Re: [PATCH v7 4/5] vfio/pci: implement get_tph and DMA_BUF_TPH feature

[sashiko-bot@kernel.org]

Thank you for your contribution! Sashiko AI review found 2 potential issue(s) to consider:
- [Medium] Unconditional validation of the flag-gated 'ph' parameter can lead to spurious -EINVAL errors, breaking the syscall trust boundary.
- [Medium] VFIO_DEVICE_FEATURE_PROBE falsely reports that VFIO_DEVICE_FEATURE_DMA_BUF_TPH is supported even if the hardware lacks TPH support.
--

commit d57637d667ca6dc95def77a0a5bd339d62d4f506
Author: Zhiping Zhang <zhipingz@meta.com>

vfio/pci: implement get_tph and DMA_BUF_TPH feature

This commit implements dma-buf get_tph for vfio-pci exported dma-bufs and
adds the VFIO_DEVICE_FEATURE_DMA_BUF_TPH feature. This allows userspace to
publish TPH metadata for a VFIO-owned device. It checks completer support
and protects the metadata under dmabuf->resv.

> diff --git a/drivers/vfio/pci/vfio_pci_dmabuf.c b/drivers/vfio/pci/vfio_pci_dmabuf.c
> index 1a177ce7de546..0a0705c8dbea6 100644
> --- a/drivers/vfio/pci/vfio_pci_dmabuf.c
> +++ b/drivers/vfio/pci/vfio_pci_dmabuf.c

[ ... ]

> @@ -333,6 +360,71 @@ int vfio_pci_core_feature_dma_buf(struct vfio_pci_core_device *vdev, u32 flags,
>  	return ret;
>  }
>  
> +int vfio_pci_core_feature_dma_buf_tph(struct vfio_pci_core_device *vdev,
> +				      u32 flags,
> +				      struct vfio_device_feature_dma_buf_tph __user *arg,
> +				      size_t argsz)
> +{
> +	struct vfio_device_feature_dma_buf_tph set_tph;
> +	struct vfio_pci_dma_buf *priv;
> +	struct dma_buf *dmabuf;
> +	u8 comp;
> +	int ret;
> +
> +	ret = vfio_check_feature(flags, argsz, VFIO_DEVICE_FEATURE_SET,
> +				 sizeof(set_tph));
> +	if (ret != 1)
> +		return ret;

[Severity: Medium]
vfio_check_feature() returns 0 for VFIO_DEVICE_FEATURE_PROBE, causing an 
early return of success. 

Should the hardware capability check (pcie_tph_completer_type) occur before
checking the feature flags, to avoid falsely reporting that the feature is
supported on devices that lack TPH?

> +
> +	if (copy_from_user(&set_tph, arg, sizeof(set_tph)))
> +		return -EFAULT;
> +
> +	if (set_tph.flags & ~(VFIO_DMA_BUF_TPH_ST | VFIO_DMA_BUF_TPH_ST_EXT))
> +		return -EINVAL;
> +
> +	if (set_tph.ph & ~0x3)
> +		return -EINVAL;

[Severity: Medium]
When userspace clears the TPH metadata by passing flags == 0, is it required
to zero-fill the unused ph field? 

Since this check is unconditional, could stack garbage in the uninitialized 
ph field cause spurious -EINVAL errors, rejecting a valid invalidation 
request?

> +
> +	comp = pcie_tph_completer_type(vdev->pdev);
> +	if (comp == PCI_EXP_DEVCAP2_TPH_COMP_NONE)
> +		return -EOPNOTSUPP;

[ ... ]

-- 
Sashiko AI review · https://sashiko.dev/#/patchset/20260610193158.2614209-1-zhipingz@meta.com?part=4