On Fri, Apr 26, 2013 at 04:34:03PM +0800, Jason Wang wrote:
> virtio-rng-ccw has zero config length, so we need validate the config length
> before trying to access it. Otherwise we may crash since vdev->config is NULL.
>
> Cc: Cornelia Huck <cornelia.h...@de.ibm.com>
> Cc: Richard Henderson <r...@twiddle.net>
> Cc: Alexander Graf <ag...@suse.de>
> Signed-off-by: Jason Wang <jasow...@redhat.com>
The real problem is dev->vdev->get_config being NULL,
isn't it? So why not validate it and be done with it?
> ---
> hw/s390x/virtio-ccw.c | 4 ++--
> 1 files changed, 2 insertions(+), 2 deletions(-)
>
> diff --git a/hw/s390x/virtio-ccw.c b/hw/s390x/virtio-ccw.c
> index 56539d3..8d0dff5 100644
> --- a/hw/s390x/virtio-ccw.c
> +++ b/hw/s390x/virtio-ccw.c
> @@ -260,7 +260,7 @@ static int virtio_ccw_cb(SubchDev *sch, CCW1 ccw)
> }
> }
> len = MIN(ccw.count, dev->vdev->config_len);
> - if (!ccw.cda) {
> + if (!ccw.cda || !len) {
> ret = -EFAULT;
> } else {
> dev->vdev->get_config(dev->vdev, dev->vdev->config);
> @@ -279,7 +279,7 @@ static int virtio_ccw_cb(SubchDev *sch, CCW1 ccw)
> }
> len = MIN(ccw.count, dev->vdev->config_len);
> hw_len = len;
> - if (!ccw.cda) {
> + if (!ccw.cda || !len) {
> ret = -EFAULT;
> } else {
> config = cpu_physical_memory_map(ccw.cda, &hw_len, 0);
> --
> 1.7.1
