Re: [PATCH] drm/virtio: fix host visible memory detection in virtio-gpu

[Dmitry Osipenko <dmitry.osipenko@collabora.com>]

On 8/27/25 11:12, Honglei Huang wrote:
> From: Honglei Huang <Honglei1.Huang@amd.com>
> 
> Commit 206cc44588f7 ("virtio: reject shm region if length is zero")
> enhanced the validation in virtio_get_shm_region() by adding a check
> for a zero-length shared memory region.
> 
> It is performed before the underlying transport's .get_shm_region()
> implementation is called. This creates an issue in the virtio-gpu
> driver, where the `region` struct is part of a larger structure
> that is zero-initialized by drmm_kzalloc().
> 
> Consequently, the `len` field is 0 at the time of the check, causing
> virtio_get_shm_region() to return false prematurely. This prevents the
> host visible memory feature from being enabled, even when the device
> supports it.
> 
> To resolve this, this patch bypasses the inline helper and calls the
> underlying vdev->config->get_shm_region() function pointer directly.
> This ensures that the region's parameters are checked only after they
> have been populated by the transport, aligning with the intended logic.
> 
> Signed-off-by: Honglei Huang <Honglei1.Huang@amd.com>
> ---
>  drivers/gpu/drm/virtio/virtgpu_kms.c | 6 ++++--
>  1 file changed, 4 insertions(+), 2 deletions(-)
> 
> diff --git a/drivers/gpu/drm/virtio/virtgpu_kms.c b/drivers/gpu/drm/virtio/virtgpu_kms.c
> index 7dfb2006c561..ed5981248302 100644
> --- a/drivers/gpu/drm/virtio/virtgpu_kms.c
> +++ b/drivers/gpu/drm/virtio/virtgpu_kms.c
> @@ -174,8 +174,10 @@ int virtio_gpu_init(struct virtio_device *vdev, struct drm_device *dev)
>  	if (virtio_has_feature(vgdev->vdev, VIRTIO_GPU_F_RESOURCE_BLOB)) {
>  		vgdev->has_resource_blob = true;
>  	}
> -	if (virtio_get_shm_region(vgdev->vdev, &vgdev->host_visible_region,
> -				  VIRTIO_GPU_SHM_ID_HOST_VISIBLE)) {
> +	if (vgdev->vdev->config->get_shm_region &&
> +	    vgdev->vdev->config->get_shm_region(
> +		    vgdev->vdev, &vgdev->host_visible_region,
> +		    VIRTIO_GPU_SHM_ID_HOST_VISIBLE)) {
>  		if (!devm_request_mem_region(&vgdev->vdev->dev,
>  					     vgdev->host_visible_region.addr,
>  					     vgdev->host_visible_region.len,

Hi, virtio_get_shm_region() change has been reverted by [1]. Don't think
anything else needs to be done.

[1]
https://git.kernel.org/pub/scm/linux/kernel/git/next/linux-next.git/commit/?h=next-20250827&id=ced17ee32a9988b8a260628e7c31a100d7dc082e

+cc Michael Tsirkin

Might be only good to send a stable kernel PR with that revert. I see
patch available only in linux-next, while stable kernels need to be
fixed sooner.

-- 
Best regards,
Dmitry