Skip to contentSkip to conversation

interface-local multicast escapes the local node

IPv6 In review Erik Hugne 4 participants 28 messages Feb 6, 2013 – Feb 11, 2013

Subsystems (5)
lore
Share & tools
lore

Erik Hugne

author
#0 lore raw

It seems that packets sent to interface-local scoped multicast addresses (ff01::/16) escape out on the network. According to RFC4291 (section 2.7):
Interface-Local scope spans only a single interface on a node and is useful only for loopback transmission of multicast.

Example program here:
https://gist.github.com/Hugne/4721151

//E

#1 lore raw
Quoted 8 lines · Erik Hugne

It seems that packets sent to interface-local scoped multicast addresses (ff01::/16) escape out on the network. According to RFC4291 (section 2.7):
Interface-Local scope spans only a single interface on a node and is useful only for loopback transmission of multicast.

Example program here:
https://gist.github.com/Hugne/4721151

Fixing the output path should be relatively straightforward, please test the following patch.

Looking at the input path, I think there is also no input protection for ff01::/16 addresses from the wire if you bind such an address.

[PATCH net-next] ipv6: don't let node/interface scoped multicast traffic escape on the wire

---
 net/ipv6/ip6_output.c | 6 ++++++
 1 file changed, 6 insertions(+)
net/ipv6/ip6_output.c +6 -0
index 906b7e6..b21ff3d 100644
@@ -118,6 +118,12 @@ static int ip6_finish_output2(struct sk_buff *skb)
 			}
 		}
 
+		if (IPV6_ADDR_MC_SCOPE(&ipv6_hdr(skb)->daddr) <=
+		    IPV6_ADDR_SCOPE_NODELOCAL) {
+			kfree_skb(skb);
+			return 0;
+		}
+
 		IP6_UPD_PO_STATS(dev_net(dev), idev, IPSTATS_MIB_OUTMCAST,
 				skb->len);
 	}
Signature · 1 line
1.8.1
author
#2 lore raw
Quoted 2 lines · Hannes Frederic Sowa

Fixing the output path should be relatively straightforward, please test the following patch.

Tested OK.

Quoted 2 lines

Looking at the input path, I think there is also no input protection for ff01::/16 addresses from the wire if you bind such an address.

Yes, this needs to be filtered on the input side aswell.

//E

#3 lore raw
Quoted 10 lines · Erik Hugne
Quoted 2 lines · Hannes Frederic Sowa

Fixing the output path should be relatively straightforward, please test the following patch.

Tested OK.

Quoted 2 lines

Looking at the input path, I think there is also no input protection for ff01::/16 addresses from the wire if you bind such an address.

Yes, this needs to be filtered on the input side aswell.

This patch should do the trick. Perhaps you could also take it for a test drive? Thanks!

[PATCH net-next] ipv6: don't accept node local multicast traffic from the wire

Cc Erik Hugne <[email protected]>
Cc YOSHIFUJI Hideaki <[email protected]>
---
 net/ipv6/ip6_input.c | 10 ++++++++++
 1 file changed, 10 insertions(+)
net/ipv6/ip6_input.c +10 -0
index 4ac5bf3..a2f71d2 100644
@@ -126,6 +126,16 @@ int ipv6_rcv(struct sk_buff *skb, struct net_device *dev, struct packet_type *pt
 	if (ipv6_addr_is_multicast(&hdr->saddr))
 		goto err;
 
+	/*
+	 * RFC4291 2.7
+	 * Interface-Local scope spans only a single interface on a node
+         * and is useful only for loopback transmission of multicast.
+	 */
+	if (ipv6_addr_is_multicast(&hdr->daddr) &&
+	    IPV6_ADDR_MC_SCOPE(&hdr->daddr) <= IPV6_ADDR_SCOPE_NODELOCAL &&
+	    skb->pkt_type != PACKET_LOOPBACK)
+		goto err;
+
 	skb->transport_header = skb->network_header + sizeof(*hdr);
 	IP6CB(skb)->nhoff = offsetof(struct ipv6hdr, nexthdr);
 
Signature · 1 line
1.8.1
author
#4 lore raw
Quoted 2 lines

This patch should do the trick. Perhaps you could also take it for a test drive? Thanks!

Verified, ff01::/16 from external hosts is blocked.

#5 lore raw
Quoted 4 lines · Erik Hugne
Quoted 2 lines

This patch should do the trick. Perhaps you could also take it for a test drive? Thanks!

Verified, ff01::/16 from external hosts is blocked.

Thanks for testing. Will do proper patch submissions later with whitespace issue fixed.

#6 lore raw
In reply to Hannes Frederic Sowa
Quoted 34 lines · Hannes Frederic Sowa
Quoted 10 lines · Erik Hugne
Quoted 2 lines · Hannes Frederic Sowa

Fixing the output path should be relatively straightforward, please test the following patch.

Tested OK.

Quoted 2 lines

Looking at the input path, I think there is also no input protection for ff01::/16 addresses from the wire if you bind such an address.

Yes, this needs to be filtered on the input side aswell.

This patch should do the trick. Perhaps you could also take it for a test drive? Thanks!

