arno-iptables-firewall / aif Goto Github PK

In AstLinux we recently updated iptables to version 1.4.21, and very rarely, usually a background plugin can generate either error below:

/usr/sbin/iptables: (4) Another app is currently holding the xtables lock.
Perhaps you want to use the -w option?

or

/usr/sbin/ip6tables: (4) Another app is currently holding the xtables lock.
Perhaps you want to use the -w option?

Digging deeper, this change occurred with iptables 1.4.20, from Changelog...
ip[6]tables: Add locking to prevent concurrent instances

Discussion and preliminary patch:
http://patchwork.ozlabs.org/patch/246619/

In my limited testing, I can forcibly generate the error, and adding -w seems to gracefully work.

We need to automatically handle this in AIF since it could mean a rule does not get applied while the firewall is being built.

Fortunately, adding a -w is only needed in 4 lines in environment.

grep 'IP[46]TABLES ' /usr/share/arno-iptables-firewall/environment

  result=`$IP4TABLES "$@" 2>&1`
  result=`$IP6TABLES "$@" 2>&1`
  $IP4TABLES "$@" >/dev/null 2>&1
  $IP6TABLES "$@" >/dev/null 2>&1

Checking for iptables-1.4.20+ is probably the most work :-)

And much testing...

Not an issue, more of a request, if you think possible.
My PC that I use as my gateway server, with arno firewall installed also acts as my dhcp server. I use mac address filtering, so I know what devices can go through the firewall on the clean side. However I can't stop anyone not in the mac-address list from getting an IP address in the first place. Apparently dhcp is at a lower level than ip, so the firewall doesn't block the request for an ip address. But I think ebtables could solve that issue, e.g. ebtables -A INPUT -s xx:xx:xx:xx:xx:xx -j DROP

Is this something you could think about adding?

I have an issue while setting it up

Which TCP ports do you want to allow from the internet? (eg. 22=SSH, 80=HTTP, etc.) (comma separate multiple ports)? 20,21,22,80,443,1935,8080,10000,10051,10050

then I get

Allowing ANYHOST for TCP port(s): 20,21,22,80,443,1935,8080.10000,10050,10051 /sbin/iptables -A EXT_INPUT_CHAIN -i + -d 0/0 -p tcp --dport 8080.10000 -j ACCEPT ERROR (2): iptables v1.6.0: invalid port/service 8080.10000' specified
Try iptables -h' or 'iptables --help' for more information. /sbin/ip6tables -A EXT_INPUT_CHAIN -i + -d 0/0 -p tcp --dport 8080.10000 -j ACCEPT ERROR (2): ip6tables v1.6.0: invalid port/service 8080.10000' specified
Try ip6tables -h' or 'ip6tables --help' for more information.

I don't know why the comma is translated to a point. As a result I cannot access webmin

Just wanted to point out that the guided configuration script fails on my system. It seems to look for /etc/init.d/arno-iptables-firewall and reports aif "is not installed on this system".

My system is an Arch Linux using systemd, so I have no init.d. To bypass the sanity_check() function I have to fake the directory like this: mkdir -p /etc/init.d/arno-iptables-firewall

Hi!
Id like to request an IPv6 nat feature if possible.

My business internet connection finally got native IPv6 using DHCPv6-PD and ran across an AIF configuration issue.

In my case, IPv4 is statically assigned, so EXT_IF_DHCP_IP=0 .

But, in order to receive the return "dhcp6 advertise" in response to my "dhcp6 solicit", AIF needs to allow -p udp --sport 547 --dport 546.

Currently the code is:

  # Here we add support for DHCP assigned IP
  ##########################################
  if [ "$EXT_IF_DHCP_IP" = "1" ]; then
    echo " Enabling support for DHCP-assigned-IP (DHCP client)"
    # Allow this host to be an DHCP client:
    ip4tables -A EXT_INPUT_CHAIN -p udp --sport 67 --dport 68 -j ACCEPT
    if [ "$IPV6_SUPPORT" = "1" ]; then
      # Allow this host to be an DHCPv6 client:
      ip6tables -A EXT_INPUT_CHAIN -p udp --sport 547 --dport 546 -j ACCEPT
    fi
  fi

It is not ideal to set EXT_IF_DHCP_IP=1 in this case.

I propose we add a EXT_IF_DHCPV6_IPV6 variable, as such:

  # Here we add support for DHCP assigned IP
  ##########################################
  if [ "$EXT_IF_DHCP_IP" = "1" ]; then
    echo " Enabling support for DHCP-assigned-IP (DHCP client)"
    # Allow this host to be an DHCP client:
    ip4tables -A EXT_INPUT_CHAIN -p udp --sport 67 --dport 68 -j ACCEPT
    if [ "$IPV6_SUPPORT" = "1" ]; then
      # Allow this host to be an DHCPv6 client:
      ip6tables -A EXT_INPUT_CHAIN -p udp --sport 547 --dport 546 -j ACCEPT
    fi
+  elif [ "$EXT_IF_DHCPV6_IPV6" = "1" ]; then
+    if [ "$IPV6_SUPPORT" = "1" ]; then
+      # Allow this host to be an DHCPv6 client via Link-Local:
+      ip6tables -A EXT_INPUT_CHAIN -s fe80::/10 -p udp --sport 547 --dport 546 -j ACCEPT
+    fi
  fi

While it would be a change, do we want to also limit EXT_IF_DHCP_IP=1 DHCPv6 client to Link-Local only as well ? If so we could combine them in one test.

Thoughts @arnova ?

Is this firewall better than csf firewall? If so, in what ways please could someone explain. Also, does it totally blocks drdos attacks?

this program is in need of ipset intervention in addition to iptables. Loading and saving times will be dramatically faster as you can load and save all sets with 1 command. ipset save -f something.txt
ipset restore -f something.txt

You could probably eliminate some the for next loops. Also the way your doing things a list of 20000 ip's, which I have 88000+ would cause the CPU to run at 100% and be slow. You add 1 iptables rule for every IP and with ipset you only need 1 iptables entry for any number of ip's.

ipset -! -N ban_ip hash:net counters
iptables -A INPUT -m set --set ban_ip src -j DROP
OR
ipset -N ban_ip hash:net counters timeout 300 comment
300 is just the default

ipset add ban_ip 1.1.1.1 adds to the list
OR
ipset add ban_ip 1.1.1.1 timeout 60
OR
ipset add ban_ip 1.1.1.1 timeout 60 comment "evil doers r us"

iptables -A ban_ip -j SET --add-set banned_hosts src

The release info on Freecode stops at version 2.0.1b on 19 March 2012. It makes the project look abandoned to the uninitiated.

