[PATCH v2 0/1] vfio: remove VFIO_GROUP_NOTIFY_SET_KVM

[PATCH v2 0/1] vfio: remove VFIO_GROUP_NOTIFY_SET_KVM

[Matthew Rosato]

As discussed in this thread:

https://lore.kernel.org/kvm/20220516172734.GE1343366@nvidia.com/

Let's remove VFIO_GROUP_NOTIFY_SET_KVM and instead assume the association
has already been established prior to device_open.  For the types today
that need a KVM (GVT, vfio-ap) these will fail if a KVM is not found.
Looking ahead, vfio-pci-zdev will optionally want the KVM association
(enable hardware assists) but it will not be a hard requirement (still
want to allow other, non-KVM userspace usage). 

This is built on top of Jason's group locking series:
https://github.com/jgunthorpe/linux/commits/vfio_group_locking

And tested with s390x-pci (zdev-kvm series) and vfio-ap (GVT changes are
compile-tested only)

Changes for v2:
- gvt no longer needs release_work, get rid of it (Christoph)
- a few compile fixes for gvt
- update commit to mention fixes gvt oops (Jason)
- s/down_write/down_read/ in a few spots (Jason)
- avoid kvm build dependency by holding group read lock over device
  open/close and put the onus on the driver to obtain a reference if
  it will actually use the kvm pointer.  Document the requirement,
  use lockdep_assert to ensure lock is held during register_notifer;
  today all callers are from driver open_device.

Matthew Rosato (1):
  vfio: remove VFIO_GROUP_NOTIFY_SET_KVM

 drivers/gpu/drm/i915/gvt/gtt.c        |  4 +-
 drivers/gpu/drm/i915/gvt/gvt.h        |  3 -
 drivers/gpu/drm/i915/gvt/kvmgt.c      | 82 ++++++---------------------
 drivers/s390/crypto/vfio_ap_ops.c     | 38 ++++---------
 drivers/s390/crypto/vfio_ap_private.h |  3 -
 drivers/vfio/vfio.c                   | 75 ++++++++----------------
 include/linux/vfio.h                  |  5 +-
 7 files changed, 56 insertions(+), 154 deletions(-)

-- 
2.27.0

[PATCH v2 1/1] vfio: remove VFIO_GROUP_NOTIFY_SET_KVM

[Matthew Rosato]

Rather than relying on a notifier for associating the KVM with
the group, let's assume that the association has already been
made prior to device_open.  The first time a device is opened
associate the group KVM with the device.

This fixes a user-triggerable oops in GVT.

Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>
Signed-off-by: Matthew Rosato <mjrosato@linux.ibm.com>
---
 drivers/gpu/drm/i915/gvt/gtt.c        |  4 +-
 drivers/gpu/drm/i915/gvt/gvt.h        |  3 -
 drivers/gpu/drm/i915/gvt/kvmgt.c      | 82 ++++++---------------------
 drivers/s390/crypto/vfio_ap_ops.c     | 38 ++++---------
 drivers/s390/crypto/vfio_ap_private.h |  3 -
 drivers/vfio/vfio.c                   | 75 ++++++++----------------
 include/linux/vfio.h                  |  5 +-
 7 files changed, 56 insertions(+), 154 deletions(-)

diff --git a/drivers/gpu/drm/i915/gvt/gtt.c b/drivers/gpu/drm/i915/gvt/gtt.c
index 9c5cc2800975..b4f69364f9a1 100644
--- a/drivers/gpu/drm/i915/gvt/gtt.c
+++ b/drivers/gpu/drm/i915/gvt/gtt.c
@@ -51,7 +51,7 @@ static int preallocated_oos_pages = 8192;
 
 static bool intel_gvt_is_valid_gfn(struct intel_vgpu *vgpu, unsigned long gfn)
 {
-	struct kvm *kvm = vgpu->kvm;
+	struct kvm *kvm = vgpu->vfio_device.kvm;
 	int idx;
 	bool ret;
 
@@ -1185,7 +1185,7 @@ static int is_2MB_gtt_possible(struct intel_vgpu *vgpu,
 
 	if (!vgpu->attached)
 		return -EINVAL;
-	pfn = gfn_to_pfn(vgpu->kvm, ops->get_pfn(entry));
+	pfn = gfn_to_pfn(vgpu->vfio_device.kvm, ops->get_pfn(entry));
 	if (is_error_noslot_pfn(pfn))
 		return -EINVAL;
 	return PageTransHuge(pfn_to_page(pfn));
diff --git a/drivers/gpu/drm/i915/gvt/gvt.h b/drivers/gpu/drm/i915/gvt/gvt.h
index 2af4c83e733c..aee1a45da74b 100644
--- a/drivers/gpu/drm/i915/gvt/gvt.h
+++ b/drivers/gpu/drm/i915/gvt/gvt.h
@@ -227,9 +227,6 @@ struct intel_vgpu {
 	struct mutex cache_lock;
 
 	struct notifier_block iommu_notifier;
-	struct notifier_block group_notifier;
-	struct kvm *kvm;
-	struct work_struct release_work;
 	atomic_t released;
 
 	struct kvm_page_track_notifier_node track_node;
diff --git a/drivers/gpu/drm/i915/gvt/kvmgt.c b/drivers/gpu/drm/i915/gvt/kvmgt.c
index 7655ffa97d51..e2f6c56ab342 100644
--- a/drivers/gpu/drm/i915/gvt/kvmgt.c
+++ b/drivers/gpu/drm/i915/gvt/kvmgt.c
@@ -228,8 +228,6 @@ static void intel_gvt_cleanup_vgpu_type_groups(struct intel_gvt *gvt)
 	}
 }
 
-static void intel_vgpu_release_work(struct work_struct *work);
-
 static void gvt_unpin_guest_page(struct intel_vgpu *vgpu, unsigned long gfn,
 		unsigned long size)
 {
@@ -761,23 +759,6 @@ static int intel_vgpu_iommu_notifier(struct notifier_block *nb,
 	return NOTIFY_OK;
 }
 
-static int intel_vgpu_group_notifier(struct notifier_block *nb,
-				     unsigned long action, void *data)
-{
-	struct intel_vgpu *vgpu =
-		container_of(nb, struct intel_vgpu, group_notifier);
-
-	/* the only action we care about */
-	if (action == VFIO_GROUP_NOTIFY_SET_KVM) {
-		vgpu->kvm = data;
-
-		if (!data)
-			schedule_work(&vgpu->release_work);
-	}
-
-	return NOTIFY_OK;
-}
-
 static bool __kvmgt_vgpu_exist(struct intel_vgpu *vgpu)
 {
 	struct intel_vgpu *itr;
@@ -789,7 +770,7 @@ static bool __kvmgt_vgpu_exist(struct intel_vgpu *vgpu)
 		if (!itr->attached)
 			continue;
 
-		if (vgpu->kvm == itr->kvm) {
+		if (vgpu->vfio_device.kvm == itr->vfio_device.kvm) {
 			ret = true;
 			goto out;
 		}
@@ -806,7 +787,6 @@ static int intel_vgpu_open_device(struct vfio_device *vfio_dev)
 	int ret;
 
 	vgpu->iommu_notifier.notifier_call = intel_vgpu_iommu_notifier;
-	vgpu->group_notifier.notifier_call = intel_vgpu_group_notifier;
 
 	events = VFIO_IOMMU_NOTIFY_DMA_UNMAP;
 	ret = vfio_register_notifier(vfio_dev, VFIO_IOMMU_NOTIFY, &events,
@@ -817,38 +797,32 @@ static int intel_vgpu_open_device(struct vfio_device *vfio_dev)
 		goto out;
 	}
 
-	events = VFIO_GROUP_NOTIFY_SET_KVM;
-	ret = vfio_register_notifier(vfio_dev, VFIO_GROUP_NOTIFY, &events,
-				     &vgpu->group_notifier);
-	if (ret != 0) {
-		gvt_vgpu_err("vfio_register_notifier for group failed: %d\n",
-			ret);
-		goto undo_iommu;
-	}
-
 	ret = -EEXIST;
 	if (vgpu->attached)
-		goto undo_register;
+		goto undo_iommu;
 
 	ret = -ESRCH;
-	if (!vgpu->kvm || vgpu->kvm->mm != current->mm) {
+	if (!vgpu->vfio_device.kvm ||
+	    vgpu->vfio_device.kvm->mm != current->mm) {
 		gvt_vgpu_err("KVM is required to use Intel vGPU\n");
-		goto undo_register;
+		goto undo_iommu;
 	}
 
