Skip to contentSkip to conversation

sent an invalid ICMP type 11, code 0 error to a broadcast: 0.0.0.0 on lo?

DISCUSSION Michael Tokarev 3 participants 9 messages Jul 6, 2005 – Jul 8, 2005

lore
Share & tools
lore
#0 lore raw

On our gateway machine, wich is running 2.6 kernel, I'm seeing quite several messages like this:

kernel: 192.168.4.2 sent an invalid ICMP type 11, code 0 error to a broadcast: 0.0.0.0 on lo last message repeated 3 times
kernel: 192.168.4.2 sent an invalid ICMP type 11, code 0 error to a broadcast: 0.0.0.0 on lo last message repeated 3 times

kernel: 81.13.90.174 sent an invalid ICMP type 11, code 0 error to a broadcast: 0.0.0.0 on lo gate last message repeated 10 times

kernel: 81.13.90.174 sent an invalid ICMP type 11, code 0 error to a broadcast: 0.0.0.0 on lo last message repeated 9 times

All the IP addresses mentioned are local to this box. It's always like this, type11 code0 to 0.0.0.0 on lo.

The box is a gateway, which have "external" interface, several "LAN" ifaces, DMZ, dialin modem pool and accepts vtund connections. And it has iptables-based firewall and NAT enabled. Here's a lis of network-related modules:

ppp_deflate             4736  0
zlib_deflate           21528  1 ppp_deflate
zlib_inflate           16640  1 ppp_deflate
ppp_async               8576  0
crc_ccitt               1920  1 ppp_async
ppp_generic            19860  2 ppp_deflate,ppp_async
slhc                    6144  1 ppp_generic
tun                     8320  13
crc32                   3968  1 tun
ipt_ECN                 2944  163
iptable_mangle          2304  1
ipt_REJECT              4096  1
ipt_LOG                 6272  3
ipt_state               1664  3
ipt_comment             1536  22
iptable_filter          2432  1
iptable_nat            19420  1
ipt_NOTRACK             1920  3
iptable_raw             1792  1
ip_tables              18304  10 ipt_ECN,iptable_mangle,ipt_REJECT,ipt_LOG,ipt_state,ipt_comment,iptable_filter,iptable_nat,ipt_NOTRACK,iptable_raw
md5                     3840  1
ipv6                  218688  12
ip_conntrack           37560  3 ipt_state,iptable_nat,ipt_NOTRACK
aes_i586               38656  1
airo                   63520  0
3c509                  11476  0

I've seen several posts about problems *like* this, but they're all not from the same box (usually it's some problem with *another* device on the network which is sending bogus packets, or with hubs/switches or cables). In this case, it looks like it is this same box who's generating those bad packets.

So far, I wasn't able to relate this message with any particular network activity. It happens "randomly".

What it can be?

Thanks.

/mjt

#1 lore raw
Quoted 56 lines · Michael Tokarev

On our gateway machine, wich is running 2.6 kernel, I'm seeing quite several messages like this:

kernel: 192.168.4.2 sent an invalid ICMP type 11, code 0 error to a broadcast: 0.0.0.0 on lo last message repeated 3 times
kernel: 192.168.4.2 sent an invalid ICMP type 11, code 0 error to a broadcast: 0.0.0.0 on lo last message repeated 3 times

kernel: 81.13.90.174 sent an invalid ICMP type 11, code 0 error to a broadcast: 0.0.0.0 on lo gate last message repeated 10 times

kernel: 81.13.90.174 sent an invalid ICMP type 11, code 0 error to a broadcast: 0.0.0.0 on lo last message repeated 9 times

All the IP addresses mentioned are local to this box. It's always like this, type11 code0 to 0.0.0.0 on lo.

The box is a gateway, which have "external" interface, several "LAN" ifaces, DMZ, dialin modem pool and accepts vtund connections. And it has iptables-based firewall and NAT enabled. Here's a lis of network-related modules:

ppp_deflate             4736  0
zlib_deflate           21528  1 ppp_deflate
zlib_inflate           16640  1 ppp_deflate
ppp_async               8576  0
crc_ccitt               1920  1 ppp_async
ppp_generic            19860  2 ppp_deflate,ppp_async
slhc                    6144  1 ppp_generic
tun                     8320  13
crc32                   3968  1 tun
ipt_ECN                 2944  163
iptable_mangle          2304  1
ipt_REJECT              4096  1
ipt_LOG                 6272  3
ipt_state               1664  3
ipt_comment             1536  22
iptable_filter          2432  1
iptable_nat            19420  1
ipt_NOTRACK             1920  3
iptable_raw             1792  1
ip_tables              18304  10 ipt_ECN,iptable_mangle,ipt_REJECT,ipt_LOG,ipt_state,ipt_comment,iptable_filter,iptable_nat,ipt_NOTRACK,iptable_raw
md5                     3840  1
ipv6                  218688  12
ip_conntrack           37560  3 ipt_state,iptable_nat,ipt_NOTRACK
aes_i586               38656  1
airo                   63520  0
3c509                  11476  0

