBERED to avoid having to add yet another value?

maintainer
#2 lore raw

From: Stephen Hemminger <[email protected]>
Date: Thu, 1 Oct 2009 09:21:00 -0700

Quoted 11 lines

home/z/w1/packages/linux_2.6.7/main/src/include/linux/sysctl.h line 447

Quoted 8 lines
home/z/w1/packages/linux_2.6.7/main/src/include/linux/sysctl.h renamed from /packages/linux_2.6.7/main/src/include/linux/sysctl.h +1 -0
@@ -444,6 +444,7 @@                                                
        NET_IPV6_IP6FRAG_TIME=23,                                  
        NET_IPV6_IP6FRAG_SECRET_INTERVAL=24,                       
        NET_IPV6_MLD_MAX_MSF=25,                                   
+       NET_IPV6_NDISC_FORCE_TLLAO=26,                             

Since numbered sysctl values are deprecated, can you use CTL_UNNUMBERED to avoid having to add yet another value?

Using CLT_UNNUMBERED is a must these days.

Also, please fix the prefixing of the paths in your patch. See Documentation/SubmittingPatches in the kernel tree.

#3 lore raw
Quoted 4 lines · On Thursday 01 October 2009 19:43:56 David Miller wrote:

Using CLT_UNNUMBERED is a must these days.

Also, please fix the prefixing of the paths in your patch. See Documentation/SubmittingPatches in the kernel tree.

Here is the new variant. Please let me know what you think.

And I apologize for using [PATCH] instead of [RFC] in the subject, I don't know much about netdev protocol (yet).

Cosmin.

From 1911a98df800cedf4c3a63b897163e2935c5f602 Mon Sep 17 00:00:00 2001
From: Cosmin Ratiu <[email protected]>
Date: Thu, 1 Oct 2009 20:27:39 +0300
Subject: [PATCH] ipv6: new sysctl for sending TLLAO with NAs

Neighbor advertisements responding to unicast Neighbor Solicitations did not include the TLLAO option. This patch makes this configurable via /proc/sys/net/ipv6/ndisc_force_tllao, which by default is off.

The need for this arose because certain routers expect the TLLAO in some situations even as a response to unicast NS packets.

Moreover, RFC 2461 recommends on page 24 sending this to avoid a race.

---
 include/net/netns/ipv6.h   |    1 +
 net/ipv6/ndisc.c           |    1 +
 net/ipv6/sysctl_net_ipv6.c |    8 ++++++++
 3 files changed, 10 insertions(+), 0 deletions(-)
include/net/netns/ipv6.h +1 -0
index dfeb2d7..dd0a95b 100644
@@ -16,6 +16,7 @@ struct netns_sysctl_ipv6 {
 	struct ctl_table_header *frags_hdr;
 #endif
 	int bindv6only;
+	int ndisc_force_tllao;
 	int flush_delay;
 	int ip6_rt_max_size;
 	int ip6_rt_gc_min_interval;
net/ipv6/ndisc.c +1 -0
index f74e4e2..f08cf65 100644
@@ -598,6 +598,7 @@ static void ndisc_send_na(struct net_device *dev, struct neighbour *neigh,
 	icmp6h.icmp6_solicited = solicited;
 	icmp6h.icmp6_override = override;
 
+	inc_opt |= dev_net(dev)->ipv6.sysctl.ndisc_force_tllao;
 	__ndisc_send(dev, neigh, daddr, src_addr,
 		     &icmp6h, solicited_addr,
 		     inc_opt ? ND_OPT_TARGET_LL_ADDR : 0);
net/ipv6/sysctl_net_ipv6.c +8 -0
index 0dc6a4e..fb423ce 100644
@@ -37,6 +37,14 @@ static ctl_table ipv6_table_template[] = {
 		.mode		= 0644,
 		.proc_handler	= proc_dointvec
 	},
+	{
+		.ctl_name	= CTL_UNNUMBERED,
+		.procname	= "ndisc_force_tllao",
+		.data		= &init_net.ipv6.sysctl.ndisc_force_tllao,
+		.maxlen		= sizeof(int),
+		.mode		= 0644,
+		.proc_handler	= proc_dointvec
+	},
 	{ .ctl_name = 0 }
 };
 
Signature · 1 line
1.6.3.3
#4 lore raw
Quoted 6 lines · On Thu, 1 Oct 2009 21:08:40 +0300 Cosmin Ratiu <[email protected]> wrote:
Quoted 4 lines · On Thursday 01 October 2009 19:43:56 David Miller wrote:

Using CLT_UNNUMBERED is a must these days.

Also, please fix the prefixing of the paths in your patch. See Documentation/SubmittingPatches in the kernel tree.

Quoted 6 lines

Here is the new variant. Please let me know what you think.