+	kvm_get_kvm(vgpu->vfio_device.kvm);
+
 	ret = -EEXIST;
 	if (__kvmgt_vgpu_exist(vgpu))
-		goto undo_register;
+		goto undo_iommu;
 
 	vgpu->attached = true;
-	kvm_get_kvm(vgpu->kvm);
 
 	kvmgt_protect_table_init(vgpu);
 	gvt_cache_init(vgpu);
 
 	vgpu->track_node.track_write = kvmgt_page_track_write;
 	vgpu->track_node.track_flush_slot = kvmgt_page_track_flush_slot;
-	kvm_page_track_register_notifier(vgpu->kvm, &vgpu->track_node);
+	kvm_page_track_register_notifier(vgpu->vfio_device.kvm,
+					 &vgpu->track_node);
 
 	debugfs_create_ulong(KVMGT_DEBUGFS_FILENAME, 0444, vgpu->debugfs,
 			     &vgpu->nr_cache_entries);
@@ -858,10 +832,6 @@ static int intel_vgpu_open_device(struct vfio_device *vfio_dev)
 	atomic_set(&vgpu->released, 0);
 	return 0;
 
-undo_register:
-	vfio_unregister_notifier(vfio_dev, VFIO_GROUP_NOTIFY,
-				 &vgpu->group_notifier);
-
 undo_iommu:
 	vfio_unregister_notifier(vfio_dev, VFIO_IOMMU_NOTIFY,
 				 &vgpu->iommu_notifier);
@@ -880,8 +850,9 @@ static void intel_vgpu_release_msi_eventfd_ctx(struct intel_vgpu *vgpu)
 	}
 }
 
-static void __intel_vgpu_release(struct intel_vgpu *vgpu)
+static void intel_vgpu_close_device(struct vfio_device *vfio_dev)
 {
+	struct intel_vgpu *vgpu = vfio_dev_to_vgpu(vfio_dev);
 	struct drm_i915_private *i915 = vgpu->gvt->gt->i915;
 	int ret;
 
@@ -898,35 +869,19 @@ static void __intel_vgpu_release(struct intel_vgpu *vgpu)
 	drm_WARN(&i915->drm, ret,
 		 "vfio_unregister_notifier for iommu failed: %d\n", ret);
 
-	ret = vfio_unregister_notifier(&vgpu->vfio_device, VFIO_GROUP_NOTIFY,
-				       &vgpu->group_notifier);
-	drm_WARN(&i915->drm, ret,
-		 "vfio_unregister_notifier for group failed: %d\n", ret);
-
 	debugfs_remove(debugfs_lookup(KVMGT_DEBUGFS_FILENAME, vgpu->debugfs));
 
-	kvm_page_track_unregister_notifier(vgpu->kvm, &vgpu->track_node);
-	kvm_put_kvm(vgpu->kvm);
+	kvm_page_track_unregister_notifier(vgpu->vfio_device.kvm,
+					   &vgpu->track_node);
 	kvmgt_protect_table_destroy(vgpu);
 	gvt_cache_destroy(vgpu);
 
 	intel_vgpu_release_msi_eventfd_ctx(vgpu);
 
-	vgpu->kvm = NULL;
 	vgpu->attached = false;
-}
-
-static void intel_vgpu_close_device(struct vfio_device *vfio_dev)
-{
-	__intel_vgpu_release(vfio_dev_to_vgpu(vfio_dev));
-}
-
-static void intel_vgpu_release_work(struct work_struct *work)
-{
-	struct intel_vgpu *vgpu =
-		container_of(work, struct intel_vgpu, release_work);
 
-	__intel_vgpu_release(vgpu);
+	if (vgpu->vfio_device.kvm)
+		kvm_put_kvm(vgpu->vfio_device.kvm);
 }
 
 static u64 intel_vgpu_get_bar_addr(struct intel_vgpu *vgpu, int bar)
@@ -1675,7 +1630,6 @@ static int intel_vgpu_probe(struct mdev_device *mdev)
 		return PTR_ERR(vgpu);
 	}
 
-	INIT_WORK(&vgpu->release_work, intel_vgpu_release_work);
 	vfio_init_group_dev(&vgpu->vfio_device, &mdev->dev,
 			    &intel_vgpu_dev_ops);
 
@@ -1713,7 +1667,7 @@ static struct mdev_driver intel_vgpu_mdev_driver = {
 
 int intel_gvt_page_track_add(struct intel_vgpu *info, u64 gfn)
 {
-	struct kvm *kvm = info->kvm;
+	struct kvm *kvm = info->vfio_device.kvm;
 	struct kvm_memory_slot *slot;
 	int idx;
 
@@ -1743,7 +1697,7 @@ int intel_gvt_page_track_add(struct intel_vgpu *info, u64 gfn)
 
 int intel_gvt_page_track_remove(struct intel_vgpu *info, u64 gfn)
 {
-	struct kvm *kvm = info->kvm;
+	struct kvm *kvm = info->vfio_device.kvm;
 	struct kvm_memory_slot *slot;
 	int idx;
 
diff --git a/drivers/s390/crypto/vfio_ap_ops.c b/drivers/s390/crypto/vfio_ap_ops.c
index e8914024f5b1..17a56bb4cf25 100644
--- a/drivers/s390/crypto/vfio_ap_ops.c
+++ b/drivers/s390/crypto/vfio_ap_ops.c
@@ -1196,6 +1196,8 @@ static int vfio_ap_mdev_set_kvm(struct ap_matrix_mdev *matrix_mdev,
 {
 	struct ap_matrix_mdev *m;
 
+	kvm_get_kvm(kvm);
+
 	if (kvm->arch.crypto.crycbd) {
 		down_write(&kvm->arch.crypto.pqap_hook_rwsem);
 		kvm->arch.crypto.pqap_hook = &matrix_mdev->pqap_hook;
@@ -1282,25 +1284,9 @@ static void vfio_ap_mdev_unset_kvm(struct ap_matrix_mdev *matrix_mdev)
 		mutex_unlock(&kvm->lock);
 		mutex_unlock(&matrix_dev->lock);
 	}
-}
 
-static int vfio_ap_mdev_group_notifier(struct notifier_block *nb,
-				       unsigned long action, void *data)
-{
-	int notify_rc = NOTIFY_OK;
-	struct ap_matrix_mdev *matrix_mdev;
-
-	if (action != VFIO_GROUP_NOTIFY_SET_KVM)
-		return NOTIFY_OK;
-
-	matrix_mdev = container_of(nb, struct ap_matrix_mdev, group_notifier);
-
-	if (!data)
-		vfio_ap_mdev_unset_kvm(matrix_mdev);
-	else if (vfio_ap_mdev_set_kvm(matrix_mdev, data))
-		notify_rc = NOTIFY_DONE;
-
-	return notify_rc;
+	if (kvm)
+		kvm_put_kvm(kvm);
 }
 
 static struct vfio_ap_queue *vfio_ap_find_queue(int apqn)
@@ -1402,11 +1388,10 @@ static int vfio_ap_mdev_open_device(struct vfio_device *vdev)
 	unsigned long events;
 	int ret;
 
-	matrix_mdev->group_notifier.notifier_call = vfio_ap_mdev_group_notifier;
-	events = VFIO_GROUP_NOTIFY_SET_KVM;
+	if (!vdev->kvm)
+		return -EPERM;
 
-	ret = vfio_register_notifier(vdev, VFIO_GROUP_NOTIFY, &events,
-				     &matrix_mdev->group_notifier);
+	ret = vfio_ap_mdev_set_kvm(matrix_mdev, vdev->kvm);
 	if (ret)
 		return ret;
 
@@ -1415,12 +1400,11 @@ static int vfio_ap_mdev_open_device(struct vfio_device *vdev)
 	ret = vfio_register_notifier(vdev, VFIO_IOMMU_NOTIFY, &events,
 				     &matrix_mdev->iommu_notifier);
 	if (ret)
-		goto out_unregister_group;
+		goto err_kvm;
 	return 0;
 
-out_unregister_group:
-	vfio_unregister_notifier(vdev, VFIO_GROUP_NOTIFY,
-				 &matrix_mdev->group_notifier);
+err_kvm:
+	vfio_ap_mdev_unset_kvm(matrix_mdev);
 	return ret;
 }
 