Hello

I can't start aif, either the package or the git I get from here. try my dedicated as well as a small vps, it's just not working

Apr 21 00:47:37 vps665030 arno-iptables-firewall[13368]: Logging of stealth scans (nmap probes etc.) enabled
Apr 21 00:47:37 vps665030 arno-iptables-firewall[13368]: Logging of INVALID TCP packets disabled
Apr 21 00:47:37 vps665030 arno-iptables-firewall[13368]: Logging of INVALID UDP packets disabled
Apr 21 00:47:37 vps665030 arno-iptables-firewall[13368]: Logging of INVALID ICMP packets disabled
Apr 21 00:47:37 vps665030 arno-iptables-firewall[13368]: Dropping and logging of IPv4 fragmented packets disabled
Apr 21 00:47:37 vps665030 arno-iptables-firewall[13368]: Logging of access from reserved nets disabled
Apr 21 00:47:37 vps665030 arno-iptables-firewall[13368]: Reading custom rules from /etc/arno-iptables-firewall/custom-rules
Apr 21 00:47:37 vps665030 arno-iptables-firewall[13368]: Checking for (user) plugins in /usr/local/share/arno-iptables-firewall/plugins...
Apr 21 00:47:37 vps665030 arno-iptables-firewall[13368]: Loaded 0 plugin(s)...
Apr 21 00:47:37 vps665030 arno-iptables-firewall[13368]: Setting up external(INET) INPUT policy
Apr 21 00:47:37 vps665030 arno-iptables-firewall[13368]: Logging of ICMP flooding enabled
Apr 21 00:47:37 vps665030 arno-iptables-firewall[13368]: Enabling support for DHCP-assigned-IP (DHCP client)
Apr 21 00:47:37 vps665030 arno-iptables-firewall[13368]: Logging of explicitly blocked hosts inbound/outbound enabled
Apr 21 00:47:37 vps665030 arno-iptables-firewall[13368]: Logging of denied local output connections enabled
Apr 21 00:47:37 vps665030 arno-iptables-firewall[13368]: Packets will NOT be checked for reserved source addresses
Apr 21 00:47:37 vps665030 arno-iptables-firewall[13368]: Allowing ANYHOST for TCP port(s): 20,21,22,25,53,80,110,143,443,465,587,993,995,1935,3306,5901,8080,8081,8082,,8112,10000,10050,10051,58846
Apr 21 00:47:37 vps665030 arno-iptables-firewall[13368]: /sbin/iptables -A EXT_INPUT_CHAIN -d 0/0 -p tcp --dport -j ACCEPT
Apr 21 00:47:37 vps665030 arno-iptables-firewall[13368]: ERROR (2): iptables v1.6.1: invalid port/service -j' specified Apr 21 00:47:37 vps665030 arno-iptables-firewall[13368]: Try iptables -h' or 'iptables --help' for more information.
Apr 21 00:47:37 vps665030 arno-iptables-firewall[13368]: Allowing ANYHOST for UDP port(s): 53
Apr 21 00:47:37 vps665030 arno-iptables-firewall[13368]: Allowing ANYHOST to send IPv4 ICMP-requests (ping)
Apr 21 00:47:37 vps665030 arno-iptables-firewall[13368]: Logging of possible stealth scans enabled
Apr 21 00:47:37 vps665030 arno-iptables-firewall[13368]: Logging of (other) packets to PRIVILEGED TCP ports enabled
Apr 21 00:47:37 vps665030 arno-iptables-firewall[13368]: Logging of (other) packets to PRIVILEGED UDP ports enabled
Apr 21 00:47:37 vps665030 arno-iptables-firewall[13368]: Logging of (other) packets to UNPRIVILEGED TCP ports enabled
Apr 21 00:47:37 vps665030 arno-iptables-firewall[13368]: Logging of (other) packets to UNPRIVILEGED UDP ports enabled
Apr 21 00:47:37 vps665030 arno-iptables-firewall[13368]: Logging of IGMP packets enabled
Apr 21 00:47:37 vps665030 arno-iptables-firewall[13368]: Logging of dropped ICMP-request(ping) packets enabled
Apr 21 00:47:37 vps665030 arno-iptables-firewall[13368]: Logging of dropped other ICMP packets enabled
Apr 21 00:47:37 vps665030 arno-iptables-firewall[13368]: Logging of other IP protocols (non TCP/UDP/ICMP/IGMP) packets enabled
Apr 21 00:47:37 vps665030 arno-iptables-firewall[13368]: Setting up external(INET) OUTPUT policy
Apr 21 00:47:37 vps665030 arno-iptables-firewall[13368]: Applying external(INET) policy to interface: ens3
Apr 21 00:47:37 vps665030 arno-iptables-firewall[13368]: Security is ENFORCED for external interface(s) in the FORWARD chain
Apr 21 00:47:37 vps665030 arno-iptables-firewall[13368]: Logging of dropped FORWARD packets enabled
Apr 21 00:47:37 vps665030 arno-iptables-firewall[13368]: Apr 21 00:47:37 WARNING: 1 firewall rules failed to apply!
Apr 21 00:47:37 vps665030 firewall[13881]: ** WARNING: 1 firewall rules failed to apply! **
Apr 21 00:47:37 vps665030 arno-iptables-firewall[13368]:
Apr 21 00:47:37 vps665030 systemd[1]: arno-iptables-firewall.service: Main process exited, code=exited, status=1/FAILURE
Apr 21 00:47:37 vps665030 systemd[1]: arno-iptables-firewall.service: Failed with result 'exit-code'.
Apr 21 00:47:37 vps665030 systemd[1]: Failed to start Arno's Iptables Firewall(AIF).
-- Subject: Unit arno-iptables-firewall.service has failed
-- Defined-By: systemd
-- Support: http://www.ubuntu.com/support

-- Unit arno-iptables-firewall.service has failed.

