Re: [PATCH v2] KVM: kvm_io_bus_unregister_dev() should never fail

[Cornelia Huck <cornelia.huck@de.ibm.com>]

On Thu, 23 Mar 2017 18:24:19 +0100
David Hildenbrand <david@redhat.com> wrote:

> No caller currently checks the return value of
> kvm_io_bus_unregister_dev(). This is evil, as all callers silently go on
> freeing their device. A stale reference will remain in the io_bus,
> getting at least used again, when the iobus gets teared down on
> kvm_destroy_vm() - leading to use after free errors.
> 
> There is nothing the callers could do, except retrying over and over
> again.
> 
> So let's simply remove the bus altogether, print an error and make
> sure no one can access this broken bus again (returning -ENOMEM on any
> attempt to access it).
> 
> Fixes: e93f8a0f821e ("KVM: convert io_bus to SRCU")
> Cc: stable@vger.kernel.org # 3.4+
> Reported-by: Dmitry Vyukov <dvyukov@google.com>
> Signed-off-by: David Hildenbrand <david@redhat.com>
> ---
> 
> Based on kvm/queue, where we just got 2a108a4e7c1 ("KVM: x86: clear bus
> pointer when destroyed"), which added a check we need here.
> 
> v1 -> v2:
> - added a check in kvm_destroy_vm()
> - added a check in virt/kvm/eventfd.c
> 
> Using 'git grep -C 4 "kvm->buses"' should help to find all users. The other
> user in virt/kvm/eventfd.c should be fine.
> 
> ---
>  include/linux/kvm_host.h |  4 ++--
>  virt/kvm/eventfd.c       |  3 ++-
>  virt/kvm/kvm_main.c      | 42 +++++++++++++++++++++++++-----------------
>  3 files changed, 29 insertions(+), 20 deletions(-)

Reviewed-by: Cornelia Huck <cornelia.huck@de.ibm.com>