Flashing router firmware through Serial Port: CFE bootloader (usually Broadcom) based routers

Here’s a summary of learnings from dd-wrt’s serial recovery instructions:

  1. Use a UART controller that signals at 3.3V (e.g. FTDI TTL-232R-3V3) to talk to the board. Regular serial RS-232 ports requires a voltage level shifter converting the signal to swing between 0V to 3.3V.
  2. You only need 3 pins: Tx, Rx and Ground. There’s voltage contention if you plug in the Vcc from TTL-232R-3V3 (It’s the USB’s 5V despite the signaling is 3.3V) to the 3.3V supply of the router. You don’t need the Vcc pin. It didn’t harm anything or do anything when I connected the Vcc pin.
  3. Stick with all serial port defaults and only set the baud to 115200 (default is 9600) and turn off flow control (default is xon/xoff). I use Putty for terminal.
  4. The terminal serves as the monitor for the computer on the router that shows a text console. Broadcom uses CFE bootloader (others use U-Boot with busybox).
  5. CFE bootloader defaults to 192.168.1.1 with subnet mask 255.255.255.0 (aka /24). Set up the network interface to have a static IP on the same subnet to talk to the board.
  6. Good habit: nvram erase
  7. The flash program relies on TFTP protocol to receive the firmware file. So get your TFTP client ready. Microsoft included a TFTP client/server since Windows 7 but usually disabled (turn it on in Windows OptionalFeatures.exe).
  8. TFTP is a simple push(put)/pull(get) design. You can either “push a file on your computer” or “get a file as filename”. You’d want to specify -i switch (binary image transfer) with Windows tftp.exe.
  9. So type this command at the command prompt but don not press enter until your router is ready to grab the file: tftp -i 192.168.1.1 put {path to whatever TRX firmware file}
  10. Go back to the serial terminal and tell the router to accept a TFTP push (in a window of a few seconds before it time out) and flash the memory region flash1.trx with this command: flash -ctheader : flash1.trx
  11. Immediately initiate the TFTP push from your computer (Windows command line example in Step #9 above)
  12. Wait for a couple of hours! The terminal might tell you that it has received the file completely, but it won’t show anything when it’s writing to the flash! It’s a painfully slow process with no feedback. Just be patient!

Some observations

  • FreshTomato firmware absolutely won’t tell you on the screen after it has done flashing (Merlin-WRT does). Just turn the router back on after a couple of hours.
  • Merlin firmware repeats (exposes) the raw passwords to the serial port!
  • FreshTomato firmware boots to a linux prompt on the serial port

Loading

Asus-wrt Merlin Firmware DDNS update interval hack

The “WAN – DDNS” page only allows users to set the DDNS updater to check as frequently as every 30 minutes. My DDNS provider does not have an update frequency limit, so I’d like to have the update client check for every 1 minute. The setting is called “Verify every”:

Attempting to set it to every 1 minute gives this error message:

I searched for the “WAN-DDNS” config webpage file (Advanced_ASUSDDNS_Content.asp) in the firmware source code, and found that it’s under /www folder in the router’s linux root.

Since “Verify every” is such generic words, and Github does not support exact phrase match in search (I use “in:file” specifier in the search box), I pick “WAN IP and hostname verification” (the closest setting which I expect the code to be in the proximity of the one corresponding to “Verify every”) so it has more unique keywords. The first jump:

Since it’s just a dictionary files, we search for the associated internal variable name
DDNS_verification_enable” which points to the this line in Advanced_ASUSDDNS_Content.asp:

Since this name appeared nowhere else, I traced the “id” attribute above, which is “check_ddns_field” and I see a Javascript (.js) file that process the data from the web page forms:

The variable check_ddns_field appears in the if-else-if branches of change_ddns_settings(), so one of the few next few variables after it is likely to correspond to “Verify with”.

The variable name showed up in 4 branches of if-elseif-else switches (switching DDNS service providers), which ddns_regular_period comes right after

Searching for the class member (or struct field)

Bingo. Here’s the entry value range check code. I’ll change the “30” minutes to “1” minute to enable checking at 1 minute intervals (which I think it’s reasonably responsive for testing and general use).


I’d prefer to check if the input range check is there out of feasibility (i.e. what is the smallest increment) or it’s just set to prevent people from getting banned by the DDNS provider for checking too frequently. I looked into the last occurrence of ddns_regular_period and found this:

Which means the web forms is updating NVRAM (environmental) variable of the same name ddns_regular_period, which appears to be called only in watchdog.c:

And Dang! The code enforces if the ddns_regular_period (on NVRAM) is set to be less than the original 30 minute minimum (invalid condition), it’d be set to the default 60 minutes (1 hr).