I've seen several posts about problems *like* this, but they're all not from the same box (usually it's some problem with *another* device on the network which is sending bogus packets, or with hubs/switches or cables). In this case, it looks like it is this same box who's generating those bad packets.

Quoted 7 lines

So far, I wasn't able to relate this message with any particular network activity. It happens "randomly".

What it can be?

Thanks.

Are you sure lo is configured properly, i.e.

ifconfig lo 127.0.0.1 netmask 255.0.0.0
route add -net 127.0.0.0 netmask 255.0.0.0 dev lo

There should be no LAN routes going through lo.
It looks as though there are:

Quoted 2 lines

kernel: 192.168.4.2 sent an invalid ICMP type 11, code 0 error to a broadcast: 0.0.0.0 on lo

                             |________ 192.168.4.2 pinged lo

Only the 127.0.0.0 network should be routed through the loop-back device.

Cheers,
Dick Johnson
Penguin : Linux version 2.6.12 on an i686 machine (5537.79 BogoMips).
Notice : All mail here is now cached for review by Dictator Bush.
98.36% of all statistics are fiction.

#2 lore raw
Quoted 2 lines · Richard B. Johnson

Only the 127.0.0.0 network should be routed through the loop-back device.

This is the normal dose of disinformation embedded into otherwise useful reply from Richard. Sometimes I wonder whether he does this for fun for for some other dark and obscure reason.

Anyway, lo is perfetly happy with packets other than 127.x.x.x going thru it.

# ip a
1: if: <BROADCAST,MULTICAST,UP> mtu 1500 qdisc pfifo_fast qlen 1000
    link/ether 00:04:23:15:ba:76 brd ff:ff:ff:ff:ff:ff
    inet 172.17.13.22/16 brd 172.17.255.255 scope global if
2: lo: <LOOPBACK,UP> mtu 16436 qdisc noqueue
    link/loopback 00:00:00:00:00:00 brd 00:00:00:00:00:00
    inet 127.0.0.1/8 scope host lo
# ip r
172.17.0.0/16 dev if  proto kernel  scope link  src 172.17.13.22
default via 172.17.0.1 dev if
# ping -c4 172.17.13.22
PING 172.17.13.22 (172.17.13.22) 56(84) bytes of data.
64 bytes from 172.17.13.22: icmp_seq=1 ttl=64 time=0.155 ms
64 bytes from 172.17.13.22: icmp_seq=2 ttl=64 time=0.113 ms
64 bytes from 172.17.13.22: icmp_seq=3 ttl=64 time=0.122 ms
64 bytes from 172.17.13.22: icmp_seq=4 ttl=64 time=0.121 ms
--- 172.17.13.22 ping statistics ---
4 packets transmitted, 4 received, 0% packet loss, time 3011ms
rtt min/avg/max/mdev = 0.113/0.127/0.155/0.021 ms

On the other console:

# tcpdump -nlilo
tcpdump: verbose output suppressed, use -v or -vv for full protocol decode
listening on lo, link-type EN10MB (Ethernet), capture size 68 bytes
11:20:37.669996 IP 172.17.13.22 > 172.17.13.22: icmp 64: echo request seq 1
11:20:37.670093 IP 172.17.13.22 > 172.17.13.22: icmp 64: echo reply seq 1
11:20:38.671066 IP 172.17.13.22 > 172.17.13.22: icmp 64: echo request seq 2
11:20:38.671125 IP 172.17.13.22 > 172.17.13.22: icmp 64: echo reply seq 2
11:20:39.672327 IP 172.17.13.22 > 172.17.13.22: icmp 64: echo request seq 3
11:20:39.672389 IP 172.17.13.22 > 172.17.13.22: icmp 64: echo reply seq 3
11:20:40.681857 IP 172.17.13.22 > 172.17.13.22: icmp 64: echo request seq 4
11:20:40.681922 IP 172.17.13.22 > 172.17.13.22: icmp 64: echo reply seq 4
--
vda
#3 lore raw
In reply to Richard B. Johnson
Quoted 4 lines · Richard B. Johnson
Quoted 2 lines · Michael Tokarev

kernel: 192.168.4.2 sent an invalid ICMP type 11, code 0 error to a
broadcast: 0.0.0.0 on lo

[]

Quoted 1 line
Quoted 1 line

All the IP addresses mentioned are local to this box.

