[Qemu-devel] a small issue with VSOCK

[Qemu-devel] a small issue with VSOCK

[Claudio Imbrenda]

Hello,

I have found a small issue in the current implementation of VSOCK: if
you load both the vhost_vsock module (or if it's compiled in) then the
vmw_vsock_virtio_transport will not load successfully (both when
compiled as a module or compiled in).

The actual error message I get is:

vmw_vsock_virtio_transport: probe of virtio2 failed with error -16

I dug into the code and it seems that vsock_core_init is called twice,
once for the vhost module and once for the virtio-transport module, the
second time returning -EBUSY (-16) since it's already initialized.

This means that if a guest loads the vhost_vsock module, or if it is
compiled in, then the guest can't use virtio-vsock. This also prevents
using vsock in a nested virtualization scenario, where both host and
guest modules could work at the same time.

Unfortunately I don't have the time to fix this, I just wanted to raise
the issue here, so that you are now aware of it, and can maybe fix it in
the future.


best regards,

Claudio Imbrenda

Re: [Qemu-devel] a small issue with VSOCK

[Stefan Hajnoczi]

[-- Attachment #1: Type: text/plain, Size: 1229 bytes --]

On Tue, Nov 15, 2016 at 07:25:59PM +0100, Claudio Imbrenda wrote:
> I have found a small issue in the current implementation of VSOCK: if
> you load both the vhost_vsock module (or if it's compiled in) then the
> vmw_vsock_virtio_transport will not load successfully (both when
> compiled as a module or compiled in).
> 
> The actual error message I get is:
> 
> vmw_vsock_virtio_transport: probe of virtio2 failed with error -16
> 
> I dug into the code and it seems that vsock_core_init is called twice,
> once for the vhost module and once for the virtio-transport module, the
> second time returning -EBUSY (-16) since it's already initialized.
> 
> This means that if a guest loads the vhost_vsock module, or if it is
> compiled in, then the guest can't use virtio-vsock. This also prevents
> using vsock in a nested virtualization scenario, where both host and
> guest modules could work at the same time.
> 
> Unfortunately I don't have the time to fix this, I just wanted to raise
> the issue here, so that you are now aware of it, and can maybe fix it in
> the future.

Thanks for letting me know.  It's definitely something that will need to
be fixed.  I've added it to my todo list.

Stefan

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 455 bytes --]