@@ -1431,8 +1415,6 @@ static void vfio_ap_mdev_close_device(struct vfio_device *vdev)
 
 	vfio_unregister_notifier(vdev, VFIO_IOMMU_NOTIFY,
 				 &matrix_mdev->iommu_notifier);
-	vfio_unregister_notifier(vdev, VFIO_GROUP_NOTIFY,
-				 &matrix_mdev->group_notifier);
 	vfio_ap_mdev_unset_kvm(matrix_mdev);
 }
 
diff --git a/drivers/s390/crypto/vfio_ap_private.h b/drivers/s390/crypto/vfio_ap_private.h
index 648fcaf8104a..a26efd804d0d 100644
--- a/drivers/s390/crypto/vfio_ap_private.h
+++ b/drivers/s390/crypto/vfio_ap_private.h
@@ -81,8 +81,6 @@ struct ap_matrix {
  * @node:	allows the ap_matrix_mdev struct to be added to a list
  * @matrix:	the adapters, usage domains and control domains assigned to the
  *		mediated matrix device.
- * @group_notifier: notifier block used for specifying callback function for
- *		    handling the VFIO_GROUP_NOTIFY_SET_KVM event
  * @iommu_notifier: notifier block used for specifying callback function for
  *		    handling the VFIO_IOMMU_NOTIFY_DMA_UNMAP even
  * @kvm:	the struct holding guest's state
@@ -94,7 +92,6 @@ struct ap_matrix_mdev {
 	struct vfio_device vdev;
 	struct list_head node;
 	struct ap_matrix matrix;
-	struct notifier_block group_notifier;
 	struct notifier_block iommu_notifier;
 	struct kvm *kvm;
 	crypto_hook pqap_hook;
diff --git a/drivers/vfio/vfio.c b/drivers/vfio/vfio.c
index cfcff7764403..272acb62facd 100644
--- a/drivers/vfio/vfio.c
+++ b/drivers/vfio/vfio.c
@@ -1083,11 +1083,22 @@ static struct file *vfio_device_open(struct vfio_device *device)
 
 	mutex_lock(&device->dev_set->lock);
 	device->open_count++;
+	down_read(&device->group->group_rwsem);
+	if (device->open_count == 1 && device->group->kvm) {
+		/*
+		 * Here we pass the KVM pointer with the group under the read
+		 * lock.  If the device driver will use it, it must obtain a
+		 * reference and release it during close_device.
+		 */
+		device->kvm = device->group->kvm;
+	}
+
 	if (device->open_count == 1 && device->ops->open_device) {
 		ret = device->ops->open_device(device);
 		if (ret)
 			goto err_undo_count;
 	}
+	up_read(&device->group->group_rwsem);
 	mutex_unlock(&device->dev_set->lock);
 
 	/*
@@ -1119,10 +1130,14 @@ static struct file *vfio_device_open(struct vfio_device *device)
 
 err_close_device:
 	mutex_lock(&device->dev_set->lock);
+	down_read(&device->group->group_rwsem);
 	if (device->open_count == 1 && device->ops->close_device)
 		device->ops->close_device(device);
 err_undo_count:
 	device->open_count--;
+	if (device->open_count == 0 && device->kvm)
+		device->kvm = NULL;
+	up_read(&device->group->group_rwsem);
 	mutex_unlock(&device->dev_set->lock);
 	module_put(device->dev->driver->owner);
 err_unassign_container:
@@ -1315,9 +1330,13 @@ static int vfio_device_fops_release(struct inode *inode, struct file *filep)
 
 	mutex_lock(&device->dev_set->lock);
 	vfio_assert_device_open(device);
+	down_read(&device->group->group_rwsem);
 	if (device->open_count == 1 && device->ops->close_device)
 		device->ops->close_device(device);
 	device->open_count--;
+	if (device->open_count == 0 && device->kvm)
+		device->kvm = NULL;
+	up_read(&device->group->group_rwsem);
 	mutex_unlock(&device->dev_set->lock);
 
 	module_put(device->dev->driver->owner);
@@ -1726,8 +1745,8 @@ EXPORT_SYMBOL_GPL(vfio_file_enforced_coherent);
  * @file: VFIO group file
  * @kvm: KVM to link
  *
- * The kvm pointer will be forwarded to all the vfio_device's attached to the
- * VFIO file via the VFIO_GROUP_NOTIFY_SET_KVM notifier.
+ * When a VFIO device is first opened the KVM will be available in
+ * device->kvm if one was associated with the group.
  */
 void vfio_file_set_kvm(struct file *file, struct kvm *kvm)
 {
@@ -1738,8 +1757,6 @@ void vfio_file_set_kvm(struct file *file, struct kvm *kvm)
 
 	down_write(&group->group_rwsem);
 	group->kvm = kvm;
-	blocking_notifier_call_chain(&group->notifier,
-				     VFIO_GROUP_NOTIFY_SET_KVM, kvm);
 	up_write(&group->group_rwsem);
 }
 EXPORT_SYMBOL_GPL(vfio_file_set_kvm);
@@ -2006,7 +2023,8 @@ static int vfio_register_iommu_notifier(struct vfio_group *group,
 	struct vfio_iommu_driver *driver;
 	int ret;
 
-	down_read(&group->group_rwsem);
+	lockdep_assert_held_read(&group->group_rwsem);
+
 	container = group->container;
 	driver = container->iommu_driver;
 	if (likely(driver && driver->ops->register_notifier))
@@ -2014,7 +2032,6 @@ static int vfio_register_iommu_notifier(struct vfio_group *group,
 						     events, nb);
 	else
 		ret = -ENOTTY;
-	up_read(&group->group_rwsem);
 
 	return ret;
 }
@@ -2026,7 +2043,8 @@ static int vfio_unregister_iommu_notifier(struct vfio_group *group,
 	struct vfio_iommu_driver *driver;
 	int ret;
 
-	down_read(&group->group_rwsem);
+	lockdep_assert_held_read(&group->group_rwsem);
+
 	container = group->container;
 	driver = container->iommu_driver;
 	if (likely(driver && driver->ops->unregister_notifier))
@@ -2034,47 +2052,10 @@ static int vfio_unregister_iommu_notifier(struct vfio_group *group,
 						       nb);
 	else
 		ret = -ENOTTY;
-	up_read(&group->group_rwsem);
 
 	return ret;
 }
 
-static int vfio_register_group_notifier(struct vfio_group *group,
-					unsigned long *events,
-					struct notifier_block *nb)
-{
-	int ret;
-	bool set_kvm = false;
-
-	if (*events & VFIO_GROUP_NOTIFY_SET_KVM)
-		set_kvm = true;
-
-	/* clear known events */
-	*events &= ~VFIO_GROUP_NOTIFY_SET_KVM;
-
-	/* refuse to continue if still events remaining */
-	if (*events)
-		return -EINVAL;
-
-	ret = blocking_notifier_chain_register(&group->notifier, nb);
-	if (ret)
-		return ret;
-
-	/*
-	 * The attaching of kvm and vfio_group might already happen, so
-	 * here we replay once upon registration.
-	 */
-	if (set_kvm) {
-		down_read(&group->group_rwsem);
-		if (group->kvm)
-			blocking_notifier_call_chain(&group->notifier,
-						     VFIO_GROUP_NOTIFY_SET_KVM,
-						     group->kvm);
-		up_read(&group->group_rwsem);
-	}
-	return 0;
-}
-
 int vfio_register_notifier(struct vfio_device *device,
 			   enum vfio_notify_type type, unsigned long *events,
 			   struct notifier_block *nb)
@@ -2090,9 +2071,6 @@ int vfio_register_notifier(struct vfio_device *device,
 	case VFIO_IOMMU_NOTIFY:
 		ret = vfio_register_iommu_notifier(group, events, nb);
 		break;
-	case VFIO_GROUP_NOTIFY:
-		ret = vfio_register_group_notifier(group, events, nb);
-		break;
 	default:
 		ret = -EINVAL;
 	}
@@ -2114,9 +2092,6 @@ int vfio_unregister_notifier(struct vfio_device *device,
 	case VFIO_IOMMU_NOTIFY:
 		ret = vfio_unregister_iommu_notifier(group, nb);
 		break;
-	case VFIO_GROUP_NOTIFY:
-		ret = blocking_notifier_chain_unregister(&group->notifier, nb);
-		break;
 	default:
 		ret = -EINVAL;
 	}
diff --git a/include/linux/vfio.h b/include/linux/vfio.h
index 45b287826ce6..5f691453e3fb 100644
--- a/include/linux/vfio.h
+++ b/include/linux/vfio.h
@@ -36,6 +36,7 @@ struct vfio_device {
 	struct vfio_device_set *dev_set;
 	struct list_head dev_set_list;
 	unsigned int migration_flags;
+	struct kvm *kvm;
 