[]

Quoted 1 line

Are you sure lo is configured properly, i.e.

Yes. More, rp_filter is activated on all interfaces.

Quoted 4 lines

ifconfig lo 127.0.0.1 netmask 255.0.0.0
route add -net 127.0.0.0 netmask 255.0.0.0 dev lo

There should be no LAN routes going through lo.

There's no.

Quoted 9 lines

It looks as though there are:

Quoted 2 lines

kernel: 192.168.4.2 sent an invalid ICMP type 11, code 0 error to a broadcast: 0.0.0.0 on lo

                             |________ 192.168.4.2 pinged lo

Only the 127.0.0.0 network should be routed through the loop-back device.

Again: All the IP addresses mentioned are local to this box.

If you ping an IP address on your eth0, the traffic will "go" over loopback. You can verify it using tcpdump:

1: lo: <LOOPBACK,UP> mtu 16436 qdisc noqueue

    link/loopback 00:00:00:00:00:00 brd 00:00:00:00:00:00
    inet 127.0.0.1/8 scope host lo
2: eth0: <BROADCAST,MULTICAST,UP> mtu 1500 qdisc pfifo_fast qlen 1000
    link/ether 00:a0:d2:1d:a7:63 brd ff:ff:ff:ff:ff:ff
    inet 192.168.1.1/24 brd 192.168.1.255 scope global eth0
# tcpdump -npi lo proto ICMP
tcpdump: verbose output suppressed, use -v or -vv for full protocol decode
listening on lo, link-type EN10MB (Ethernet), capture size 96 bytes
15:55:13.679234 IP 192.168.1.1 > 192.168.1.1: icmp 64: echo request seq 1
15:55:13.679269 IP 192.168.1.1 > 192.168.1.1: icmp 64: echo reply seq 1
$ ping 192.168.1.1
PING 192.168.1.1 (192.168.1.1) 56(84) bytes of data.
64 bytes from 192.168.1.1: icmp_seq=1 ttl=64 time=0.060 ms
--- 192.168.1.1 ping statistics ---
1 packets transmitted, 1 received, 0% packet loss, time 0ms
rtt min/avg/max/mdev = 0.060/0.060/0.060/0.000 ms

/mjt

#4 lore raw
Quoted 27 lines · Michael Tokarev
Quoted 4 lines · Richard B. Johnson
Quoted 2 lines · Michael Tokarev

kernel: 192.168.4.2 sent an invalid ICMP type 11, code 0 error to a
broadcast: 0.0.0.0 on lo

[]

Quoted 1 line
Quoted 1 line

All the IP addresses mentioned are local to this box.

[]

Quoted 1 line

Are you sure lo is configured properly, i.e.

Yes. More, rp_filter is activated on all interfaces.

Quoted 4 lines

ifconfig lo 127.0.0.1 netmask 255.0.0.0
route add -net 127.0.0.0 netmask 255.0.0.0 dev lo

There should be no LAN routes going through lo.

There's no.

Quoted 7 lines

It looks as though there are:

Quoted 2 lines

kernel: 192.168.4.2 sent an invalid ICMP type 11, code 0 error to a broadcast: 0.0.0.0 on lo

                             |________ 192.168.4.2 pinged lo
Quoted 8 lines
Quoted 2 lines

Only the 127.0.0.0 network should be routed through the loop-back device.

Again: All the IP addresses mentioned are local to this box.

If you ping an IP address on your eth0, the traffic will "go" over loopback. You can verify it using tcpdump:

If you ping an IP address on your computer, the traffic will go through lo. However, I think that the IP address shown is the result of an instrumentation error because it is impossible to put, for instance your 192.168.1.1, through a 127.0.0.0 network, the ONLY route through lo. This shows that 'local' traffic bypasses the lo route filtering altogether. You can verify this by deleting the lo route altogether, you can still ping the local addresses.

Somebody else mentioned that lo was 'perfectly happy' to carry whatever. The fact that something bogus appears on lo can be a sign of a misconfiguration error, just as the reserved 127.0.0.0 network must never appear on ethernet.

In the case of 0.0.0.0 (a possible broadcast), there is no "local" address that could cause a bypass via lo. Instead, any such traffic should have been on the ethernet wire. This shows the possible configuration error that I mentioned.

