RootMyTV is a user-friendly exploit for rooting/jailbreaking LG webOS smart TVs

RootMyTV is a user-friendly exploit for rooting/jailbreaking LG webOS smart TVs.

It bootstraps the installation of the webOS Homebrew Channel, and allows it to run with elevated privileges. The Homebrew Channel is a community-developed open source app, that makes it easier to develop and install 3rd party software. Find out more about it here.

If you want the full details of how the exploit works, skip ahead to our writeup.

Is my TV vulnerable?

At the time of writing the original exploit (RootMyTV v1 - 2021-05-15), all webOS versions between 3.4 and 6.0 we tested (TVs released between mid-2017 and early-2021) are supported by this exploit chain. Around June-July 2021 LG started rolling out updates which added some minor mitigations that broke our original exploit chain.

At the time of writing (RootMyTV v2 - 2022-01-05), all webOS versions between 4.x and 6.2+ we tested (TVs released between early-2018 and late-2021) are supported by the new exploit chain.

Some versions between 3.4 and 3.9 may be supported by RootMyTV v2, but your mileage may vary.

Note: this versioning refers to the "webOS TV Version" field in the settings menu, not the "Software Version" field.

If you want to protect your TV against remote exploitation, please see the relevant section of our writeup and/or await an update from LG.

Usage Instructions

Step Zero (disclaimer): Be aware of the risks. Rooting your TV is (unfortunately) not supported by LG, and although we've done our best to minimise the risk of damage, we cannot make any guarantees. This may void your warranty.

1. (Pre-webOS 4.0) Make sure "Settings → Network → LG Connect Apps" feature is enabled.
2. Developer Mode app must be uninstalled before rooting. Having this application installed will interfere with RootMyTV v2 exploit, and its full functionality is replaced by Homebrew Channel built-in SSH server.
3. Open the TV's web browser app and navigate to https://rootmy.tv
4. "Slide to root" using a Magic Remote or press button "5" on your remote.
5. Accept the security prompt.
6. The exploit will proceed automatically. The TV will reboot itself once during this process, and optionally a second time to finalize the installation of the Homebrew Channel. On-screen notifications will indicate the exploit's progress. On webOS 6.x Home Screen needs to be opened for notifications/prompts to show up.

Your TV should now have Homebrew Channel app installed.

By default system updates and remote root access are disabled on install. If you want to change these settings go to Homebrew Channel → Settings. Options there are applied after a reboot.

For exploiting broken TVs, check out the information here.

Why rooting

FAQ

Is it safe?

While we cannot take any responsibility for Your actions, we have not encountered any bricks due to rooting. If you only use trusted software from official Homebrew Channel repository, then you should be safe.

Will this void my warranty?

This is not a legal advice. At least in the EU, rooting and other software modifications are generally deemed to be legal and should not be a basis for voiding your warranty.

How do I get rid of this?

Factory reset should remove all root-related configuration files.

We don't have a convenient tool for root removal without factory reset, though a knowledgable person may be able to remove our customizations manually.

Are system updates possible?

While updates are technically possible, if LG patches the exploit, you might end up "locked out" and unable to re-root your TV if you somehow lose access. We also can't predict how future updates will affect our techniques used to elevate and operate the Homebrew Channel app.

Will this break Netflix/YouTube/AmazonVideo?

No. This does not break or limit access to subscription services or other DRMed content.

However, staying on very old firmware version (which may be required for keeping root access persistent) may limit Your access to LG Content Store application installs, updates, or (rarely) launches. Workarounds for this are in the works.

How do I update from RootMyTV v1? (released 2021/05)

If you are not going to update your TV Software Version to the one that is already patched (most 4.x+ released after 2021/06) there is no need to update. New chain does not bring any new features - the most sensible thing you can do is to update your Homebrew Channel app.

If you are already rooted on downgraded/pre-2021-06 firmware version and want to upgrade further, doing an official software update will remove existing root files and homebrew applications. Running RootMyTV v2 then will reenable root access again. You will need to reinstall removed applications yourself.

