2 cents (Score 1)

I have been making my living writing software for various automotive applications from BCMs to instrument clusters for about 15 years. I currently do advanced r&d for a tier 1. A recent project was to evaluate and document the major OSes and their applications in the telematics realm. The findings were what we usually find in selecting supporting software. Different OSes work in different situations. There is no clear path and different scenarios require different solutions much like the PC application domain. Outside of telematics like an instrument cluster our OSes consist of a while loop and not much more. The requirements from the OEMs dictate that the resources for auto apps are pretty small in order to keep the costs down. If they can shave a penny from cost you can bet an engineering change is coming your way. For instance, the last BCM we did had 256k of rom, 12k of ram. This includes compliance to foreign regulations on the same module. Meaning headlamp and horn behavior will be different outside the U.S.

I think the upshot for Linux is that it is a solid low cost alternative that will become more attractive as the cost to manufacture a higher end chip that supports a real OS goes down. Most people don't realize that your modern automobile has about 17 to 20 different computer systems. They all run apps of some sort. It would seem like a logical step to eventually combine some of these systems and control them with a real OS.
We'll see.