-- The result is RESULT.

For many years I have had an annoying, but hard to reproduce issue with a SIP ATA (Analog Terminal Adapter) behind NAT.

The ATA is a Grandstream (not my choice) specifically configured for my home alarm system for remote monitoring. The alarm provider automatically provisions it.

The problem is the ATA looses it's register when AstLinux using AIF reboots, but only 1 (or 2) out of 10 times. This has been an issue for many years, various AIF versions and iptables versions.

The problem seems to be an outbound UDP packet is generated by the ATA before the firewall is completely built, causing a stuck inappropriate conntrack state.

Normally the connection state should look like this with a 3:00 TTL

10.10.22.40  5060  184.73.xxx.xxx  5060  UDP  2:47

When it fails the state only has a 1:00 TTL

10.10.22.40  5060  184.73.xxx.xxx  5060  UDP  0:57

My hunch is this occurs when the UDP packet is allowed into the INPUT chain but before the NAT POSTROUTING rules are applied.

With the incorrect state, fixing it is a pain, either unplugging the ATA for 1 minute, (wait for the conntrack state to expire) or playing iptables games to block 10.10.22.40 UDP, reboot the ATA, wait 1 minute, then allow 10.10.22.40 UDP in.

I suspect this issue is only noticed for devices that force a fixed source and destination port so any retries always hits the same conntrack state.

Anyone see this ?

I have managed to get OpenVPN working with AIF by setting INT_IF to tun+ as below. The AIF machine is sitting in the DMZ of consumer router. I also want to forward port 55538 (for incoming connection from host 1.2.3.4) to port 22 on a LAN host (10.0.0.14), which I thought I'd be able to do using NAT_FORWARD_TCP as below:

EXT_IF="eth0"
EXT_IF_DHCP_IP="0"
INT_IF="tun+"
INTERNAL_NET="10.8.0.0/24"
NAT=1
NAT_INTERNAL_NET="10.8.0.0/24"
NAT_LOCAL_REDIRECT=1
NAT_FORWARD_TCP="1.2.3.4~55538>10.0.0.14~22"
IF_TRUSTS="tun+"
FULL_ACCESS_HOSTS="10.0.0.0/24,1.2.3.4"
OPEN_UDP="1194"

However, I'm getting Dropped FORWARD packet in the logs:
AIF:Dropped FORWARD packet: IN=eth0 OUT=eth0 MAC=themacaddress SRC=1.2.3.4 DST=10.0.0.14 LEN=60 TOS=0x00 PREC=0x00 TTL=51 ID=15573 DF PROTO=TCP SPT=35616 DPT=55538 WINDOW=29200 RES=0x00 SYN URGP=0.

Please can anyone offer any advice about getting this to work?

Thanks

Nick

Currently we set the environment & plugin bin path from firewall.conf. This causes problems when eg. switching between deb & tar.gz versions. I can't recall why we ever had this as an option but I think it's safe to assume we can autodetect where to find those files?

Just though I'd drop a note to say there seems to be some issues with AIF and Red Hat EL 8. For example, when an IP from the FULL_ACCESS_HOSTS list is processed the call to "get_ifs " results in "interfaces" being set to "+". On CentOS 7 (basically Red Hat EL 7) the following rule is set:

iptables -A EXT_INPUT_CHAIN -i + -s -d 0/0 -j ACCEPT

and "iptables -S" shows the following so that is allowed through:

-A EXT_INPUT_CHAIN -s /32 -j ACCEPT

In this case the version of iptables is 1.4.21.

On Red Hat EL 8 (iptables 1.8.0) the same procedure as above results in "iptables -S" showing the following:

-A EXT_INPUT_CHAIN -s /32 -i + -j ACCEPT

where is just some non-ASCII jumble which does not allow through. This was tested using AIF 2.0.3.

It appears that AIF is relying on the fact that "-i +" used to behave as "any interface", but it seems this is no longer the case. Indeed, the documentation for iptables is ambiguous in this respect:

[!] -i, --in-interface name
Name of an interface via which a packet was received (only
for packets entering the INPUT, FORWARD and PREROUTING chains).
When the "!" argument is used before the interface name, the
sense is inverted. If the interface name ends in a "+", then
any interface which begins with this name will match. If this
option is omitted, any interface name will match.

so it's not clear what should happen if the interface name is null (the man page hasn't changed in this regard between 1.4.21 and 1.8.0).

It's true that Red Hat EL 8 is still in beta, but this seems to be more of an issue with how iptables is now parsing the "-i" setting, and might be worth looking into.

Is there any config options to open for example 80,443 to the world, but only on a specific interface?
In the case where you have eth0 eth0:1 eth0:2 with 3 different IPs, I'd like only eth0:2 and its IP to be exposed to the world, not the two other IPs.

So basically have a OPEN_TCP="21,22,1919,6363,7777,19922" as usual with an addition of:
OPEN_TCP_INT=80,443-local.ip2 1200-local.ip0"

Or any other current way of doing this?

-Osi

@arnova A confusing issue can up with an AstLinux user, using AIF as a router:

It was desired to NAT port TCP 8443 to port 443 on the "Local" router, so the destination address was set to 192.168.0.1, a LAN gateway interface ...

NAT_LOCAL_REDIRECT=1
NAT_FORWARD_TCP="1.2.3.4~8443>192.168.0.1~443"

This worked as expected, but this LOG was always generated:

Jul 25 10:06:35 pbx3 user.info kernel: AIF:PRIV TCP packet: IN=eth0 OUT= MAC=00:30:18:... SRC=1.2.3.4 DST=192.168.0.1 LEN=64 TOS=0x00 PREC=0x00 TTL=64 ID=16302 DF PROTO=TCP SPT=58305 DPT=443 WINDOW=65535 RES=0x00 SYN URGP=0

Normally NAT_FORWARD rewrites the packets to pass through the FORWARD chain, but in this case the IPv4 address "192.168.0.1" is associated with "Local" so it is passed via the INPUT chain. A rule added to the EXT_FORWARD_IN_CHAIN chain to normally allow this rewritten packet is not used.

