Tesla infotainment system upgradeable from MCU1 to MCU2

[rooter]

Possibly? But the way software updates used to work is that the cars would actually make the request from the Mothership. They send along their current configuration and then a "custom" firmware gets built and sent back based on the hardware manifest. Doing it this way ensures that if Tesla's info is out of date (not sure why that would be), then reality takes precedence.

The way it's worked since I released the eMMC rooting/replacement (Swissbit) method ~2 years ago is a full firmware saveset is downloaded to the car and written into partition 1 or 2, whichever is -not- active. All ECM components for all cars are contained therein, located in 1or2/deploy/seed_artifacts_v2/ (attached).

cid-updater (née service-redeploy) then uses the car's specific hardware manifest to stage the relevant ECM components for that car, into 3/spool/cid-updater/staged. These are then (code-signing) checked and written into the car's 16 or so ECMs, stepwise. This is one reason that you must get some reasonable facsimile of your own car's partitions 3 and 4. (not to mention its carkeys.tar, vin, gateway.cfg, etc) You can partially recover many damaged parts using my ddrescue method.

My sig line is full due to forum limitations (max characters, no links) so I can't link to my DIYElectriCar articles, and I'm tired of repeating myself. I'll move them to Wikiversity when I get time.

 

[viper2ko]

I'm surprised no one in the community still didn't try to retrofit MCU1 with MCU2-IC2. If you could get around the wiring then I don't think programming it would be an issue. Although it is difficult because we don't know (At least I don't know) if MCU2 still has an IP for IC, even if it is living in MCU2 case now. (IC2 is just a display) Also AFAIK you can't SSH let alone root in the newer versions without a whitelisted rsa id so things are getting more difficult by the day. As a fiddler myself (not the discoverer-tier like green or wk but as a hobby-est let's say) my ultimate dream would be to get my 2017 S90 to HW3 and MCU2. So I can get sentry, netflix, visualizations and all other goodies. Still waiting on it to be confirmed.

Would it end there? I doubt it. I'd probably want to upgrade to a low cycle count 100 pack after my pack warranty is done after 2025.

IMO, just a matter of time before aftermarket parties start offering it. There just aren't enough old MCU2 units on the market yet

 

[nimrooz]

The way it's worked since I released the eMMC rooting/replacement (Swissbit) method ~2 years ago is a full firmware saveset is downloaded to the car and written into partition 1 or 2, whichever is -not- active. All ECM components for all cars are contained therein, located in 1or2/deploy/seed_artifacts_v2/ (attached).

cid-updater (or service-redeploy) then uses the car's specific hardware manifest to stage the relevant ECM components for that car, into 3/spool/cid-updater/staged. These are then (code-signing) checked and written into the car's 16 or so ECMs, stepwise. This is one reason that you must get some reasonable facsimile of your own car's partitions 3 and 4. (not to mention its carkeys.tar, vin, etc) You can partially recover many damaged parts using my ddrescue method.

My sig line is full due to forum limitations (max characters, no links) so I can't link to my DIYElectriCar articles, and I'm tired of repeating myself. I'll move them to Wikiversity when I get time.

So, actually, in theory, it would work if you mounted an MCU2 in the car, at least regarding to updates, if it gets in the hardware manifest.
So the question is, does it update the hardware manifest on every startup?
Another question, does it alternate the downloads between partition 1 and 2?
-if a update fails its restarting on the working partition? Much like other computer components in an example, like a network router?

However if first question holds, the remaining question will be if it works with the rest of the car and wiring.

 

[rooter]

Yes, in fact to make it easy you could just move the Tegra board (CID) from your MCU1 to 2.

One of my working theories is that the "gateway" is actually the boot coprocessor inside the Tegra chip, which in fact runs the firmware in the Spansion chip on the obverse of the CID. If this is the case the gateway config would move with the CID. I know that the whizkid claims the gateway is some other nebulous chip, but I'll believe it when proven.

And I suspect that the hardware manifest is actually gateway.cfg. This has most if not all the requirements, and so no it's not updated on every startup. All of us are exploring, and as I'm the only one to publicize info many have built on my work.

Yes the CID is an nVidia board (bespoke for Tesla), so it uses their method of alternating partitions as is used in their high-end video cards. This is smart because you can upgrade and verify the new firmware, before committing to booting to it. Only after it is confirmed good (using code-signing) is it allowed to be booted to. This bit is flipped in the Spansion chip.

