Saturday, 5. December 2020Week 48

Using your own router on a Sunrise fiber line

Sunrise does not like when people run their own router on a fiber line. While they do not directly forbid it, they don't provide any of the required configuration parameters which makes it quite hard to use your own router.

Below you'll find the required configuration parameters to make your own router connect with IPv4 and IPv6 on a Sunrise fiber line.
Beware though that especially the VLAN configuration might be different depending on your city, the following worked for me in Zürich.

Also, please note that I do not recommend Sunrise as Internet provider (as a matter of fact I'm on the way out of their contract and switching to SolNet).
Besides not supporting to bring your own router, they also like to make up additional early-termination fees (the contract states 100CHF early termination fees, but once you call them to initiate the process they tell you that it's gonna cost 300CHF as they decided to change their pricing structure unilaterally).

Enough of the ranting, now to the interesting part :-)

The Sunrise line has multiple VLANs to differentiate between Internet, Phone and TV services.
To receive an IPv4 address it requires a special value for the Client Identifier DHCP option.
For IPv6 6rd is employed, for which we need to know the prefix and gateway address.

The following configuration was tested with a MikroTik CRS125 router starting from the default settings.
For simplicity I've named the network interfaces according to their intended usage (eg. LAN, sunrise and 6rd).

The first step is to configure the VLAN on top of your fiber interface. In my case it was VLAN ID 131, others were also successful with VLAN ID 10.

/interface vlan add interface=sfp1-gateway name=sunrise vlan-id=131

Next let's put in place some basic firewall rules to make sure we're not exposing our LAN to the Internet once the connection comes up.

/ip firewall filter
add action=accept chain=forward connection-state=established,related
add action=accept chain=forward in-interface=LAN out-interface=sunrise
add action=drop chain=forward
add action=accept chain=input connection-state=established,related
add action=accept chain=input protocol=icmp
add action=drop chain=input in-interface=!LAN
/ip firewall nat
add action=masquerade chain=srcnat out-interface=sunrise

Now we can configure the special value for the Client Identifier DHCP option and configure the DHCP client on the VLAN interface.

/ip dhcp-client option add code=61 name=clientid-sunrise value="'dslforum.org,Fast5360-sunrise'"
/ip dhcp-client add dhcp-options=clientid-sunrise disabled=no interface=sunrise

This should now give us IPv4 Internet connectivity. We can test this by checking that we received an IPv4 address, have an IPv4 default route and that we can ping a host in the Internet.

/ip dhcp-client print
Flags: X - disabled, I - invalid, D - dynamic
 #   INTERFACE       USE-PEER-DNS ADD-DEFAULT-ROUTE STATUS        ADDRESS
 0   sunrise         yes          yes               bound         198.51.100.123/25
/ip route check 1.1
     status: ok
  interface: sunrise
    nexthop: 198.51.100.1
/ping count=1 1.1
  SEQ HOST                                     SIZE TTL TIME  STATUS
    0 1.0.0.1                                    56  59 1ms
    sent=1 received=1 packet-loss=0% min-rtt=1ms avg-rtt=1ms max-rtt=1ms

Sunrise doesn't offer native IPv6 connectivity but employs 6rd (which defines how to create a 6to4 tunnel based on the public IPv4 address, an IPv6 prefix and the tunnel gateway).

Before we setup the 6rd tunnel, it's important to put in place firewall rules for IPv6 as afterwards all devices on the local network will receive a public IPv6 address.

/ipv6 firewall filter
add action=accept chain=forward connection-state=established,related
add action=accept chain=forward in-interface=LAN
add action=drop chain=forward
add action=accept chain=input connection-state=established,related
add action=accept chain=input protocol=icmpv6
add action=drop chain=input

To setup the 6rd tunnel, I've modified an existing script with the specific parameters for Sunrise (namely the 2001:1710::/28 prefix and the 212.161.209.1 tunnel gateway address).
The script creates the tunnel interface, configures an IPv6 address on the external interface, configures an IPv6 address on the internal interface (which also enables SLAAC to provide IPv6 addresses to the clients on the local network) and configures an IPv6 default route over the 6rd tunnel.

The script itself will be run via the scheduler, thus let's save it under the name 6rd-script.