After some headscratching, the NAT_LOCAL_REDIRECT=1 adds this to the EXT_INPUT_CHAIN chain:

ACCEPT     all  --  0.0.0.0/0            0.0.0.0/0            ctstate DNAT

which allows DNAT'ed packets to be allowed in the INPUT chain, but this rule is added after the 'AIF:PRIV TCP packet' log (and others) so even if the DNAT'ed packet is allowed the log is generated.

I'm proposing to move the NAT_LOCAL_REDIRECT=1 conditional up, After "# Handle multicast traffic" and before "# Log packets to privileged TCP ports?"

Seem reasonable ?

Here is the dump:

[root@rees log]# service arno-iptables-firewall start
Starting Arno's Iptables Firewall...
/sbin/iptables -A HOST_BLOCK_DST_DROP -m limit --limit 1/m --limit-burst 1 -j LOG --log-level info --log-prefix AIF:Blocked host(s) Outbound:
ERROR (2): iptables v1.4.7: Maximum prefix length 29 for --log-prefix
Try `iptables -h' or 'iptables --help' for more information.
Nov 23 17:26:11 WARNING: 1 firewall rules failed to apply!

Hello

setting up a new server and after I installed aif, I can't connect via ftp anymore. it says Error: Failed to retrieve directory listing

I opened port 21 though, what is the cause of this

thanks

I am trying to perform an unattended install (i.e. without user interaction) of arno-iptables-firewall.
I have setup predefined answers using debconf-set-selections before running the apt-get install -q -y arno-iptables-firewall but the debconf answer are ignored.
Is it on purpose? what is the supported way to perform unattended setup on Ubuntu?

Hello,

Is there any way to block p2p traffic?

thank you!

I installed arno-iptables-firewall & answered the questions during setup, but its status says it's dead. Is there more I need to do?

$ sudo service arno-iptables-firewall status
● arno-iptables-firewall.service - Arno's Iptables Firewall                                                                                            
   Loaded: loaded (/lib/systemd/system/arno-iptables-firewall.service; disabled; vendor preset: enabled)                                               
   Active: inactive (dead)

I have a multi-interface firewall that I use a .conf file to configure the rules. I'm trying to setup something like this:

Instead of using 10.10.10.0/24>192.168.1.0/27~3389

I'd like to use eth0>eth1~3389. Is that possible? How would I go about setting this up to use interface aliases instead of network addresses?

Currently all LAN traffic (INT_IF interfaces) is allowed to the DMZ network(s)

# Apply policy for LAN->DMZ (allow all)
iptables -A FORWARD -i $iif -o $dif -j ACCEPT

There are situations where a particular LAN should not have access to the DMZ.

One solution is with interface granularity, add a DMZ_DENY_INT_IF variable that contains INT_IF interfaces that should not be allowed to the DMZ, example pseudo-code:

# Apply policy for LAN->DMZ
if $iif is in $DMZ_DENY_INT_IF
iptables -A FORWARD -i $iif -o $dif -j DROP
else
iptables -A FORWARD -i $iif -o $dif -j ACCEPT

Another approach is general filtering granularity, add a LAN_DMZ_FORWARD_CHAIN which defaults to one rule of ACCEPT

# Apply policy for LAN->DMZ (default to allow)
iptables -A FORWARD -i $iif -o $dif -j LAN_DMZ_FORWARD_CHAIN

then add support for the following rules:

LAN_DMZ_HOST_DENY_TCP=""
LAN_DMZ_HOST_DENY_UDP=""
LAN_DMZ_HOST_DENY_IP=""

which would be applied with a new function setup_lan_dmz_forward_chain() adding defined DROP rules before the default ACCEPT in the new LAN_DMZ_FORWARD_CHAIN.

@arnova Thoughts ?

using dhcpcd with prefix delegation set for internal interfaces causes ipv6nd_sendadvertisement: Operation not permitted error.

example: (enp3s0 is the public interface)

interface enp3s0
ipv6rs # enable routing solicitation get the
ia_na 0 # request an IPv6 address
ia_pd 1 enp3s0/1
ia_pd 2 enp2s0/2 wlx9cefd5febaca/3

private interfaces enp2s0 and wlx9cefd5febaca do NOT receive a pd.

ip6 prefixes delegated by dhcpcd to two internal networks do not appear to function correctly.

ETH1,ETH0 = Internal
ETH2=external.

only one interface can properly route ipv6 through the firewall. ipv4 traffic functions as expected.

For example, the OPEN_TCP setting allows settings like 1.2.3.4,5.6.7.8#80 to open port 80 on the local IPs 1.2.3.4 and 5.6.7.8.
However, the equivalent for IPv6 (eg. fe80::1:1#80, ignore that this is a link-local, I was using real public addresses when trying this) doesn't appear to work and generates commands like /sbin/ip6tables -A EXT_INPUT_CHAIN -i fe80::1:1 -d 0/0 -p tcp --dport 80 -j ACCEPT, where it seems to be confused about what the interface and destination IPs are.

More examples to make the format of IPv6 settings clear would be great.

Thanks.

An AstLinux integrator reported a German customer moved from ADSL to VDSL and is now logging every 30 secs the following:

May 23 10:27:18 router user.info kernel: AIF:Spoofed packet: IN=eth0 OUT= MAC=01:00:5e:00:xx:xx:xx:80:df:43:cf:xx:xx:xx SRC=192.168.4.2 DST=224.0.0.1 LEN=36 TOS=0x00 PREC=0x00 TTL=1 ID=2733 PROTO=2

A solution is to add this to custom-rules:

echo "[CUSTOM RULE] Disable antispoof checking for IGMP packets"
ip4tables -I SPOOF_CHK -p 2 -j RETURN

and that stops the spoof logging.

@arnova is this something we should add by default to the SPOOF_CHK chain ?

Hi,

yesterday the iptables update 1.8.1 came in to debian buster. With the new iptables version, something in the rule behaviour seem to have changed.