And I apologize for using [PATCH] instead of [RFC] in the subject, I don't know much about netdev protocol (yet).

Cosmin.

Probably this should be a per interface property rather than per namespace.

#5 lore raw
Quoted 2 lines · On Thursday 01 October 2009 21:14:50 you wrote:

Probably this should be a per interface property rather than per namespace.

In our case, where we have lots of interfaces active, it would be nice to have the per namespace property as well.

But, as Cosmin suggested, perhaps it would be better to just send this options by default? (its a RFC SHOULD after all...)

Thanks,
tavi

#6 lore raw
Quoted 6 lines · On Thu, 1 Oct 2009 21:39:32 +0300 Octavian Purdila <[email protected]> wrote:
Quoted 2 lines · On Thursday 01 October 2009 21:14:50 you wrote:

Probably this should be a per interface property rather than per namespace.

In our case, where we have lots of interfaces active, it would be nice to have the per namespace property as well.

The ipv6 control infrastructure already has that option. If you changed your patch to use a per-interface control then there would be:

/proc/sys/net/ipv6/conf/all/force_tllao

maintainer
#7 lore raw

From: Stephen Hemminger <[email protected]>
Date: Thu, 1 Oct 2009 11:56:11 -0700

Quoted 7 lines
Quoted 4 lines · On Thu, 1 Oct 2009 21:39:32 +0300 Octavian Purdila <[email protected]> wrote:
Quoted 2 lines · On Thursday 01 October 2009 21:14:50 you wrote:

Probably this should be a per interface property rather than per namespace.

Quoted 7 lines
Quoted 2 lines

In our case, where we have lots of interfaces active, it would be nice to have the per namespace property as well.

The ipv6 control infrastructure already has that option. If you changed your patch to use a per-interface control then there would be:

/proc/sys/net/ipv6/conf/all/force_tllao

Right, this would work a lot better.

#8 lore raw
Quoted 13 lines · On Thursday 01 October 2009 22:37:40 you wrote:

From: Stephen Hemminger <[email protected]>
Date: Thu, 1 Oct 2009 11:56:11 -0700

Quoted 10 lines

On Thu, 1 Oct 2009 21:39:32 +0300

Quoted 6 lines · Octavian Purdila
Quoted 2 lines · On Thursday 01 October 2009 21:14:50 you wrote:

Probably this should be a per interface property rather than per namespace.

In our case, where we have lots of interfaces active, it would be nice to have the per namespace property as well.

Quoted 7 lines
Quoted 4 lines

The ipv6 control infrastructure already has that option. If you changed your patch to use a per-interface control then there would be:

/proc/sys/net/ipv6/conf/all/force_tllao

Right, this would work a lot better.

Here is v3 which also updates Documentation/networking/ip-sysctl.txt.

Thanks,
tavi

From fa7fcaca49da1a34ada7b4d3e0e2046148ffa337 Mon Sep 17 00:00:00 2001
From: Octavian Purdila <[email protected]>
Date: Fri, 2 Oct 2009 00:51:15 +0300
Subject: [PATCH] ipv6: new sysctl for sending TLLAO with unicast NAs

Neighbor advertisements responding to unicast neighbor solicitations did not include the target link-layer address option. This patch adds a new sysctl option (disabled by default) which controls whether this option should be sent even with unicast NAs.

The need for this arose because certain routers expect the TLLAO in some situations even as a response to unicast NS packets.

Moreover, RFC 2461 recommends sending this to avoid a race condition (section 4.4, Target link-layer address)

---
 Documentation/networking/ip-sysctl.txt |   18 ++++++++++++++++++
 include/linux/ipv6.h                   |    1 +
 net/ipv6/addrconf.c                    |    8 ++++++++
 net/ipv6/ndisc.c                       |    1 +
 4 files changed, 28 insertions(+), 0 deletions(-)
Documentation/networking/ip-sysctl.txt +18 -0
index fbe427a..301e19e 100644
@@ -1086,6 +1086,24 @@ accept_dad - INTEGER
 	2: Enable DAD, and disable IPv6 operation if MAC-based duplicate
 	   link-local address has been found.
 
