Hi,
I’m using JP 4.5.1, TX2 NX module.
After received the module from store, I do first flashing JP 4.5.1 image.
But I found that the EEPROM is totally corrupted.
By modifying flash-helper.sh script, I got eeprom binary likes this:
may I know how you flash the target? flash script will have USB communication to check eeprom for the board information before image flashing.
so, it should be okay if you’re able to flash the target.
BTW,
please try adding -f options to force access to the device even it’s busy.
for example, $ sudo i2cdump -f -y 7 0x50
actually, there’re two eeprom for parsing board info, one is cvb, the other is cvm.
they’re based-on different i2c address. they’ve contain some board-specific information.
for example,
are you using TX2 NX DevKits?
also, where’s your cvm.bin came from? it’ll dump and store the info as cvm.bin during image flash. flash may failed if the content is corrupted.
you may try adding SKIP_EEPROM_CHECK=1;into flash commands, this will ignore the checking.
Of course we are talking about CVM EEPROM with address 0x50. My cvm.bin came from flash.sh script (see log above Saved platform info in /work/works/tools/nvidia_sdk/sdk/JetPack_4.5.1_Linux_JETSON_TX2_NX/Linux_for_Tegra/bootloader/cvm.bin)
By adding BOARDID=3636 FAB="300" BOARDSKU="0001" BOARDREV="" before flash.sh in command, it is also ignore EEPROM checking. That why I’m able to flash the firmware.
But my concern is, why the EEPROM reading by I2C is so strange? I have many TX2NX (more than 5), but only one of them has this issue.
Is there any customer having similar problem? Just to check if it can happens for other TX2NX modules in future use or not.
it’s strange, and I’ve never see this before.
my point is… it is flash script to communicate with the board to dump and store the info as cvm.bin. if there’s failure, it might be something wrong on hardware side.
it looks like hardware issue since you’ve mentioned only one of them has this issue.
please contact the NVIDIA Customer Care team for the RMA process.
thanks
Hi Jerry,
Yes, it is very strange.
The cvm.bin looks quite correct since the hardware information “699-…” string is fine and matching TX2NX info.
The I2C write looks good also (no error message, and when readback by flash.sh into cvm.bin got written value).
Just I2C reading in Linux has problem :(.
My code reads EEPROM to check if it is TX2NX or Nano and behave differently.
My company actually bought some hundreds TX2NX for production, I will check if this happens again on other boards.
We bought via distributor, RMA process takes time and need invoice/order batch. So will not go with that process.