All outgoing packets fall into the the DROP policy and never reach the EXT_OUTPUT_CHAIN (like before).
Chain OUTPUT (policy DROP 0 packets, 0 bytes)
pkts bytes target prot opt in out source destination
15 1221 BASE_OUTPUT_CHAIN all -- * * 0.0.0.0/0 0.0.0.0/0
1 60 TCPMSS tcp -- * wlp3s0 0.0.0.0/0 0.0.0.0/0 tcp flags:0x06/0x02 TCPMSS clamp to PMTU
15 1221 OUTPUT_CHAIN all -- * * 0.0.0.0/0 0.0.0.0/0
15 1221 HOST_BLOCK_DST all -- * * 0.0.0.0/0 0.0.0.0/0
5 389 LOG all -f * * 0.0.0.0/0 0.0.0.0/0 limit: avg 3/min burst 5 LOG flags 0 level 6 prefix "AIF:Fragment packet: "
13 1117 DROP all -f * * 0.0.0.0/0 0.0.0.0/0
0 0 EXT_OUTPUT_CHAIN all -- * wlp3s0 0.0.0.0/0 0.0.0.0/0

Reverting back to iptables 1.6.2-1.1, everything worked fine again.

Affects the debian stock package for arno-iptables-firewall, as well as the git master of this repository.

As of Kernel 4.7, AIF users are seeing this message:

nf_conntrack: default automatic helper assignment has been turned off for security reasons and CT-based  firewall rule not found. Use the iptables CT target to attach helpers instead.

Ref: https://askubuntu.com/questions/906197/weird-iptables-log-messages-after-upgraded-to-server-zesty-17-04

It appears by default the only nf_conntrack_<helper> AIF enables is "ftp". "irc" is a firewall.conf option and "sip" is in an optional plugin.

An excellent reference is here:

"Secure use of iptables and connection tracking helpers"
https://home.regit.org/netfilter-en/secure-use-of-helpers/

BTW, this should not be confused with the nf_nat_* helpers, AFAIK.

I don't have any commit ideas at this point, but we need to get a handle on this.

Is it possible to add nf_nat_pptp module in load_modules function of main arno-iptables-firewall? Because without this module, users behind the FW could not connect to MS VPN server with MS VPN Client 😞.

When the multiroute.conf is enabled, I received an error in /var/log/message.log
Error: either "to" is duplicate, or "equalize" is garbage

According : https://rocky.eld.leidenuniv.nl/pipermail/firewall/2013-July/002357.html
In 50multiroute.plugin I changed the line
ip route add default equalize
to
ip route add default scope global \

It appears to work perfectly, but not sure it's correct ?

Daniel

Hey, I have similar problem.
https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=778845;msg=2

Hi!
Is it possible somehow to block a range of IPs assigned to a specific country(ies), using your script, most preferable by supplying a list in cfg? There are examples of such :
https://www.cyberciti.biz/faq/block-entier-country-using-iptables/ , not involving dependence on GeoIP.

Due to the implementation of our current blocked-hosts batch loader, we drop all firewall traffic on force-reload because we temporarely set the fw policy to drop while reloading. Especially with large lists of blocked hosts this is undesireable. Perhaps we should fix this by using a passive and an active chain?

Seems my server is visited by a korean guy with a port scanner. The scanner goes at a very low pace (like 2 seconds/connection) but still is annoying to see the packets logged on journald. These are the packets:

Oct 23 23:54:17 host kernel: AIF:UNPRIV TCP packet: IN=vmbr0 OUT= MAC=52:ec:95:9f:fc:28:18:55:0f:b3:ce:71:08:00 SRC=112.175.124.2 DST=192.168.178.4 LEN=40 TOS=0x08 PREC=0x20 TTL=67 ID=47713 DF PROTO=TCP SPT=45986 DPT=3000 WINDOW=29200 RES=0x00 SYN URGP=0
Oct 23 23:54:21 host kernel: AIF:PRIV TCP packet: IN=vmbr0 OUT= MAC=52:ec:95:9f:fc:28:18:55:0f:b3:ce:71:08:00 SRC=112.175.124.2 DST=192.168.178.4 LEN=40 TOS=0x08 PREC=0x20 TTL=63 ID=64892 DF PROTO=TCP SPT=49086 DPT=53 WINDOW=29200 RES=0x00 SYN URGP=0
Oct 23 23:54:22 host kernel: AIF:UNPRIV TCP packet: IN=vmbr0 OUT= MAC=52:ec:95:9f:fc:28:18:55:0f:b3:ce:71:08:00 SRC=112.175.124.2 DST=192.168.178.4 LEN=40 TOS=0x08 PREC=0x20 TTL=60 ID=29017 DF PROTO=TCP SPT=40306 DPT=3000 WINDOW=29200 RES=0x00 SYN URGP=0
Oct 23 23:54:30 host kernel: AIF:PRIV TCP packet: IN=vmbr0 OUT= MAC=52:ec:95:9f:fc:28:18:55:0f:b3:ce:71:08:00 SRC=112.175.127.189 DST=192.168.178.4 LEN=40 TOS=0x08 PREC=0x20 TTL=59 ID=28435 DF PROTO=TCP SPT=37831 DPT=53 WINDOW=29200 RES=0x00 SYN URGP=0
Oct 23 23:54:37 host kernel: AIF:UNPRIV TCP packet: IN=vmbr0 OUT= MAC=52:ec:95:9f:fc:28:18:55:0f:b3:ce:71:08:00 SRC=112.175.127.189 DST=192.168.178.4 LEN=40 TOS=0x08 PREC=0x20 TTL=55 ID=8847 DF PROTO=TCP SPT=37568 DPT=3000 WINDOW=29200 RES=0x00 SYN URGP=0

Is there any way to prevent those packets from being logged? I have already set to zero all LOG settings, but I am still getting them.

lan-inet rules should have nolog settings for reject/deny

Unlike DHCPv4, DHCPv6 does not provision the default gateway for a client. Devices running DHCPv6 on the WAN need to be learning and installing a default gateway based on router advertisements.

Currently AIF sets accept_ra=0 when forwarding=1, which is not needed since with forwarding=1 the kernel ignores RA's when accept_ra=0 or accept_ra=1

After Linux kernel 2.6.35, accept_ra=2 was supported to override this behavior when forwarding=1 to selectively allow a DHCPv6 client on a WAN interface.