If you know what you are doing and want to persist installed applications, you need to remove /media/cryptofs/apps/usr/palm/services/com.palmdts.devmode.service/start-devmode.sh file right before an update (without rebooting inbetween), and then run RootMyTV v2 right on first boot after software update.

I quickly turned my TV on and off and it's really angry about Failsafe Mode

If "Failsafe Mode" got tripped on your TV and it's showing angry notifications, go to Homebrew Channel → Settings, switch "Failsafe Mode" off and press "Reboot".

"Failsafe Mode" is a mode where none of our system customizations are enabled and only an emergency remote access server gets started up.

This mode gets enabled automatically when the TV crashes, gets its power removed or is shut down during early system startup. In order to reduce chances of that happening we recommend enabling "Quick Start+" setting in webOS System Settings General tab. This will make the TV only go to "sleep mode" (which doesn't take much more power) instead of doing a full shutdown, and will not need to restart our services on every suspend. This will also make TV startup much faster.

I want to run some commands as root during boot!

Our startup script runs all executable files in /var/lib/webosbrew/init.d on boot (via run-parts - filenames may only contain a-zA-Z0-9-_ letters!) - create your own scripts there.

Create any customizations there and do not modify existing RootMyTV/Homebrew Channel scripts, since these may be overwritten on future updates.

If you are a homebrew developer - create a symlink to a script in your own app path there, and do not copy over anything there.

I want to support you financially!

If you want, you can support this project via GitHub Sponsors - see "Sponsor" button in upper right corner.

Post-Installation Advice (IMPORTANT!)

1. Don't update your TV. While updates are technically possible, if LG patches the exploit, you might end up "locked out" and unable to re-root your TV if you somehow lose access. We also can't predict how future updates will affect our techniques used to elevate and operate the Homebrew Channel app. "Block system updates" option in Homebrew Channel will disable firmware update checks. Make sure "Automatic system updates" option in webOS System Settings is disabled as well.

2. It is required to remove "Developer Mode" app before rooting. Otherwise it will interfere with the startup script used to bootstrap the jailbreak. SSH service exposed by Homebrew Channel is compatible with webOS SDK tooling.

3. If you need remote root shell access and know how to use SSH, you can enable it in Homebrew Channel settings. Default password is alpine, but we recommend setting up SSH Public Key authentication by copying your SSH Public Key over to /home/root/.ssh/authorized_keys on the TV. This will disable password authentication after a reboot.

   GitHub user registered keys can be installed using the following snippet:

   mkdir -p ~/.ssh && curl https://github.com/USERNAME.keys > ~/.ssh/authorized_keys

   Alternative option is Telnet (can be enabled in Homebrew Channel → Settings → Telnet) though it is highly discouraged, since this gives unauthenticated root shell to anyone on a local network.

4. It is recommended to have "Quick Start+" functionality enabled. This will make shutdown button on a remote not do a full system shutdown. If you quickly turn the TV on and off without Quick Start+, our "Failsafe Mode" may get triggered (which is there to prevent startup scripts bricking the TV) which will go away after switching relevant switch in Homebrew Channel Settings.

Troubleshooting

In case of any problems join the OpenLGTV Discord server and ask for help on #rootmytv channel, ask on our #openlgtv:netserve.live Matrix channel, or file a GitHub issue.

Before asking for support, please consult our Troubleshooting guide.

Research Summary and Timeline

RootMyTV is a chain of exploits. The discovery and development of these exploits has been a collaborative effort, with direct and indirect contributions from multiple researchers.

On October 05, 2020, Andreas Lindh reported a root file overwrite vulnerability to LG. On February 03, 2021, Andreas published his findings, demonstrating a local root exploit against the webOS Emulator (a part of LG's development SDK). LG had boldly claimed that this issue did not affect their devices, and that they were going to patch their emulator.