+ndisc_force_tllao - BOOLEAN
+	Enable sending the target link-layer address option even when
+	responding to a unicast neighbor solicitation.
+	Default: FALSE
+
+	Quoting from RFC 2461, section 4.4, Target link-layer address:
+
+	"The option MUST be included for multicast solicitations in order to
+	avoid infinite Neighbor Solicitation "recursion" when the peer node
+	does not have a cache entry to return a Neighbor Advertisements
+	message.  When responding to unicast solicitations, the option can be
+	omitted since the sender of the solicitation has the correct link-
+	layer address; otherwise it would not have be able to send the unicast
+	solicitation in the first place. However, including the link-layer
+	address in this case adds little overhead and eliminates a potential
+	race condition where the sender deletes the cached link-layer address
+	prior to receiving a response to a previous solicitation."
+
 icmp/*:
 ratelimit - INTEGER
 	Limit the maximal rates for sending ICMPv6 packets.
include/linux/ipv6.h +1 -0
index c662efa..3d7b2ca 100644
@@ -167,6 +167,7 @@ struct ipv6_devconf {
 #endif
 	__s32		disable_ipv6;
 	__s32		accept_dad;
+	__s32		ndisc_force_tllao;
 	void		*sysctl;
 };
 
net/ipv6/addrconf.c +8 -0
index 1fd0a3d..ffa4747 100644
@@ -4353,6 +4353,14 @@ static struct addrconf_sysctl_table
 			.proc_handler	=	proc_dointvec,
 		},
 		{
+			.ctl_name       = CTL_UNNUMBERED,
+			.procname       = "ndisc_force_tllao",
+			.data           = &ipv6_devconf.ndisc_force_tllao,
+			.maxlen         = sizeof(int),
+			.mode           = 0644,
+			.proc_handler   = proc_dointvec
+		},
+		{
 			.ctl_name	=	0,	/* sentinel */
 		}
 	},
net/ipv6/ndisc.c +1 -0
index f74e4e2..27af229 100644
@@ -598,6 +598,7 @@ static void ndisc_send_na(struct net_device *dev, struct neighbour *neigh,
 	icmp6h.icmp6_solicited = solicited;
 	icmp6h.icmp6_override = override;
 
+	inc_opt |= ifp->idev->cnf.ndisc_force_tllao;
 	__ndisc_send(dev, neigh, daddr, src_addr,
 		     &icmp6h, solicited_addr,
 		     inc_opt ? ND_OPT_TARGET_LL_ADDR : 0);
Signature · 1 line
1.5.6.5
#9 lore raw
Quoted 19 lines · On Fri, 2 Oct 2009 01:19:47 +0300 Octavian Purdila <[email protected]> wrote:
Quoted 18 lines · On Thursday 01 October 2009 22:37:40 you wrote:

From: Stephen Hemminger <[email protected]>
Date: Thu, 1 Oct 2009 11:56:11 -0700

Quoted 14 lines

On Thu, 1 Oct 2009 21:39:32 +0300

Quoted 6 lines · Octavian Purdila
Quoted 2 lines · On Thursday 01 October 2009 21:14:50 you wrote:

Probably this should be a per interface property rather than per namespace.

In our case, where we have lots of interfaces active, it would be nice to have the per namespace property as well.

The ipv6 control infrastructure already has that option. If you changed your patch to use a per-interface control then there would be:

/proc/sys/net/ipv6/conf/all/force_tllao

Quoted 9 lines
Quoted 2 lines

Right, this would work a lot better.

Here is v3 which also updates Documentation/networking/ip-sysctl.txt.

Thanks,
tavi

This is good although I would have shortened the name.

#10 lore raw
Quoted 1 line · On Friday 02 October 2009 20:53:51 you wrote:

This is good although I would have shortened the name.

Ah, I knew I forgot something :) Here is v4.

tavi

From 24d96d825b9fa832b22878cc6c990d5711968734 Mon Sep 17 00:00:00 2001
From: Octavian Purdila <[email protected]>
Date: Fri, 2 Oct 2009 00:51:15 +0300
Subject: [PATCH] ipv6: new sysctl for sending TLLAO with unicast NAs

Neighbor advertisements responding to unicast neighbor solicitations did not include the target link-layer address option. This patch adds a new sysctl option (disabled by default) which controls whether this option should be sent even with unicast NAs.

The need for this arose because certain routers expect the TLLAO in some situations even as a response to unicast NS packets.

Moreover, RFC 2461 recommends sending this to avoid a race condition (section 4.4, Target link-layer address)

---
 Documentation/networking/ip-sysctl.txt |   18 ++++++++++++++++++
 include/linux/ipv6.h                   |    1 +
 net/ipv6/addrconf.c                    |    8 ++++++++
 net/ipv6/ndisc.c                       |    1 +
 4 files changed, 28 insertions(+), 0 deletions(-)
Documentation/networking/ip-sysctl.txt +18 -0
index fbe427a..a0e134d 100644
@@ -1086,6 +1086,24 @@ accept_dad - INTEGER
 	2: Enable DAD, and disable IPv6 operation if MAC-based duplicate
 	   link-local address has been found.
 
