On Thu, Sep 12, 2024 at 06:17:29AM -0700, Yi Liu wrote:
> @@ -1242,6 +1245,28 @@ int iommufd_get_hw_info(struct iommufd_ucmd *ucmd)
> if (device_iommu_capable(idev->dev, IOMMU_CAP_DIRTY_TRACKING))
> cmd->out_capabilities |= IOMMU_HW_CAP_DIRTY_TRACKING;
>
> + cmd->out_max_pasid_log2 = 0;
> +
> + if (dev_is_pci(idev->dev)) {
> + struct pci_dev *pdev = to_pci_dev(idev->dev);
> + int ctrl;
> +
> + if (pdev->is_virtfn)
> + pdev = pci_physfn(pdev);
Don't we do this twice now?
Let's just keep it in the pci core?
It looks Ok otherwise
Reviewed-by: Jason Gunthorpe <j...@nvidia.com>
Jason
