On 11/12/24 04:39, Jan Beulich wrote:
> On 12.11.2024 10:02, Roger Pau Monné wrote:
>> On Mon, Nov 11, 2024 at 03:07:28PM -0500, Stewart Hildebrand wrote:
>>> On 10/28/24 14:41, Roger Pau Monné wrote:
>>>> if ( !pdev->info.is_virtfn && !list_empty(&pdev->vf_list) )
>>>> {
>>>> struct pci_dev *vf_pdev;
>>>>
>>>> while ( (vf_pdev = list_first_entry_or_null(&pdev->vf_list,
>>>> struct pci_dev,
>>>> vf_list)) != NULL )
>>>> {
>>>> list_del(&vf_pdev->vf_list);
>>>> vf_pdev->virtfn.pf_pdev = NULL;
>>>> vf_pdev->broken = true;
>>>> }
>>>>
>>>> printk(XENLOG_WARNING "PCI SR-IOV PF %pp removed with VFs still
>>>> present\n",
>>>> &pdev->sbdf);
>>>> }
>>>
>>> Yeah. Given that the consensus is leaning toward keeping the PF and
>>> returning an error, here's my suggestion:
>>>
>>> if ( !pdev->info.is_virtfn && !list_empty(&pdev->vf_list) )
>>> {
>>> struct pci_dev *vf_pdev;
>>>
>>> list_for_each_entry(vf_pdev, &pdev->vf_list, vf_list)
>>> vf_pdev->broken = true;
>>>
>>> pdev->broken = true;
>>
>> Do you need to mark the devices as broken? My expectation would be
>> that returning -EBUSY here should prevent the device from being
>> removed, and hence there would be no breakage, just failure to fulfill
>> the (possible) hot-unplug request.
>
> That very much depends on Dom0 kernels then actually respecting the error,
> and not considering the underlying hypercall a mere notification.
All dom0 Linux does is print a warning:
# echo 1 > /sys/bus/pci/devices/0000\:01\:00.0/sriov_numvfs
# echo 1 > /sys/bus/pci/devices/0000\:01\:00.0/remove
[ 56.738750] 0000:01:00.0: driver left SR-IOV enabled after remove
(XEN) Attempted to remove PCI SR-IOV PF 0000:01:00.0 with VFs still present
[ 56.749904] pci 0000:01:00.0: Failed to delete - passthrough or MSI/MSI-X
might fail!
# echo $?
0
Subsequently, lspci reveals no entry for 0000:01:00.0. I think it's
appropriate to mark them broken.
