NHTSA needed to decide that Tesla had to replace its MCUs for the company to take action. Sadly, it disputed the need for a recall arguing that these computers were wear parts but accepted to replace them anyway. The issue was that the eMMC (embedded multimedia card) did not stand so much overwriting and eventually failed. The hacker GreenTheOnly just revealed that Tesla did not eliminate the cause: excessive logging still happens with the MCUs.
According to GreenTheOnly, it seems Tesla engineers decided that the eMMC with more capacity can stand endless write cycles like the ones that led the original eMMC 8 GB to fail. One of these logs was the computer registering every 30 seconds that the eMMC was in good shape. Ironically, these logs are being stored on the eMMC in question, which should lead them to fail as well, even if after more time than the ones Tesla first presented.
As software gets more sophisticated, its requirements also increase. The first Tesla firmware image size went from 300 MB to 1 GB in recent releases. That shows that hardware tends to age poorly with increasing demands from multiple updates, especially if it has to log stuff that IT people such as GreenTheOnly repute to be useless.
The MCU had two versions, and the last one died with the latest refresh the Model S and Model X received. These vehicles now use the same ICE computer that Tesla introduced with the Model 3. Despite that, and if logging is not controlled, it would be no surprise if the MCUv2 also became a wear part.
Tesla may take GreenTheOnly’s warning seriously and work to only log critical information, or it can ignore it until new MCUs start failing again. If they do so when warranty coverage is no longer valid, Tesla may charge owners for replacing these computers as it used to do with the MCUv1 when it could not cope with logging anymore.
Hopefully, Tesla will at least erase the replaced computers or properly destroy them not to show their data. GreenTheOnly also discovered that MCUs and ICEs were ending up on websites with all their previous owners’ information on them.
As software gets more sophisticated, its requirements also increase. The first Tesla firmware image size went from 300 MB to 1 GB in recent releases. That shows that hardware tends to age poorly with increasing demands from multiple updates, especially if it has to log stuff that IT people such as GreenTheOnly repute to be useless.
The MCU had two versions, and the last one died with the latest refresh the Model S and Model X received. These vehicles now use the same ICE computer that Tesla introduced with the Model 3. Despite that, and if logging is not controlled, it would be no surprise if the MCUv2 also became a wear part.
Tesla may take GreenTheOnly’s warning seriously and work to only log critical information, or it can ignore it until new MCUs start failing again. If they do so when warranty coverage is no longer valid, Tesla may charge owners for replacing these computers as it used to do with the MCUv1 when it could not cope with logging anymore.
Hopefully, Tesla will at least erase the replaced computers or properly destroy them not to show their data. GreenTheOnly also discovered that MCUs and ICEs were ending up on websites with all their previous owners’ information on them.
Hm, this "let's log how healthy our EMMC is to the EMMC itself". every 30 seconds... that's lovely too
— green (@greentheonly) August 9, 2021
I am sure it's super important to do it this frequently
and then new updater-envoy that busyloops trying to connect to down hermes (when car is asleep), backing off for 0.1 sec pic.twitter.com/KsMFB5HUev