Welcome to Tesla Motors Club
Discuss Tesla's Model S, Model 3, Model X, Model Y, Cybertruck, Roadster and More.
Register

Preventive eMMC replacement on MCU1

This site may earn commission on affiliate links.
Can someone here give me some procedure for this job? I'm a DIY guy.

Tools for de-soldering.
Tools for copying the old eMMc to the new one.
What brand of the eMMc is good to use?
@keon my HowTo's may help. I'm in the process of moving to Wikiversity, since DIYE uses Cloudflare which stupidly bans me for daring to deconstruct the shadow file.

De/resoldering the eMMC requires a good deal of skill, and:
- JBC infrared soldering station
- Kester solder balls
- NC-256 flux
 
I'm moving forward in my emmc recovery.
P3 and P4 clean, but P1 and P2 I am not sure how to check.
here an image
I think I'm not checking well p1 and p2.
Anyone know how to get it right?
I try it with fsck
 

Attachments

  • IMG_1458.JPG
    IMG_1458.JPG
    564.9 KB · Views: 211
p1 and p2 are squashfs and can’t me fsck.

There are only used Read Only. Can you mount p1 or p2?
Right, if you can mount them they're Ok. If you can do that check the firmware version in each one:
# cat /mnt/deploy/platform.ver

You must not change versions on these respective partitions, as one or the other is active and is keyed to the boot flash using code-signing. If the system is running you can get the running version by pressing 'T' at the top.

For parts 3 & 4, set aside copies of the files and use e2fsck to check them. Forget fsck for ext* filesystems.

PS - widodh are you the author of that image-building script? If so have you tested it in practice?
 
Last edited:
98% memory usage is normal. That's what Linux (what runs on the MCU) does. It will use up all available memory.

This has nothing to do with the eMMC which is the permanent storage of the MCU.

My Linux laptop which I'm typing this on uses 95% of it's 16GB of memory. Mainly caches. Memory (DRAM) is much faster then the NVMe SSD in my laptop and therefor Linux caches as much as it can.

The wear level of the eMMC is something different. A SSD/eMMC has a finite number of writes it can accept. Usually there is a counter which keeps track of the health. Some other people report the SeC can read out this value.
I understand what you are saying. The SC indicated this was not normal but said on exit it was operating normally. Who knows what they put in they replaced it for yellow border (before UV trick).
 
Right, if you can mount them they're Ok. If you can do that check the firmware version in each one:
# cat /mnt/deploy/platform.ver

You must not change versions on these respective partitions, as one or the other is active and is keyed to the boot flash using code-signing. If the system is running you can get the running version by pressing 'T' at the top.

For parts 3 & 4, set aside copies of the files and use e2fsck to check them. Forget fsck for ext* filesystems.

PS - widodh are you the author of that image-building script? If so have you tested it in practice?
Yes, I wrote that script and it works. It only contains a empty image. You need to write the contents to it.
 
I'm moving forward in my emmc recovery.
P3 and P4 clean, but P1 and P2 I am not sure how to check.
here an image
I think I'm not checking well p1 and p2.
Anyone know how to get it right?
I try it with fsck

You can check the integrity of p1/p2 very easy on a Windows pc with the tool 7-zip, use the "Test archive" function.
You should get only one warning/error "There are some data after the end of the payload data", this is good/normal:
7zipok.png


If you get any other errors like these then I would strongly suggest to search for a copy of the firmware file so you can use that instead of the corrupted dump.
7zipnotok.png
 
bad news
P1 and P2 cannot be opened so I assume they are wrong.
I only have two possibilities:
1-transfer the car 700km to the sc and 3000 EUR.
2- someone would be kind enough to share their p1 and p2 files with firmware 2019.40.2.3 40ef2d4d1432
 

Attachments

  • IMG_1459.JPG
    IMG_1459.JPG
    1.2 MB · Views: 176
  • IMG_1460.JPG
    IMG_1460.JPG
    1.2 MB · Views: 162
If you're sure that's the version you need, it's easy. There's a complete library of firmware images (which I don't have access to anymore). Maybe someone will front you the image.

However do -not- try to write just any dd'ed image. They must be trimmed first using a hex editor. The images are only signed, and the signature is 64 bytes of non-zero data close to the end of the first or second partition. Cid-updater validates the signature after download. To be valid as an installed image, all you have to do is trim the file -AFTER- the signature. There is an area with 00 before the signature, and then after the 64 byte signature... cut this latter.

Also remember, on the new chip you can't just create partitions starting at block 0. There is null space before partition 1. wido has made a script to set up the parts, although there's a mystery block of gibberish within that null space. Best to restore a full image of the chip.
 
If you're sure that's the version you need, it's easy. There's a complete library of firmware images (which I don't have access to anymore). Maybe someone will front you the image.

However do -not- try to write just any dd'ed image. They must be trimmed first using a hex editor. The images are only signed, and the signature is 64 bytes of non-zero data close to the end of the first or second partition. Cid-updater validates the signature after download. To be valid as an installed image, all you have to do is trim the file -AFTER- the signature. There is an area with 00 before the signature, and then after the 64 byte signature... cut this latter.

Also remember, on the new chip you can't just create partitions starting at block 0. There is null space before partition 1. wido has made a script to set up the parts, although there's a mystery block of gibberish within that null space. Best to restore a full image of the chip.
ok
If there is someone who can help me with these 2 partitions and a friend will win forever in Spain.