I propose we simply remove the line where accept_ra is set to 0 when forwarding=1, this would allow external agents to set accept_ra=2 as needed for DHCPv6-client. Sysctl accept_ra=2 should only be done for interfaces that require it (not all EXT_IF's) and is often needed before AIF is called, this should be left to an outside agent, and AIF should simply leave accept_ra alone for the forwarding=1 case.

Ref: http://www.mattb.net.nz/blog/2011/05/12/linux-ignores-ipv6-router-advertisements-when-forwarding-is-enabled/

I try to make openSUSE RPM for Arno's iptables firewall script

But get than nemysis user for openSUSE Build Service can't make RPM. Tthat works only that root. This is not usual.

Please make one Makefile for installation and configure when needed.

Hello,

i'am Running some VPS with Ubuntu 14.04. I've Installed this Firewall with a script to fill the blocked-host file. Everything runs Perfect.

So now i have tried it on Ubuntu 15.10 an 16.04 64bit Server.
Everything runs Perfect, till i filled up the blocked-host file. I grep much sites like blocklist.de all IPs or Ranges, so the File has 30K IPs. On Ubuntu 14.04 restart takes under a minute. In 15.10 or 16.04, the newest, the start/restart takes 15 Minutes. After reading the blocked-hosts file. It takes much time to start. The System is on 100% cpu.

I Hope anyone understand me and can help

I was reading the description for this commit where Eric Dumazet explains why /proc/sys/net/ipv4/ip_local_port_range was changed to 32768-60999 instead of 32768-61000.

LOCAL_PORT_RANGE in AIF's .conf file still defaults to 61000 for the last local port number, should we change it accordingly?

Hi guys, I've recently purchased a Netgear Nighthawk D7000v2 modem/router, and it's fantastic. I've logged into it via telnet and added the following:

/etc/arno-iptables-firewall/plugins/nat_loopback.conf, and configured it with ENABLED=1, set the proper IPv4 internal net.

I added the binary for the nat_loopback to the binary_dir set in firewall.conf:

PLUGIN_BIN_PATH="/usr/local/share/arno-iptables-firewall/plugins" << here
PLUGIN_CONF_PATH="/etc/arno-iptables-firewall/plugins" (here is where the .conf is)

I restart the firewall, and it flushes everything. It says Unloaded 0 user plugins, even after it states that it has loaded it:

Checking for (user) plugins in /usr/local/share/arno-iptables-firewall/plugins...
NAT Loopback plugin v1.02
NAT Loopback default IPv4: 169.254.25.54
NAT Loopback internal net(s): 192.168.1.0/24
NAT Loopback local forwards are enabled.
Loaded 1 plugin(s)...

Every time I restart the firewall, it says Unloaded 0 user plugins. However, every time it starts, it states the above message saying the Plugin has been loaded.

status-plugins doesn't show the plugin status. It just shows:

/etc/arno-iptables-firewall # arno-iptables-firewall status-plugins
Arno's Iptables Firewall Script v2.0.1f

Showing status of (user) plugins:

/etc/arno-iptables-firewall #

And consequently, the loopback-nat functionality isn't working. When I try to open the URL that links to my externally forwarded server, it is still trying to connect to the router (because it says connection refused, and this specific port isn't a port that would be used by my router).

Could you please give me some pointers as to where I'm going wrong? thank you so very much.
Kindest regards, Anthony

By default, LAN's are isolated from each other, a good thing for example for a guest WiFi VLAN.

But in special situations it may be desirable to allow certain LAN's access to specific network resources, such as network printers, possibly advertised using mDNS using avahi and 'enable-reflector=yes'.

The problem is currently AIF only supports TRUSTED_IF and IF_TRUSTS beyond the isolated default. All or nothing.

I propose an additional chain LAN_LAN_FORWARD_CHAIN and set of rules LAN_LAN_HOST_OPEN_xxx, modeled after DMZ_LAN_HOST_OPEN_xxx to selectively allow LAN to LAN traffic by src, dst and port.

I propose a new function setup_lan_lan_forward_chain() implementing the new rules which would be called...

    # Setup helper chain for the LAN:
    setup_lan_inet_forward_chain;
    setup_lan_lan_forward_chain;  <<< added <<<

    IFS=' ,'
    for iif in $INT_IF; do
      echo "Applying internal(LAN) FORWARD policy to interface: $iif"

      # Always make subnets on the SAME interface trust each other
      iptables -A FORWARD -i $iif -o $iif -j ACCEPT

      for eif in $EXT_IF; do
        iptables -A FORWARD -i $iif -o $eif -j LAN_INET_FORWARD_CHAIN
      done

    <<< hook in LAN_LAN_FORWARD_CHAIN using INT_IF permutations >>>
    done

@arnova do you agree ?

I'm running Raspbian Jessie with systemd on board, and I noticed that install.sh fails to install arno-iptables-firewall.service into /etc/systemd/system.

Shouldn't
# Install service file if systemd directory is available if [ -d "/usr/lib/systemd/system/" ]; then copy_ask_if_exist ./lib/systemd/system/arno-iptables-firewall.service /usr/lib/systemd/system/ fi
be
# Install service file if systemd directory is available if [ -d "/etc/systemd/system/" ]; then copy_ask_if_exist ./lib/systemd/system/arno-iptables-firewall.service /etc/systemd/system/ fi
(at least for Raspbian)

Thanks, otherwise works marvellously. A way out of CSF...

Nick

Starting with Linux kenel 3.7 ip6tables gained support for IPv6 NAT, iptables added support with v1.4.17 in 2012-Dec-25.

Regardless of the good/bad of IPv6 NAT, AIF should be flushing and deleting all ip6tables -t nat rules as we do for ip4tables -t nat.

BTW, the NETMAP target is very useful for IPv6 unique local address (ULA) routing. A work in progress NPTv6 AIF plugin is here:
https://sourceforge.net/p/astlinux/code/HEAD/tree/branches/1.0/package/arnofw/net-prefix-translation/

So, what is the best way to handle this, do we test for IPv6 NAT support and if/then in various places, or simply use try_ip6tables so any errors are ignored. These actions only occur when AIF is started/stopped/restarted.

