CM4: Booting from NVMe works fine, but reboot got stuck with "failed to open nvme" #378
Comments
|
Sounds quite similar to raspberrypi/Raspberry-Pi-OS-64bit#188 ? |
|
I've tried to reproduce this but it seems to work for me. I've tried a LITE and 8GB EMMC CM4. I'm using the waveshare board mentioned in raspberrypi/Raspberry-Pi-OS-64bit#188 that has the resistor fix, but I don't know why that would be relevant. I'm using a different make of NVME disk. |
|
@ctschach if I sent you a version of the bootloader with some extra logging - would you be prepared to update the eeprom on your CM4 using USBBOOT and repro the problem? |
Absolutely.... |
|
Ok, first attempt... pieeprom.original1.zip cd usbboot\recovery switch [RPI]boot switch on and connect your PC to the USB type-C interface to program the eeprom with this image. Switch off the boot switch and power cycle and repro the problem. If you can attach any UART logging that would be helpful |
|
So here is what happened: I flashed the new EEPROM and booted the RPi from the NVMe card. This ended up in a kernel panic (see below). Since then the NVMe became unusable. If I boot from an SD card, I see the NVMe disk, but an fsck giving me error. Putting it into an external enclosure and attaching it to another Raspi did not work either. The interesting thing: It worked directly before changing the EEPROM. I don't think that this could have destroyed the NVMe disk - maybe just some unfortunate coincidences. I've now ordered two new (and different) disks which should arrive early next week. I'll test them and report back. ´´´ |
|
Oh dear... |
|
I've got the SSD you mentioned and can reproduce the problem you were seeing. |
|
I can repro the issue on a RPI IO board so it doesn't seem to be an issue with the Waveshare board. |
No sure if I'm allowed to say: "great that you are facing the same problem" :-) But definitely good that it's wasn't my fault. My other two disks are still in transit. I'll definitely test them both and will report back. |
|
The hardware is indicating a pcie error which seems to "appear" after initialisation. To recover we have to re-initialise the hardware. I have a change that fixes the issue for me - can you give it a try to confirm? |
|
Are you sure this is the correct EEPROM? I I do see an "Unknown boot mode 0x6. Trying SD." in the output on the console. I though the I need boot-mode "6" for NVME boot, but looks like this eeprom you just gave me do not support it. |
|
So sorry to waste your time. I picked an old branch. Here's the right one, and I've actually tested it! It should show up as version e0c6d67c |
|
Okay, that one works better...much better :-) But this is a different brand of NVMe card. The other one is already on it's way back to the vendor. Not sure it I get a replacement or a refund. |
|
Thanks, much appreciated. I'll keep this open until a fix is released. |
|
The fix for this should be in released now. |
Mandatory information
CM4 with 4 GB Memory and no eMMC - attached to https://www.waveshare.com/wiki/CM4-IO-BASE-A
Revision : c03140
Rasbian Bullseye
Describe the bug
The system boots without any trouble from the attached NVMe disk. No problem in running the machine, no hiccups, everything works fine. However, a simple "reboot now" got stuck which the following error on the boot console:
This loops endless. The only way to restart the machine is by power-cycling.
Bootloader version and configuration
Please include the bootloader version and config.
NVMe boot (please complete the following information):
Additional context
A dump of the successful boot log when turning on the power:
The text was updated successfully, but these errors were encountered: