gnrc_netif: add option to always join solicited-node multicast group (for connectivity across border router restart) #16988

benpicco · 2021-10-14T13:27:27Z

Contribution description

A border router will send a neighbor solicitation to the solicited-node multicasts if it encounters a packet from a global address that is not registered with the border router.

This happens when the border router is restarted while some nodes were online.

As the nodes were not in the solicited-node multicast group, they would drop the neighbor solicitation which resulted in them not being able to route any packets through the border router until they would reboot themself (or their prefix expired).

To fix this, always join the solicited-node multicast address as a 6LN.
This allows to have a connection across BR restarts.

Testing procedure

To make sure the border router retains it's address across reboots, we have to provide it with a EUI-64:

--- a/examples/gnrc_border_router/Makefile.native.conf
+++ b/examples/gnrc_border_router/Makefile.native.conf
@@ -23,3 +23,4 @@ TERMPROG ?= sudo $(RIOTTOOLS)/zep_dispatch/start_network.sh $(TERMPROG_FLAGS)
 
 # -z [::1]:$PORT for each ZEP device
 TERMFLAGS ?= $(patsubst %,-z [::1]:%, $(shell seq $(ZEP_PORT_BASE) $(ZEP_PORT_MAX)))
+TERMFLAGS += -U 7A:37:FC:7D:1A:AF

ARSM is enabled for routers, so to reproduce this our node must be a simple 6LN:

--- a/examples/gnrc_networking/Makefile
+++ b/examples/gnrc_networking/Makefile
@@ -14,7 +14,7 @@ USEMODULE += auto_init_gnrc_netif
 # Activate ICMPv6 error messages
 USEMODULE += gnrc_icmpv6_error
 # Specify the mandatory networking modules for IPv6 and UDP
-USEMODULE += gnrc_ipv6_router_default
+USEMODULE += gnrc_ipv6_default
 USEMODULE += gnrc_udp
 # Add a routing protocol
 USEMODULE += gnrc_rpl

start the (native) border router:

$ make -C examples/gnrc_border_router BOARD=native all term

> ifconfig
Iface  7  HWaddr: 00:00  Channel: 26  NID: 0x23 
          Long HWaddr: 7A:37:FC:7D:1A:AF:00:00 
          L2-PDU:102  MTU:1280  HL:64  RTR  
          RTR_ADV  6LO  IPHC  
          Source address length: 8
          Link type: wireless
          inet6 addr: fe80::7837:fc7d:1aaf:0  scope: link  VAL
          inet6 addr: 2001:db8::7837:fc7d:1aaf:0  scope: global  VAL
          inet6 group: ff02::2
          inet6 group: ff02::1
          inet6 group: ff02::1:ffaf:0
          
Iface  6  HWaddr: 7A:37:FC:7D:1A:AF 
          L2-PDU:1500  MTU:1500  HL:64  RTR  
          Source address length: 6
          Link type: wired
          inet6 addr: fe80::7837:fcff:fe7d:1aaf  scope: link  VAL
          inet6 addr: fe80::2  scope: link  VAL
          inet6 group: ff02::2
          inet6 group: ff02::1
          inet6 group: ff02::1:ff7d:1aaf
          inet6 group: ff02::1:ff00:2

start the 6LoWPAN node

$ make -C examples/gnrc_networking USE_ZEP=1 all term

> ifconfig
Iface  7  HWaddr: 34:18  Channel: 26  NID: 0x23 
          Long HWaddr: DA:A0:EF:4A:D1:76:34:18 
          L2-PDU:102  MTU:1280  HL:64  6LO  
          IPHC  
          Source address length: 8
          Link type: wireless
          inet6 addr: fe80::d8a0:ef4a:d176:3418  scope: link  VAL
          inet6 addr: 2001:db8::d8a0:ef4a:d176:3418  scope: global  VAL
          inet6 group: ff02::1
          inet6 group: ff02::1:ff76:3418
          
          Statistics for Layer 2
            RX packets 4  bytes 286
            TX packets 3 (Multicast: 1)  bytes 171
            TX succeeded 6 errors 0
          Statistics for IPv6
            RX packets 3  bytes 304
            TX packets 3 (Multicast: 1)  bytes 224
            TX succeeded 3 errors 0

> ping fdea:dbee:f::1   
12 bytes from fdea:dbee:f::1: icmp_seq=0 ttl=63 rssi=-512 dBm time=0.649 ms
12 bytes from fdea:dbee:f::1: icmp_seq=1 ttl=63 rssi=-512 dBm time=0.600 ms
12 bytes from fdea:dbee:f::1: icmp_seq=2 ttl=63 rssi=-512 dBm time=0.723 ms

--- fdea:dbee:f::1 PING statistics ---
3 packets transmitted, 3 packets received, 0% packet loss
round-trip min/avg/max = 0.600/0.657/0.723 ms

We can now reboot the border router, our node keeps it's packets routed:

> ping -c 100 fdea:dbee:f::1
ping -c 100 fdea:dbee:f::1
12 bytes from fdea:dbee:f::1: icmp_seq=0 ttl=63 rssi=-512 dBm time=0.488 ms
12 bytes from fdea:dbee:f::1: icmp_seq=1 ttl=63 rssi=-512 dBm time=0.529 ms
12 bytes from fdea:dbee:f::1: icmp_seq=2 ttl=63 rssi=-512 dBm time=0.539 ms
# border router restart
nib: Creating NCE for (ipv6 = 2001:db8::7837:fc7d:1aaf:0, iface = 7, nud_state = STALE)
nib: Neighbor cache full
12 bytes from fdea:dbee:f::1: icmp_seq=3 ttl=63 rssi=-512 dBm time=1.338 ms
12 bytes from fdea:dbee:f::1: icmp_seq=4 ttl=63 rssi=-512 dBm time=0.528 ms
12 bytes from fdea:dbee:f::1: icmp_seq=5 ttl=63 rssi=-512 dBm time=0.590 ms
12 bytes from fdea:dbee:f::1: icmp_seq=6 ttl=63 rssi=-512 dBm time=0.590 ms
12 bytes from fdea:dbee:f::1: icmp_seq=7 ttl=63 rssi=-512 dBm time=0.592 ms
# border router restart
nib: Creating NCE for (ipv6 = 2001:db8::7837:fc7d:1aaf:0, iface = 7, nud_state = STALE)
nib: Neighbor cache full
12 bytes from fdea:dbee:f::1: icmp_seq=8 ttl=63 rssi=-512 dBm time=1.357 ms
12 bytes from fdea:dbee:f::1: icmp_seq=9 ttl=63 rssi=-512 dBm time=0.602 ms
12 bytes from fdea:dbee:f::1: icmp_seq=10 ttl=63 rssi=-512 dBm time=0.598 ms

compared to the situation on master:

> ifconfig
Iface  7  HWaddr: 08:64  Channel: 26  NID: 0x23 
          Long HWaddr: 7A:30:7B:E2:5D:42:88:64 
          L2-PDU:102  MTU:1280  HL:64  6LO  
          IPHC  
          Source address length: 8
          Link type: wireless
          inet6 addr: fe80::7830:7be2:5d42:8864  scope: link  VAL
          inet6 addr: 2001:db8::7830:7be2:5d42:8864  scope: global  VAL
          inet6 group: ff02::1
          
          Statistics for Layer 2
            RX packets 4  bytes 286
            TX packets 3 (Multicast: 1)  bytes 171
            TX succeeded 6 errors 0
          Statistics for IPv6
            RX packets 3  bytes 304
            TX packets 3 (Multicast: 1)  bytes 224
            TX succeeded 3 errors 0

> ping -c 100 fdea:dbee:f::1
ping -c 100 fdea:dbee:f::1
12 bytes from fdea:dbee:f::1: icmp_seq=0 ttl=63 rssi=3584 dBm time=0.508 ms
12 bytes from fdea:dbee:f::1: icmp_seq=1 ttl=63 rssi=3584 dBm time=0.662 ms
12 bytes from fdea:dbee:f::1: icmp_seq=2 ttl=63 rssi=3584 dBm time=0.330 ms
12 bytes from fdea:dbee:f::1: icmp_seq=3 ttl=63 rssi=3584 dBm time=0.765 ms
12 bytes from fdea:dbee:f::1: icmp_seq=4 ttl=63 rssi=3584 dBm time=0.748 ms
# border router restart - node never recovers

Issues/PRs references

A border router will send a neighbor solicitation to the solicited-node multicasts if it encounters a packet from a global address that is not registered with the border router. This happens when the border router is restarted while some nodes were online. As the nodes were not in the solicited-node multicast group, they would drop the neighbor solicitation which resulted in them not being able to route any packets through the border router until they would reboot themself (or their prefix expired). To fix this, always join the solicited-node multicast address as a 6LN. This allows to have a connection across BR restarts.

miri64 · 2021-10-14T13:39:04Z

sys/net/gnrc/netif/gnrc_netif.c

-    }
-#else  /* CONFIG_GNRC_IPV6_NIB_ARSM */
-    if (!gnrc_netif_is_6ln(netif)) {
+    if (IS_ACTIVE(CONFIG_GNRC_IPV6_NIB_ARSM) || gnrc_netif_is_6ln(netif)) {


See https://datatracker.ietf.org/doc/html/rfc6775#section-5.2

There is no need to join the solicited-node multicast address, since
nobody multicasts NSs in this type of network. A host MUST join the
all-nodes multicast address.

IMHO this should be made configurable for 6LNs (and default turned off), so a user can set it in accordance with their respective use case.

Fine, I just don't see why this should not be enabled. To me this appears like a case of 'broken defaults'.

nobody multicasts NSs in this type of network

Is this a bug on the border router then? Because it clearly does.

5.1. Forbidden Actions A host MUST NOT multicast an NS message.

so if BR does it is a bug in BR

i like how ipv6 avoids broadcast by doing multicast -> then some extensions happens and there should not be any multicast NS but there is no broadcast -> so it is an all-multicast which by the wording is no multicast

It it close to some philosophy discussion.

rfc6775 seems to define a essential part
of
rfc4861 illegal
rfc4861 sec 4.1

destination address: Either the solicited-node multicast address corresponding to the target address, or the target address.

sys/net/gnrc/netif/gnrc_netif.c

sys/include/net/gnrc/ipv6/nib/conf.h

sys/net/gnrc/network_layer/ipv6/nib/Kconfig

Co-authored-by: Martine Lenders <mail@martine-lenders.eu>

sys/net/gnrc/netif/gnrc_netif.c

benpicco · 2021-11-02T10:53:20Z

The thing is: why not have this as default when gnrc_sixlowpan_border_router expects the address to be present?
So this is either a bug in the border router that it tries to send to the solicited-node address to resolve a 6LoWPAN address or a bug in the node that it does not join the solicited-node multicast group.

OlegHahm · 2022-03-09T12:17:53Z

@benpicco, where are we at here?

benpicco · 2022-03-09T12:18:40Z

Waiting for a response from @miri64

fabian18 · 2022-04-18T19:48:49Z

I think I have also recently discovered this. My fix was CFLAGS+=-DCONFIG_GNRC_IPV6_NIB_ARSM=1
Because in conf.h it was explicitly disabled for hosts.

#if !CONFIG_GNRC_IPV6_NIB_6LR
# ifndef CONFIG_GNRC_IPV6_NIB_ARSM
# define CONFIG_GNRC_IPV6_NIB_ARSM                    0
# endif
# ifndef CONFIG_GNRC_IPV6_NIB_NUMOF
/* only needs to store default router */
# define CONFIG_GNRC_IPV6_NIB_NUMOF                  (1)
# endif
#endif

I think it can actually be enabled by default, even when we are a host (RFC4861)

On Linux, an interface also joins the solicited-node multicast address, even if it only has a link local address:

$ sysctl -w net.ipv6.conf.all.forwarding=0
$ ip link set eth0 down
# replug cable
$ ip link set eth0 up

$ ip a
5: eth0: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qdisc fq_codel state UP group default qlen 1000
    link/ether 00:0e:c6:b4:b4:f4 brd ff:ff:ff:ff:ff:ff
    altname enp0s20f0u2u4u1u2
    inet6 fe80::3478:9157:9fd5:81c3/64 scope link noprefixroute 
       valid_lft forever preferred_lft forever

$ ip maddr show
5:      eth0
        link  01:00:5e:00:00:01
        link  33:33:00:00:00:01
        link  33:33:00:00:00:fb
        link  33:33:ff:d5:81:c3
        inet  224.0.0.1
        inet6 ff02::1:ffd5:81c3
        inet6 ff02::fb
        inet6 ff02::1
        inet6 ff01::1

fabian18 · 2022-04-18T20:09:38Z

See https://datatracker.ietf.org/doc/html/rfc6775#section-5.2

There is no need to join the solicited-node multicast address, since
nobody multicasts NSs in this type of network. A host MUST join the
all-nodes multicast address.

IMHO this should be made configurable for 6LNs (and default turned off), so a user can set it in accordance with their respective use case.

Ah true, 6LN host ...

benpicco · 2022-04-18T20:15:34Z

#11038 suggests Linux here behaves the same as our gnrc_border_router.
I'm not sure what the BR is supposed to do instead, but if just joining the the solicited-node multicast group is such an easy fix, why not do it on our side?

fabian18 · 2022-04-18T21:33:28Z

but if just joining the solicited-node multicast group is such an easy fix, why not do it on our side?

hm, because @miri64 is, right. Technically the RFC clearly says that a 6LN host is not supposed to join that group.
I think the ping from the BR to the host simply does not work until the host has re-registered at the router.

Linux somehow does join the group, also as a 6LN host. Maybe the kernel does not fully support RFC6775.

$ sudo sysctl -w net.ipv6.conf.all.forwarding=0
$ sudo ip a a fd00:a:a:a:c4b0:b9f6:fefa:c8d3/64 dev lowpan0

8:      lowpan0
        inet  224.0.0.1
        inet6 ff02::fb
        inet6 ff02::1:fffa:c8d3 users 2
        inet6 ff02::1
        inet6 ff01::1

I am not saying that I am against a configuration flag to force-join the group.

miri64 · 2022-05-23T12:52:29Z

Linux somehow does join the group, also as a 6LN host. Maybe the kernel does not fully support RFC6775.

Yes, Linux does not implement RFC 6775 and uses normal (energy-intensive) NDP with 6LoWPAN.

miri64 · 2022-05-23T12:53:35Z

The compromise, if it yields a problem in communication with Linux: Deactivate 6Lo-ND completely. This of course increases multicast messages within the network.

miri64 · 2023-03-16T13:26:01Z

With the description in https://forum.riot-os.org/t/telnet-server-example/3893/2:

[…] so addresses not registered with the border router will be ignored.

Is this really a bug then? If a (6LN) node did not register its address, the address is not valid anyways, so why would the border router need to send to those addresses?

benpicco · 2024-02-08T20:38:18Z

With #20046 merged this PR does not make sense anymore.

benpicco requested review from cgundogan, haukepetersen, miri64 and PeterKietzmann as code owners October 14, 2021 13:27

benpicco added the Type: bug The issue reports a bug / The PR fixes a bug (including spelling errors) label Oct 14, 2021

github-actions bot added Area: network Area: Networking Area: sys Area: System labels Oct 14, 2021

benpicco requested a review from chrysn October 14, 2021 13:29

miri64 requested changes Oct 14, 2021

View reviewed changes

fixup! gnrc_netif: always join solicited-node multicast group if 6LN

82ed5a4

benpicco requested review from jia200x and leandrolanzieri as code owners October 14, 2021 13:57

github-actions bot added the Area: Kconfig Area: Kconfig integration label Oct 14, 2021