PS -- Hate the title of this thread, BTW. "Musk", like he's the enemy or something. I believe that he will be the most influential person of the 21st Century. He's already changed the destiny of the auto and space industries.

 

[tyson]

@SmartElectric
Please elaborate.

 

[bmah]

Yes the CID is an nVidia board (bespoke for Tesla), so it uses their method of alternating partitions as is used in their high-end video cards. This is smart because you can upgrade and verify the new firmware, before committing to booting to it. Only after it is confirmed good (using code-signing) is it allowed to be booted to. This bit is flipped in the Spansion chip.

PS -- Hate the title of this thread, BTW. "Musk", like he's the enemy or something. I believe that he will be the most influential person of the 21st Century. He's already changed the destiny of the auto and space industries.

Two random comments:

nVidia didn't invent the idea of alternating partitions for holding firmware. I'm sure you know that, but just getting that out there for people who might not. We did that at my work, before Tesla Motors even existed, and it was not a new idea then.

The original thread title was "Musk confirms upgradeability from MCU1 to MCU2". A moderator (possibly me) edited the thread title at one point because people were getting excited about Elon announcing MCU2 upgrades, but then getting their hopes dashed when service centers said "nope". The intent was not to disparage Elon but to temper expectations.

Thanks...now back to lurking...

Bruce.

 

[NoirMS]

Yes, in fact to make it easy you could just move the Tegra board (CID) from your MCU1 to 2.

One of my working theories is that the "gateway" is actually the boot coprocessor inside the Tegra chip, which in fact runs the firmware in the Spansion chip on the obverse of the CID. If this is the case the gateway config would move with the CID. I know that the whizkid claims the gateway is some other nebulous chip, but I'll believe it when proven.

And I suspect that the hardware manifest is actually gateway.cfg. This has most if not all the requirements, and so no it's not updated on every startup. All of us are exploring, and as I'm the only one to publicize info many have built on my work.

Yes the CID is an nVidia board (bespoke for Tesla), so it uses their method of alternating partitions as is used in their high-end video cards. This is smart because you can upgrade and verify the new firmware, before committing to booting to it. Only after it is confirmed good (using code-signing) is it allowed to be booted to. This bit is flipped in the Spansion chip.

PS -- Hate the title of this thread, BTW. "Musk", like he's the enemy or something. I believe that he will be the most influential person of the 21st Century. He's already changed the destiny of the auto and space industries.

your PS is dead on.

 

[emir-t]

Yes, in fact to make it easy you could just move the Tegra board (CID) from your MCU1 to 2.

One of my working theories is that the "gateway" is actually the boot coprocessor inside the Tegra chip, which in fact runs the firmware in the Spansion chip on the obverse of the CID. If this is the case the gateway config would move with the CID. I know that the whizkid claims the gateway is some other nebulous chip, but I'll believe it when proven.

And I suspect that the hardware manifest is actually gateway.cfg. This has most if not all the requirements, and so no it's not updated on every startup. All of us are exploring, and as I'm the only one to publicize info many have built on my work.

Yes the CID is an nVidia board (bespoke for Tesla), so it uses their method of alternating partitions as is used in their high-end video cards. This is smart because you can upgrade and verify the new firmware, before committing to booting to it. Only after it is confirmed good (using code-signing) is it allowed to be booted to. This bit is flipped in the Spansion chip.

PS -- Hate the title of this thread, BTW. "Musk", like he's the enemy or something. I believe that he will be the most influential person of the 21st Century. He's already changed the destiny of the auto and space industries.

Haven't seen an Intel board of an MCU2 but is it using the same method of alternating between partitions and does it also have an emmc for all and a rom chip for bootloader? If so do they share the same BGA? If so getting an MCU2 and soldering it on with your MCU1 hynix & spansion chips would make this too easy.

Also I thought gateway files were in the SD card? (not the nav sd card)

However we still don't know;

- If antennas and all other connections would work with this new setup.
- If the .101 IP still exists (maybe virtually) with MCU2 (because MCU2 IC is now only a display.)
- If partitions and structure is still the same.

Very little people working on MCU2 cars with root access these days and all are keeping their cards close to their chest. So we will remain with unknowns until someone proves it. Miss the 2014-2016 days for the Tesla tinkerer community.