[PATCH net-next] ipv6: don't accept node local multicast traffic from the wire

Cc Erik Hugne <[email protected]>
Cc YOSHIFUJI Hideaki <[email protected]>
---
 net/ipv6/ip6_input.c | 10 ++++++++++
 1 file changed, 10 insertions(+)
net/ipv6/ip6_input.c +2 -0
index 4ac5bf3..a2f71d2 100644
@@ -126,6 +126,16 @@ int ipv6_rcv(struct sk_buff *skb, struct net_device *dev, struct packet_type *pt
 	if (ipv6_addr_is_multicast(&hdr->saddr))
 		goto err;
 
+	/*
+	 * RFC4291 2.7

net/ipv6/ip6_input.c int ipv6_rcv(struct sk_buff *skb, struct net_device *dev, struct packet_type *pt · line 140

Quoted 10 lines
net/ipv6/ip6_input.c +8 -0
+	 * Interface-Local scope spans only a single interface on a node
+         * and is useful only for loopback transmission of multicast.
+	 */
+	if (ipv6_addr_is_multicast(&hdr->daddr) &&
+	    IPV6_ADDR_MC_SCOPE(&hdr->daddr) <= IPV6_ADDR_SCOPE_NODELOCAL &&
+	    skb->pkt_type != PACKET_LOOPBACK)
+		goto err;
+
 	skb->transport_header = skb->network_header + sizeof(*hdr);
 	IP6CB(skb)->nhoff = offsetof(struct ipv6hdr, nexthdr);

NAK.

Well, do you have relevant RFC?
RFC4291 says that we should drop ff00::/16, but not ff01::/16.

--yoshfuji

#7 lore raw
Quoted 48 lines · YOSHIFUJI Hideaki
Quoted 34 lines · Hannes Frederic Sowa
Quoted 10 lines · Erik Hugne
Quoted 2 lines · Hannes Frederic Sowa

Fixing the output path should be relatively straightforward, please test the following patch.

Tested OK.

Quoted 2 lines

Looking at the input path, I think there is also no input protection for ff01::/16 addresses from the wire if you bind such an address.

Yes, this needs to be filtered on the input side aswell.

This patch should do the trick. Perhaps you could also take it for a test drive? Thanks!

[PATCH net-next] ipv6: don't accept node local multicast traffic from the wire

Cc Erik Hugne <[email protected]>
Cc YOSHIFUJI Hideaki <[email protected]>
---
 net/ipv6/ip6_input.c | 10 ++++++++++
 1 file changed, 10 insertions(+)
net/ipv6/ip6_input.c +2 -0
index 4ac5bf3..a2f71d2 100644
@@ -126,6 +126,16 @@ int ipv6_rcv(struct sk_buff *skb, struct net_device *dev, struct packet_type *pt
 	if (ipv6_addr_is_multicast(&hdr->saddr))
 		goto err;
 
+	/*
+	 * RFC4291 2.7

net/ipv6/ip6_input.c int ipv6_rcv(struct sk_buff *skb, struct net_device *dev, struct packet_type *pt · line 140

Quoted 10 lines
net/ipv6/ip6_input.c +8 -0
+	 * Interface-Local scope spans only a single interface on a node
+         * and is useful only for loopback transmission of multicast.
+	 */
+	if (ipv6_addr_is_multicast(&hdr->daddr) &&
+	    IPV6_ADDR_MC_SCOPE(&hdr->daddr) <= IPV6_ADDR_SCOPE_NODELOCAL &&
+	    skb->pkt_type != PACKET_LOOPBACK)
+		goto err;
+
 	skb->transport_header = skb->network_header + sizeof(*hdr);
 	IP6CB(skb)->nhoff = offsetof(struct ipv6hdr, nexthdr);

NAK.

Quoted 2 lines

Well, do you have relevant RFC?
RFC4291 says that we should drop ff00::/16, but not ff01::/16.

I know what you mean, the RFC does not state it directly. Hm, the BSDs seem to drop such destination addresses, too, if they don't originate from a loopback interface. Or did you mean that there is a flaw in the skb->pkt_type != PACKET_LOOPBACK condition?

#8 lore raw
Quoted 9 lines · Hannes Frederic Sowa
Quoted 2 lines · YOSHIFUJI Hideaki

Well, do you have relevant RFC?
RFC4291 says that we should drop ff00::/16, but not ff01::/16.

I know what you mean, the RFC does not state it directly. Hm, the BSDs seem to drop such destination addresses, too, if they don't originate from a loopback interface. Or did you mean that there is a flaw in the skb->pkt_type != PACKET_LOOPBACK condition?

RFC 4291 states on page 13 :

Interface-Local scope spans only a single interface on a node and is useful only for loopback transmission of multicast.

ff01::/16 are Interface-Local, or am I missing something ?

#9 lore raw
Quoted 11 lines · Eric Dumazet
Quoted 9 lines · Hannes Frederic Sowa
Quoted 2 lines · YOSHIFUJI Hideaki

Well, do you have relevant RFC?
RFC4291 says that we should drop ff00::/16, but not ff01::/16.

I know what you mean, the RFC does not state it directly. Hm, the BSDs seem to drop such destination addresses, too, if they don't originate from a loopback interface. Or did you mean that there is a flaw in the skb->pkt_type != PACKET_LOOPBACK condition?

Quoted 6 lines

RFC 4291 states on page 13 :

Interface-Local scope spans only a single interface on a node and is useful only for loopback transmission of multicast.

ff01::/16 are Interface-Local, or am I missing something ?

No, that's correct.

#10 lore raw
In reply to Hannes Frederic Sowa
Quoted 49 lines · Hannes Frederic Sowa
Quoted 48 lines · YOSHIFUJI Hideaki
Quoted 34 lines · Hannes Frederic Sowa
Quoted 10 lines · Erik Hugne
Quoted 2 lines · Hannes Frederic Sowa

Fixing the output path should be relatively straightforward, please test the following patch.

