cross-posted from: https://programming.dev/post/368257
Thoughts?
Honestly just anti-foss rambling. Nothing is stopping them to make a custom hardened kernel with what they need. What they want is someone else to cater for them.
ELISA (Enabling Linux in Safety Applications) Project announced that Boeing has joined as a Premier member, marking its commitment to Linux and its effective use in safety critical applications. Hosted by the Linux Foundation, ELISA is an open source initiative that aims to create a shared set of tools and processes to help companies build and certify Linux-based safety-critical applications and systems
I imagine this means they’re contributing both actively and financially to Linux.
A bunch of bullshit so that Boeing can sell the shit they want to sell.
I’d be interested in hearing what they are using for safety-critical OS. Notice they said “software engineering” and not “OS”, which makes me think they’re running on Windows.
Most Windows drivers also run in kernel mode.[1]
For avionics, I doubt that they would use a traditional os. As far as I’m aware, Microsoft doesn’t have safety-certified builds of Windows with a real time kernel. Certifying a Linux build would also be a huge and costly endeavor. What they are likely using is a certified RTOS, like Vxworks, RTEMS, ThreadX, SafeRTOS, etc., or even Ada with the Ravenscar profile. You don’t really “develop” applications for these, you instead incorporate them as libraries inside your application and compile the RTOS into your application, and then run it on bare metal. Infotainment systems on the other hand will use more traditional OSes.
A lot of the presentation seems to be rather typical of the aerospace industry, which is all about safety. Im not too convinced that this is due to Boeing being Boeing, but rather that DO-178 compliance is a bitch, ITAR can be another bitch, and certifying not only a single build of the Linux kernel but also an entire distro build will be a superhuman effort. At best it’ll take a long time with a sizeable team. Not sure that would Boeing be filling to fund that.
I doubt they run on windows tbh. If they take issue with with monolithic design of Linux, then windows would be an even bigger problem.
Also, most of the devices in question are probably small controllers, incapable of running windows. (Microsoft are struggling to run it on arm so…)
I agree that a small, special purpose OS would probably be more suitable for safety-critical systems. On the other hand I highly doubt that the safety-culture is better at Boeing than in the Linux ecosystem.
I’d expect it to be about the same, with 737 MAX, yes, on one side and too many examples on the other.
You forget to take into account that every Boeing employee knows they are building systems that can kill people if things go wrong. Meanwhile on Linux a lot of bugs really don’t matter that much, especially in -rc and otherwise non LTS versions.
Taking that into account their safety culture is much worse.
Linux and the open source community may be chaotic … but companies like Boeing are completely corporate and they will risk or even sacrifice safety if it means making a few extra million or preventing the loss of millions in profits. They’ll calculate how much it will cost to make settlements with the families of the dead or in to issuing changes or recalls and figure out which is cheaper … pay off the dead or fix the problem. If paying off the dead is cheaper, they don’t mind watching the body count.
The only real “problem” would be the lack of certifications, which are quite hard to get.
Real Time Operating Systems (RTOS) are normally used for these tasks, but, AFAIK there are already projects using linux with patches to make it run a RTOS kernel.
In my opinion, I think it all depends on what part of the plane it is running. If it is a core sensor, providing real time data, it makes a lot of sense to use a RTOS. It needs to prove it can run its tasks on time, and the scheduler needs to be understandable. There’s also a lot of overhead with running a full OS with processes, which don’t make sense for a sensor which only function is to provide data over a CAN/LIN bus.
But, for other things, like dashboard visualizations, music for the aircraft, entertainment, and those non-critical-realtime needs, then it makes a lot of sense to run linux. After all, you’d get access to a lot of already built software and a working dev environment.
And don’t get me wrong, this is clearly BS from boeing to keep selling their closed source software. There are already open source RTOS systems, like FreeRTOS. I do not mean to keep those real time systems closed, but to use a full OS where it makes sense and a RTOS where that makes more sense. Both open source!
That makes sense, and yeah I imagine the problem isn’t the entertainment system.
I just don’t get the the last paragraph. I don’t know if using Linux affects their code being OS or not. If they’re just running it on top of Linux and not modifying it, it probably won’t be a GPL violation to keep it closed.
I remember there were talks about merging the patches and making it an option when building. I don’t know the current status of that.
On real time operating systems, like freertos, not only the kernel is real time but everything else is too. Like: you can guarantee your call on the I2C and SPI won’t take more than 5ms, for example, even with hardware issues. The whole environment is built around the hardware realtime concept.
LinuS does not have a software engineering culture
What?
Something something personal attack on Linus, whom still manages to this day every merge in master.
You know, he has no software engineering culture developing his software engineering masterpiece over 20 years, as opposed to the impeccable software engineering culture at Boeing.
Absolute clowns.