It’s actually sloppy coding because the defaults are specified in struct fields in defaults.c:

yet that 60 minutes is hard-coded in watchdog.c. That means if I don’t catch it and only changed the default in one place, the behavior will not be what I expected given the right conditions. This is an example of why software feature expansion are likely to break things. If you have solid code, bugs on updates are likely to happen.

I was curious why it says (period*2)

and suspected the ddns_check_count is incremented in 30 second (half-minute) interval. Since it’s watchdog.c, my natural guess is that the watchdog checks every 30 seconds for these event hooks. Turns out the notes (comments) in the code has “30 seconds periods” noted everywhere.

I searched a little bit more about linux watchdogs and found this useful webpage which explained how it works. I didn’t see /dev/watchdog in my router’s rootfs (root file system) so I assumed it’s a hardware watchdog (embedded linux, so duh).

I was about to dig up the hardware manual for the chipset for my router, but I search for they string HW_RTC_WATCHDOG  first and it showed up in linux kernel code (duh):

Note that the HW_RTC_WATCHDOG is a register in this code base, not the number of seconds from Christian’s Blog. i.e. they are completely different things, but it provided a good keyword lead for me to start digging.

The code seems to be the same for various kernel version so I picked any one of them to understand the behavior. First occurrence is in wdt_enable():

The other places are suspend/resume, so I’ll ignore those for now. Note that wdt_enable() is a static function, so only need to search within the same file. The only active place that calls it is wdt_ping():

So there are only 2 things I’ll need to find out: heartbeat and WDOG_COUNTER_RATE:

…. unfinished

https://bitsum.com/firmware_mod_kit.htm


While it’s a lot of useful learning about Embedded Linux and hunting down source code, for the meantime, given that namecheap does not care if you blindly update every minute, it’s easier to just set up a cron job that runs at every N minutes using curl/wget.

dd-wrt has a place for you to enter the cron scripts with the web interface, but you might need to log into the router using SSH and register the cron job yourself:

The core command is called ‘cru‘, which typing it in the command prompt will show you the very simple usage:


admin@Router:/tmp/home/root# cru

Cron Utility
add:    cru a <unique id> <"min hour day month week command">
delete: cru d <unique id>
list:   cru l

<unique id> is just a tag that you make up to name your task. Again the one-liner command needs to be direct absolute path. My ‘curl‘ program is located in /usr/sbin, so the command is:

cru a ncddns * * * * * /usr/sbin/curl "https://dynamicdns.park-your-domain.com/update?host={subdomain or @ for root}&domain={registered domain name}&password={DDNS-specific password generated by namecheap's domain administration page under Advanced DNS}"

The “* * * * *” refers to run at every “minute, hour, date of month, month, date of week”, in other words, run at every minute in every waking moment. The wild card * means ALL-OF.

Cron job registration through CRU is not persistent, so to make it survive reboots, add the above cru command as a line to /jffs/scripts/services-start script. It should be executable by default, if not, make sure you set it to be executable or it won’t run.

Loading

Namecheap Dynamic DNS setup in dd-wrt

Namecheap support page explained the process of configuring your dd-wrt router firmware to use Namecheap’s REST (HTTP URL) update interface to dynamically update the IP of your (sub-)domain. The instruction works, but there are few items which doesn’t quite make sense to me as a programmer, and I did a few experiments to figured that it’s bogus and developed a few insights about what’s necessary and why they do it.

Their instructions looks like this:

and specific verbal instructions are:

  • DDNS Service: Custom
  • DYNDNS Server: dynamicdns.park-your-domain.com – the name of the server should not be changed
  • Username: yourdomain.com – replace it with your domain name
  • Password: Dynamic DNS password for your domain (Domain List >> click on the Manage button next to the domain >> the Advanced DNS tab >> Dynamic DNS)
  • Hostname: Your subdomain (@ for yourdomain.comwww for www.yourdomain.com, etc.)
  • URL/update?domain=yourdomain.com&password=DynamicDNSPassword&host=

I stroke out Username and Password fields because they are not used!

If you look at the URL, namecheap’s instructions are asking you to re-enter the domain and the password key-value pair AGAIN, which means Username and Password fields are not used.