+force_tllao - BOOLEAN
+	Enable sending the target link-layer address option even when
+	responding to a unicast neighbor solicitation.
+	Default: FALSE
+
+	Quoting from RFC 2461, section 4.4, Target link-layer address:
+
+	"The option MUST be included for multicast solicitations in order to
+	avoid infinite Neighbor Solicitation "recursion" when the peer node
+	does not have a cache entry to return a Neighbor Advertisements
+	message.  When responding to unicast solicitations, the option can be
+	omitted since the sender of the solicitation has the correct link-
+	layer address; otherwise it would not have be able to send the unicast
+	solicitation in the first place. However, including the link-layer
+	address in this case adds little overhead and eliminates a potential
+	race condition where the sender deletes the cached link-layer address
+	prior to receiving a response to a previous solicitation."
+
 icmp/*:
 ratelimit - INTEGER
 	Limit the maximal rates for sending ICMPv6 packets.
include/linux/ipv6.h +1 -0
index c662efa..ae74ede 100644
@@ -167,6 +167,7 @@ struct ipv6_devconf {
 #endif
 	__s32		disable_ipv6;
 	__s32		accept_dad;
+	__s32		force_tllao;
 	void		*sysctl;
 };
 
net/ipv6/addrconf.c +8 -0
index 1fd0a3d..bdcee69 100644
@@ -4353,6 +4353,14 @@ static struct addrconf_sysctl_table
 			.proc_handler	=	proc_dointvec,
 		},
 		{
+			.ctl_name       = CTL_UNNUMBERED,
+			.procname       = "force_tllao",
+			.data           = &ipv6_devconf.force_tllao,
+			.maxlen         = sizeof(int),
+			.mode           = 0644,
+			.proc_handler   = proc_dointvec
+		},
+		{
 			.ctl_name	=	0,	/* sentinel */
 		}
 	},
net/ipv6/ndisc.c +1 -0
index f74e4e2..3507cfe 100644
@@ -598,6 +598,7 @@ static void ndisc_send_na(struct net_device *dev, struct neighbour *neigh,
 	icmp6h.icmp6_solicited = solicited;
 	icmp6h.icmp6_override = override;
 
+	inc_opt |= ifp->idev->cnf.force_tllao;
 	__ndisc_send(dev, neigh, daddr, src_addr,
 		     &icmp6h, solicited_addr,
 		     inc_opt ? ND_OPT_TARGET_LL_ADDR : 0);
Signature · 1 line
1.5.6.5
maintainer Applied
#11 lore raw

From: Octavian Purdila <[email protected]>
Date: Sat, 3 Oct 2009 00:39:15 +0300

Quoted 7 lines

Subject: [PATCH] ipv6: new sysctl for sending TLLAO with unicast NAs

Neighbor advertisements responding to unicast neighbor solicitations did not include the target link-layer address option. This patch adds a new sysctl option (disabled by default) which controls whether this option should be sent even with unicast NAs.

Quoted 8 lines

The need for this arose because certain routers expect the TLLAO in some situations even as a response to unicast NS packets.

Moreover, RFC 2461 recommends sending this to avoid a race condition (section 4.4, Target link-layer address)

Applied, thanks!

#12 lore raw
In reply to Stephen Hemminger
Quoted 24 lines · On Thursday 01 October 2009 19:21:00 Stephen Hemminger wrote:

On Thu, 1 Oct 2009 19:16:40 +0300

Quoted 21 lines · Cosmin Ratiu

Hello,

This is a patch that adds a sysctl to control the sending of the Target Link Layer Address Option (TLLAO) with Neighbor Advertisements responding to unicast NS. The patch was made for kernel 2.6.7 (yes it is ancient), but the code is similar with the current kernel and I can rework it if you want it in.

RFC 2461, page 24 suggests that this option should be included with NAs to avoid a race with the sender clearing its cache after sending an unicast NS, but before receiving a NA.

It seems there are some Juniper routers (MX series) that expect this option to be included with all NAs.

Another solution is to always send this option, as it has little overhead.

Please let me know what you think,
Cosmin.

Quoted 11 lines

home/z/w1/packages/linux_2.6.7/main/src/include/linux/sysctl.h line 447

Quoted 8 lines
home/z/w1/packages/linux_2.6.7/main/src/include/linux/sysctl.h renamed from /packages/linux_2.6.7/main/src/include/linux/sysctl.h +1 -0
@@ -444,6 +444,7 @@
        NET_IPV6_IP6FRAG_TIME=23,
        NET_IPV6_IP6FRAG_SECRET_INTERVAL=24,
        NET_IPV6_MLD_MAX_MSF=25,
+       NET_IPV6_NDISC_FORCE_TLLAO=26,

Since numbered sysctl values are deprecated, can you use CTL_UNNUMBERED to avoid having to add yet another value?

Of course, but that is a detail.
If you decide on the sysctl solution, I'll do it.
If you decide on making this the default behavior, it's even better.

Cosmin.