:global ipv6localinterface "LAN"
:global uplinkinterface "sunrise"

:global IPv4addr [/ip address get [find interface=$uplinkinterface] address];
:global IPv4addr [:pick $IPv4addr 0 [:find $IPv4addr "/"]]
:global IPv4addr2 [:pick $IPv4addr 0 30]
:global IPv6temp [:toip6 ("1::" . $IPv4addr2)]

:global IPv4hex1 [:pick $IPv6temp 3 4]
:global IPv4hex2 [:pick $IPv6temp 4 7]
:global IPv4hex3 [:pick $IPv6temp 8 9]
:global IPv4hex4 [:pick $IPv6temp 9 12]
:global IPv6addr [("2001:171" . $IPv4hex1 . ":". $IPv4hex2 .$IPv4hex3 . ":" . $IPv4hex4 . "0::1/64")]
:global IPv6addrLoc [("2001:171" . $IPv4hex1 . ":". $IPv4hex2 . $IPv4hex3 . ":" . $IPv4hex4 . "1::1/64")]

#6to4 interface
:global 6to4id [/interface 6to4 find where name="6rd"]
:if ($6to4id!="") do={
:global 6to4addr [/interface 6to4 get $6to4id local-address]
if ($6to4addr != $IPv4addr) do={ :log warning "Updating local-address for 6to4 tunnel '6rd' from '$6to4addr' to '$IPv4addr'."; /interface 6to4 set [find name="6rd"] local-address=$IPv4addr }
} else { :log warning "Creating 6to4 interface '6rd'. "; /interface 6to4 add !keepalive local-address=$IPv4addr mtu=1480 name="6rd" remote-address=212.161.209.1 }

#ipv6 for uplink
:global IPv6addrnumber [/ipv6 address find where comment="6rd" and interface="6rd"]
:if ($IPv6addrnumber!="") do={
:global oldip ([/ipv6 address get $IPv6addrnumber address])
if ($oldip != $IPv6addr) do={ :log warning "Updating 6rd IPv6 from '$oldip' to '$IPv6addr'."; /ipv6 address set number=$IPv6addrnumber address=$IPv6addr disabled=no }
} else {:log warning "Setting up 6rd IPv6 '$IPv6addr' to '6rd'. "; /ipv6 address add address=$IPv6addr interface="6rd" comment="6rd" advertise=no }

#ipv6 for local
:global IPv6addrnumberLocal [/ipv6 address find where comment=("6rd_local") and interface=$ipv6localinterface]
:if ($IPv6addrnumberLocal!="") do={
:global oldip ([/ipv6 address get $IPv6addrnumberLocal address])
if ($oldip != $IPv6addrLoc) do={ :log warning "Updating 6rd LOCAL IPv6 from '$oldip' to '$IPv6addrLoc'."; /ipv6 address set number=$IPv6addrnumberLocal address=$IPv6addrLoc disabled=no }
} else {:log warning "Setting up 6rd LOCAL IPv6 '$IPv6addrLoc' na '$ipv6localinterface'. "; /ipv6 address add address=$IPv6addrLoc interface=$ipv6localinterface comment="6rd_local" advertise=yes }

#ipv6 route
:global routa [/ipv6 route find where dst-address="2000::/3" and gateway="6rd"]
if ($routa="") do={ :log warning "Setting IPv6 route '2000::/3' pres '6rd'. "; /ipv6 route add distance=1 dst-address="2000::/3" gateway="6rd" }

Once we've added the script we also need to create the scheduler entry to run it periodically (as it needs to re-configure the tunnel and addresses whenever the public IPv4 address changes).

/system scheduler add interval=1m name=schedule1 on-event=6rd-script

After the first run of the script we should now have IPv6 connectivity. Let's test this again by checking that we have a public IPv6 address, an IPv6 default route and can ping an IPv6 host in the Internet.

/ipv6 address print where interface=6rd and global 
Flags: X - disabled, I - invalid, D - dynamic, G - global, L - link-local 
 #    ADDRESS                                     FROM-POOL INTERFACE        ADVERTISE
 0  G ;;; 6rd
      2001:171c:6336:47b0::1/64                             6rd              no
/ipv6 route check 2600::
     status: ok
  interface: 6rd
    nexthop: 2600::
/ping count=1 2600::
  SEQ HOST                                     SIZE TTL TIME  STATUS
    0 2600::                                     56  50 118ms echo reply
    sent=1 received=1 packet-loss=0% min-rtt=118ms avg-rtt=118ms max-rtt=118ms

And that's how you can configure and validate IPv4 and IPv6 connectivity with your own router on a Sunrise fiber line despite them not liking it very much ;-)

Saturday, 21. November 2020Week 46

MTA-STS

Recently I added MTA-STS support to one of my domains, and it turns out that this was easier than expected.