 	/* Members below here are private, not for driver use */
 	refcount_t refcount;
@@ -155,15 +156,11 @@ extern int vfio_dma_rw(struct vfio_device *device, dma_addr_t user_iova,
 /* each type has independent events */
 enum vfio_notify_type {
 	VFIO_IOMMU_NOTIFY = 0,
-	VFIO_GROUP_NOTIFY = 1,
 };
 
 /* events for VFIO_IOMMU_NOTIFY */
 #define VFIO_IOMMU_NOTIFY_DMA_UNMAP	BIT(0)
 
-/* events for VFIO_GROUP_NOTIFY */
-#define VFIO_GROUP_NOTIFY_SET_KVM	BIT(0)
-
 extern int vfio_register_notifier(struct vfio_device *device,
 				  enum vfio_notify_type type,
 				  unsigned long *required_events,
-- 
2.27.0

RE: [PATCH v2 1/1] vfio: remove VFIO_GROUP_NOTIFY_SET_KVM

[Tian, Kevin]

> From: Matthew Rosato
> Sent: Thursday, May 19, 2022 5:26 AM
> 
> Rather than relying on a notifier for associating the KVM with
> the group, let's assume that the association has already been
> made prior to device_open.  The first time a device is opened
> associate the group KVM with the device.
> 
> This fixes a user-triggerable oops in GVT.

No changelog.

> 
> Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>
> Signed-off-by: Matthew Rosato <mjrosato@linux.ibm.com>

Not sure whether Christoph wants a s-o-b here when he wrote
the snippet to remove the release work of gvt...

> @@ -1083,11 +1083,22 @@ static struct file *vfio_device_open(struct
> vfio_device *device)
> 
>  	mutex_lock(&device->dev_set->lock);
>  	device->open_count++;
> +	down_read(&device->group->group_rwsem);
> +	if (device->open_count == 1 && device->group->kvm) {
> +		/*
> +		 * Here we pass the KVM pointer with the group under the
> read
> +		 * lock.  If the device driver will use it, it must obtain a
> +		 * reference and release it during close_device.
> +		 */
> +		device->kvm = device->group->kvm;
> +	}
> +
>  	if (device->open_count == 1 && device->ops->open_device) {

Merge the two branches so both are under if (device->open_count == 1) {}
(and group_rwsem can be also moved inside)

> @@ -1315,9 +1330,13 @@ static int vfio_device_fops_release(struct inode
> *inode, struct file *filep)
> 
>  	mutex_lock(&device->dev_set->lock);
>  	vfio_assert_device_open(device);
> +	down_read(&device->group->group_rwsem);
>  	if (device->open_count == 1 && device->ops->close_device)
>  		device->ops->close_device(device);
>  	device->open_count--;
> +	if (device->open_count == 0 && device->kvm)
> +		device->kvm = NULL;

This can be moved out of group_rwsem as there is no reference to
vfio_group.

> +	up_read(&device->group->group_rwsem);

otherwise,

Reviewed-by: Kevin Tian <kevin.tian@intel.com>

Thanks
Kevin

Re: [Intel-gfx] [PATCH v2 1/1] vfio: remove VFIO_GROUP_NOTIFY_SET_KVM

[Tian, Kevin]

> From: Matthew Rosato
> Sent: Thursday, May 19, 2022 5:26 AM
> 
> Rather than relying on a notifier for associating the KVM with
> the group, let's assume that the association has already been
> made prior to device_open.  The first time a device is opened
> associate the group KVM with the device.
> 
> This fixes a user-triggerable oops in GVT.

No changelog.

> 
> Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>
> Signed-off-by: Matthew Rosato <mjrosato@linux.ibm.com>

Not sure whether Christoph wants a s-o-b here when he wrote
the snippet to remove the release work of gvt...

> @@ -1083,11 +1083,22 @@ static struct file *vfio_device_open(struct
> vfio_device *device)
> 
>  	mutex_lock(&device->dev_set->lock);
>  	device->open_count++;
> +	down_read(&device->group->group_rwsem);
> +	if (device->open_count == 1 && device->group->kvm) {
> +		/*
> +		 * Here we pass the KVM pointer with the group under the
> read
> +		 * lock.  If the device driver will use it, it must obtain a
> +		 * reference and release it during close_device.
> +		 */
> +		device->kvm = device->group->kvm;
> +	}
> +
>  	if (device->open_count == 1 && device->ops->open_device) {

Merge the two branches so both are under if (device->open_count == 1) {}
(and group_rwsem can be also moved inside)

> @@ -1315,9 +1330,13 @@ static int vfio_device_fops_release(struct inode
> *inode, struct file *filep)
> 
>  	mutex_lock(&device->dev_set->lock);
>  	vfio_assert_device_open(device);
> +	down_read(&device->group->group_rwsem);
>  	if (device->open_count == 1 && device->ops->close_device)
>  		device->ops->close_device(device);
>  	device->open_count--;
> +	if (device->open_count == 0 && device->kvm)
> +		device->kvm = NULL;

This can be moved out of group_rwsem as there is no reference to
vfio_group.

> +	up_read(&device->group->group_rwsem);

otherwise,

Reviewed-by: Kevin Tian <kevin.tian@intel.com>

Thanks
Kevin

Re: [PATCH v2 1/1] vfio: remove VFIO_GROUP_NOTIFY_SET_KVM

[hch]

On Thu, May 19, 2022 at 06:43:06AM +0000, Tian, Kevin wrote:
> > This fixes a user-triggerable oops in GVT.
> 
> No changelog.

??

the cover latter clearly states what has changed since v1, and this
patch has a good commit log.  This is exactly how it is supposed to
be done.

> 
> > 
> > Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>
> > Signed-off-by: Matthew Rosato <mjrosato@linux.ibm.com>
> 
> Not sure whether Christoph wants a s-o-b here when he wrote
> the snippet to remove the release work of gvt...

That's just tivial code removal, so no.

> 
> > @@ -1083,11 +1083,22 @@ static struct file *vfio_device_open(struct
> > vfio_device *device)
> > 
> >  	mutex_lock(&device->dev_set->lock);
> >  	device->open_count++;
> > +	down_read(&device->group->group_rwsem);
> > +	if (device->open_count == 1 && device->group->kvm) {
> > +		/*
> > +		 * Here we pass the KVM pointer with the group under the
> > read
> > +		 * lock.  If the device driver will use it, it must obtain a
> > +		 * reference and release it during close_device.
> > +		 */
> > +		device->kvm = device->group->kvm;
> > +	}
> > +
> >  	if (device->open_count == 1 && device->ops->open_device) {
> 
> Merge the two branches so both are under if (device->open_count == 1) {}
> (and group_rwsem can be also moved inside)

Yeah.  And we don't really need the device->group->kvm check, as
it would otherwise assign NULL which is perfectly fine.

But otherwise this also looks good to me:

Reviewed-by: Christoph Hellwig <hch@lst.de>

Re: [Intel-gfx] [PATCH v2 1/1] vfio: remove VFIO_GROUP_NOTIFY_SET_KVM

[hch]

On Thu, May 19, 2022 at 06:43:06AM +0000, Tian, Kevin wrote:
> > This fixes a user-triggerable oops in GVT.
> 
> No changelog.

??

the cover latter clearly states what has changed since v1, and this
patch has a good commit log.  This is exactly how it is supposed to
be done.

> 
> > 
> > Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>
> > Signed-off-by: Matthew Rosato <mjrosato@linux.ibm.com>
> 
> Not sure whether Christoph wants a s-o-b here when he wrote
> the snippet to remove the release work of gvt...

That's just tivial code removal, so no.

> 
> > @@ -1083,11 +1083,22 @@ static struct file *vfio_device_open(struct
> > vfio_device *device)
> > 
> >  	mutex_lock(&device->dev_set->lock);
> >  	device->open_count++;
> > +	down_read(&device->group->group_rwsem);
> > +	if (device->open_count == 1 && device->group->kvm) {
> > +		/*
> > +		 * Here we pass the KVM pointer with the group under the
> > read
> > +		 * lock.  If the device driver will use it, it must obtain a
> > +		 * reference and release it during close_device.
> > +		 */
> > +		device->kvm = device->group->kvm;
> > +	}
> > +
> >  	if (device->open_count == 1 && device->ops->open_device) {
> 
> Merge the two branches so both are under if (device->open_count == 1) {}
> (and group_rwsem can be also moved inside)

Yeah.  And we don't really need the device->group->kvm check, as
it would otherwise assign NULL which is perfectly fine.