On February 15th, 2021, David Buchanan reported a vulnerability in LG's "ThinQ login" app, which allowed the app to be hijacked via a specific sequence of user inputs, allowing an attacker to call privileged APIs. On March 23rd 2021, David published a proof-of-concept exploit, which enabled users to gain root privileges on their LG smart TVs. This was made possible by combining it with the local root vulnerability previously reported by Andreas (Yes, the same one that LG said did not affect their devices!).

Around March 28th 2021, Piotr Dobrowolski discovered a similar vulnerability in the "Social login" app, which is present across a wider range of webOS versions. More importantly, this exploit could be easily triggered over the local network, using SSAP (details below), making it much more reliable and user-friendly.

At time of writing, the code in this repo is the combined work of David Buchanan (Web design, initial PoC exploit) and Piotr Dobrowolski (Improved "v1" exploit implementation, writeup, and "v2" research and implementation).

We would like to thank:

• Andreas Lindh for publishing his webOS research.

• The wider webOS community, particularly the XDA forums and the OpenLGTV discord.

• All the contributors (present and future) to the Homebrew Channel, and development of other homebrew apps and software.

• LG, for patching symptoms of bugs rather than underlying causes...

The Technical Details

Background

webOS, as the name suggests, is a Smart TV operating system mostly based on web technologies. Applications, both system and external are either run in a stripped down Chromium-based web browser ("WebAppMgr") or in Qt QML runtime. Almost all system and external applications run in chroot-based jails as an additional security layer.

"Web apps", outside of standard web technologies, also get access to an API for communicating with "Luna Service Bus". This is a bus, similar to D-Bus, used to exchange messages and provide various services across different security domains. Bus clients can expose some RPC methods to other applications (identified by URIs luna://service-name/prefix-maybe/method-name) which accept JSON object message as their call parameters, and then can return one or many messages. (depending on the call being "subscribable" or not)

While Luna bus seems to have extensive ACL handling, considering the history of webOS IP transfers, seems like not many engineers fully understand its capabilities. Part of the bus is marked as "private", which is only accessible by certain system applications, while most of the other calls are "public" and can be accessed by all apps.

Unexpectedly, one of the internal services exposed on a bus is "LunaDownloadMgr" which provides a convenient API for file download, progress tracking, etc... Said service has been researched in the past and an identity confusion bug leading to an arbitrary unjailed root file write vulnerability has been publicly documented.

This in and of itself was not very helpful in production hardware, thus we needed to find a way of calling an arbitrary Luna service from an application with a com.webos. / com.palm. / com.lge. application ID.

Step #0 - Getting in (index.html)

In order to gain initial programmatic control of the TV GUI, an interface called "LG Connect Apps" can be used. Its protocol, called "SSAP" (Simple Service Access Protocol), is a simple websocket-based RPC mechanism that can be used to indirectly interact with Luna Service bus, and has been extensively documented in various home-automation related contexts. We use that to launch a vulnerable system application which is not easily accessible with normal user interaction.

Step #0.1 - Escaping the origins

SSAP API is meant to be used from an external mobile app. For the sake of simplicity, though, we wanted to serve our exploit as a web page. This lead us to notice that, understandably, the SSAP server explicitly rejects any connections from (plaintext) HTTP origins. However, there was an additional exception to that rule, and seemingly the authors wanted to allow file:// origins, which present themselves to the server as null. Turns out there's one other origin that can be used that is also reprted as null, and that is data: URIs.

In order to exploit this, we've created a minimal WebSocket API proxy implementation that opens a hidden iframe with a javascript payload (which is now running in a data:/null origin) and exchanges the messages with the main browser frame. This has been released as a separate library.

Step #0.2 - General Data Protocol Redirection

There's a minor problem with establishing the connection with the SSAP websocket server. While we all believe in utter chaos, we don't feel very comfortable with serving our exploit over plaintext HTTP, which would be the only way of avoiding Mixed Content prevention policies. (by default, https origins are not allowed to communicate with plaintext http endpoints)

While some newer Chromium versions do allow Mixed Content communication with localhost, that was not the case when Chromium 38 was released (used in webOS 3.x). Thankfully, it seems like the system browser on webOS 3.x is also vulnerable to something that has been considered a security issue in most browsers for a while now - navigation to data: URIs. Thus, when applicable, our exploits attempts to open itself as a data: base64-encoded URI. This makes our browser no longer consider the origin being secure, and we can again access the plain-http WebSocket server.