Tested OK.

Quoted 2 lines

Looking at the input path, I think there is also no input protection for ff01::/16 addresses from the wire if you bind such an address.

Yes, this needs to be filtered on the input side aswell.

This patch should do the trick. Perhaps you could also take it for a test drive? Thanks!

[PATCH net-next] ipv6: don't accept node local multicast traffic from the wire

Cc Erik Hugne <[email protected]>
Cc YOSHIFUJI Hideaki <[email protected]>
---
 net/ipv6/ip6_input.c | 10 ++++++++++
 1 file changed, 10 insertions(+)
net/ipv6/ip6_input.c +2 -0
index 4ac5bf3..a2f71d2 100644
@@ -126,6 +126,16 @@ int ipv6_rcv(struct sk_buff *skb, struct net_device *dev, struct packet_type *pt
 	if (ipv6_addr_is_multicast(&hdr->saddr))
 		goto err;
 
+	/*
+	 * RFC4291 2.7

net/ipv6/ip6_input.c int ipv6_rcv(struct sk_buff *skb, struct net_device *dev, struct packet_type *pt · line 140

Quoted 10 lines
net/ipv6/ip6_input.c +8 -0
+	 * Interface-Local scope spans only a single interface on a node
+         * and is useful only for loopback transmission of multicast.
+	 */
+	if (ipv6_addr_is_multicast(&hdr->daddr) &&
+	    IPV6_ADDR_MC_SCOPE(&hdr->daddr) <= IPV6_ADDR_SCOPE_NODELOCAL &&
+	    skb->pkt_type != PACKET_LOOPBACK)
+		goto err;
+
 	skb->transport_header = skb->network_header + sizeof(*hdr);
 	IP6CB(skb)->nhoff = offsetof(struct ipv6hdr, nexthdr);

NAK.

Quoted 7 lines
Quoted 2 lines

Well, do you have relevant RFC?
RFC4291 says that we should drop ff00::/16, but not ff01::/16.

I know what you mean, the RFC does not state it directly. Hm, the BSDs seem to drop such destination addresses, too, if they don't originate from a loopback interface. Or did you mean that there is a flaw in the skb->pkt_type != PACKET_LOOPBACK condition?

We do not drop ff01::/16, because RFC is silent about it.

--yoshfuji

#11 lore raw
Quoted 6 lines · YOSHIFUJI Hideaki
Quoted 6 lines
Quoted 5 lines

NAK.

Well, do you have relevant RFC?
RFC4291 says that we should drop ff00::/16, but not ff01::/16.

Quoted 6 lines
Quoted 4 lines

I know what you mean, the RFC does not state it directly. Hm, the BSDs seem to drop such destination addresses, too, if they don't originate from a loopback interface. Or did you mean that there is a flaw in the skb->pkt_type != PACKET_LOOPBACK condition?

We do not drop ff01::/16, because RFC is silent about it.

I just did a little bit of research on this topic and found this:

RFC4541 3. IPv6 Considerations:

MLD messages are also not sent regarding groups with addresses in the range FF00::/15 (which encompasses both the reserved FF00::/16 and node-local FF01::/16 IPv6 address spaces). These addresses should never appear in packets on the link.

It gives a strong indication that we should drop these packets. What do you think?

Thanks,

Hannes

author
#12 lore raw
Quoted 9 lines · Hannes Frederic Sowa

RFC4541 3. IPv6 Considerations:

MLD messages are also not sent regarding groups with addresses in the range FF00::/15 (which encompasses both the reserved FF00::/16 and node-local FF01::/16 IPv6 address spaces). These addresses should never appear in packets on the link.

It gives a strong indication that we should drop these packets. What do you think?

That was my understanding aswell. I asked for some clarifications on the ietf list:
<quote>
> The key is that the packets should be dropped without generating > any errors or traffic.
Yep, you are right - I assumed that implicitly, but
I should have mentioned that explicitly :-)
</quote>
http://www.ietf.org/mail-archive/web/ipv6/current/msg17126.html

//E

#13 lore raw
Quoted 11 lines · Erik Hugne
Quoted 9 lines · Hannes Frederic Sowa

RFC4541 3. IPv6 Considerations:

MLD messages are also not sent regarding groups with addresses in the range FF00::/15 (which encompasses both the reserved FF00::/16 and node-local FF01::/16 IPv6 address spaces). These addresses should never appear in packets on the link.

It gives a strong indication that we should drop these packets. What do you think?

Quoted 8 lines

That was my understanding aswell. I asked for some clarifications on the ietf list:
<quote>
> The key is that the packets should be dropped without generating > any errors or traffic.
Yep, you are right - I assumed that implicitly, but
I should have mentioned that explicitly :-)
</quote>
http://www.ietf.org/mail-archive/web/ipv6/current/msg17126.html

Please file errata on RFC4291. Thanks.

--yoshfuji

#14 lore raw
In reply to Hannes Frederic Sowa
Quoted 34 lines · Hannes Frederic Sowa
Quoted 8 lines · Erik Hugne

It seems that packets sent to interface-local scoped multicast addresses (ff01::/16) escape out on the network. According to RFC4291 (section 2.7):
Interface-Local scope spans only a single interface on a node and is useful only for loopback transmission of multicast.

Example program here:
https://gist.github.com/Hugne/4721151

Fixing the output path should be relatively straightforward, please test the following patch.

Looking at the input path, I think there is also no input protection for ff01::/16 addresses from the wire if you bind such an address.

