Can't boot existing Unraid

[carlosparamio]

Hello everyone,

 

I had a running Unraid server on a PC with a Gigabyte GA-970A-DS3P motherboard and an AMD Athlon II X4 640, with a purchased license associated to a SanDisk SDCZ33-016G-B35 device. Then several hard drives with a couple parity drives and a cache drive, and several docker services running.

 

This system has been working fine, but lately I've noticed that it struggled with Plex running at Docker when playing 4K videos (the 4 cores went to 100% easily, and still the video stops continuously). So I decided to move to a Ryzen 5 2600 processor on a Gigabyte GA-AX370-Gaming K7 where I'll possibly consider the setup of a VM for gaming. I purchased all the hardware, did the substitutions, and then realised that the motherboard doesn't recognise the new processor until it receives a BIOS upgrade (doh!), for which it is required an older AMD Athlon processor that uses AM4 socket, or the first generation of Ryzen. I've purchased a cheap compatible old AMD processor just for this purpose, but now I need to wait to receive it so I can start to use the new hardware.

 

Then I simply rolled back my hardware upgrade and installed again my previous motherboard. And the hell started. I can't make UNRAID to boot from the Sandisk USB drive anymore. It always ends up with a few "device descriptor read/64, error -32" messages, to end up not finding the drive with a "UNRAID does not exist", and due to not being able to mount the volume, it isn't capable of loading network modules and such.

 

I've tried everything. I tried to unplug all drives and unnecessary cards and the only other USB device I have (the UPS). I've tried to plug in the USB device on every single USB port at the motherboard (and even the external ones). I've tried to reconfigure the BIOS to force usage of UEFI and renaming the "EFI-" folder at the drive to "EFI" (I was using legacy before, because my first Unraid version was older from this EFI support if I remember well). I've checked that the drive has "UNRAID" as label. I've passed fsck and diskutil checks / repairs to the drive from my Mac (no corruption, no repairing needed). I've even used a different USB drive with a brand new setup of UNRAID using the USB creator. Nothing of this has helped. A USB portable keyboard/mouse didn't work after boot (it works fine at the EFI bios), so I ended up plugging a PS2 keyboard to operate a bit at the command line. I can see there are no disks listed at /dev.

 

I don't know what else to do. But I'm a bit desperate, because it looks like I can't revert to the previous setup and get access to my files. Is there anything else do you think I might have missed?

 

PD: I've run diagnostics, but since the file is created at what seems to be a ram virtual drive, and I have no network or anything, I wasn't able to copy it anywhere else.

[carlosparamio]

I just tried something else to discard issues with the UNRAID system that I can't explain: I prepared the secondary USB drive with a copy of a bootable Ubuntu system. And guess what... I got the same "device descriptor read/64, error -32" messages! So it sounds like somehow the chipset in charge of managing the USB devices on this board stopped working somehow with the transplant.

[carlosparamio]

Oh wow... I switched to a search about this error with the Gigabyte motherboard, and found this: https://unix.stackexchange.com/questions/72625/why-is-usb-not-working-in-linux-when-it-works-in-uefi-bios#answer-72698

 

I enabled "IOMMU Controller" at the BIOS, and suddenly everything started to work again!

 

Now it's time to research what this option means. I still wonder how is that suddenly needed this to boot (perhaps it was enabled before and somehow got disabled?).

 

Well, at least everything is solved now.

[freezing57]

I'm so happy you posted this thread.  I upgraded my UNRAID server mobo to a Gigabyte GA-970A-DSP3 also and could not for the life of me figure out why I kept getting an APIPA address.  I must have tried 35 different things until I ran across this thread.  That was it.  IOMMU Controller needed to be ENABLED in my mobo BIOS.

Edited September 4, 2020 by freezing57