Red Hat in-vehicle operating system in modern and future vehicles | Hacker News Reader

Red Hat in-vehicle operating system in modern and future vehicles

redhat.com

78 pointsohjeez39 comments

This seems to target the "bigger" cpu's in a car, but there are tens of ECU's that are far smaller and that I think could benefit from a better platform. A much better one than AUTOSAR, which is a nightmare to work with. I swear I don't know what automakers are thinking with pushing AUTOSAR.

Maybe zephyr [1] is a good option for those ECUs? or hubris [2]?

[1] https://docs.zephyrproject.org/latest/introduction/index.htm...

[2] https://github.com/oxidecomputer/hubris

Zephyr hasn't finished their certification package, to the best of my knowledge. It was supposed to be finished late last year, but I'm not sure what the current status is. That's typically a bare minimum for RTOS vendor selection, and the turnaround time between selection and deployment is typically 3+ years.

Unfortunately still a bit premature right now.

Ok, color me confused, but what does this actually mean? It mentions fedora IoT, via the centos link, which is AFAIK just another Linux distribution (something like coreos, but not coreos?). Where does that fit in a car? Is linux real-time enough, for any of the actual control plane/etc? Is it certifiable? Which leaves what, the entertainment system? What does this provide that a android fork doesn't, longer support lifetimes? It can't actually be Linux can it? I mean RHEL is considered an ancient distro and it only does ~10 years. 10 years old is basically brand new in some car circles, where people are regularly driving 20-30 year old cars. The Linux of 2042 will likely be as unrecognizable as the Linux of 1999 is today, so this has to be something else right? I guess there is a midplane for non driving, non entertainment? AKA the electric windows, door locks, etc? Does that need a full blown heavyweight OS?

Confused...

Besides the dashboard and the infotainment, more recent ADAS units usually have some sort of POSIX on them. But, yes, most of the systems are RTOS'es.

You mention an android fork and complain that RHEL is only supported for 10 years. How long is that fork supported for?

Confused...

If it's an android fork with a 20 year lifetime, then I understand what it is.

Is it?

edit: I guess another way to put this is, looking at car's I wouldn't expect it to be linux. Maybe something entirely new, designed for real time, safety, and long term support. Maybe with an android style layer bolted on somewhere for the infotainment systems. Being redhat, open source of course.

So, we can expect a Fuchsia like code drop at some point? Why not just say that and post it?

Linux has kernels patched for real time support. I don't know how it compares to other RTOS offerings, but I know Red Hat has a real time distribution they put out based on RHEL 7.

I'm not sure where you're getting all the Android stuff from. They could just let the developers develop for whatever on top of it (Qt, Gtk, whatever) or they could implement some Android compatibility layer (but why bother, it's a moving target that will just be very obsolete within a couple years). I expect the automotive companies that use this would just port whatever GUIs they already have over to it at least initially.

  They could just let the developers develop for whatever on top of it (Qt, Gtk, whatever) or they could implement some Android compatibility layer (but why bother, it's a moving target that will just be very obsolete within a couple years).

No, doubt on the obsolete part, but that is already the problem they have with their existing proprietary solutions as well (see also TV, and game machine companies). Spotify isn't jumping up and down and making app updates for 10 year old Toyota's and I don't imagine the next media darling will be interested in porting their app to a 2025 era car in 2030 either.

Hence android, where in theory the car manufacturer could at least provide an app compatibility shim to run the latest entertainment apps. Its a moving target, but its a target. The other choice is just to standardize an android auto/carplay, and start treating the cars like dumb handset/display devices. Which I predict is what will happen since the car companies are even less capable than the LG/Samsungs of the world of creating and maintaining a media platform. And everyone is already in the google/apple ecosystem which doesn't show any signs of being replaced, at least not over the next 5-10 years. Basically they have already lost the war, and don't even know it. Sure they can reskin a navsystem, but no one actually uses the ones the have now they just aren't as good as asking siri to direct you to the local restaurant. Worse the car manufactures are deluded if they think they can sell map upgrades and a bunch of other connected services. I would be curious to see actual dealer surveys about why people are purchasing upgrades, because I'm betting its a case of, someone wants the leather, and they end up buying the fancy infotainment system to get it. As I mentioned elsewhere, no one uses those systems except as bluetooth handsets these days, and just about everyone I know thinks they are garbage. They are the modern equivalent of a CD player in a car and the automobile manufactures just haven't noticed. Same with remote starters and the other long list of "services" the car companies think they can charge extra for which are just fee harvesting until the next whatsapp or whatever shows up with a better freemium model.

This is actually Linux. Linux can be real time enough for parts of the safety-critical side, but it's usually not run all the way through and Linux itself has certification difficulties. You'll typically either have an RTOS side or have a high integrity super/hypervisor that can guarantee your system properties.

