The vfio_ap device driver registers a group notifier with VFIO when the file descriptor for a VFIO mediated device attached to a KVM guest is opened. The group notifier is registered to receive notification that the KVM pointer for the guest is set (VFIO_GROUP_NOTIFY_SET_KVM event). When the KVM pointer is set to a non-NULL value, the vfio_ap driver takes the following actions: 1. Stashes the KVM pointer in the vfio_ap_mdev struct that holds the state of the mediated device. 2. Calls the kvm_get_kvm() function to increment its reference counter. 3. Sets the function pointer to the function that handles interception of the instruction that enables/disables interrupt processing for the KVM guest. 4. Plugs the AP devices assigned to the mediated device into the KVM guest.
These actions are reversed by the release callback which is invoked when userspace closes the mediated device's file descriptor. In this case, the group notifier does not get called to invalidate the KVM pointer because the notifier is unregistered by the release callback; however, there are no guarantees that userspace will do the right thing before shutting down. To ensure that there are no resource leaks should the group notifier get called to set the KVM pointer to NULL, the notifier should also reverse the actions taken when it was called to set the pointer. This patch series ensures proper clean up is done via the group notifier.
Tony Krowiak (2): s390/vfio-ap: No need to disable IRQ after queue reset s390/vfio-ap: reverse group notifier actions when KVM pointer invalidated
drivers/s390/crypto/vfio_ap_drv.c | 1 - drivers/s390/crypto/vfio_ap_ops.c | 80 +++++++++++++++++---------- drivers/s390/crypto/vfio_ap_private.h | 1 - 3 files changed, 50 insertions(+), 32 deletions(-)