Hacker News new | past | comments | ask | show | jobs | submit login

This is rather disingenuous. Of course Tesla doesn't replace the eMMC; that's a soldered on chip. No car manufacturer in the world does chip level electronics repairs. Instead you replace the entire board, and I have no doubt you can get this repair from Tesla.

That's not to entirely excuse Tesla. If you have a modern operating system running from flash that needs to work for 20+ years as people expect from a car, it needs to be very carefully designed - all logging and runtime data written only to a RAM disk, system and user data on entirely separate partitions, etc.

You are right but at the same time one does not expect a component of the dashboard to actually brick a car. I guess.

While counter-intuitive, I have actually encountered this issue with other cars as well, specifically a Ford f150 where an internal problem in the APIM module (accessory protocol interface module, essentially the dash touchscreen controller) essentially bricked the truck by spamming the canbus with erroneous signals. Remember that modern automobiles are rolling networks with multiple interconnected controllers, some of which are required for the vehicle to function.

Note that this doesn't excuse Tesla here, since the situation I discussed is very rare, normally if that module fails the vehicle will still start and run. Tesla engineers should absolutely have been aware of this issue, as pointed out up thread there are multiple tutorials for ras-pi SD memory preservation, and I have trouble believing a competent EE shouldn't be aware of life issues due to eMMC. It also shouldn't brick the car, normally automotive electronics are designed very carefully to avoid single points if failure, with fallback routines and safety "limp-home" modes in case of problems.

bricked the truck by spamming the canbus with erroneous signals.

Wasn't preventing this one of the design goals/selling points of CAN?

Yes, that is one of the strong points of CANBUS. Like I said in the original post, this was a very rare failure. However, the APIM managed to spam the bus in just the right way where it de-synced modules; when I initially connected to the vehicle w/ a snap-on scan tool it was throwing codes for BCM and TCM non-comm, as well as codes that implied the ECM was seeing different speeds on the CKP & CMP (crankshaft & camshaft sensors). The CKP/CMP disagreement was what caused the vehicle to be 'bricked', since the engine management had no idea where the crankshaft & valves were in relation to each other.

Near as I could tell from my scope, the APIM was spamming the bus with exactly the right frequency to interrupt the ECM during it's scan of critical sensors. It was an extremely rare failure, and to Ford's credit they covered both the repair as well as my shop's diagnostic time.

edit: To make it clear, I have seen 2 vehicles that still operated with a direct CANBUS short to ground, as well as a vehicle that had CANBUS shorted to 12V+. In these cases, aside from expected failures (such as the BCM systems not responding, or transmission limp-home), modules were able to fall back into either safe states (limp-home, in the case of the TCM) or just a dashboard warning light (in the case of BCM no-comms).

Wow! Unbelievable. I mean, just SEPARATION.

Thanks for the anecdote!

“Bricked” is supposed to mean “permanently and unfixably ceasing to function”. That’s absolutely not what is happening here. The article is completely and inexcusably wrong on this point.

The car won’t run when the eMMC chip fails, and Tesla solution is a new MCU board which costs $2,700 out of warranty. Not surprisingly Tesla is not desoldering and reworking just the eMMC chip.

There are any number of components that can disable a car, from the battery to the starter to something with the ignition, electronics, anti-theft, etc.

Sometimes the repair is as simple as a new battery, sometimes the repair is an expensive piece of hardware.

Yet those things are crucial to the fundamental operation of the vehicle. We're talking about logs no user cares about here. It's just sloppy and a silly failure mode.

Anti-theft is not crucial to the operation of the vehicle, and that had certainly immobilized plenty of vehicles, sometimes in expensive to fix manors.

Now if the article was that Tesla Model S has a chip which wears out and the board holding the chip is expensive to replace, and maybe even getting into why don’t they push a software update to lower the writes to that chip — I would not disagree.

The article falsely claimed the cars are “bricked”. As it’s the main thrust of the article, it should be retracted.

> However, until the company starts stocking parts like the eMMC chip, as well as release detailed service manuals to the public, Tesla is going to be looking at a number of newish cars dead in a junkyard real soon.

They should stock a chip which is soldered to the board, and what, do reworks? That’s asinine.

Newish cars dead in a junkyard? Totally false. It’s an expensive repair for a problem that could have been avoided, and hopefully Tesla will remedy with a firmware update.

I'm not sure you understand the situation.

This is the board the chip is on, which Tesla does not offer to replace: https://share.icloud.com/photos/0u78AylGb9fv8QHQnyr9IC3nA

This is what Tesla will offer to replace: https://share.icloud.com/photos/0MVG8edJheaHYiEp7yWKkVePg

Is that what you were imagining?

According to the video, the log is not used for anything and is basically unnecessary for the most part.

Why a full memory chip logging data which is never used should brick a car is simply price gouging.

There is no need for it whatsoever.

Thousands of dollars to replace a board because of a full memory chip? That is even worse than Apples repair racket, at least in Apple's case some other circuitry besides memory is faulty - https://www.youtube.com/watch?v=2yJKix17yYE

Erm, not full exactly. This is more like a rechargable battery that cannot handle being charged anymore. The old battery is not "empty", it's dead. This chip is not "full", it is dead. The lifetime physical endurance limits are exceeded, and functionality is lost.

Except Apple batteries die because of physics, Tesla has this issue because Tesla decided to:

1. Log something they never read back

2. Use a soldered down chip (they use a SD card in other locations)

3. Crash the media controls when the unused logs cannot be written

4. Disable certain car features, potentially immobilizing the vehicle because the media center is off.

There are several ways that Tesla could have prevented this issue, and the fact that they've never bothered to resolve the issue in later iterations is just baffling.

I read your argument, but I do not understand what you are taking exception to.

Smart phone batteries degrading is inevitable, these issues with Tesla’s cars aren’t, therefore it’s a bad comparison.

> all logging and runtime data written only to a RAM disk, system and user data on entirely separate partitions, etc.

I suspect "we can fix it in a software update" lowers the priority of actually shipping this design.

and then add in "autopilot this year" and "model y coming soon", you get indefinite postponement.

Guidelines | FAQ | Support | API | Security | Lists | Bookmarklet | Legal | Apply to YC | Contact