sent an invalid ICMP type 11, code 0 error to a broadcast: 0.0.0.0 on lo?
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
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 timeskernel: 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 0I'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.
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
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
lois 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 loThere 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 loOnly 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
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
lois configured properly, i.e.Yes. More,
rp_filteris 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 loThere 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:1This is a possible configuration error.
Quoted 10 lines
2: eth0: <BROADCAST,MULTICAST,UP> mtu 1500 qdisc
pfifo_fastqlen 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.
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 loThis 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:1This 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
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
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)), ordump_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.
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 loThis 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:1This 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