[PATCH net-next] ipv6: don't let node/interface scoped multicast traffic escape on the wire

---
 net/ipv6/ip6_output.c | 6 ++++++
 1 file changed, 6 insertions(+)
net/ipv6/ip6_output.c +0 -0
index 906b7e6..b21ff3d 100644
@@ -118,6 +118,12 @@ static int ip6_finish_output2(struct sk_buff *skb)
 			}
 		}

net/ipv6/ip6_output.c static int ip6_finish_output2(struct sk_buff *skb) · line 129

Quoted 10 lines
net/ipv6/ip6_output.c +6 -0
 
+		if (IPV6_ADDR_MC_SCOPE(&ipv6_hdr(skb)->daddr) <=
+		    IPV6_ADDR_SCOPE_NODELOCAL) {
+			kfree_skb(skb);
+			return 0;
+		}
+
 		IP6_UPD_PO_STATS(dev_net(dev), idev, IPSTATS_MIB_OUTMCAST,
 				skb->len);
 	}

NAK. I think we should select routes via loopback device here.

--yoshfuji

#15 lore raw
Quoted 1 line · YOSHIFUJI Hideaki

NAK. I think we should select routes via loopback device here.

Will try your idea, thanks.

#16 lore raw
Quoted 4 lines · Hannes Frederic Sowa
Quoted 1 line · YOSHIFUJI Hideaki

NAK. I think we should select routes via loopback device here.

Will try your idea, thanks.

Does this patch look reasonable? Btw. i am pleased to see this kind of things work out as expected most of the time (addrtype checking etc. all in place). :)

I have a few questions:

a) Do we have a way to lock down routes in the kernel that they are untouchable from userspace?

b) We don't set IFF_MULTICAST in the loopback interface driver by default. Is there a reason we couldn't do so? This patch tries to also bind the interface local all router multicast address on lo, but will only do so if multicast is enabled on lo before changing the sysctl.

c) I will send a patch as soon as the errata thing is resolved to drop ff00::/15 from entering the host from the wire. In case of outgoing traffic to ff00::/16, should this be stopped with a net-unreachable route?

Btw, I saw that the errata on RFC 4291 is already submitted:
http://www.rfc-editor.org/errata_search.php?rfc=4291
Thanks, Erik!

[PATCH net-next RFC] ipv6: let loopback join ff01::{1,2} and create a loopback route to ff00::/16

This patch ensures that traffic to ff01::/16 is directed to the loopback interface by default and does not leave the host. The interface-local all nodes and all routers (in case forwarding is enabled) are now joined by default, too.

---
 include/linux/in6.h |  6 ++++++
 net/ipv6/addrconf.c | 37 +++++++++++++++++++++++++++++++++++--
 2 files changed, 41 insertions(+), 2 deletions(-)
include/linux/in6.h +6 -0
index a16e193..fd0e86d 100644
@@ -36,4 +36,10 @@ extern const struct in6_addr in6addr_linklocal_allnodes;
 extern const struct in6_addr in6addr_linklocal_allrouters;
 #define IN6ADDR_LINKLOCAL_ALLROUTERS_INIT \
 		{ { { 0xff,2,0,0,0,0,0,0,0,0,0,0,0,0,0,2 } } }
+extern const struct in6_addr in6addr_interfacelocal_allnodes;
+#define IN6ADR_INTERFACELOCAL_ALLNODES_INIT \
+		{ { { 0xff,1,0,0,0,0,0,0,0,0,0,0,0,0,0,1 } } }
+extern const struct in6_addr in6addr_interfacelocal_allrouters;
+#define IN6ADR_INTERFACELOCAL_ALLROUTERS_INIT \
+		{ { { 0xff,1,0,0,0,0,0,0,0,0,0,0,0,0,0,2 } } }
 #endif
net/ipv6/addrconf.c +35 -2
index bd9f936..c15b98f 100644
@@ -244,6 +244,8 @@ const struct in6_addr in6addr_any = IN6ADDR_ANY_INIT;
 const struct in6_addr in6addr_loopback = IN6ADDR_LOOPBACK_INIT;
 const struct in6_addr in6addr_linklocal_allnodes = IN6ADDR_LINKLOCAL_ALLNODES_INIT;
 const struct in6_addr in6addr_linklocal_allrouters = IN6ADDR_LINKLOCAL_ALLROUTERS_INIT;
+const struct in6_addr in6addr_interfacelocal_allnodes = IN6ADR_INTERFACELOCAL_ALLNODES_INIT;
+const struct in6_addr in6addr_interfacelocal_allrouters = IN6ADR_INTERFACELOCAL_ALLROUTERS_INIT;
 
 /* Check if a valid qdisc is available */
 static inline bool addrconf_qdisc_ok(const struct net_device *dev)
@@ -611,10 +613,17 @@ static void dev_forward_change(struct inet6_dev *idev)
 	if (idev->cnf.forwarding)
 		dev_disable_lro(dev);
 	if (dev->flags & IFF_MULTICAST) {
-		if (idev->cnf.forwarding)
+		if (idev->cnf.forwarding) {
 			ipv6_dev_mc_inc(dev, &in6addr_linklocal_allrouters);
-		else
+			if (dev->flags & IFF_LOOPBACK)
+				ipv6_dev_mc_inc(dev,
+						&in6addr_interfacelocal_allrouters);
+		} else {
 			ipv6_dev_mc_dec(dev, &in6addr_linklocal_allrouters);
+			if (dev->flags & IFF_LOOPBACK)
+				ipv6_dev_mc_dec(dev,
+						&in6addr_interfacelocal_allrouters);
+		}
 	}
 
 	list_for_each_entry(ifa, &idev->addr_list, if_list) {
@@ -2518,6 +2527,29 @@ static void sit_add_v4_addrs(struct inet6_dev *idev)
 }
 #endif
 