 

[SmartElectric]

MCU1 to MCU2 is plug and play as well for AP2.0+ vehicles. Now this assumes your ok with not having the additional Bluetooth and 5Ghz WiFi.

If you are implying that MCU1 and MCU2 are a compatible upgrade step, that has not been substantiated. Evidence from the technical community has been the opposite, namely that MCU2 and it’s IC are of a different design entirely than MCU1 and it’s IC.

There have been no substantiated upgrades performed either by Tesla or the aftermarket.

 

[tyson]

If you are implying that MCU1 and MCU2 are a compatible upgrade step, that has not been substantiated. Evidence from the technical community has been the opposite, namely that MCU2 and it’s IC are of a different design entirely than MCU1 and it’s IC.

There have been no substantiated upgrades performed either by Tesla or the aftermarket.

Define substantiated? In this very thread several people with root access and the physical hardware have stated that there is compatibility for AP 2.0+ vehicles. That includes @verygreen and @BigD0g. Who are you waiting for to say it's possible?

 

[SmartElectric]

Define substantiated? In this very thread several people with root access and the physical hardware have stated that there is compatibility for AP 2.0+ vehicles

Let’s see the practical advancement of an actual upgrade before claiming victory. The hardware design has fundamental differences that cannot be simply hand waved away without seeing a valid working upgrade. That’s my point. Hope is not a plan. Looking to see hard evidence.

 

[tyson]

Fair enough, I don't think it will be too long. Keep an eye on @greentheonly on twitter....

 

[Tesomega]

Yea, but they are already turning profit away by discouraging annual services, so why invite more people to service centers for the $500 upgrade?

At least for now, every service visit with Tesla has been 1 week for me, that's about $700-$800 on UBER Bucks. So, I am thinking money earned by Tesla should be the cost of the upgrade + rental/loaner/uber credits + some margin, if not they are not going to offer it.

 

[Jedi2155]

Well then, tell my why Ford motors don't have a problem selling me parts for a focus if I own a mondeo?
-it is that straight forward, I have money and want to buy a part with partnumber xxxx.
Never ever had a problem with buying parts to whatever, even if I don't own a car that the part fits on.
........
Ahh, but if I trade in my S, the package would get cheaper I hope

I think you mean "core charge" instead of trade in value

Two random comments:

nVidia didn't invent the idea of alternating partitions for holding firmware. I'm sure you know that, but just getting that out there for people who might not. We did that at my work, before Tesla Motors even existed, and it was not a new idea then.

Bruce.

I was kind of under the impression this came from the aerospace field as late as the 70s with the Voyager probes if not earlier. I always recall space probes having a Side A and Side B for redundancy and I suspect it being the same in our vehicles.

 

[whitex]

PS -- Hate the title of this thread, BTW. "Musk", like he's the enemy or something. I believe that he will be the most influential person of the 21st Century. He's already changed the destiny of the auto and space industries.

The headline is just stating a fact, not judging or making anyone out to be an enemy. Elon's prior achievements in life don't change facts - in this case he tweeted something but reality didn't match.

 

[HankLloydRight]

The headline is just stating a fact, not judging or making anyone out to be an enemy. Elon's prior achievements in life don't change facts - in this case he tweeted something but reality didn't match.

Totally agree. Thread title is accurate statement of fact.

 

[BigD0g]

Define substantiated? In this very thread several people with root access and the physical hardware have stated that there is compatibility for AP 2.0+ vehicles. That includes @verygreen and @BigD0g. Who are you waiting for to say it's possible?

*cough* *cough*. Twitter

 

[BigD0g]

Please also note, no pin swaps or harness retrofits were required as expected.

 

[tyson]

Let’s see the practical advancement of an actual upgrade before claiming victory. The hardware design has fundamental differences that cannot be simply hand waved away without seeing a valid working upgrade. That’s my point. Hope is not a plan. Looking to see hard evidence.

*cough* *cough*. Twitter

Well I did say it wouldn't be long right.... There you have it. His X was an MCU1 AP2.5 car prior to the hardware swaps.

 

[viper2ko]

Per green, "I cannot imagine why anybody would opt to NOT do the mcu2 retrofit after this experience (current component pricing around ~$1600 from junkyards). You only need: MCU2, IC2 and the new tuner (can skip that even, e.g. I don't have it yet, but then updates are complicated)."