My programmer instinct immediate screams the updater is assuming certain REST API syntax that are not properly substituted so they need to be entered manually, exposing the password without the benefit of the masks (forget about keeping the password top secret, router firmware guys aren’t top security engineers. Just re-generate one in Namecheap’s admin interface if it gets compromised).

I checked by entering bogus Username and Password fields (the web’s GUI/forms checks if they are blank, so you can’t get away with not entering). It worked as expected. This means the two fields are dummies with Namecheap’s instructions.

Based on the fact that Namecheap’s instructions being unable to substitute Username and Password fields and the host key must be put at the end for Hostname field to substitute correctly, I can safely speculate that the one who wrote this couldn’t find out what the syntax for the variables are, and exploit that the last parameter hostname gets attached at the end in the absence of substitution variables in the URL syntax.

Apparent people are doing something stupid like this because nobody in the chain remember to document the substitution variable names! It’s not in dd-wrt’s user interface (should have that printed the ‘usage’ info next to the URL box) and neither it’s in INADYN’s github readmes!

I decided to dig deeper and go after the dynamic DNS updater package in question. dd-wrt is using inadyn package to do the dynamic DNS update, as “INADYN” is shown in “DDNS status” box gives it away (confirmed by dd-wrt’s docs):

The service itself is called ddns though.

I ended up reading the /examples folder on the repository and found this:

Bingo! Here it is:

URL substitution stringdd-wrt fieldsNamecheap REST key
%uUsernamedomain
%pPasswordpassword
%hHostnamehost

generic.c is a plugin also shows the above table as well


Since namecheap’s dynamic DNS service do not mandate how frequently you can update nor they charge per update, it’s easiest and most reliable to just blindly update the IP every N minutes instead of checking against a local cache to see if the external IP has really changed before updating at each poll interval

This user interface does not have the option to set the updater to run every 1 minute, so why bother since it’s just a simple program that creates a simple URL and do a curl/wget? At the end of the day, I decided to just do a cron job:

PATH=/sbin:/usr/sbin:/bin:/usr/bin
* * * * * root curl "https://dynamicdns.park-your-domain.com/update?host={subdomain or @}&domain={domain name bought}&password={generated by namecheap's account management page}"

There are 3 things that you will need to know:

  • Paths is from a clean slate. Need to define it first
  • * * * * * means every minute. Specify numbers/range for each time unit (minute, hour, day of month, month, day of week) if desired. Asterisk means ON EVERY.
  • Need to specify the user after the time syntax and before the actual command

Loading

dd-wrt gotchas

dd-wrt is very a powerful firmware compared to ASUS Merlin, but the UI leaves a lot to be desired. It’s very close to editing a config file and there’s little help to what each setting. The developers of dd-wrt didn’t invest time in designing the web administration interface and used the most basic primitive HTML forms so there are no tooltip that explains the features and the interaction between different settings.

There are also some confusing (nonsensical) UI design that are a lot less work to the developer but confused users to no end. Here are the examples I’ve found so far:

  • Enabling remote admin through SSH (for embedded linux command prompt) is a two step process out of the box. You’ll need to first enable SSHd from Services -> Secure Shell before enabling SSH Management from Administration -> Management (otherwise it’s greyed out)
  • The router username (user modifiable) for dd-wrt applies to web UI only. SSH’s username remains root. They share the same password though (so login and password are decoupled in dd-wrt, they are effectively two passwords in practice except they don’t put asterisk over the username as you type). ASUS Merlin firmware’s login is consistent across both web page and SSH
  • Cron jobs is from a bare environment which means you need to manually define the paths and specify the user in the cron job syntax. e.g.
    PATH=/sbin:/usr/sbin:/bin:/usr/bin
    * * * * * root {command_to_execute}

Loading

Remote Desktop INTO Linux GUI (xrdp)

To serve Linux Desktop just like other Windows computer through Windows Remote Desktop (formerly Terminal Services), so far I have found xrdp (xorgxrdp). VNCs, NX (NoMachine)/TeamViewer does not count because they share the screen of an existing session, instead of creating a new one for you.

Xrdp does not follow the use pattern as Microsoft’s RDP. When you log in to a Xrdp host (server) through a RDP client, you go into an intermediary (welcome interface) called sesman (Session Manager), which is a multi-protocol remote graphical session client (think of it as a very rudimentary Remmina).

“Session” is the roughly same as protocol, which formerly and internally it’s called “Module” as the client programs are in implemented lib*.so object module file.

