M.2 nvme SSD detection issue

Hi,

I asked you to add “nvidia,enable-power-down” to access the register, without it kernel panic is expected.
Register dump in failure case tell us that link is stuck in compliance mode.

Yes, nvidia,enable-power-down” will not impact PCIe link status. You can continue with disabling NVMe in bootloader as WAR for you work. I will continue to track this internally.

I am marking comment #21 as the solution.

Thanks,
Manikanta

Hi Manikanta,

Thanks for your reply. As you said before “I asked you to add “nvidia,enable-power-down” to access the register, without it kernel panic is expected.”. The interesting thing is the kernel panic happened when “nvidia,enable-power-down” existed and NVMe (Kingston OM8PDP3256B-AB1) was on. It seems the this NVMe made PCIe link disappeared.
The setting “nvidia,enable-power-down” is default existed in the node “pcie@141a0000” within dtsi file. But once I removed the setting “nvidia,enable-power-down” in the node “pcie@141a0000”, we can see the PCIe link and dump the register without kernel panic and the NVMe device (Kingston OM8PDP3256B-AB1) was still on. I am not sure the method of removing NVMe device from booting option is a solution or workaround.

Thanks,

Kunyang

Hi,

“nvidia,enable-power-down” has no impact on functionality. This options should be removed to let controller be in power up state to dump the register. You can ignore kernel panic error when “nvidia,enable-power-down” is present, this is expected.

Are you seeing any issue after disabling NVMe in bootloader?

Thanks,
Manikanta

Hi Manikanta,

No, it seems all good after disabling NVMe in bootloader and we can use the NVMe in OS. But it may remain the limitation of no boot option with NVMe I guess.

lspci

0005:00:00.0 PCI bridge: NVIDIA Corporation Device 1ad0 (rev a1)
0005:01:00.0 Non-Volatile memory controller: Kingston Technology Company, Inc. Device 500d (rev 01)

Thanks,

Kunyang

Hi Manikanta,

Thanks for your help. The NVMe device “(Kingston OM8PDP3256B-AB1)” can work in the Ubuntu after removing NVMe deice as a boot option. But we still hope NVMe device can be a boot option since the eMMC has storage size capacity limitation. Moreover, this device we have already purchased a lot. We are not sure how many NVMe device may have the similar interoperability problem in the bootloader. Could you help check the root cause of why this device “(Kingston OM8PDP3256B-AB1)” can’t work in the bootloader ?

Thanks,

Kunyang

Hi~ Manikanta & Wayne,
Add more information for this issue discussion.

Thanks
Ken

Hi,

  1. Are you saying that this issue only happened to emmc module? Have you done enough tests on multiple boards + modules?

  2. Are you sure these two modules are under the same test environment? For example, both have nvme in cboot enabled/disabled.

Hi~Wayne,

  1. Are you saying that this issue only happened to emmc module? Have you done enough tests on multiple boards + modules?
    Q1–>Yes, this issue just happend to eMMC mdule Xavier NX PN: 180-13668-DAAA-A03.
    Q2 → Please check wayne_liao’s reply on Jan 11

  2. Are you sure these two modules are under the same test environment? For example, both have nvme in cboot enabled/disabled.

Q1 & Q2 → Please check wayne_liao’s reply on Jan 11

Thanks
Ken

Hi,

Since bootloader is getting involved, could you share the uart log from both devices?

What you should do is

  1. (NTFS case) Disable nvme in cboot for both boards, share the uart log and dmesg.

  2. (NTFS case) Enable nvme in cboot for both boards, share the uart log and dmesg.

  3. (ext4 case) Enable nvme in cboot for both boards, share the uart log and dmesg.

Also, could you probe that de-assert signal for each case here?

Hi~Wayne,
Please refer to test results as below:
M.2 nvme SSD detection issue test result_20220314A.xlsx (9.5 KB)
Xavier SSD issue UART log.zip (119.0 KB)
Xavier SSD issu dmesg log.zip (115.0 KB)

Thank
Ken

Hi~ Manikanta & Wayne,
Sorry to bother you. Do you have any progress on this issue?

Thanks
Ken

No update yet.

I’m also having an issue with M.2 NVMe SSD using the latest SDK Manager.
My NVMe PCIe Gen 4.0 (Samsung 980 Pro) is not being not detected on the Jetson Xavier NX when attempting flashing using command:

sudo ./nvsdkmanager_flash.sh --storage nvme0n1p1

The script runs without errors for a long time but then it quits when it should start flashing the disk.
What should I do?

Hi~ Manikanta & Wayne,
Sorry to bother you. Do you have any progress on this issue?

Thanks
Ken

Hi,

Please apply these two to your cboot.

9eeaf45.diff.zip (1.4 KB)
1b5b11a.diff.zip (2.8 KB)

Hi~Wayne,
I test it on Xavier Devkit. The problem still exists.
Xavier_SSD_issue_dmesg_UART log_20220401.txt (107.2 KB)

Thanks
Ken

Hi,

Just want to clarify this.

Is this issue only reproducible with Kingston OM8PDP3256B-AB1 NVMe drive?

Hi~Wayne.
Yes, It just happened on Kingston series SSD.

KST/OM8PDP3128B-AB1/EDFK0SA6
KST/OM8PDP3256B-AB1/EDFK0SA6

Thanks
Ken

Hi~ Wayne,
Sorry to bother you. Do you have any progress on this issue?

Thanks
Ken