+static void init_loopback_mc(struct net_device *dev)
+{
+	struct fib6_config cfg = {
+		.fc_table = RT6_TABLE_LOCAL,
+		.fc_metric = IP6_RT_PRIO_ADDRCONF,
+		.fc_dst_len = 16,
+		.fc_ifindex = dev->ifindex,
+		.fc_flags = RTF_UP,
+		.fc_protocol = RTPROT_KERNEL,
+		.fc_type = RTN_MULTICAST,
+		.fc_nlinfo.nl_net = dev_net(dev),
+	};
+
+	ipv6_addr_set(&cfg.fc_dst, htonl(0xFF010000), 0, 0, 0);
+	if (ip6_route_add(&cfg))
+		pr_debug("%s: ip6_route_add failed for node local multicast\n",
+			 __func__);
+
+	if (ipv6_dev_mc_inc(dev, &in6addr_interfacelocal_allnodes))
+		pr_debug("%s: failed to join interface-local all nodes mc\n",
+			 __func__);
+}
+
 static void init_loopback(struct net_device *dev)
 {
 	struct inet6_dev  *idev;
@@ -2532,6 +2564,7 @@ static void init_loopback(struct net_device *dev)
 	}
 
 	add_addr(idev, &in6addr_loopback, 128, IFA_HOST);
+	init_loopback_mc(dev);
 }
 
 static void addrconf_add_linklocal(struct inet6_dev *idev, const struct in6_addr *addr)
Signature · 1 line
1.8.1.2
#17 lore raw

Hi,

Quoted 10 lines · Hannes Frederic Sowa
Quoted 4 lines · Hannes Frederic Sowa
Quoted 1 line · YOSHIFUJI Hideaki

NAK. I think we should select routes via loopback device here.

Will try your idea, thanks.

Does this patch look reasonable? Btw. i am pleased to see this kind of things work out as expected most of the time (addrtype checking etc. all in place). :)

Well, I rethink of what "interface-local" means.

It seems applications will join ff01::/16%eth0 instead of ff01::/16%lo. If so, your original patch seems better. My bad, sorry.

Would you update original one, with minor modification that defers kfree_skb() after incrementing MIB, please?

If you think we should join ff01::{1,2} by default, you can send another patch for it. (BTW, why don't we join ff05::2, then? ;-))

Thanks.

--yoshfuji

#18 lore raw
Quoted 14 lines · YOSHIFUJI Hideaki

Hi,

Quoted 10 lines · Hannes Frederic Sowa
Quoted 4 lines · Hannes Frederic Sowa
Quoted 1 line · YOSHIFUJI Hideaki

NAK. I think we should select routes via loopback device here.

Will try your idea, thanks.

Does this patch look reasonable? Btw. i am pleased to see this kind of things work out as expected most of the time (addrtype checking etc. all in place). :)

Quoted 4 lines

Well, I rethink of what "interface-local" means.

It seems applications will join ff01::/16%eth0 instead of ff01::/16%lo. If so, your original patch seems better. My bad, sorry.

No problem, will do. I have not checked carefully, but this would mean we have to do changes to glibc to have a correct behaving getaddrinfo?

Quoted 2 lines

Would you update original one, with minor modification that defers kfree_skb() after incrementing MIB, please?

Yes. Will send the patch tomorrow at the latest.

Quoted 3 lines

If you think we should join ff01::{1,2} by default, you can send another patch for it. (BTW, why don't we join ff05::2, then? ;-))

Ok, I'll split it up. You are right about ff05::2, seems like the right thing to do, somehow. I hope this won't impact any multicast routing daemons.

Thanks,

Hannes

#19 lore raw
In reply to YOSHIFUJI Hideaki
Quoted 5 lines · YOSHIFUJI Hideaki

It seems applications will join ff01::/16%eth0 instead of ff01::/16%lo. If so, your original patch seems better. My bad, sorry.

Would you update original one, with minor modification that defers kfree_skb() after incrementing MIB, please?

I would add another constraint to the if "&& !(dev->flags & IFF_LOOPBACK)", so it becomes:

                if (IPV6_ADDR_MC_SCOPE(&ipv6_hdr(skb)->daddr) <=
                    IPV6_ADDR_SCOPE_NODELOCAL &&
                    !(dev->flags & IFF_LOOPBACK))
                        kfree_skb(skb);
                        return 0;
                }

Otherwise ff01::/16%lo would not work because the multicast mirroring through dev_loopback_xmit won't be taken and the packet would be dropped after that.

Can you confirm? Thanks.

#20 lore raw
Quoted 18 lines · Hannes Frederic Sowa
Quoted 5 lines · YOSHIFUJI Hideaki

It seems applications will join ff01::/16%eth0 instead of ff01::/16%lo. If so, your original patch seems better. My bad, sorry.

Would you update original one, with minor modification that defers kfree_skb() after incrementing MIB, please?

I would add another constraint to the if "&& !(dev->flags & IFF_LOOPBACK)", so it becomes:

                if (IPV6_ADDR_MC_SCOPE(&ipv6_hdr(skb)->daddr) <=
                    IPV6_ADDR_SCOPE_NODELOCAL &&
                    !(dev->flags & IFF_LOOPBACK))
                        kfree_skb(skb);
                        return 0;
                }
Quoted 4 lines

Otherwise ff01::/16%lo would not work because the multicast mirroring through dev_loopback_xmit won't be taken and the packet would be dropped after that.

