Using a Samsung 960 EVO 250GB, the device works as expected after flashing.
I am curious, would someone from NVIDIA know of a method (perhaps via “/sys”) for adding PCIe bus enumeration logging at an extremely verbose level? I’d be curious if some sort of logging might show evidence that the NVMe presence was noted, but rejected, versus not even being noticed at a low level.
We are looking into this issue.
Kangalow,
Could you share the full dmesg when ssd is not detected? I saw there was a log “coldbootdmesg.txt”, but seems a ok case.
It’s here in post #9: Lost NVMe SSD on L4T 31.1.0 upgrade - Jetson AGX Xavier - NVIDIA Developer Forums
This is the boot log after the flash. A warm restart log does not appear to be different.
Can you please try patches mentioned in
https://devtalk.nvidia.com/default/topic/1042559/jetson-agx-xavier/no-pcie-link-with-some-devices/post/5291033/#5291033 ?
Kangalow,
Our team also tried WD SSD (256GB), but cannot see any error.
Was this with the patch or without?
This is with default release codebase. Have you tried getting the link up without really touching the card between old release and new release? I’m wondering if there is any physical movement of the card in between because of which it might be appearing as if the issue with release??
I have one of the early engineering samples of the Xavier, so it could very well be that there are issues that have been resolved after going into production. It is unlikely there is an issue with physical movement of the card, as the card is captured with a screw and has been reseated multiple times during testing. Marked as closed issue.