Quoted 2 lines

1: lo: <LOOPBACK,UP> mtu 16436 qdisc noqueue
link/loopback 00:00:00:00:00:00 brd 00:00:00:00:00:00

                    ^^^^^^^^^^^^^^^^
Quoted 1 line
   inet 127.0.0.1/8 scope host lo

This looks as though there is no netmask set. My configuration shows:

lo Link encap:Local Loopback

          inet addr:127.0.0.1  Mask:255.0.0.0
          inet6 addr: ::1/128 Scope:Host
          UP LOOPBACK RUNNING  MTU:16436  Metric:1

This is a possible configuration error.

Quoted 10 lines

2: eth0: <BROADCAST,MULTICAST,UP> mtu 1500 qdisc pfifo_fast qlen 1000
link/ether 00:a0:d2:1d:a7:63 brd ff:ff:ff:ff:ff:ff
inet 192.168.1.1/24 brd 192.168.1.255 scope global eth0

# tcpdump -npi lo proto ICMP
tcpdump: verbose output suppressed, use -v or -vv for full protocol decode
listening on lo, link-type EN10MB (Ethernet), capture size 96 bytes
15:55:13.679234 IP 192.168.1.1 > 192.168.1.1: icmp 64: echo request seq 1
15:55:13.679269 IP 192.168.1.1 > 192.168.1.1: icmp 64: echo reply seq 1

[SNIPPED rest]

Quoted 3 lines

/mjt

Cheers,
Dick Johnson
Penguin : Linux version 2.6.12 on an i686 machine (5537.79 BogoMips).
Notice : All mail here is now cached for review by Dictator Bush.
98.36% of all statistics are fiction.

#5 lore raw

Richard B. Johnson wrote:
[]

Quoted 8 lines

If you ping an IP address on your computer, the traffic will go through lo. However, I think that the IP address shown is the result of an instrumentation error because it is impossible to put, for instance your 192.168.1.1, through a 127.0.0.0 network, the ONLY route through lo. This shows that 'local' traffic bypasses the lo route filtering altogether. You can verify this by deleting the lo route altogether, you can still ping the local addresses.

Hmm. I can't parse this. ;)
Well, I can bring loopback down, but in this case I can't even ping my local IP addresses anymore, ping reports 'Invalid argument' error (there's only one route left after deleting lo on my test machine - 192.168.1.0/24).

Quoted 4 lines

Somebody else mentioned that lo was 'perfectly happy' to carry whatever. The fact that something bogus appears on lo can be a sign of a misconfiguration error, just as the reserved 127.0.0.0 network must never appear on ethernet.

I'm not denying it may be some misconfiguration. The problem is that I don't see what/where it is. All the stuff looks pretty normal.

Quoted 4 lines

In the case of 0.0.0.0 (a possible broadcast), there is no "local" address that could cause a bypass via lo. Instead, any such traffic should have been on the ethernet wire. This shows the possible configuration error that I mentioned.

Again, I don't understand what do you mean. The message in $subj says that something *on this box* sent that bogus ICMP packet, with source address on this same box. It may be a reply to bogus packet sent with src=0.0.0.0 somehow, or maybe not. Maybe the host is trying to reply to a packet sent to one of its local IPs -- eg, imagine I send packet with src=0.0.0.0 to another host to non-listening port (rp_filter should be activated in that case I think, but IF that packet comes from the interface where the default route goes, rp_filter may not trigger) - and kernel is trying to send an ICMP reply... back to 0.0.0.0...

Even that does not explain everything still. My 'default route' interface - the one which goes to my ISP - I doubt it's possible from the ISP side to send a packet destined to 192.168.4.2 so that the packet will come to us - unless their routing is hosed.

The problem is, I can't see what is causing this misconfiguration or whatever. I wasn't able to capture such a packet so far either -- it never happened while tcpdump was running.

Note the local IP address mentioned is different, I've
seen 3 so far, all 3 are local on this box and are on 3 different (ethernet) interfaces (but the ICMP always comes from lo).

Quoted 11 lines
Quoted 3 lines

1: lo: <LOOPBACK,UP> mtu 16436 qdisc noqueue
link/loopback 00:00:00:00:00:00 brd 00:00:00:00:00:00

                    ^^^^^^^^^^^^^^^^
Quoted 1 line
   inet 127.0.0.1/8 scope host lo

This looks as though there is no netmask set. My configuration shows:

The netmask is perfect - it's 127.255.255.255. The thing you quoted is *ethernet* address, not IP address - and for loopback, it's ok to have that as all-zeros.

