Passthrough of PCI devices to VM in TrueNAS Core causes checksum errors and SMART self test inability.
Description
Problem/Justification
Impact
Attachments
- 17 Dec 2023, 05:59 PM
- 17 Dec 2023, 05:59 PM
Activity
Chris Cox December 17, 2023 at 5:59 PM
@William Gryzbowski I migrated over to SCALE and can confirm that passthrough is working as expected. However, now I cannot get anything on the data protection page to load. It just pinwheels forever. I know there are tasks in there because scrubs and snapshots have been running fine but I am unable to make any adjustments to them.
William Gryzbowski November 30, 2023 at 3:07 PM
VM support is a hundred times better in SCALE one way or another, so I think you will be doing yourself a favor if you care about that use case.
Let us know what you find, please.
Chris Cox November 29, 2023 at 8:40 PM
@William Gryzbowski I was planning on having that be a rainy day project but if that’s what I have to do to get working passthrough I guess it is what it is.
William Gryzbowski November 29, 2023 at 7:27 PM
@Chris Cox Yes, GELI does not work in SCALE.
One way or another we suggest people to move away from GELI even in CORE, as users are no longer able to create that kind of pool nowadays.
Chris Cox November 29, 2023 at 7:21 PM
@William Gryzbowski I have a GELI encrypted pool with no spare storage for getting that data off of there. It’s my understanding that GELI encryption cannot migrate over from CORE to SCALE?
* Device: /dev/ada1, not capable of SMART self-check. * Device: /dev/ada0, not capable of SMART self-check.
The above error only happens when a PCI device is passed through to a VM and the VM is started. It does not matter which PCI device is passed through, nor does it matter which PCI port it is used in (or indeed, onboard). I don’t really have an explanation for this because in my mind it shouldn’t be related but I’ve tested everything I can think of and the conclusion is inescapable. No Passthrough → No Error. Passthrough → Error within 1 minute of VM start.
I ran my drives for 3 months with no PCI passthrough on any VM just to verify and there was no issues. My disks are fine, so far as I can tell.
Some background on ada1 and ada0: These two disks are in a Z2 pool with 6 disks in it total. The pool has legacy encryption. The VM has storage within this pool (but I’ve already tried migrating it to another unencrypted pool and it didn’t help).
Happy hunting.