I'm suggesting using try_ip6tables, ex.

try_ip6tables -t nat -F
try_ip6tables -t nat -X

@arnova agree ?

I am using a Debian 4.9 based virtual machine and arno does not start with the system. It was installed with apt.

I have it setup to use an OpenVPN tunnel interface for it's upstream port and the second ethernet device as the downstream port.

Is there an issue with the tunnel not being setup by the time the firewall goes up so it fails?

I started getting an error when starting my firewall.

Arno's Iptables Firewall Script v2.0.0c
-------------------------------------------------------------------------------
Platform: Linux 3.8.0-32-generic i686
WARNING: External interface eth0 does NOT exist (yet?)
WARNING: External interface tun0 does NOT exist (yet?)

My interfaces are are up and running.

When I isolate the check_interfaces and run a simple test it returns 1 not 0.

When I check

sudo bash -x /usr/sbin/arno-iptables-firewall restart 2>&1 |grep check_interface

it shows

+ check_interface eth0
+ check_interface tun0

here is my full verbose output of a restart:

�[40m�[1;32mArno's Iptables Firewall Script v2.0.0c�[0m
-------------------------------------------------------------------------------
Platform: Linux 3.8.0-32-generic i686
�[40m�[1;31mWARNING: External interface tun0 does NOT exist (yet?)�[0m
Stopping (user) plugins...
 SSH Brute-Force Protection plugin v1.1a
Checking/probing Iptables modules:
 Loaded kernel module ip_tables. 
 Loaded kernel module nf_conntrack. 
 Loaded kernel module nf_conntrack_ftp. 
 Loaded kernel module xt_conntrack. 
 Loaded kernel module xt_limit. 
 Loaded kernel module xt_state. 
 Loaded kernel module xt_multiport. 
 Loaded kernel module iptable_filter. 
 Loaded kernel module iptable_mangle. 
 Loaded kernel module ipt_REJECT. 
 Loaded kernel module ipt_LOG. 
 Loaded kernel module xt_TCPMSS. 
 Loaded kernel module xt_DSCP. 
 Loaded kernel module iptable_nat. 
 Module check done...
Setting the kernel ring buffer to only log panic messages to the console
Configuring general kernel parameters:
 Setting the max. amount of simultaneous connections to 16384
Configuring kernel parameters:
 Disabling send redirects
 Enabling protection against source routed packets
 Enabling packet forwarding
 Setting some kernel performance options
 Enabling reduction of the DoS'ing ability
 Enabling anti-spoof with rp_filter
 Enabling SYN-flood protection via SYN-cookies
 Disabling the logging of martians
 Disabling the acception of ICMP-redirect messages
 Setting default TTL=64
 Disabling ECN (Explicit Congestion Notification)
 Enabling kernel support for dynamic IPs
 Enabling PMTU discovery
 Flushing route table
 Kernel setup done...
Initializing firewall chains
 Setting all default policies to DROP while "setting up firewall rules"
IPv4 mode selected but IPv6 available, DROP all IPv6 packets
Using loglevel "info" for syslogd

Setting up firewall rules:
-------------------------------------------------------------------------------
Enabling setting the maximum packet size via MSS
Enabling mangling TOS
Logging of stealth scans (nmap probes etc.) enabled
Logging of packets with bad TCP-flags enabled
Logging of INVALID TCP packets disabled
Logging of INVALID UDP packets disabled
Logging of INVALID ICMP packets disabled
Logging of fragmented packets enabled
Logging of access from reserved addresses enabled
Reading custom rules from /etc/arno-iptables-firewall/custom-rules
Checking for (user) plugins in /usr/share/arno-iptables-firewall/plugins...
 SSH Brute-Force Protection plugin v1.1a
  Loaded kernel module xt_recent. 
  Allowing bypass of SSH protection checks for: malevich
  Protecting TCP port(s): 22
 Loaded 1 plugin(s)...
Setting up external(INET) INPUT policy
 Logging of ICMP flooding enabled
 Enabling support for DHCP-assigned-IP (DHCP client)
 Logging of explicitly blocked hosts enabled
 Logging of denied local output connections enabled
 Packets will NOT be checked for private source addresses
 Allowing ANYHOST for TCP port(s): 80
 Allowing ANYHOST for TCP port(s): 21
 Allowing ANYHOST for TCP port(s): 22
 Allowing ANYHOST for TCP port(s): 1194
 Allowing ANYHOST for TCP port(s): 8112
 Allowing ANYHOST for TCP port(s): 4040
 Allowing ANYHOST for TCP port(s): 10000
 Allowing ANYHOST for TCP port(s): 53
 Allowing ANYHOST for TCP port(s): 137
 Allowing ANYHOST for TCP port(s): 138
 Allowing ANYHOST for TCP port(s): 139
 Allowing ANYHOST for TCP port(s): 443
 Allowing ANYHOST for TCP port(s): 445
 Allowing ANYHOST for TCP port(s): 631
 Allowing ANYHOST for TCP port(s): 58846
 Allowing ANYHOST for TCP port(s): 873
 Allowing ANYHOST for TCP port(s): 17500
 Allowing ANYHOST for TCP port(s): 6566
 Allowing ANYHOST for TCP port(s): 50000:50200
 Allowing ANYHOST for TCP port(s): 8094
 Allowing ANYHOST for TCP port(s): 4444
 Allowing ANYHOST for TCP port(s): 23423
 Allowing ANYHOST for TCP port(s): 8895
 Allowing ANYHOST for TCP port(s): 8228
 Allowing ANYHOST for UDP port(s): 80
 Allowing ANYHOST for UDP port(s): 21
 Allowing ANYHOST for UDP port(s): 22
 Allowing ANYHOST for UDP port(s): 1194
 Allowing ANYHOST for UDP port(s): 8112
 Allowing ANYHOST for UDP port(s): 4040
 Allowing ANYHOST for UDP port(s): 10000
 Allowing ANYHOST for UDP port(s): 53
 Allowing ANYHOST for UDP port(s): 137
 Allowing ANYHOST for UDP port(s): 138
 Allowing ANYHOST for UDP port(s): 139
 Allowing ANYHOST for UDP port(s): 443
 Allowing ANYHOST for UDP port(s): 445
 Allowing ANYHOST for UDP port(s): 631
 Allowing ANYHOST for UDP port(s): 58846
 Allowing ANYHOST for UDP port(s): 873
 Allowing ANYHOST for UDP port(s): 17500
 Allowing ANYHOST for UDP port(s): 6566
 Allowing ANYHOST for UDP port(s): 50000:50200
 Allowing ANYHOST for UDP port(s): 8094
 Allowing ANYHOST for UDP port(s): 4444
 Allowing ANYHOST for UDP port(s): 1900
 Allowing ANYHOST for UDP port(s): 8228
 Allowing ANYHOST to send IPv4 ICMP-requests (ping)
 Logging of possible stealth scans enabled
 Logging of (other) packets to PRIVILEGED TCP ports enabled
 Logging of (other) packets to PRIVILEGED UDP ports enabled
 Logging of (other) packets to UNPRIVILEGED TCP ports enabled
 Logging of (other) packets to UNPRIVILEGED UDP ports enabled
 Logging of IGMP packets enabled
 Logging of dropped ICMP-request(ping) packets enabled
 Logging of dropped other ICMP packets enabled
 Logging of other IP protocols (non TCP/UDP/ICMP/IGMP) packets enabled
Setting up external(INET) OUTPUT policy
Applying external(INET) policy to interface: eth0 (without an external subnet specified)
Applying external(INET) policy to interface: tun0 (without an external subnet specified)
Security is LOOSENED for external interface(s) in the FORWARD chain!
 Logging of dropped FORWARD packets enabled