But otherwise this also looks good to me:

Reviewed-by: Christoph Hellwig <hch@lst.de>

RE: [PATCH v2 1/1] vfio: remove VFIO_GROUP_NOTIFY_SET_KVM

[Tian, Kevin]

> From: hch@infradead.org <hch@infradead.org>
> Sent: Thursday, May 19, 2022 2:48 PM
> 
> On Thu, May 19, 2022 at 06:43:06AM +0000, Tian, Kevin wrote:
> > > This fixes a user-triggerable oops in GVT.
> >
> > No changelog.
> 
> ??
> 
> the cover latter clearly states what has changed since v1, and this
> patch has a good commit log.  This is exactly how it is supposed to
> be done.

sigh... don't know why I missed the coverletter.

Re: [Intel-gfx] [PATCH v2 1/1] vfio: remove VFIO_GROUP_NOTIFY_SET_KVM

[Tian, Kevin]

> From: hch@infradead.org <hch@infradead.org>
> Sent: Thursday, May 19, 2022 2:48 PM
> 
> On Thu, May 19, 2022 at 06:43:06AM +0000, Tian, Kevin wrote:
> > > This fixes a user-triggerable oops in GVT.
> >
> > No changelog.
> 
> ??
> 
> the cover latter clearly states what has changed since v1, and this
> patch has a good commit log.  This is exactly how it is supposed to
> be done.

sigh... don't know why I missed the coverletter.

Re: [PATCH v2 1/1] vfio: remove VFIO_GROUP_NOTIFY_SET_KVM

[Jason Gunthorpe]

On Wed, May 18, 2022 at 05:26:07PM -0400, Matthew Rosato wrote:
> Rather than relying on a notifier for associating the KVM with
> the group, let's assume that the association has already been
> made prior to device_open.  The first time a device is opened
> associate the group KVM with the device.
> 
> This fixes a user-triggerable oops in GVT.
> 
> Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>
> Signed-off-by: Matthew Rosato <mjrosato@linux.ibm.com>
> ---
>  drivers/gpu/drm/i915/gvt/gtt.c        |  4 +-
>  drivers/gpu/drm/i915/gvt/gvt.h        |  3 -
>  drivers/gpu/drm/i915/gvt/kvmgt.c      | 82 ++++++---------------------
>  drivers/s390/crypto/vfio_ap_ops.c     | 38 ++++---------
>  drivers/s390/crypto/vfio_ap_private.h |  3 -
>  drivers/vfio/vfio.c                   | 75 ++++++++----------------
>  include/linux/vfio.h                  |  5 +-
>  7 files changed, 56 insertions(+), 154 deletions(-)

I'm fine with this, thanks for finishing it

> diff --git a/drivers/s390/crypto/vfio_ap_ops.c b/drivers/s390/crypto/vfio_ap_ops.c
> index e8914024f5b1..17a56bb4cf25 100644
> --- a/drivers/s390/crypto/vfio_ap_ops.c
> +++ b/drivers/s390/crypto/vfio_ap_ops.c
> @@ -1196,6 +1196,8 @@ static int vfio_ap_mdev_set_kvm(struct ap_matrix_mdev *matrix_mdev,
>  {
>         struct ap_matrix_mdev *m;
>
> +       kvm_get_kvm(kvm);
> +

[..]

> +       if (kvm)
> +               kvm_put_kvm(kvm);
>  }

This extra ref traffic is not necessary, the kvm will have a valid ref
because we are under the group_rwsem here and then the function will
obtain a ref when it stores it in matrix_mdev->kvm.

> diff --git a/drivers/vfio/vfio.c b/drivers/vfio/vfio.c
> index cfcff7764403..272acb62facd 100644
> --- a/drivers/vfio/vfio.c
> +++ b/drivers/vfio/vfio.c
> @@ -1083,11 +1083,22 @@ static struct file *vfio_device_open(struct vfio_device *device)
>  
>  	mutex_lock(&device->dev_set->lock);
>  	device->open_count++;
> +	down_read(&device->group->group_rwsem);
> +	if (device->open_count == 1 && device->group->kvm) {
> +		/*
> +		 * Here we pass the KVM pointer with the group under the read
> +		 * lock.  If the device driver will use it, it must obtain a
> +		 * reference and release it during close_device.
> +		 */
> +		device->kvm = device->group->kvm;
> +	}

But it is a bit ugly to keep the device->kvm in the struct device
without holding a reference count.

It is probably worth adding a comment to the definition in the struct
as well that the device driver must reference it or never touch it.

Jason

Re: [Intel-gfx] [PATCH v2 1/1] vfio: remove VFIO_GROUP_NOTIFY_SET_KVM

[Jason Gunthorpe]

On Wed, May 18, 2022 at 05:26:07PM -0400, Matthew Rosato wrote:
> Rather than relying on a notifier for associating the KVM with
> the group, let's assume that the association has already been
> made prior to device_open.  The first time a device is opened
> associate the group KVM with the device.
> 
> This fixes a user-triggerable oops in GVT.
> 
> Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>
> Signed-off-by: Matthew Rosato <mjrosato@linux.ibm.com>
> ---
>  drivers/gpu/drm/i915/gvt/gtt.c        |  4 +-
>  drivers/gpu/drm/i915/gvt/gvt.h        |  3 -
>  drivers/gpu/drm/i915/gvt/kvmgt.c      | 82 ++++++---------------------
>  drivers/s390/crypto/vfio_ap_ops.c     | 38 ++++---------
>  drivers/s390/crypto/vfio_ap_private.h |  3 -
>  drivers/vfio/vfio.c                   | 75 ++++++++----------------
>  include/linux/vfio.h                  |  5 +-
>  7 files changed, 56 insertions(+), 154 deletions(-)

I'm fine with this, thanks for finishing it

> diff --git a/drivers/s390/crypto/vfio_ap_ops.c b/drivers/s390/crypto/vfio_ap_ops.c
> index e8914024f5b1..17a56bb4cf25 100644
> --- a/drivers/s390/crypto/vfio_ap_ops.c
> +++ b/drivers/s390/crypto/vfio_ap_ops.c
> @@ -1196,6 +1196,8 @@ static int vfio_ap_mdev_set_kvm(struct ap_matrix_mdev *matrix_mdev,
>  {
>         struct ap_matrix_mdev *m;
>
> +       kvm_get_kvm(kvm);
> +

[..]

> +       if (kvm)
> +               kvm_put_kvm(kvm);
>  }

This extra ref traffic is not necessary, the kvm will have a valid ref
because we are under the group_rwsem here and then the function will
obtain a ref when it stores it in matrix_mdev->kvm.

> diff --git a/drivers/vfio/vfio.c b/drivers/vfio/vfio.c
> index cfcff7764403..272acb62facd 100644
> --- a/drivers/vfio/vfio.c
> +++ b/drivers/vfio/vfio.c
> @@ -1083,11 +1083,22 @@ static struct file *vfio_device_open(struct vfio_device *device)
>  
>  	mutex_lock(&device->dev_set->lock);
>  	device->open_count++;
> +	down_read(&device->group->group_rwsem);
> +	if (device->open_count == 1 && device->group->kvm) {
> +		/*
> +		 * Here we pass the KVM pointer with the group under the read
> +		 * lock.  If the device driver will use it, it must obtain a
> +		 * reference and release it during close_device.
> +		 */
> +		device->kvm = device->group->kvm;
> +	}

But it is a bit ugly to keep the device->kvm in the struct device
without holding a reference count.

It is probably worth adding a comment to the definition in the struct
as well that the device driver must reference it or never touch it.