Mitigation note

An observant reader may have noticed that the service we use is meant to be used remotely. While the connection itself needs a confirmation using a remote we highly recommend to disable LG Connect Apps functionality in order to prevent remote exploitation. However, this option seems to only be present on webOS versions older than webOS 4.x - in such cases the only solutions are to either keep the TV on a separate network, or disable SSAP service manually using the following command after rooting:

luna-send -n 1 'palm://com.webos.settingsservice/setSystemSettings' '{"category":"network","settings":{"allowMobileDeviceAccess":false}}'

Step #1 - Social login escape (stage1.html)

Having some initial programmatic control of the TV via SSAP, we can execute any application present on the TV. All cross-application launches can contain an extra JSON object called launchParams. This is used to eg. open a system browser with specific site open, or launch a predetermined YouTube video. Turns out this functionality is also used to select which social website to use in com.webos.app.facebooklogin, which is the older sibling of com.webos.app.iot-thirdparty-login used in initial exploit, present on all webOS versions up until (at least) 3.x.

When launching social login via LG Account Management, this application accepts an argument called server. This turns out to be a part of URL that "web app" browser is navigated to. Thus, using a properly prepared launchParams we are able to open an arbitrary web page (with the only requirement being that it's served over https) running as a system app that is considered by LunaDownloadMgr a "system" app.

Step #2 - Download All The Things (stage2.html)

Since we are already running as a system application, we can download files (securely over https!) into arbitrary unjailed filesystem locations as root.

We use that to download following files:

• stage3.sh → /media/cryptofs/apps/usr/palm/services/com.palmdts.devmode.service/start-devmode.sh - this is the script executed at startup by /etc/init/devmode.conf as root, in order to run developer mode jailed SSH daemon.
• hbchannel.ipk → /media/internal/downloads/hbchannel.ipk - since our end goal is intalling the Homebrew Channel app, we can also just download it during the earlier stages of an exploit and confirm it's actually downloaded.
• devmode_enabled → /var/luna/preferences/devmode_enabled - this is the flag checked before running start-devmode.sh script, and is just a dummy file.

Step #3 - Homebrew Channel Deployment (stage3.sh)

stage3.sh script is a minimal tool that, after opening an emergency telnet shell and removing itself (in case something goes wrong and the user needs to reboot a TV - script keeps running but will no longer be executed on next startup), installs the homebrew channel app via standard devmode service calls and elevates its service to run unjailed as root as well.

2021/06: The Old-New Chain (RootMyTV v2)

Around 2021/06 LG started rolling out a patched version which involved some fixes for the tricks we used in this chain:

• Certain applications we used for private bus access have their permissions limited to public
• LunaDownloadMgr now checks target paths against a list of regular expressions in /etc/palm/luna-downloadmgr/download.json
• start-devmode.sh script is now shipped with a signature and is now verified using openssl on each boot
  • This one had an interesting side effect - it took approximately a month for LG to roll out a new Developer Mode application with signed start-devmode.sh, during which time updated TVs were unable to use developer mode at all.

Most of these mitigations are too trivial to work around, thus we still consider this chain unfixed.

• There are still applications on the system that are vulnerable to XSS attacks with private bus permissions
• Regular expressions used to verify target paths are too broad, and thus still allow us to write to relevant paths
• There are multiple paths that are executed during bootup, so we don't even need to use start-devmode.sh

Our initial estimate for fixing these issues in our chain were "a couple of hours" - patches theorized on our side on 2021/05/27 turned out to be correct, but due to some strategic choices and lack of personal time, we decided to postpone testing and release for a couple of months. Sorry. :)

from Hacker News https://ift.tt/EiCtMd4