Re: [PATCH v2 2/5] vfio/pci: Release the VGA arbiter client on register_device() failure

[sashiko-bot@kernel.org]

Thank you for your contribution! Sashiko AI review found 1 potential issue(s) to consider:

Pre-existing issues:
- [Low] Unconditional assignment of PM callbacks pointer without pm_ptr() wrapper.
--

commit 009bb147bb78eda6b9cfb1c508c98573c30df2f1
Author: Alex Williamson <alex.williamson@nvidia.com>

vfio/pci: Release the VGA arbiter client on register_device() failure

This commit adds the required VGA unwind callout to the failure path
of vfio_pci_core_register_device. A previous re-order displaced the VGA
initialization without introducing an unwind.

> diff --git a/drivers/vfio/pci/vfio_pci_core.c b/drivers/vfio/pci/vfio_pci_core.c
> index 9f71eae0cc947..d2d3fddec6107 100644
> --- a/drivers/vfio/pci/vfio_pci_core.c
> +++ b/drivers/vfio/pci/vfio_pci_core.c
> @@ -2254,6 +2254,7 @@ int vfio_pci_core_register_device(struct vfio_pci_core_device *vdev)
>  		pm_runtime_get_noresume(dev);
>  
>  	pm_runtime_forbid(dev);

[Severity: Low]
This is a pre-existing issue, but I noticed an unconditional assignment of
the PM callback pointer earlier in vfio_pci_core_register_device() in
drivers/vfio/pci/vfio_pci_core.c:

dev->driver->pm = &vfio_pci_core_pm_ops;

Does this code need to use the pm_ptr() wrapper? When CONFIG_PM is disabled,
assigning the callbacks unconditionally can produce dead code because the
functions get linked into the kernel even though they can never execute.

> +	vfio_pci_vga_uninit(vdev);
>  out_vf:
>  	vfio_pci_vf_uninit(vdev);
>  	return ret;

-- 
Sashiko AI review · https://sashiko.dev/#/patchset/20260611213539.4100590-1-alex.williamson@nvidia.com?part=2