Can you confirm? Thanks.

Ack.

--yoshfuji

#21 lore raw
In reply to YOSHIFUJI Hideaki
Quoted 14 lines · YOSHIFUJI Hideaki

Hi,

Quoted 10 lines · Hannes Frederic Sowa
Quoted 4 lines · Hannes Frederic Sowa
Quoted 1 line · YOSHIFUJI Hideaki

NAK. I think we should select routes via loopback device here.

Will try your idea, thanks.

Does this patch look reasonable? Btw. i am pleased to see this kind of things work out as expected most of the time (addrtype checking etc. all in place). :)

Quoted 4 lines

Well, I rethink of what "interface-local" means.

It seems applications will join ff01::/16%eth0 instead of ff01::/16%lo. If so, your original patch seems better. My bad, sorry.

I was looking at getpeername et. al. where we should report the scope back to the user. A common pattern is:

                        if (ipv6_addr_type(&sin->sin6_addr) & IPV6_ADDR_LINKLOCAL)
                                sin->sin6_scope_id = IP6CB(skb)->iif;

I propose to introduce something like 'bool ipv6_addr_intf_scoped(in6_addr)' and let it check for ll addresses and interface scoped addresses.

#22 lore raw
Quoted 20 lines · Hannes Frederic Sowa
Quoted 14 lines · YOSHIFUJI Hideaki

Hi,

Quoted 10 lines · Hannes Frederic Sowa
Quoted 4 lines · Hannes Frederic Sowa
Quoted 1 line · YOSHIFUJI Hideaki

NAK. I think we should select routes via loopback device here.

Will try your idea, thanks.

Does this patch look reasonable? Btw. i am pleased to see this kind of things work out as expected most of the time (addrtype checking etc. all in place). :)

Quoted 4 lines

Well, I rethink of what "interface-local" means.

It seems applications will join ff01::/16%eth0 instead of ff01::/16%lo. If so, your original patch seems better. My bad, sorry.

Quoted 8 lines

I was looking at getpeername et. al. where we should report the scope back to the user. A common pattern is:

                        if (ipv6_addr_type(&sin->sin6_addr) & IPV6_ADDR_LINKLOCAL)
                                sin->sin6_scope_id = IP6CB(skb)->iif;

I propose to introduce something like 'bool ipv6_addr_intf_scoped(in6_addr)' and let it check for ll addresses and interface scoped addresses.

Hmm, maybe, we might want to say:

__u32 __ipv6_iface_scope_id(int type, unsigned int iface)

{
	if (type == IPV6_ADDR_ANY ||
	    type & IPV6_ADDR_LOOPBACK ||
	    __ipv6_addr_src_scope(type) > IPV6_ADDR_SCOPE_LINKLOCAL)
		return 0;
	return iface;
}

__u32 ipv6_iface_scope_id(const struct in6_addr *addr, unsigned int iface)

{
	return __ipv6_iface_scope_id(__ipv6_addr_type(addr), iface);
}

And then,

	sin->sin6_scope_id = __ipv6_iface_scope_id(__ipv6_addr_type(&sin->sin6_addr),
						   IP6CB(skb)->iif);
or
	sin->sin6_scope_id = ipv6_iface_scope_id(&sin->sin6_addr, IP6CB(skb)->iif);

--yoshfuji

#23 lore raw
Quoted 26 lines · YOSHIFUJI Hideaki
Quoted 8 lines · Hannes Frederic Sowa

I was looking at getpeername et. al. where we should report the scope back to the user. A common pattern is:

                        if (ipv6_addr_type(&sin->sin6_addr) & IPV6_ADDR_LINKLOCAL)
                                sin->sin6_scope_id = IP6CB(skb)->iif;

I propose to introduce something like 'bool ipv6_addr_intf_scoped(in6_addr)' and let it check for ll addresses and interface scoped addresses.

Hmm, maybe, we might want to say:

__u32 __ipv6_iface_scope_id(int type, unsigned int iface)

{
	if (type == IPV6_ADDR_ANY ||
	    type & IPV6_ADDR_LOOPBACK ||
	    __ipv6_addr_src_scope(type) > IPV6_ADDR_SCOPE_LINKLOCAL)
		return 0;
	return iface;
}

__u32 ipv6_iface_scope_id(const struct in6_addr *addr, unsigned int iface)

{
	return __ipv6_iface_scope_id(__ipv6_addr_type(addr), iface);
}
Quoted 5 lines

And then,

	sin->sin6_scope_id = __ipv6_iface_scope_id(__ipv6_addr_type(&sin->sin6_addr),
						   IP6CB(skb)->iif);
or
	sin->sin6_scope_id = ipv6_iface_scope_id(&sin->sin6_addr, IP6CB(skb)->iif);

I like it. Will try to convert some checks and let you know how it turned out.

#24 lore raw

RFC PATCH 1/2 ipv6: introdcue __ipv6_addr_needs_scope_id and ipv6_iface_scope_id helper functions

This strictly a RFC only (I also left out my signed-off). This patch introduces helper functions to set and check for the need of scope ids.

