On Thu, 9 Mar 2017 12:12:00 +0100 Paolo Bonzini <pbonz...@redhat.com> wrote:
> On 09/03/2017 12:07, Cornelia Huck wrote: > >>> - Driver #2 uses the device without doing a reset or proper setup > >> Without this patch, even if driver #2 do a reset, it can still use the > >> old map if it don't set queue pfn. > > > > Yes, the cleanup-on-reset is definetly needed. > > It is good to have for defensiveness, but it would still cause a > segfault so we should also add the checks on vq->vring.desc throughout. Agreed.