Quoted 6 lines

lo Link encap:Local Loopback

          inet addr:127.0.0.1  Mask:255.0.0.0
          inet6 addr: ::1/128 Scope:Host
          UP LOOPBACK RUNNING  MTU:16436  Metric:1

This is a possible configuration error.

Here's how ifconfig shows my interface:

lo Link encap:Local Loopback
inet addr:127.0.0.1 Mask:255.0.0.0 UP LOOPBACK RUNNING MTU:16436 Metric:1

The above output is from `ip addr'.

/mjt

#8 lore raw
In reply to Richard B. Johnson
Quoted 18 lines · Richard B. Johnson
Quoted 8 lines
Quoted 2 lines

Only the 127.0.0.0 network should be routed through the loop-back device.

Again: All the IP addresses mentioned are local to this box.

If you ping an IP address on your eth0, the traffic will "go" over loopback. You can verify it using tcpdump:

If you ping an IP address on your computer, the traffic will go through lo. However, I think that the IP address shown is the result of an instrumentation error because it is impossible to put, for instance your 192.168.1.1, through a 127.0.0.0 network, the ONLY route through lo. This shows that 'local' traffic bypasses the lo route filtering altogether. You can verify this by deleting the lo route altogether, you can still ping the local addresses.

Quoted 4 lines

Somebody else mentioned that lo was 'perfectly happy' to carry whatever. The fact that something bogus appears on lo can be a sign of a misconfiguration error, just as the reserved 127.0.0.0 network must never appear on ethernet.

Care to tcpdump your own lo?

Quoted 6 lines

In the case of 0.0.0.0 (a possible broadcast), there is no "local" address that could cause a bypass via lo. Instead, any such traffic should have been on the ethernet wire. This shows the possible configuration error that I mentioned.

Quoted 7 lines
Quoted 2 lines

1: lo: <LOOPBACK,UP> mtu 16436 qdisc noqueue
link/loopback 00:00:00:00:00:00 brd 00:00:00:00:00:00

                     ^^^^^^^^^^^^^^^^
Quoted 1 line
   inet 127.0.0.1/8 scope host lo

This looks as though there is no netmask set. My configuration shows:

BS. 00:00:00:00:00:00's above aren't netmasks.

Quoted 6 lines

lo Link encap:Local Loopback

           inet addr:127.0.0.1  Mask:255.0.0.0
           inet6 addr: ::1/128 Scope:Host
           UP LOOPBACK RUNNING  MTU:16436  Metric:1

This is a possible configuration error.

Yours is ifconfig output, whereas "link/loopback 00:00:00:00:00:00 brd 00:00:00:00:00:00" line above was from ip, not ifconfig.
--
vda

#6 lore raw
In reply to Michael Tokarev
Quoted 3 lines

The problem is, I can't see what is causing this misconfiguration or whatever. I wasn't able to capture such a packet so far either -- it never happened while tcpdump was running.

You may try to add printk("bad boy is: %s\n", current()->comm)), or dump_stack(), or something like that in icmp path of IP stack. (I am currently tracking an intermittent network problem on my home box in similar way).

Quoted 4 lines

Note the local IP address mentioned is different, I've
seen 3 so far, all 3 are local on this box and are on 3 different (ethernet) interfaces (but the ICMP always comes from lo).

BTW what tcpdump actually shows?
--
vda

#7 lore raw
Quoted 9 lines · Denis Vlasenko
Quoted 3 lines

The problem is, I can't see what is causing this misconfiguration or whatever. I wasn't able to capture such a packet so far either -- it never happened while tcpdump was running.

You may try to add printk("bad boy is: %s\n", current()->comm)), or dump_stack(), or something like that in icmp path of IP stack. (I am currently tracking an intermittent network problem on my home box in similar way).

Quoted 9 lines
Quoted 4 lines

Note the local IP address mentioned is different, I've
seen 3 so far, all 3 are local on this box and are on 3 different (ethernet) interfaces (but the ICMP always comes from lo).

BTW what tcpdump actually shows?
--
vda

Please read line 1148 of /usr/src/linux-uname -r/net/ipv4/icmp.c. The error message is in response to a bogus frame. The NETMASK must match at both ends (a configuration error). If the mask matches, then you can turn off the messages with 'sysctl', i.e., .... echo "1" >/proc/sys/net/icmp_ignore_bogus_error_responses. Otherwise, find the interface netmask that's broken and fix it.

Cheers,
Dick Johnson
Penguin : Linux version 2.6.12 on an i686 machine (5537.79 BogoMips).
Notice : All mail here is now cached for review by Dictator Bush.
98.36% of all statistics are fiction.