Jason

Re: [PATCH v2 1/1] vfio: remove VFIO_GROUP_NOTIFY_SET_KVM

[Tony Krowiak]

I made a few comments, but other than that this looks good to
me:

Reviewed-by: Tony Krowiak <akrowiak@linux.ibm.com>

On 5/18/22 5:26 PM, Matthew Rosato wrote:
> Rather than relying on a notifier for associating the KVM with
> the group, let's assume that the association has already been
> made prior to device_open.  The first time a device is opened
> associate the group KVM with the device.
>
> This fixes a user-triggerable oops in GVT.
>
> Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>
> Signed-off-by: Matthew Rosato <mjrosato@linux.ibm.com>
> ---
>   drivers/gpu/drm/i915/gvt/gtt.c        |  4 +-
>   drivers/gpu/drm/i915/gvt/gvt.h        |  3 -
>   drivers/gpu/drm/i915/gvt/kvmgt.c      | 82 ++++++---------------------
>   drivers/s390/crypto/vfio_ap_ops.c     | 38 ++++---------
>   drivers/s390/crypto/vfio_ap_private.h |  3 -
>   drivers/vfio/vfio.c                   | 75 ++++++++----------------
>   include/linux/vfio.h                  |  5 +-
>   7 files changed, 56 insertions(+), 154 deletions(-)
>
> diff --git a/drivers/gpu/drm/i915/gvt/gtt.c b/drivers/gpu/drm/i915/gvt/gtt.c
> index 9c5cc2800975..b4f69364f9a1 100644
> --- a/drivers/gpu/drm/i915/gvt/gtt.c
> +++ b/drivers/gpu/drm/i915/gvt/gtt.c
> @@ -51,7 +51,7 @@ static int preallocated_oos_pages = 8192;
>   
>   static bool intel_gvt_is_valid_gfn(struct intel_vgpu *vgpu, unsigned long gfn)
>   {
> -	struct kvm *kvm = vgpu->kvm;
> +	struct kvm *kvm = vgpu->vfio_device.kvm;
>   	int idx;
>   	bool ret;
>   
> @@ -1185,7 +1185,7 @@ static int is_2MB_gtt_possible(struct intel_vgpu *vgpu,
>   
>   	if (!vgpu->attached)
>   		return -EINVAL;
> -	pfn = gfn_to_pfn(vgpu->kvm, ops->get_pfn(entry));
> +	pfn = gfn_to_pfn(vgpu->vfio_device.kvm, ops->get_pfn(entry));
>   	if (is_error_noslot_pfn(pfn))
>   		return -EINVAL;
>   	return PageTransHuge(pfn_to_page(pfn));
> diff --git a/drivers/gpu/drm/i915/gvt/gvt.h b/drivers/gpu/drm/i915/gvt/gvt.h
> index 2af4c83e733c..aee1a45da74b 100644
> --- a/drivers/gpu/drm/i915/gvt/gvt.h
> +++ b/drivers/gpu/drm/i915/gvt/gvt.h
> @@ -227,9 +227,6 @@ struct intel_vgpu {
>   	struct mutex cache_lock;
>   
>   	struct notifier_block iommu_notifier;
> -	struct notifier_block group_notifier;
> -	struct kvm *kvm;
> -	struct work_struct release_work;
>   	atomic_t released;
>   
>   	struct kvm_page_track_notifier_node track_node;
> diff --git a/drivers/gpu/drm/i915/gvt/kvmgt.c b/drivers/gpu/drm/i915/gvt/kvmgt.c
> index 7655ffa97d51..e2f6c56ab342 100644
> --- a/drivers/gpu/drm/i915/gvt/kvmgt.c
> +++ b/drivers/gpu/drm/i915/gvt/kvmgt.c
> @@ -228,8 +228,6 @@ static void intel_gvt_cleanup_vgpu_type_groups(struct intel_gvt *gvt)
>   	}
>   }
>   
> -static void intel_vgpu_release_work(struct work_struct *work);
> -
>   static void gvt_unpin_guest_page(struct intel_vgpu *vgpu, unsigned long gfn,
>   		unsigned long size)
>   {
> @@ -761,23 +759,6 @@ static int intel_vgpu_iommu_notifier(struct notifier_block *nb,
>   	return NOTIFY_OK;
>   }
>   
> -static int intel_vgpu_group_notifier(struct notifier_block *nb,
> -				     unsigned long action, void *data)
> -{
> -	struct intel_vgpu *vgpu =
> -		container_of(nb, struct intel_vgpu, group_notifier);
> -
> -	/* the only action we care about */
> -	if (action == VFIO_GROUP_NOTIFY_SET_KVM) {
> -		vgpu->kvm = data;
> -
> -		if (!data)
> -			schedule_work(&vgpu->release_work);
> -	}
> -
> -	return NOTIFY_OK;
> -}
> -
>   static bool __kvmgt_vgpu_exist(struct intel_vgpu *vgpu)
>   {
>   	struct intel_vgpu *itr;
> @@ -789,7 +770,7 @@ static bool __kvmgt_vgpu_exist(struct intel_vgpu *vgpu)
>   		if (!itr->attached)
>   			continue;
>   
> -		if (vgpu->kvm == itr->kvm) {
> +		if (vgpu->vfio_device.kvm == itr->vfio_device.kvm) {
>   			ret = true;
>   			goto out;
>   		}
> @@ -806,7 +787,6 @@ static int intel_vgpu_open_device(struct vfio_device *vfio_dev)
>   	int ret;
>   
>   	vgpu->iommu_notifier.notifier_call = intel_vgpu_iommu_notifier;
> -	vgpu->group_notifier.notifier_call = intel_vgpu_group_notifier;
>   
>   	events = VFIO_IOMMU_NOTIFY_DMA_UNMAP;
>   	ret = vfio_register_notifier(vfio_dev, VFIO_IOMMU_NOTIFY, &events,
> @@ -817,38 +797,32 @@ static int intel_vgpu_open_device(struct vfio_device *vfio_dev)
>   		goto out;
>   	}
>   
> -	events = VFIO_GROUP_NOTIFY_SET_KVM;
> -	ret = vfio_register_notifier(vfio_dev, VFIO_GROUP_NOTIFY, &events,
> -				     &vgpu->group_notifier);
> -	if (ret != 0) {
> -		gvt_vgpu_err("vfio_register_notifier for group failed: %d\n",
> -			ret);
> -		goto undo_iommu;
> -	}
> -
>   	ret = -EEXIST;
>   	if (vgpu->attached)
> -		goto undo_register;
> +		goto undo_iommu;
>   
>   	ret = -ESRCH;
> -	if (!vgpu->kvm || vgpu->kvm->mm != current->mm) {
> +	if (!vgpu->vfio_device.kvm ||
> +	    vgpu->vfio_device.kvm->mm != current->mm) {
>   		gvt_vgpu_err("KVM is required to use Intel vGPU\n");
> -		goto undo_register;
> +		goto undo_iommu;
>   	}
>   
> +	kvm_get_kvm(vgpu->vfio_device.kvm);
> +
>   	ret = -EEXIST;
>   	if (__kvmgt_vgpu_exist(vgpu))
> -		goto undo_register;
> +		goto undo_iommu;
>   
>   	vgpu->attached = true;
> -	kvm_get_kvm(vgpu->kvm);
>   
>   	kvmgt_protect_table_init(vgpu);
>   	gvt_cache_init(vgpu);
>   
>   	vgpu->track_node.track_write = kvmgt_page_track_write;
>   	vgpu->track_node.track_flush_slot = kvmgt_page_track_flush_slot;
> -	kvm_page_track_register_notifier(vgpu->kvm, &vgpu->track_node);
> +	kvm_page_track_register_notifier(vgpu->vfio_device.kvm,
> +					 &vgpu->track_node);
>   
>   	debugfs_create_ulong(KVMGT_DEBUGFS_FILENAME, 0444, vgpu->debugfs,
>   			     &vgpu->nr_cache_entries);
> @@ -858,10 +832,6 @@ static int intel_vgpu_open_device(struct vfio_device *vfio_dev)
>   	atomic_set(&vgpu->released, 0);
>   	return 0;
>   
> -undo_register:
> -	vfio_unregister_notifier(vfio_dev, VFIO_GROUP_NOTIFY,
> -				 &vgpu->group_notifier);
> -
>   undo_iommu:
>   	vfio_unregister_notifier(vfio_dev, VFIO_IOMMU_NOTIFY,
>   				 &vgpu->iommu_notifier);
> @@ -880,8 +850,9 @@ static void intel_vgpu_release_msi_eventfd_ctx(struct intel_vgpu *vgpu)
>   	}
>   }
>   
> -static void __intel_vgpu_release(struct intel_vgpu *vgpu)
> +static void intel_vgpu_close_device(struct vfio_device *vfio_dev)
>   {
> +	struct intel_vgpu *vgpu = vfio_dev_to_vgpu(vfio_dev);
>   	struct drm_i915_private *i915 = vgpu->gvt->gt->i915;
>   	int ret;
>   
> @@ -898,35 +869,19 @@ static void __intel_vgpu_release(struct intel_vgpu *vgpu)
>   	drm_WARN(&i915->drm, ret,
>   		 "vfio_unregister_notifier for iommu failed: %d\n", ret);
>   
> -	ret = vfio_unregister_notifier(&vgpu->vfio_device, VFIO_GROUP_NOTIFY,
> -				       &vgpu->group_notifier);
> -	drm_WARN(&i915->drm, ret,
> -		 "vfio_unregister_notifier for group failed: %d\n", ret);
> -
>   	debugfs_remove(debugfs_lookup(KVMGT_DEBUGFS_FILENAME, vgpu->debugfs));
>   
> -	kvm_page_track_unregister_notifier(vgpu->kvm, &vgpu->track_node);
> -	kvm_put_kvm(vgpu->kvm);
> +	kvm_page_track_unregister_notifier(vgpu->vfio_device.kvm,
> +					   &vgpu->track_node);
>   	kvmgt_protect_table_destroy(vgpu);
>   	gvt_cache_destroy(vgpu);
>   
>   	intel_vgpu_release_msi_eventfd_ctx(vgpu);
>   
> -	vgpu->kvm = NULL;
>   	vgpu->attached = false;
> -}
> -
> -static void intel_vgpu_close_device(struct vfio_device *vfio_dev)
> -{
> -	__intel_vgpu_release(vfio_dev_to_vgpu(vfio_dev));
> -}
> -
> -static void intel_vgpu_release_work(struct work_struct *work)
> -{
> -	struct intel_vgpu *vgpu =
> -		container_of(work, struct intel_vgpu, release_work);
>   
> -	__intel_vgpu_release(vgpu);
> +	if (vgpu->vfio_device.kvm)
> +		kvm_put_kvm(vgpu->vfio_device.kvm);
>   }
>   
>   static u64 intel_vgpu_get_bar_addr(struct intel_vgpu *vgpu, int bar)
> @@ -1675,7 +1630,6 @@ static int intel_vgpu_probe(struct mdev_device *mdev)
>   		return PTR_ERR(vgpu);
>   	}
>   
> -	INIT_WORK(&vgpu->release_work, intel_vgpu_release_work);
>   	vfio_init_group_dev(&vgpu->vfio_device, &mdev->dev,
>   			    &intel_vgpu_dev_ops);
>   
> @@ -1713,7 +1667,7 @@ static struct mdev_driver intel_vgpu_mdev_driver = {
>   
>   int intel_gvt_page_track_add(struct intel_vgpu *info, u64 gfn)
>   {
> -	struct kvm *kvm = info->kvm;
> +	struct kvm *kvm = info->vfio_device.kvm;
>   	struct kvm_memory_slot *slot;
>   	int idx;
>   
> @@ -1743,7 +1697,7 @@ int intel_gvt_page_track_add(struct intel_vgpu *info, u64 gfn)
>   
>   int intel_gvt_page_track_remove(struct intel_vgpu *info, u64 gfn)
>   {
> -	struct kvm *kvm = info->kvm;
> +	struct kvm *kvm = info->vfio_device.kvm;
>   	struct kvm_memory_slot *slot;
>   	int idx;
>   
> diff --git a/drivers/s390/crypto/vfio_ap_ops.c b/drivers/s390/crypto/vfio_ap_ops.c
> index e8914024f5b1..17a56bb4cf25 100644
> --- a/drivers/s390/crypto/vfio_ap_ops.c
> +++ b/drivers/s390/crypto/vfio_ap_ops.c
> @@ -1196,6 +1196,8 @@ static int vfio_ap_mdev_set_kvm(struct ap_matrix_mdev *matrix_mdev,
>   {
>   	struct ap_matrix_mdev *m;
>   
> +	kvm_get_kvm(kvm);
> +

