Re: [PATCH] s390/vfio-ap: fix memory leak in mdev remove callback

From: Tony Krowiak
Date: Mon May 10 2021 - 13:44:27 EST

On 5/6/21 6:22 AM, Cornelia Huck wrote:
On Wed, 5 May 2021 13:28:26 -0400
Tony Krowiak <akrowiak@xxxxxxxxxxxxx> wrote:

The mdev remove callback for the vfio_ap device driver bails out with
-EBUSY if the mdev is in use by a KVM guest. The intended purpose was
to prevent the mdev from being removed while in use; however, returning a
non-zero rc does not prevent removal. This could result in a memory leak
of the resources allocated when the mdev was created. In addition, the
KVM guest will still have access to the AP devices assigned to the mdev
even though the mdev no longer exists.

To prevent this scenario, cleanup will be done - including unplugging the
AP adapters, domains and control domains - regardless of whether the mdev
is in use by a KVM guest or not.

Fixes: 258287c994de ("s390: vfio-ap: implement mediated device open callback")
Cc: stable@xxxxxxxxxxxxxxx
Signed-off-by: Tony Krowiak <akrowiak@xxxxxxxxxxx>
Signed-off-by: Tony Krowiak <akrowiak@xxxxxxxxxxxxx>
drivers/s390/crypto/vfio_ap_ops.c | 39 +++++++++++++++++++++++--------
1 file changed, 29 insertions(+), 10 deletions(-)

diff --git a/drivers/s390/crypto/vfio_ap_ops.c b/drivers/s390/crypto/vfio_ap_ops.c
index b2c7e10dfdcd..757166da947e 100644
--- a/drivers/s390/crypto/vfio_ap_ops.c
+++ b/drivers/s390/crypto/vfio_ap_ops.c
@@ -335,6 +335,32 @@ static void vfio_ap_matrix_init(struct ap_config_info *info,
matrix->adm_max = info->apxa ? info->Nd : 15;
+static bool vfio_ap_mdev_has_crycb(struct ap_matrix_mdev *matrix_mdev)
+ return (matrix_mdev->kvm && matrix_mdev->kvm->arch.crypto.crycbd);
+static void vfio_ap_mdev_clear_apcb(struct ap_matrix_mdev *matrix_mdev)
+ /*
+ * If the KVM pointer is in the process of being set, wait until the
+ * process has completed.
+ */
+ wait_event_cmd(matrix_mdev->wait_for_kvm,
+ !matrix_mdev->kvm_busy,
+ mutex_unlock(&matrix_dev->lock),
+ mutex_lock(&matrix_dev->lock));
+ if (vfio_ap_mdev_has_crycb(matrix_mdev)) {
+ matrix_mdev->kvm_busy = true;
+ mutex_unlock(&matrix_dev->lock);
+ kvm_arch_crypto_clear_masks(matrix_mdev->kvm);
+ mutex_lock(&matrix_dev->lock);
+ matrix_mdev->kvm_busy = false;
+ wake_up_all(&matrix_mdev->wait_for_kvm);
+ }
Looking at vfio_ap_mdev_unset_kvm(), do you need to unhook the kvm here
as well?

(Or can you maybe even combine the two functions into one?)

I contemplated just calling the vfio_ap_unset_kvm() function from
the vfio_ap_mdev_remove() function, but my thinking at the time
was that some of the other things done in the unset function, such
as kvm_put_kvm() etc., might cause problems.
After thinking about it some more, the vfio_ap_mdev_remove()
function will not get called until the vfio_ap_mdev_release()
callback is invoked unless something weird happens. Since the
remove callback gets rid of the mdev and the release callback
calls the vfio_ap_unset_kvm() function anyway, I now see no harm
in just calling the unset function from the remove callback and
eliminating the function above.

static int vfio_ap_mdev_create(struct mdev_device *mdev)
struct ap_matrix_mdev *matrix_mdev;
@@ -366,16 +392,9 @@ static int vfio_ap_mdev_remove(struct mdev_device *mdev)
struct ap_matrix_mdev *matrix_mdev = mdev_get_drvdata(mdev);
- /*
- * If the KVM pointer is in flux or the guest is running, disallow
- * un-assignment of control domain.
- */
- if (matrix_mdev->kvm_busy || matrix_mdev->kvm) {
- mutex_unlock(&matrix_dev->lock);
- return -EBUSY;
- }
+ WARN(vfio_ap_mdev_has_crycb(matrix_mdev),
+ "Removing mdev leaves KVM guest without any crypto devices");
+ vfio_ap_mdev_clear_apcb(matrix_mdev);