MTA-STS is used to tell mail senders that your server supports TLS. And then you can define the policy for your server and tell them that they should only use TLS (resp. STARTTLS) when connecting to you and not fall back to unencrypted SMTP.

The way this works is with two components:

  • a special _mta-sts.<your-site.com> TXT DNS entry indicating that your domain supports MTA-STS and the version number of your MTA-STS policy
  • a mta-sts.txt file served under a specific well-known URL https://mta-sts.<your-site.com>/.well-known/mta-sts.txt containing your MTA-STS policy (which mx hosts it is valid for, should it be run in enforcing or testing mode, max-age etc.)

The idea is that a mail sender checks your MTA-STS policy through protected channels (DNSSEC, HTTPS) and then never sends mails to you in plaintext (similar approach as HSTS for HTTP but this time between mail servers).

To setup the MTA-STS configuration, I followed this Enable MTA-STS in 5 Minutes with NGINX guide from Yoonsik Park.

Then to check my configuration I used this MTA-STS validator (which is an opensource project available on GitHub), the classic checktls.com //email/testTo: tool (MTA-STS checking needs to be explicitly enabled under 'More Options') and the free testing service provided by Hardenize.

Sunday, 1. November 2020Week 43

15 years of o5

15 years ago this weblog received the current o5 design (or theme as it would be called nowadays).
During this time the design has aged quite well and also survived the move of the backend from a self-written PHP blog-engine to Jekyll.

Although it still works surprisingly well and presents the content nicely every day, there are some parts where better usage of contemporary technologies would be desirable.
It has no mobile version nor a responsive layout as the design was created before the now omnipresent smartphones were invented. Similar is the font-size hardcoded and not very adequate for todays retina displays. And yes, it uses the XHTML 1.0 strict standard with all its quirks and CSS tricks from 2002 (which luckily are still supported in current browsers).

Overall I'm quite happy that the o5 design has turned out to be so timeless and that I did not have to come up with a new one every other year (btw: I don't remember where the o5 name came from, likely the 5 is a reference to 2005 when it was created).

With the current Corona situation forcing me to spend more time at home again, I have the feeling that some things might change around the weblog (not quite sure what or when exactly, first I need to re-learn how websites are built in 2020 :-).

Saturday, 31. October 2020Week 43

NAT Slipstreaming (NAT traversal part 2)

Compared to the previous post where intentional NAT traversal was discussed, here now comes an article about 'unintentional' (malicious) NAT traversal.

Samy Kamar describes in his NAT Slipstreaming article how a combination of TCP packet segmentation and smuggling SIP requests in HTTP, can be used to trick the NAT ALG of your router into opening arbitrary ports for inbound connections from the Internet to your computer.

The article analyses in detail the SIP ALG of the Linux netfilter stack in it's default configuration, but likely similar attacks could also be possible with ALGs of other protocols and vendors.

Important to note: the Linux SIP ALG module has two parameters (sip_direct_media and sip_direct_signalling), which restrict the IP address for which additional ports are opened to the one sending the original SIP packet. By default they are set to 1, but if any of these is set to 0 in a router's configuration, the described NAT Slipstreaming attack will not only allow to make inbound connections to your computer, but also to any other device in the local network!

Sunday, 23. August 2020Week 33
Sunday, 7. June 2020Week 22

Replace the root disk

Recently the disk holding the root (/) filesystem on one of my linux systems started to report increased SMART raw read error rates, seek error rates and ECC recovered hardware errors.

As these are early indications of a failing disk, it became time to replace the disk.

Normally replacing a disk comes down to plugging in the new one, coyping over the data, umount the old disk, mount the new one in place, unplug the old disk.
But when it is the disk with the root filesystem a couple extra steps are needed.

The steps below worked for my Debian system without problems (even used the opportunity to upgrade to an SSD :-)

(source is this thread on StackExchange)

The following makes some assumptions:

  • All commands ran as root when possible
  • You are on a physical console to the host (need to type in grub commands to boot up the new disk!)
  • You want an ext4 files system
  • You are loosely familiar on a basic level with all commands run
  • You are NOT booting from a RAID device

So here we go.

  1. Physically install new disk into computer and connect to available port leaving old disk in existing position.
  2. Boot computer into old OS.
  3. Prepare and mount new disk; first identify new disk
    fdisk -l
  4. Partition new disk
    fdisk /dev/(newdisk)
    Make partition primary partition with type "83" file system type.
  5. Create filesystem
    mkfs.ext4 /dev/(newpartition)
  6. Mount new filesystem
    mkdir /mnt/(newpartitionmountpoint)
    mount /dev/(newpartition) /mnt/(newpartitionmountpoint)
  7. Copy disk:
    /sbin/init 1 (drop to single user mode)
    rsync -avxHAX / /mnt/(newpartitionmountpoint)
  8. Update FSTAB on newdisk
    blkid (note UUID of new partition)
    vi /mnt/(newpartitionmountpoint)/etc/fstab
    Replace existing UUID of / in FSTAB to new disk UUID
  9. Configure grub and install to new disk boot loader:
    grub-mkconfig
    update-grub
    grub-install /dev/(newdisk)
  10. Copy grub.cfg from old disk to new
    cp -ax /boot/grub/grub.cfg /mnt/(newpartitionmountpoint)/boot/grub/grub.cfg
  11. Open grub.cfg on new disk and replace all UUIDs with new disk
    vi /mnt/(newpartitionmountpoint)/boot/grub/grub.cfg
    Replace all old UUIDs with the UUID of the new disk
  12. Shut down computer
    shutdown
  13. Physically move the new drive to the 1st drive location and remove old drive
  14. Start computer and grub should present:
    error: no such device: xxxxxxxx-xxxx-xxxx-xxxx-xxxxxxxxxxxx
    
    GRUB rescue>
  15. Manually boot new OS from grub; first identify the drive and partition of the boot files
    ls [to identify your drive and partition options]
    ls (hdx,p)/ [to identify which partition has the /boot folder]
  16. Then, you can load the boot menu manually from the drive and partition you found above. Typically this would be (hd0,msdos1).
    set prefix="(hdx,p)/boot/grub"
    set root="(hdx,p)"
    insmod normal
    normal
  17. Login to OS on new drive
  18. Configure grub again
    fdisk -l (note dev of newdisk)
    grub-mkconfig
    update-grub
    grub-install /dev/newdisk

And that should be it!

Sunday, 24. May 2020Week 20

rkhunter CRLF confusion

On my Linux hosts I'm running rkhunter. On a newly configured host it lately reported the following warning:

Warning: The SSH and rkhunter configuration options should be the same:
        SSH configuration option 'PermitRootLogin': no
	Rkhunter configuration option 'ALLOW_SSH_ROOT_USER': no

On first sight the warning does not seem to make much sense, as both configuration options seem to be set to the same value (no).
But digging further reveals that they are stored slightly different:

# file /etc/rkhunter.conf
/etc/rkhunter.conf: ASCII text
# file /etc/ssh/sshd_config
/etc/ssh/sshd_config: ASCII text, with CRLF line terminators

Turns out that rkhunter is also checking the line terminators as part of the configuration values, and warns because they are different.

Knowing this, the fix is simple: run dos2unix on the CRLF file

Thursday, 21. May 2020Week 20

ipaddr CLI tool

While doing some maintenance on my server, I got tired of searching through the output of ip addr show to find the IP addresses configured on the interfaces.
Thus I wrote a simple CLI tool to display the information I needed in a concise and human friendly form: ipaddr

$ ipaddr
lo          127.0.0.1/8
ens5        198.51.100.160/24
tun24008    10.123.199.78/32
tun71991639 10.200.123.5/32
tun26724    10.100.100.235/32
tun3883710  10.123.111.7/32

A nice side-effect of writing this in Go is that it works out-of-the-box also on non-Linux systems :-)

Saturday, 18. April 2020Week 15

Poor man's reboot notification

Sometimes you need to be notified about reboots of a machine without having the luxury of a proper monitoring system.

The following crontab entry triggers an e-mail when the host has been rebooted in the last 5 minutes.

*/5 * * * * [ $(sed -e 's/\..*//' /proc/uptime) -lt 540 ] && echo "Host has been rebooted! Uptime: $(uptime)"
Sunday, 12. April 2020Week 14
Saturday, 28. March 2020Week 12
Saturday, 21. March 2020Week 11

ip_compact and ip_diff

Somehow I always end up working with lists of IP networks and needing to minimize and compare them.

Some of my Perl scripts for this might still be hidden in a corporate source repository, and somewhere in the backups of my old Linux laptop should be even earlier attempts in Bash.

Both of them are not very useful to me where they are, thus I've written yet another version.
This time in Go using the ipaddr package.

Say hello to ip_compact and ip_diff :-)

Tuesday, 17. March 2020Week 11