Anyway... The point of Linux is to help automotive software teams escape the crushing burden of traditional embedded development. The current situation is such that it's impossible to adhere to modern development best practices with typical team sizes and development workloads. Adopting commodity/standardized platforms is a huge part of how companies like GM are trying to escape that reality and the software issues automakers have become known for.

I don't think that Linux can be "realtime enough" for safety-critical functions.

But I expect that it can work just fine for interacting with those systems and monitoring, configuring, and logging them and communicating state between them.

Sort of like how you will pull your hand away from a hot stove before you become aware that you are burning it. But then once I am aware of it I can take further steps to deal with the damage.. like turning off the stove, treating the hand, etc.

Bit late to the table, Linux is already used in safety critical systems and has been for years. This LWN article [1] gives a good overview of the various approaches and some of their tradeoffs. I can provide talks from groups within major automakers talking about their own usage too, if you'd prefer.

But Linux is fine given all the caveats, limitations, and system designs are done properly. This is where most systems are headed.

[1] https://lwn.net/Articles/734694/

I really wish Red Hats Marketing was less bullshitty sometimes. I’ve been following the RHEL for Edge and RHEL Automotive development and Fedora IOT for a while now, and the technology is really nice, but this article is seemingly buzzword soup…

> Automakers are moving toward providing more personalized and intuitive driving experiences while becoming more embedded within a driver’s personal ecosystem. Doing so creates opportunities for additional revenue streams and improved customer loyalty. The Red Hat In-Vehicle Operating System provides automakers with a common, updatable platform that enables automakers to better meet customer expectations via frequent and seamless feature upgrades and deployment of new applications and services.

Honestly, this sounds horrible. This is PR speak for "Red Hat In-Vehicle Operating System will make it easier to serve ads and collect user data that can be sold."

How much can this damn' data be worth in order to risk the loyalty of people paying tens of thousands of dollars on your product? I assume the _actual_ profit they are making are in "addons" and extras that can be unlocked remotely.

Because there is no loyalty risk. They are betting on a favorable regulatory environment and apathetic consumers for the next several years.

> apathetic consumers

All automakers have to do is require access to that data for features, like maps, traffic updates, or entertainment.

I hate the idea of in-car advertising, and I don't like the thought my data being sold. But I do give up some privacy for functionality. My biggest annoyance with my car is I don't have Android Auto for the maps, I use google maps when driving, and video from my ADAS is used as training data.

Yeah I understand why they market it that way to automakers, but I wish there was a push to show what it could do for the consumer and create demand. I would love to have an open source OS that I could customize in my car. I understand there is stuff that for regulatory reasons can't change, but all the maps and music and whatnot I'd love control over

Meh, it seems like all infotainment computers get outdated a lot faster than the car, and they don't even have an advantage by being brand new because the manufacturer just slaps on something "the cheapest and slowest computer you can find, and just tolerable enough" to save costs.

The car should just offer HDMI input for the display and speakers, and allow the computer to be upgradable. Even the display should be a module (say someone wants to upgrade to 4K OLED...), which makes me think the DIN got it right: https://www.bestcaraudio.com/what-is-a-double-din-car-radio/

Yeah my first choice would be a car that doesn't have a display or lets me project what's on my phone's screen. But if manufacturers insist on some OS with a screen as an output device, and they do, having an open one would at least have the potential to stimulate innovation in available software. As it is, everything I've seem is just an afterthought or lazy attempt at a revenue stream. And it will stay that way as long as manufacturers can lock in car buyers to a proprietary closed system

I think there was a small window around 2010 where people actually used the onboard maps/GPS/etc. But every single person I know with a fancy new car + nav system/etc just uses their phone tethered over bluetooth.

So, none of that matters, in the end. All you need is a volume control, a push to talk to siri/etc button on your steering wheel and a decently placed audio pickup.

Then the upgrade happens every couple years with a new phone.

There’s a brief window after you purchase a new vehicle where the maps aren’t hopelessly out of date… and then what?

When I went looking at how to update the maps in my car I had to dig so far I ended up in the basement in a locked room with a sign that said “beware the tiger” digging through the bottom drawer of a dusty filing cabinet. I’m fairly certain that nobody had ever purchased the updates before.

And they were like $500.

Alternatively, you don’t even need a new phone… just use Google/Apple/etc maps and they’re evergreen _and_ generally contain live information on road closures, traffic density, etc.

So you get more functionality and updates with a phone for free. It’s no wonder people don’t want to use the built in nav. (Never mind that it doesn’t pause my podcasts when it starts talking over them…)

I have nav in my car because it’s basically bundled with a bunch of other creature comforts I did want. The only use I’ve ever found for it was on long road trips where I’d like a rough idea how much further to go / etc, but don’t want to have my phone screen on for dozens of hours or have to be fiddling with my phone to get progress updates. Highways don’t change often so it’s been fairly competent at that.