I'm not sure what version of the code on which the patch was rebased, 
but in the
latest master branch from our repository the kvm_get_kvm(kvm) function is
called inside of the if block below. I'm fine with moving outside of the 
block, but
I don't see a corresponding removal of it from inside the block.

>   	if (kvm->arch.crypto.crycbd) {
>   		down_write(&kvm->arch.crypto.pqap_hook_rwsem);
>   		kvm->arch.crypto.pqap_hook = &matrix_mdev->pqap_hook;
> @@ -1282,25 +1284,9 @@ static void vfio_ap_mdev_unset_kvm(struct ap_matrix_mdev *matrix_mdev)
>   		mutex_unlock(&kvm->lock);
>   		mutex_unlock(&matrix_dev->lock);
>   	}
> -}
>   
> -static int vfio_ap_mdev_group_notifier(struct notifier_block *nb,
> -				       unsigned long action, void *data)
> -{
> -	int notify_rc = NOTIFY_OK;
> -	struct ap_matrix_mdev *matrix_mdev;
> -
> -	if (action != VFIO_GROUP_NOTIFY_SET_KVM)
> -		return NOTIFY_OK;
> -
> -	matrix_mdev = container_of(nb, struct ap_matrix_mdev, group_notifier);
> -
> -	if (!data)
> -		vfio_ap_mdev_unset_kvm(matrix_mdev);
> -	else if (vfio_ap_mdev_set_kvm(matrix_mdev, data))
> -		notify_rc = NOTIFY_DONE;
> -
> -	return notify_rc;

Similarly, there is a call to kvm_put_kvm(kvm) inside the if block.
Again, I'm okay with moving this outside of the if block, but I would
think the call inside the if block ought to be removed.

> +	if (kvm)
> +		kvm_put_kvm(kvm);
>   }
>   
>   static struct vfio_ap_queue *vfio_ap_find_queue(int apqn)
> @@ -1402,11 +1388,10 @@ static int vfio_ap_mdev_open_device(struct vfio_device *vdev)
>   	unsigned long events;
>   	int ret;
>   
> -	matrix_mdev->group_notifier.notifier_call = vfio_ap_mdev_group_notifier;
> -	events = VFIO_GROUP_NOTIFY_SET_KVM;
> +	if (!vdev->kvm)
> +		return -EPERM;

Perhaps -EINVAL or -EFAULT?

>   
> -	ret = vfio_register_notifier(vdev, VFIO_GROUP_NOTIFY, &events,
> -				     &matrix_mdev->group_notifier);
> +	ret = vfio_ap_mdev_set_kvm(matrix_mdev, vdev->kvm);
>   	if (ret)
>   		return ret;
>   
> @@ -1415,12 +1400,11 @@ static int vfio_ap_mdev_open_device(struct vfio_device *vdev)
>   	ret = vfio_register_notifier(vdev, VFIO_IOMMU_NOTIFY, &events,
>   				     &matrix_mdev->iommu_notifier);
>   	if (ret)
> -		goto out_unregister_group;
> +		goto err_kvm;
>   	return 0;
>   
> -out_unregister_group:
> -	vfio_unregister_notifier(vdev, VFIO_GROUP_NOTIFY,
> -				 &matrix_mdev->group_notifier);
> +err_kvm:
> +	vfio_ap_mdev_unset_kvm(matrix_mdev);
>   	return ret;
>   }
>   
> @@ -1431,8 +1415,6 @@ static void vfio_ap_mdev_close_device(struct vfio_device *vdev)
>   
>   	vfio_unregister_notifier(vdev, VFIO_IOMMU_NOTIFY,
>   				 &matrix_mdev->iommu_notifier);
> -	vfio_unregister_notifier(vdev, VFIO_GROUP_NOTIFY,
> -				 &matrix_mdev->group_notifier);
>   	vfio_ap_mdev_unset_kvm(matrix_mdev);

I'm not sure if this matters, but the vfio_ap_mdev_unset_kvm(matrix_mdev)
function uses the KVM pointer stored in matrix_mdev->kvm. I can't imagine
the KVM pointer stored in vdev->kvm being different than matrix_mdev->kvm,
but thought I should point it out. Previously, this function was called 
by the
notifier handler which did not have access to the KVM pointer which is 
why it
was retrieved from matrix_mdev->kvm. Even if the vdev->kvm and
matrix_mdev->kvm did not match, we should probably go ahead and call
the unset function anyway to remove access to AP resources for the guest and
reset the queues.