Dec 01 15:03:04 �[40m�[1;32mAll firewall rules applied.�[0m

I just committed some stuff which changes AIF's behavior at start. Previously we would start before network was up which caused all kinds of problems with eg. plugins which need to setup routes or try to resolve names etc. I've now changed the service file so that AIF is started right after network is up, but this leaves the host exposed temporarily. This window of opportunity is very small but we may want to fix this by eg. adding a helper service which simply inits all builtin iptables chains to DROP.

I was doing EDNS Compliance Tester today (https://ednscomp.isc.org/ednscomp) while IPv4 part was doing fine with tos mangling on or off, IPv6 was failing until i turned it off. Error was in PREROUTING chain. Also i notice PREROUTING and OUTPUT chains are using different set-tos options, not sure if its intentional or not, but anyway if i set ANY kind of ToS mangling for IPv6 on PREROUTING chain for udp 53 it fails last test with large options.

This is output with TOS mangling disabled:

domain.net. @84.255.xxx.xxx (ns.domain.net.): dns=ok edns=ok edns1=ok edns@512=ok ednsopt=ok edns1opt=ok do=ok ednsflags=ok docookie=ok,cookie edns512tcp=ok optlist=ok,expire,cookie,subnet
domain.net. @2a01:260:xxxx:xxxx::x (ns.domain.net.): dns=ok edns=ok edns1=ok edns@512=ok ednsopt=ok edns1opt=ok do=ok ednsflags=ok docookie=ok,cookie edns512tcp=ok optlist=ok,expire,cookie,subnet

While with TOS mangling enabled for DNS i get following:

domain.net. @84.255.xxx.xxx (ns.domain.net.): dns=ok edns=ok edns1=ok edns@512=ok ednsopt=ok edns1opt=ok do=ok ednsflags=ok docookie=ok,cookie edns512tcp=ok optlist=ok,expire,cookie,subnet
domain.net. @2a01:260:xxxx:xxxx::x (ns.domain.net.): dns=ok edns=ok edns1=ok edns@512=ok ednsopt=ok edns1opt=ok do=ok ednsflags=ok docookie=ok,cookie edns512tcp=ok optlist=timeout

It's now reasonably common for machines (e.g managed servers, virtual servers etc.) to get allocated a single IPv6 /64. Running containers on these servers is typically done with IPv4 NAT with port-forwarding, and IPv6 (with firewalling but without port forwarding). Ideally, the ISP would allocate larger than a /64, but unfortunately many don't.

A relatively straightforward way around this issue for IPv6 is to use a bridge to the physical network to pass IPv6 traffic. Loading the br_netfilter module then allows this bridged traffic to be firewalled. Arno currently doesn't support this configuration, but I use the following to workaround in custom-rules, but it would be great to have some support in aif. If there's interest I can look into developing a plugin?

# Allow all icmpv6 to allow IPv6 neighbour discovery to the machine behind the firewalled bridge (this should be improved on).
/sbin/ip6tables -A FORWARD -i br0 -o br0 -p icmpv6 -j ACCEPT
# Treat other traffic appropriately
/sbin/ip6tables -m physdev -A FORWARD -i br0  --physdev-in 'veth+' -j INT_FORWARD_OUT_CHAIN
/sbin/ip6tables -m physdev -A FORWARD -i br0  --physdev-out eth0 -j INT_FORWARD_IN_CHAIN
#  Stateful related will catch inbound replies...
/sbin/ip6tables -m physdev -A FORWARD -i br0  --physdev-in 'veth+' -j LAN_INET_FORWARD_CHAIN

Hi,

It's issue is probably due to my fault but, i suppose that my configuration is right.
ICMP requests does not work at all.

Here is my configuration file in case of mistake :

EXT_IF="eth0"
EXT_IF_DHCP_IP=0
INT_IF="lxcbr0"
INTERNAL_NET="172.4.1.0/24"
OPEN_ICMP=1
ECHO_IGNORE=0
ICMP_REDIRECT=1
OPEN_TCP="80, 443, 587, 143, 5000, 22, 9001"
OPEN_UDP="80, 443, 587, 143, 5000, 22, 9001"
NAT=1
NAT_INTERNAL_NET="172.24.1.0/24"
NAT_TCP_FORWARD="143,587>172.24.1.3 80,443>172.24.1.2"
NAT_UDP_FORWARD="143,587>172.24.1.3 80,443>172.24.1.2"  

If my fault, it could be interesting to add a section for ICMP configuration, because with a monitoring in front of a server, it's very usefull to allow it, especially without iptables skills.

Tell me if you need more info about this problem.

Great work, thanx.