miri64 reviewed Oct 14, 2021

View reviewed changes

sys/net/gnrc/netif/gnrc_netif.c Outdated Show resolved Hide resolved

miri64 requested changes Oct 14, 2021

View reviewed changes

sys/include/net/gnrc/ipv6/nib/conf.h Outdated Show resolved Hide resolved

sys/include/net/gnrc/ipv6/nib/conf.h Outdated Show resolved Hide resolved

sys/include/net/gnrc/ipv6/nib/conf.h Show resolved Hide resolved

sys/net/gnrc/network_layer/ipv6/nib/Kconfig Outdated Show resolved Hide resolved

benpicco and others added 3 commits October 14, 2021 17:26

Update sys/net/gnrc/network_layer/ipv6/nib/Kconfig

5bfef0a

Co-authored-by: Martine Lenders <mail@martine-lenders.eu>

Update sys/include/net/gnrc/ipv6/nib/conf.h

495cefa

Co-authored-by: Martine Lenders <mail@martine-lenders.eu>

fixup! Update sys/include/net/gnrc/ipv6/nib/conf.h

d6355e7

miri64 reviewed Oct 14, 2021

View reviewed changes

sys/net/gnrc/netif/gnrc_netif.c Outdated Show resolved Hide resolved

fixup! fixup! Update sys/include/net/gnrc/ipv6/nib/conf.h

2d5ed7f

benpicco changed the title ~~gnrc_netif: always join solicited-node multicast group if 6LN~~ gnrc_netif: add option to always join solicited-node multicast group (for connectivity across border router restart) Oct 17, 2021

benpicco added this to the Release 2022.01 milestone Dec 2, 2021

benpicco requested a review from biboc December 21, 2021 11:18

benpicco requested a review from kfessel January 4, 2022 09:54

emmanuelsearch modified the milestones: Release 2022.01, Release 2022.04 Jan 27, 2022

benpicco mentioned this pull request Apr 18, 2022

Neighbor Discovery not working after router reboot when using SLAAC #11038

Open

benpicco removed this from the Release 2022.04 milestone May 18, 2022

benpicco mentioned this pull request Nov 20, 2022

gnrc_ipv6_nib: clean up _resolve_addr() #18939

Merged

benpicco mentioned this pull request Feb 28, 2023

ipv6/nib: 6LBR should not send RS on their downstream interface #19335

Merged

benpicco mentioned this pull request Nov 3, 2023

gnrc/ipv6/nib: don't do multicast address resolution on 6LoWPAN #20046

Merged

benpicco closed this Feb 8, 2024

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

gnrc_netif: add option to always join solicited-node multicast group (for connectivity across border router restart) #16988

gnrc_netif: add option to always join solicited-node multicast group (for connectivity across border router restart) #16988

benpicco commented Oct 14, 2021

miri64 Oct 14, 2021

benpicco Oct 14, 2021

benpicco Oct 26, 2021 •

edited

Loading

kfessel Jan 4, 2022

kfessel Jan 4, 2022 •

edited

Loading

kfessel Jan 4, 2022 •

edited

Loading

benpicco commented Nov 2, 2021

OlegHahm commented Mar 9, 2022

benpicco commented Mar 9, 2022

fabian18 commented Apr 18, 2022

fabian18 commented Apr 18, 2022

benpicco commented Apr 18, 2022 •

edited

Loading

fabian18 commented Apr 18, 2022

miri64 commented May 23, 2022

miri64 commented May 23, 2022

miri64 commented Mar 16, 2023

benpicco commented Feb 8, 2024

gnrc_netif: add option to always join solicited-node multicast group (for connectivity across border router restart) #16988

gnrc_netif: add option to always join solicited-node multicast group (for connectivity across border router restart) #16988

Conversation

benpicco commented Oct 14, 2021

Contribution description

Testing procedure

Issues/PRs references

miri64 Oct 14, 2021

Choose a reason for hiding this comment

benpicco Oct 14, 2021

Choose a reason for hiding this comment

benpicco Oct 26, 2021 • edited Loading

Choose a reason for hiding this comment

kfessel Jan 4, 2022

Choose a reason for hiding this comment

kfessel Jan 4, 2022 • edited Loading

Choose a reason for hiding this comment

kfessel Jan 4, 2022 • edited Loading

Choose a reason for hiding this comment

benpicco commented Nov 2, 2021

OlegHahm commented Mar 9, 2022

benpicco commented Mar 9, 2022

fabian18 commented Apr 18, 2022

fabian18 commented Apr 18, 2022

benpicco commented Apr 18, 2022 • edited Loading

fabian18 commented Apr 18, 2022

miri64 commented May 23, 2022

miri64 commented May 23, 2022

miri64 commented Mar 16, 2023

benpicco commented Feb 8, 2024

benpicco Oct 26, 2021 •

edited

Loading

kfessel Jan 4, 2022 •

edited

Loading

kfessel Jan 4, 2022 •

edited

Loading

benpicco commented Apr 18, 2022 •

edited

Loading