Re: [PATCH] vfio/pci: clear vdev->msi_perm after freeing it on init failure

[sashiko-bot@kernel.org]

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

Pre-existing issues:
- [Critical] Heap buffer overflow in vfio_cap_init() and vfio_ecap_init() due to missing bounds check on capability length.
--

commit 5867deb8ee022404d313dea6794c1dc0a44460cb
Author: Xiang Mei <xmei5@asu.edu>

vfio/pci: clear vdev->msi_perm after freeing it on init failure

This commit clears the vdev->msi_perm pointer after freeing it during an
init_pci_cap_msi_perm() failure. This prevents a use-after-free and a double
free of the MSI permission table when vfio_config_init() fails.

> diff --git a/drivers/vfio/pci/vfio_pci_config.c b/drivers/vfio/pci/vfio_pci_config.c
> index a10ed733f0e3a..9914f3ac69aef 100644
> --- a/drivers/vfio/pci/vfio_pci_config.c
> +++ b/drivers/vfio/pci/vfio_pci_config.c
> @@ -1272,6 +1272,7 @@ static int vfio_msi_cap_len(struct vfio_pci_core_device *vdev, u8 pos)
>  	ret = init_pci_cap_msi_perm(vdev->msi_perm, len, flags);
>  	if (ret) {
>  		kfree(vdev->msi_perm);
> +		vdev->msi_perm = NULL;
>  		return ret;
>  	}

[Severity: Critical]
This is a pre-existing issue, but there appears to be a heap buffer overflow
due to missing bounds checks on capability lengths.

In vfio_cap_init(), if a PCI device advertises a capability where the offset
'pos' plus length 'len' exceeds 'pdev->cfg_size', can this memset() write out
of bounds on the heap?

drivers/vfio/pci/vfio_pci_config.c:vfio_cap_init() {
    ...
		memset(map + pos, cap, len);
		ret = vfio_fill_vconfig_bytes(vdev, pos, len);
    ...
}

Similarly, in vfio_ecap_init(), extended capabilities are not bounds-checked
against the buffer size:

drivers/vfio/pci/vfio_pci_config.c:vfio_ecap_init() {
    ...
		memset(map + epos, ecap, len);
		ret = vfio_fill_vconfig_bytes(vdev, epos, len);
    ...
}

This out-of-bounds write would then continue into vfio_fill_vconfig_bytes():

drivers/vfio/pci/vfio_pci_config.c:vfio_fill_vconfig_bytes() {
    ...
			*dwordp = cpu_to_le32(dword);
    ...
}

Should we validate that the capability length fits within the configuration
space buffer before performing these writes?

-- 
Sashiko AI review · https://sashiko.dev/#/patchset/20260705014010.1297885-1-xmei5@asu.edu?part=1