linux-kernel - [REGRESSION] virtio: reject shm region if length is zero

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <kgasjsq2s4pshravsinycfihdfjhdts5iz2fox42aejs4xqhce@frurksambnk3>
Date: Fri, 15 Aug 2025 21:09:52 +0200
From: Alyssa Ross <hi@...ssa.is>
To: SamiUddinsami.md.ko@...il.com
Cc: mst@...hat.com, jasowang@...hat.com, xuanzhuo@...ux.alibaba.com, 
	eperezma@...hat.com, virtualization@...ts.linux.dev, linux-kernel@...r.kernel.org, 
	Sami Uddin <sami.md.ko@...il.com>, regressions@...ts.linux.dev
Subject: [REGRESSION] virtio: reject shm region if length is zero

On Mon, May 12, 2025 at 07:51:53AM +0930, SamiUddinsami.md.ko@...il.com wrote:
> From: Sami Uddin <sami.md.ko@...il.com>
>
> Prevent usage of shared memory regions where the length is zero,
> as such configurations are not valid and may lead to unexpected behavior.
>
> Signed-off-by: Sami Uddin <sami.md.ko@...il.com>
> ---
> v3:
> - Use idiomatic 'if (!region->len)' as suggested by reviewer
> v2:
> - Fixed coding style issue: added space after 'if' statement
>
>  include/linux/virtio_config.h | 2 ++
>  1 file changed, 2 insertions(+)

Hi, I'm sorry to be the bearer of bad news, but since this patch my VM
no longer works.  The system is running wayland-proxy-virtwl[1] inside
a crosvm[2] VM, using crosvm's virtio-gpu device to do cross-domain
Wayland forwarding.

Since this change, wayland-proxy-virtwl crashes with the following log
message:

	wl-proxy [WARNING]: Error handling client: Unix.Unix_error(Unix.EINVAL, "DRM_IOCTL_VIRTGPU_RESOURCE_CREATE_BLOB", "")

I'm pretty confused by what this change was supposed to do in the first
place…  Looking at how virtio_get_shm_region() is used in
virtio_gpu_init(), it's called with a pointer to zeroed memory, and then
the get_shm_region() implementation is supposed to write to the region,
without ever reading from it as far as I can tell.  Why is the initial
value of an out parameter being checked at all?  How does this prevent
using zero-length shared memory regions?

[1]: https://crosvm.dev/
[2]: https://github.com/talex5/wayland-proxy-virtwl

#regzbot introduced: 206cc44588f72b49ad4d7e21a7472ab2a72a83df

> diff --git a/include/linux/virtio_config.h b/include/linux/virtio_config.h
> index 169c7d367fac..b3e1d30c765b 100644
> --- a/include/linux/virtio_config.h
> +++ b/include/linux/virtio_config.h
> @@ -329,6 +329,8 @@ static inline
>  bool virtio_get_shm_region(struct virtio_device *vdev,
>  			   struct virtio_shm_region *region, u8 id)
>  {
> +	if (!region->len)
> +		return false;
>  	if (!vdev->config->get_shm_region)
>  		return false;
>  	return vdev->config->get_shm_region(vdev, region, id);
> --
> 2.34.1
>

Download attachment "signature.asc" of type "application/pgp-signature" (229 bytes)