Re: [PATCH v2 5/7] vfio: Use new interfaces for MSI/MSI-X enablement

[Alex Williamson <alex.williamson@redhat.com>]

On Wed, 2014-01-08 at 08:57 +0100, Alexander Gordeev wrote:
> Signed-off-by: Alexander Gordeev <agordeev@redhat.com>
> ---
>  drivers/vfio/pci/vfio_pci_intrs.c |    8 ++++----
>  1 files changed, 4 insertions(+), 4 deletions(-)
> 
> diff --git a/drivers/vfio/pci/vfio_pci_intrs.c b/drivers/vfio/pci/vfio_pci_intrs.c
> index 641bc87..66d1746 100644
> --- a/drivers/vfio/pci/vfio_pci_intrs.c
> +++ b/drivers/vfio/pci/vfio_pci_intrs.c
> @@ -482,15 +482,15 @@ static int vfio_msi_enable(struct vfio_pci_device *vdev, int nvec, bool msix)
>  		for (i = 0; i < nvec; i++)
>  			vdev->msix[i].entry = i;
>  
> -		ret = pci_enable_msix(pdev, vdev->msix, nvec);
> -		if (ret) {
> +		ret = pci_enable_msix_range(pdev, vdev->msix, nvec, nvec);
> +		if (ret < 0) {
>  			kfree(vdev->msix);
>  			kfree(vdev->ctx);
>  			return ret;
>  		}
>  	} else {
> -		ret = pci_enable_msi_block(pdev, nvec);
> -		if (ret) {
> +		ret = pci_enable_msi_range(pdev, nvec, nvec);
> +		if (ret < 0) {
>  			kfree(vdev->ctx);
>  			return ret;
>  		}

Based on your description, this is a user visible API change.  We now
return success so long as we allocated at least a single vector and the
user has no way to know that they didn't get all the vectors they
requested.  That's unacceptable, userspace expects the old API - setup
the requested vectors or setup none and tell me how many to retry with.
To maintain the same API exposed to userspace, I'd think we need
something like:

if (ret != nvec) {
	if (ret > 0)
		pci_disable...
	kfree(...
	kfree(...
	return ret;
}

Thanks,
Alex