Mailing List Archive: enable msix get err / PCI frontend error when trying to pass through NVMe

Hi All,

I am trying to pass through an NVMe device using pci-passthrough.
All seems to go fine until I load the "nvme" module.

Check "dmesg" I see the following errors:

[Mon Mar 18 07:59:19 2024] nvme 0000:81:00.0: Xen PCI mapped GSI56 to IRQ269
[Mon Mar 18 07:59:19 2024] nvme nvme0: Shutdown timeout set to 8 seconds
[Mon Mar 18 07:59:19 2024] nvme nvme0: allocated 64 MiB host memory buffer.
[Mon Mar 18 07:59:20 2024] nvme 0000:81:00.0: enable msix get err ffffff8e
[Mon Mar 18 07:59:20 2024] nvme 0000:81:00.0: Xen PCI frontend error: -114!
[Mon Mar 18 07:59:20 2024] nvme nvme0: 1/0/0 default/read/poll queues

Any advice on how to resolve this would be greatly appreciated.
The devices do show in /dev/... , but with that error, I'm not certain
I can reliably use the drive.

Many thanks,

Joost Roeleveld

Hi Joost and everyone,

On 2024-03-18 08:16, Joost Roeleveld wrote:
> Hi All,
>
> I am trying to pass through an NVMe device using pci-passthrough.
> All seems to go fine until I load the "nvme" module.
>
> Check "dmesg" I see the following errors:
>
> [Mon Mar 18 07:59:19 2024] nvme 0000:81:00.0: Xen PCI mapped GSI56 to
> IRQ269
> [Mon Mar 18 07:59:19 2024] nvme nvme0: Shutdown timeout set to 8
> seconds
> [Mon Mar 18 07:59:19 2024] nvme nvme0: allocated 64 MiB host memory
> buffer.
> [Mon Mar 18 07:59:20 2024] nvme 0000:81:00.0: enable msix get err
> ffffff8e
> [Mon Mar 18 07:59:20 2024] nvme 0000:81:00.0: Xen PCI frontend error:
> -114!
> [Mon Mar 18 07:59:20 2024] nvme nvme0: 1/0/0 default/read/poll queues

I get exactly the same errors in the logs for my NVME drives. I have not
encountered any
data issues so far (not production yet but have performed extensive
bench testing using
FIO).

> Any advice on how to resolve this would be greatly appreciated.
> The devices do show in /dev/... , but with that error, I'm not certain
> I can reliably use the drive.

It would be good to at least know it is nothing to worry about :-)

> Many thanks,
>
> Joost Roeleveld

Regards,

Leigh.