The two session modules we are interested in here is

  • Xorg (libxup.so): Xorgxrdp is the MS-RDP-like mode that starts a new X session without first attaching to a screen.
  • Xvnc (libvnc.so): basically a VNC client. You start a VNC server (like X11vnc) with a display/screen (can be started in any X session you logged in, or the local user screen if you set the VNC server as a service) and connect to it in this RDP intermediary (welcome interface) without installing VNC client software.

In Windows, RDP do not distinguish between local and remote users and sessions with the same login account will take over other existing sessions. If you want each session to start fresh and leave other sessions alone, disable this in Group Policy Object editor under Computer Configuration > Administrative Templates > Windows Components > Remote Desktop Services > Remote Desktop Session Host > Connections -> "Restrict Remote Desktop Services user to a single Remote Desktop Services session“.

I am usually fine with this arrangement as well, but often prefer to connect to my remote sessions work in the background leaving the local user alone (i.e. if I want things to show up on the local monitor screen, I’ll use VNC instead). I’d also like to resume my remote sessions if I log in from another computer instead of starting from scratch with each new RDP connection. Turns out given a bunch of quirks of xrdp, this is much easier to do so than reproducing MS-RDP’s default behavior.


Allow overtaking of locally logged-in session

First of all, out of the box, the same remote user cannot overtake locally logged-in desktops nor be simultaneously logged in! It’s either one way or the other! I got bumped out immediately after logging in through sesman, or if I logged in remotely first, I get bumped out when I try to log in locally.

Found somebody suggested that certain desktop environment might have added code to prevent the second session from opening. And this blog suggested you edit the windows manager launch script

sudo nano /etc/xrdp/startwm.sh

to add EITHER

unset DBUS_SESSION_BUS_ADDRESS
unset XDG_RUNTIME_DIR

OR

export $(dbus-launch)

RIGHT BEFORE the last lines which checks and calls the Xsession

test -x /etc/X11/Xsession && exec /etc/X11/Xsession
exec /bin/sh /etc/X11/Xsession

This only solves the part of simultaneous local & remote logons


In the newer version as of writing, the default behavior is that locally logged in sessions are independent of remotely logged in sessions, yet the remotely logged in sessions resumes by default (if you log in as the same user).

Turns out this is what I preferred as the local sessions should be reached with VNC instead and I’d prefer my remote sessions happen at the background without showing it on the local screen.

As of now (2020-03-17), the author of xrdp mentioned in his blog that a remote session cannot take over a locally logged in session like Windows RDP does:


Generic VNC client hosted on RDP server

The interface also offer a built-in VNC client that you can use it to connect to your current session if you have VNC server enabled so you don’t have to install a separate VNC client from the client machine you are connecting from. In other words, xorgxrdp self-hosts a VNC client within xrdp so you can use your RDP as a VNC client.

Since the VNC server (like x11vnc) only has a password (not using system’s active directory or user management system), there is no user name. I went to [Xvnc] section of xrdp.ini and replaced username=ask with username=na. The port number -1 no longer applies as we aren’t emulating RDP with VNC anymore (where sesman creates a new VNC server instance if not previously done). Given that I’m running VNC as a service at default port 5900, I also changed it to port=5900.


Other less important features in xorgxrdp

Session/Module “vnc-any” uses the same libvnc.so as Xvnc before it, and they are pretty much the same thing except it exposes ip:port entry so you can use it as a gateway to connect to VNC servers hosted on other machines (can be used to connect to the VNC server on the current machine you just connected to through RDP if you stick with 127.0.0.1:5900). It’s more like a convenience thing that hosts the VNC client software that you can RDP into (so you don’t need to install a VNC client from where you are).


There is also a RDP client module/session called ‘neutrinordp-any’, which basically uses the linux machine you just connected to as a gateway to visit another machine hosting RDP. It’s rarely useful and it doesn’t work out of the box when I tried it (does nothing after I press OK despite entering all the info correctly). So I removed it from my xrdp.ini


Minor annoyances to deal with in xrdp

There’s also a minor annoyance that if you connect remotely, “Authentication Required…” message box will show up on start since remote user is a little more restrictive than local users. This can be solved by creating this file with nano

sudo nano /etc/polkit-1/localauthority/50-local.d/46-allow-update-repo.pkla

and paste the contents there and save it:

[Allow Package Management all Users]
Identity=unix-user:*
Action=org.freedesktop.packagekit.system-sources-refresh
ResultAny=yes
ResultInactive=yes
ResultActive=yes

Loading