Re: [PATCH v3 3/3] iommu/ipmmu-vmsa: Implement basic PCIE-IPMMU OSID support

[Andrew Cooper <andrew.cooper3@xxxxxxxxxx>]

On 21/08/2025 11:15 am, Mykyta Poturai wrote:
> diff --git a/xen/drivers/passthrough/arm/ipmmu-vmsa.c 
> b/xen/drivers/passthrough/arm/ipmmu-vmsa.c
> index dac0dd6d46..49f149e222 100644
> --- a/xen/drivers/passthrough/arm/ipmmu-vmsa.c
> +++ b/xen/drivers/passthrough/arm/ipmmu-vmsa.c
> @@ -1138,8 +1155,38 @@ static int ipmmu_assign_device(struct domain *d, u8 
> devfn, struct device *dev,
>      if ( !to_ipmmu(dev) )
>          return -ENODEV;
>  
> -    spin_lock(&xen_domain->lock);
> +#ifdef CONFIG_HAS_PCI
> +    if ( dev_is_pci(dev) )
> +    {
> +        struct pci_dev *pdev = dev_to_pci(dev);
> +        struct domain *old_d = pdev->domain;
> +
> +        /* Ignore calls for phantom functions */
> +        if ( devfn != pdev->devfn )
> +            return 0;
> +
> +        write_lock(&d->pci_lock);
> +        list_move(&pdev->domain_list, &d->pdev_list);
> +        write_unlock(&d->pci_lock);
> +        pdev->domain = d;
> +
> +        /* dom_io is used as a sentinel for quarantined devices */
> +        if ( d == dom_io )
> +        {
> +            int ret;
> +
> +            /*
> +             * Try to de-assign: do not return error if it was already
> +             * de-assigned.
> +             */
> +            ret = ipmmu_deassign_device(old_d, dev);
> +
> +            return ret == -ESRCH ? 0 : ret;
> +        }
> +    }
> +#endif
>  
> +    spin_lock(&xen_domain->lock);

Minor point.  You've removed a newline here, which is why the hunk is
not cleanly ahead of the spin_lock() call.

~Andrew