I leave an ancient tomtom (which had lifetime map updates that stopped after ~10 years when they said the HW wasn't supported anymore) in the glove box of my truck when i'm traveling the back roads of the USA, where cellphone coverage doesn't exist and google doesn't know the difference between a fire road, and a rancher's driveway. Its self contained and more than once I've fired it up when google was non responsive or tried to route me through a farmer's pasture with a locked gate or through an unmaintained fire road.

So self contained is definitely a strength, and I suspect that old tomtom will still be tossing out routes in 10-15 years when we all have 7G cell modems in our cars leaving all the 4G/5G ones disconnected as has been happening with the 2G ones from 10 years ago.

But the other end of the problem here is that these systems simply aren't designed properly anymore. Hence the reliance on cell modems, updates, etc. 22 years ago the two primary Linux Architectures in use today didn't even exist (amd64 and aarch64). Will the linux of 2040 even run on anything we have available today? Will the reactjs of 2040 fit in the puny 8G of RAM (or whatever) a modern car/linux machine might ship with?

I would say no, nothing seems to be turning the less efficient software boat, as we steadily head down into ever more layers of software abstraction using ever more inefficient technologies.

Anyway, my point is that the phone upgrade is basically required to run the lastest google/apple maps because that is the treadmill we are on. If you stop upgrading eventually you get dropped. I tried booting an ipad2 a couple years back, and while it booted, it had stopped getting OS updates, and the first thing it did was basically wipe out pretty much every single app, or in some cases leave them there in various dysfunctional states. I might have been able to repurpose it as a ebook reader if I was willing to crack the DRM/etc on them and load them manually.. Other than that, paperweight.

PS: Delorme Atlases are the real answer to shady GPS routing, I keep them in the glove box too :)

> just uses their phone tethered over bluetooth

Carplay/Android auto are great too (which require a physical connection or wifi). I love using GPS this way because some apps take advantage of both screens -- the the phone itself can show a list of the turn by turn directions while the large vehicle screen can show a large map and lane guidance.

Would is also make it easier to swap parts like sensors? Imagine upgrading your lidar, radar, or computer yourself because the drivers are upstreamed and standardized.

This isn’t Red Hat - already the auto makers are reselling location data from every vehicle (many/most new ones) with a cellular modem built in. GM has specifically written about monetizing vehicle occupants. No GM vehicle in my future, that’s for sure.

One could probaby whack or shield the antenna with aluminium tape though.

Is this an alternative to Android Automotive [1]?

[1] https://source.android.com/devices/automotive/start/what_aut...

No, this is related to IEEE/ASME functional safety.

This seems to be in an exploratory phase with many questions still on how to implement. I do see this as positive in general that big car makers such as GM are recognizing the benefits and willing to partner with IBM on building open source solutions.

I recently evaluated several cars for purchase and had real difficulty with all of them because of the mysterious proprietary operating systems that I know are collecting data and sending to somewhere. Cars have become Iphones on wheels which sounds great to some people, but not to me. I ultimately decided not to purchase a new car and stick with an older more analogue one until an open source modern car would be available. I might be waiting a while, but I don't think I am alone in this predicament.

Looking at android, the open source data-probe is every bit as effective as the closed-source one. The idea of controlling heavy machinery with red hat seems nearly as inadvisable as controlling it with windows.

"Computers can do better than ever what needn't be done at all. Making sense is still a human monopoly."

- Marshall McLuhan

I feel like if there was a press release about IBM repackaging Linux to sell to manufacturers, the post here would have no positive response, maybe no negative response either. I feel like either apathy or cynicism. This is probably part of the value of the Red Hat purchase.

I really don’t understand red hat. Are they just offering good enterprise support or something? I have never heard of anyone choosing Red Hat as their Linux flavor for their servers (or clients) or even in their top 10 flavors

> I have never heard of anyone choosing Red Hat as their Linux flavor for their servers (or clients) or even in their top 10 flavors

Do you mean you've never seen someone running CentOS or Rocky? It sounds like that's what you're saying, and if so, that's really surprising.

Yes. They're "the norm" in some countries and business niches ( e.g. banking, but not Germany, there it'd be SUSE), software only comes packaged and tested for their OSes, etc.

For everyone else, there's Debian/Ubuntu/whatever.

I can only add my anecdotal experience. When the Linux servers are running proprietary code that has a high support cost it would often make sense to have a paid Linux distro as well to avoid finger pointing games. For example, even if I knew what the problem was I could put Redhat and Oracle on the phone together and let them hash it out. They would have to find the root cause and that would be provided to our customers regardless of what they find. I know it sounds weak but that was a good way to shield our business from paying out downtime credits when a vendor had a serious bug and tried to blame us. We had clauses to protect ourselves when we were not negligent. I know this will be an unpopular answer but I know that many large businesses follow this logic when writing their contracts.

I’m in as long as they make the buttons perform different on every update like the touchpad in my beloved Fedora running laptop.

Oh, and turn on auto-update checking no matter how many times I manually turn it off.

Oh great, now my car can be built with node microservices /s

nice this sounds great!

Reply on news.ycombinator.com