Flaminis
Member
And I did not understand all this fuss with the supply voltage of the emmk, pull-up resistors and expensive programmers.
Above I met sound ideas, let me add.
1. Voltage.
5 volts, preferably from a power bank - to the input power capacitors of the Tegra board.
2. Wires.
The length is not particularly critical, the maximum on which I tried 20 centimeters. But shorter is better, and with good wires. For in-circuit work, I use the so-called MGTF with a cross section of 0.2 - a copper stranded wire in fluoroplastic insulation.
3. Pull-up resistors.
Not required as they are available on the Tegra board.
4. Connection points.
There are 5 of them. CLK, CMD, GND, D0, D1 (connect this pin in parallel with GND)
5. Blocking the Tegra processor.
Jumper to the quartz resonator.
And jumper D1-GND forbids the processor to access EMMC.
6. Reader.
Any USB SD card reader.
7. Content of sections.
FW versions for p1 and p2 partitions can be found in /deploy/platform.ver.
Workable images of these two partitions are best taken from the IC - they are exactly the same as on the MCU. If you know exactly which firmware version is relevant for your MCU, it is enough to write one image to both partitions.
Section p3 strongly recommend reading and saving file-by-file, and not as an image - with a check of what has been preserved - especially certificates and the mcu command key. The most important thing is to keep the /etc folder completely, the rest is secondary.
Section p4 does not have to be written to a new EMMC, but if it is read stably, then it is enough to save and transfer only the tesla folder. The system will add the rest to p4 itself.
8. Magazines
The reasons for MCC failures can be viewed in /log/syslog by searching for the keywords squashfs, mmcblk0p, error with a command like <cat syslog | grep squashfs> (execute in file directory)
The D0 exchange rate is low, reading a full EMMC dump takes about 20 minutes.
Of course, I use the programmer for in-circuit work, but functionally it is completely identical to the SD reader. This circuit has been successfully used for about 5 years, several dozen MCU have been repaired.
Alternative connection - CLK, CMD, GND, D0-D3, jumper to quartz (do not connect D1 to GND) time not less than 20 minutes) - and you can freely access EMMC via 4 data lines
Above I met sound ideas, let me add.
1. Voltage.
5 volts, preferably from a power bank - to the input power capacitors of the Tegra board.
2. Wires.
The length is not particularly critical, the maximum on which I tried 20 centimeters. But shorter is better, and with good wires. For in-circuit work, I use the so-called MGTF with a cross section of 0.2 - a copper stranded wire in fluoroplastic insulation.
3. Pull-up resistors.
Not required as they are available on the Tegra board.
4. Connection points.
There are 5 of them. CLK, CMD, GND, D0, D1 (connect this pin in parallel with GND)
5. Blocking the Tegra processor.
Jumper to the quartz resonator.
And jumper D1-GND forbids the processor to access EMMC.
6. Reader.
Any USB SD card reader.
7. Content of sections.
FW versions for p1 and p2 partitions can be found in /deploy/platform.ver.
Workable images of these two partitions are best taken from the IC - they are exactly the same as on the MCU. If you know exactly which firmware version is relevant for your MCU, it is enough to write one image to both partitions.
Section p3 strongly recommend reading and saving file-by-file, and not as an image - with a check of what has been preserved - especially certificates and the mcu command key. The most important thing is to keep the /etc folder completely, the rest is secondary.
Section p4 does not have to be written to a new EMMC, but if it is read stably, then it is enough to save and transfer only the tesla folder. The system will add the rest to p4 itself.
8. Magazines
The reasons for MCC failures can be viewed in /log/syslog by searching for the keywords squashfs, mmcblk0p, error with a command like <cat syslog | grep squashfs> (execute in file directory)
The D0 exchange rate is low, reading a full EMMC dump takes about 20 minutes.
Of course, I use the programmer for in-circuit work, but functionally it is completely identical to the SD reader. This circuit has been successfully used for about 5 years, several dozen MCU have been repaired.
Alternative connection - CLK, CMD, GND, D0-D3, jumper to quartz (do not connect D1 to GND) time not less than 20 minutes) - and you can freely access EMMC via 4 data lines