---
 include/net/ipv6.h | 13 +++++++++++++
 1 file changed, 13 insertions(+)
include/net/ipv6.h +13 -0
index 851d541..1f61dc9 100644
@@ -319,6 +319,19 @@ static inline int ipv6_addr_src_scope(const struct in6_addr *addr)
 	return __ipv6_addr_src_scope(__ipv6_addr_type(addr));
 }
 
+static inline bool __ipv6_addr_needs_scope_id(int type)
+{
+	return type != IPV6_ADDR_ANY &&
+	       !(type & IPV6_ADDR_LOOPBACK && type & IPV6_ADDR_UNICAST) &&
+	       __ipv6_addr_src_scope(type) <= IPV6_ADDR_SCOPE_LINKLOCAL;
+}
+
+static inline __u32 ipv6_iface_scope_id(const struct in6_addr *addr,
+					unsigned int iface)
+{
+	return __ipv6_addr_needs_scope_id(__ipv6_addr_type(addr)) ? iface : 0;
+}
+
 static inline int ipv6_addr_cmp(const struct in6_addr *a1, const struct in6_addr *a2)
 {
 	return memcmp(a1, a2, sizeof(struct in6_addr));
Signature · 1 line
1.8.1.2
#25 lore raw

Re: [PATCH RFC 1/2] ipv6: introdcue __ipv6_addr_needs_scope_id and ipv6_iface_scope_id helper functions

Quoted 24 lines · Hannes Frederic Sowa

This strictly a RFC only (I also left out my signed-off). This patch introduces helper functions to set and check for the need of scope ids.

---
 include/net/ipv6.h | 13 +++++++++++++
 1 file changed, 13 insertions(+)
include/net/ipv6.h +6 -0
index 851d541..1f61dc9 100644
@@ -319,6 +319,19 @@ static inline int ipv6_addr_src_scope(const struct in6_addr *addr)
 	return __ipv6_addr_src_scope(__ipv6_addr_type(addr));
 }
 
+static inline bool __ipv6_addr_needs_scope_id(int type)
+{
+	return type != IPV6_ADDR_ANY &&
+	       !(type & IPV6_ADDR_LOOPBACK && type & IPV6_ADDR_UNICAST) &&
+	       __ipv6_addr_src_scope(type) <= IPV6_ADDR_SCOPE_LINKLOCAL;
+}

include/net/ipv6.h static inline int ipv6_addr_src_scope(const struct in6_addr *addr) · line 337

Quoted 10 lines
include/net/ipv6.h +7 -0
+
+static inline __u32 ipv6_iface_scope_id(const struct in6_addr *addr,
+					unsigned int iface)
+{
+	return __ipv6_addr_needs_scope_id(__ipv6_addr_type(addr)) ? iface : 0;
+}
+
 static inline int ipv6_addr_cmp(const struct in6_addr *a1, const struct in6_addr *a2)
 {
 	return memcmp(a1, a2, sizeof(struct in6_addr));

Yes, something like that.

--yoshfuji

#26 lore raw

RFC PATCH 2/2 ipv6: use newly introduced helper functions __ipv6_addr_needs_scope_id and ipv6_iface_scope_id

In reply to Hannes Frederic Sowa

This patch is strictly an RFC (I also left out my signed-off). I intended post it just to show the usage of the new helper functions.

With this patch applied my test programs reported that interface scoped multicast finally works over loopback. :)

Cc YOSHIFUJI Hideaki <[email protected]>
---
 net/ipv6/af_inet6.c              |  6 +++---
 net/ipv6/datagram.c              | 12 ++++++------
 net/ipv6/icmp.c                  |  2 +-
 net/ipv6/inet6_connection_sock.c |  6 ++----
 net/ipv6/ip6_input.c             |  2 +-
 net/ipv6/udp.c                   |  6 +++---
 6 files changed, 16 insertions(+), 18 deletions(-)
net/ipv6/af_inet6.c +3 -3
index 6b793bf..f56277f 100644
@@ -323,7 +323,7 @@ int inet6_bind(struct socket *sock, struct sockaddr *uaddr, int addr_len)
 			struct net_device *dev = NULL;
 
 			rcu_read_lock();
-			if (addr_type & IPV6_ADDR_LINKLOCAL) {
+			if (__ipv6_addr_needs_scope_id(addr_type)) {
 				if (addr_len >= sizeof(struct sockaddr_in6) &&
 				    addr->sin6_scope_id) {
 					/* Override any existing binding, if another one
@@ -471,8 +471,8 @@ int inet6_getname(struct socket *sock, struct sockaddr *uaddr,
 
 		sin->sin6_port = inet->inet_sport;
 	}
-	if (ipv6_addr_type(&sin->sin6_addr) & IPV6_ADDR_LINKLOCAL)
-		sin->sin6_scope_id = sk->sk_bound_dev_if;
+	sin->sin6_scope_id = ipv6_iface_scope_id(&sin->sin6_addr,
+						 sk->sk_bound_dev_if);
 	*uaddr_len = sizeof(*sin);
 	return 0;
 }
net/ipv6/datagram.c +6 -6
index f5a5478..c05a2c9 100644
@@ -124,7 +124,7 @@ ipv4_connected:
 		goto out;
 	}
 
-	if (addr_type&IPV6_ADDR_LINKLOCAL) {
+	if (__ipv6_addr_needs_scope_id(addr_type)) {
 		if (addr_len >= sizeof(struct sockaddr_in6) &&
 		    usin->sin6_scope_id) {
 			if (sk->sk_bound_dev_if &&
@@ -362,8 +362,8 @@ int ipv6_recv_error(struct sock *sk, struct msghdr *msg, int len)
 			sin->sin6_addr = ip6h->daddr;
 			if (np->sndflow)
 				sin->sin6_flowinfo = ip6_flowinfo(ip6h);
-			if (ipv6_addr_type(&sin->sin6_addr) & IPV6_ADDR_LINKLOCAL)
-				sin->sin6_scope_id = IP6CB(skb)->iif;
+			sin->sin6_scope_id = ipv6_iface_scope_id(&sin->sin6_addr,
+								 IP6CB(skb)->iif);
 		} else {
 			ipv6_addr_set_v4mapped(*(__be32 *)(nh + serr->addr_offset),
 					       &sin->sin6_addr);
@@ -381,8 +381,8 @@ int ipv6_recv_error(struct sock *sk, struct msghdr *msg, int len)
 			sin->sin6_addr = ipv6_hdr(skb)->saddr;
 			if (np->rxopt.all)
 				ip6_datagram_recv_ctl(sk, msg, skb);
-			if (ipv6_addr_type(&sin->sin6_addr) & IPV6_ADDR_LINKLOCAL)
-				sin->sin6_scope_id = IP6CB(skb)->iif;
+			sin->sin6_scope_id = ipv6_iface_scope_id(&sin->sin6_addr,
+								 IP6CB(skb)->iif);
 		} else {
 			struct inet_sock *inet = inet_sk(sk);
 
@@ -653,7 +653,7 @@ int ip6_datagram_send_ctl(struct net *net, struct sock *sk,
 					rcu_read_unlock();
 					return -ENODEV;
 				}
-			} else if (addr_type & IPV6_ADDR_LINKLOCAL) {
+			} else if (__ipv6_addr_needs_scope_id(addr_type)) {
 				rcu_read_unlock();
 				return -EINVAL;
 			}
net/ipv6/icmp.c +1 -1
index fff5bdd..71b900c 100644
@@ -434,7 +434,7 @@ void icmpv6_send(struct sk_buff *skb, u8 type, u8 code, __u32 info)
 	 *	Source addr check
 	 */
 
-	if (addr_type & IPV6_ADDR_LINKLOCAL)
+	if (__ipv6_addr_needs_scope_id(addr_type))
 		iif = skb->dev->ifindex;
 
 	/*
net/ipv6/inet6_connection_sock.c +2 -4
index b386a2c..9f1020b 100644
@@ -170,10 +170,8 @@ void inet6_csk_addr2sockaddr(struct sock *sk, struct sockaddr * uaddr)
 	sin6->sin6_port	= inet_sk(sk)->inet_dport;
 	/* We do not store received flowlabel for TCP */
 	sin6->sin6_flowinfo = 0;
-	sin6->sin6_scope_id = 0;
-	if (sk->sk_bound_dev_if &&
-	    ipv6_addr_type(&sin6->sin6_addr) & IPV6_ADDR_LINKLOCAL)
-		sin6->sin6_scope_id = sk->sk_bound_dev_if;
+	sin6->sin6_scope_id = ipv6_iface_scope_id(&sin6->sin6_addr,
+						  sk->sk_bound_dev_if);
 }
 
 EXPORT_SYMBOL_GPL(inet6_csk_addr2sockaddr);
net/ipv6/ip6_input.c +1 -1
index 521d9fd..f10bd9f 100644
@@ -279,7 +279,7 @@ int ip6_mc_input(struct sk_buff *skb)
 	 *      IPv6 multicast router mode is now supported ;)
 	 */
 	if (dev_net(skb->dev)->ipv6.devconf_all->mc_forwarding &&
-	    !(ipv6_addr_type(&hdr->daddr) & IPV6_ADDR_LINKLOCAL) &&
+	    !__ipv6_addr_needs_scope_id(__ipv6_addr_type(&hdr->daddr)) &&
 	    likely(!(IP6CB(skb)->flags & IP6SKB_FORWARDED))) {
 		/*
 		 * Okay, we try to forward - split and duplicate
net/ipv6/udp.c +3 -3
index 599e1ba6..97d5ad0 100644
@@ -457,8 +457,8 @@ try_again:
 					       &sin6->sin6_addr);
 		else {
 			sin6->sin6_addr = ipv6_hdr(skb)->saddr;
-			if (ipv6_addr_type(&sin6->sin6_addr) & IPV6_ADDR_LINKLOCAL)
-				sin6->sin6_scope_id = IP6CB(skb)->iif;
+			sin6->sin6_scope_id = ipv6_iface_scope_id(&sin6->sin6_addr,
+								  IP6CB(skb)->iif);
 		}
 
 	}
@@ -1118,7 +1118,7 @@ do_udp_sendmsg:
 
 		if (addr_len >= sizeof(struct sockaddr_in6) &&
 		    sin6->sin6_scope_id &&
-		    ipv6_addr_type(daddr)&IPV6_ADDR_LINKLOCAL)
+		    __ipv6_addr_needs_scope_id(__ipv6_addr_type(daddr)))
 			fl6.flowi6_oif = sin6->sin6_scope_id;
 	} else {
 		if (sk->sk_state != TCP_ESTABLISHED)
Signature · 1 line
1.8.1.2
#27 lore raw
In reply to Hannes Frederic Sowa
Quoted 3 lines · Hannes Frederic Sowa

c) I will send a patch as soon as the errata thing is resolved to drop ff00::/15 from entering the host from the wire. In case of outgoing traffic to ff00::/16, should this be stopped with a net-unreachable route?

No, check should placed around ipv6_addr_loopback(), like you have posted. Would you cook a patch for ffx0::/16 anyway, please?

--yoshfuji