>   }
>   
> diff --git a/drivers/s390/crypto/vfio_ap_private.h b/drivers/s390/crypto/vfio_ap_private.h
> index 648fcaf8104a..a26efd804d0d 100644
> --- a/drivers/s390/crypto/vfio_ap_private.h
> +++ b/drivers/s390/crypto/vfio_ap_private.h
> @@ -81,8 +81,6 @@ struct ap_matrix {
>    * @node:	allows the ap_matrix_mdev struct to be added to a list
>    * @matrix:	the adapters, usage domains and control domains assigned to the
>    *		mediated matrix device.
> - * @group_notifier: notifier block used for specifying callback function for
> - *		    handling the VFIO_GROUP_NOTIFY_SET_KVM event
>    * @iommu_notifier: notifier block used for specifying callback function for
>    *		    handling the VFIO_IOMMU_NOTIFY_DMA_UNMAP even
>    * @kvm:	the struct holding guest's state
> @@ -94,7 +92,6 @@ struct ap_matrix_mdev {
>   	struct vfio_device vdev;
>   	struct list_head node;
>   	struct ap_matrix matrix;
> -	struct notifier_block group_notifier;
>   	struct notifier_block iommu_notifier;
>   	struct kvm *kvm;
>   	crypto_hook pqap_hook;
> diff --git a/drivers/vfio/vfio.c b/drivers/vfio/vfio.c
> index cfcff7764403..272acb62facd 100644
> --- a/drivers/vfio/vfio.c
> +++ b/drivers/vfio/vfio.c
> @@ -1083,11 +1083,22 @@ static struct file *vfio_device_open(struct vfio_device *device)
>   
>   	mutex_lock(&device->dev_set->lock);
>   	device->open_count++;
> +	down_read(&device->group->group_rwsem);
> +	if (device->open_count == 1 && device->group->kvm) {
> +		/*
> +		 * Here we pass the KVM pointer with the group under the read
> +		 * lock.  If the device driver will use it, it must obtain a
> +		 * reference and release it during close_device.
> +		 */
> +		device->kvm = device->group->kvm;
> +	}
> +
>   	if (device->open_count == 1 && device->ops->open_device) {
>   		ret = device->ops->open_device(device);
>   		if (ret)
>   			goto err_undo_count;
>   	}
> +	up_read(&device->group->group_rwsem);
>   	mutex_unlock(&device->dev_set->lock);
>   
>   	/*
> @@ -1119,10 +1130,14 @@ static struct file *vfio_device_open(struct vfio_device *device)
>   
>   err_close_device:
>   	mutex_lock(&device->dev_set->lock);
> +	down_read(&device->group->group_rwsem);
>   	if (device->open_count == 1 && device->ops->close_device)
>   		device->ops->close_device(device);
>   err_undo_count:
>   	device->open_count--;
> +	if (device->open_count == 0 && device->kvm)
> +		device->kvm = NULL;
> +	up_read(&device->group->group_rwsem);
>   	mutex_unlock(&device->dev_set->lock);
>   	module_put(device->dev->driver->owner);
>   err_unassign_container:
> @@ -1315,9 +1330,13 @@ static int vfio_device_fops_release(struct inode *inode, struct file *filep)
>   
>   	mutex_lock(&device->dev_set->lock);
>   	vfio_assert_device_open(device);
> +	down_read(&device->group->group_rwsem);
>   	if (device->open_count == 1 && device->ops->close_device)
>   		device->ops->close_device(device);
>   	device->open_count--;
> +	if (device->open_count == 0 && device->kvm)
> +		device->kvm = NULL;
> +	up_read(&device->group->group_rwsem);
>   	mutex_unlock(&device->dev_set->lock);
>   
>   	module_put(device->dev->driver->owner);
> @@ -1726,8 +1745,8 @@ EXPORT_SYMBOL_GPL(vfio_file_enforced_coherent);
>    * @file: VFIO group file
>    * @kvm: KVM to link
>    *
> - * The kvm pointer will be forwarded to all the vfio_device's attached to the
> - * VFIO file via the VFIO_GROUP_NOTIFY_SET_KVM notifier.
> + * When a VFIO device is first opened the KVM will be available in
> + * device->kvm if one was associated with the group.
>    */
>   void vfio_file_set_kvm(struct file *file, struct kvm *kvm)
>   {
> @@ -1738,8 +1757,6 @@ void vfio_file_set_kvm(struct file *file, struct kvm *kvm)
>   
>   	down_write(&group->group_rwsem);
>   	group->kvm = kvm;
> -	blocking_notifier_call_chain(&group->notifier,
> -				     VFIO_GROUP_NOTIFY_SET_KVM, kvm);
>   	up_write(&group->group_rwsem);
>   }
>   EXPORT_SYMBOL_GPL(vfio_file_set_kvm);
> @@ -2006,7 +2023,8 @@ static int vfio_register_iommu_notifier(struct vfio_group *group,
>   	struct vfio_iommu_driver *driver;
>   	int ret;
>   
> -	down_read(&group->group_rwsem);
> +	lockdep_assert_held_read(&group->group_rwsem);
> +
>   	container = group->container;
>   	driver = container->iommu_driver;
>   	if (likely(driver && driver->ops->register_notifier))
> @@ -2014,7 +2032,6 @@ static int vfio_register_iommu_notifier(struct vfio_group *group,
>   						     events, nb);
>   	else
>   		ret = -ENOTTY;
> -	up_read(&group->group_rwsem);
>   
>   	return ret;
>   }
> @@ -2026,7 +2043,8 @@ static int vfio_unregister_iommu_notifier(struct vfio_group *group,
>   	struct vfio_iommu_driver *driver;
>   	int ret;
>   
> -	down_read(&group->group_rwsem);
> +	lockdep_assert_held_read(&group->group_rwsem);
> +
>   	container = group->container;
>   	driver = container->iommu_driver;
>   	if (likely(driver && driver->ops->unregister_notifier))
> @@ -2034,47 +2052,10 @@ static int vfio_unregister_iommu_notifier(struct vfio_group *group,
>   						       nb);
>   	else
>   		ret = -ENOTTY;
> -	up_read(&group->group_rwsem);
>   
>   	return ret;
>   }
>   
> -static int vfio_register_group_notifier(struct vfio_group *group,
> -					unsigned long *events,
> -					struct notifier_block *nb)
> -{
> -	int ret;
> -	bool set_kvm = false;
> -
> -	if (*events & VFIO_GROUP_NOTIFY_SET_KVM)
> -		set_kvm = true;
> -
> -	/* clear known events */
> -	*events &= ~VFIO_GROUP_NOTIFY_SET_KVM;
> -
> -	/* refuse to continue if still events remaining */
> -	if (*events)
> -		return -EINVAL;
> -
> -	ret = blocking_notifier_chain_register(&group->notifier, nb);
> -	if (ret)
> -		return ret;
> -
> -	/*
> -	 * The attaching of kvm and vfio_group might already happen, so
> -	 * here we replay once upon registration.
> -	 */
> -	if (set_kvm) {
> -		down_read(&group->group_rwsem);
> -		if (group->kvm)
> -			blocking_notifier_call_chain(&group->notifier,
> -						     VFIO_GROUP_NOTIFY_SET_KVM,
> -						     group->kvm);
> -		up_read(&group->group_rwsem);
> -	}
> -	return 0;
> -}
> -
>   int vfio_register_notifier(struct vfio_device *device,
>   			   enum vfio_notify_type type, unsigned long *events,
>   			   struct notifier_block *nb)
> @@ -2090,9 +2071,6 @@ int vfio_register_notifier(struct vfio_device *device,
>   	case VFIO_IOMMU_NOTIFY:
>   		ret = vfio_register_iommu_notifier(group, events, nb);
>   		break;
> -	case VFIO_GROUP_NOTIFY:
> -		ret = vfio_register_group_notifier(group, events, nb);
> -		break;
>   	default:
>   		ret = -EINVAL;
>   	}
> @@ -2114,9 +2092,6 @@ int vfio_unregister_notifier(struct vfio_device *device,
>   	case VFIO_IOMMU_NOTIFY:
>   		ret = vfio_unregister_iommu_notifier(group, nb);
>   		break;
> -	case VFIO_GROUP_NOTIFY:
> -		ret = blocking_notifier_chain_unregister(&group->notifier, nb);
> -		break;
>   	default:
>   		ret = -EINVAL;
>   	}
> diff --git a/include/linux/vfio.h b/include/linux/vfio.h
> index 45b287826ce6..5f691453e3fb 100644
> --- a/include/linux/vfio.h
> +++ b/include/linux/vfio.h
> @@ -36,6 +36,7 @@ struct vfio_device {
>   	struct vfio_device_set *dev_set;
>   	struct list_head dev_set_list;
>   	unsigned int migration_flags;
> +	struct kvm *kvm;
>   
>   	/* Members below here are private, not for driver use */
>   	refcount_t refcount;
> @@ -155,15 +156,11 @@ extern int vfio_dma_rw(struct vfio_device *device, dma_addr_t user_iova,
>   /* each type has independent events */
>   enum vfio_notify_type {
>   	VFIO_IOMMU_NOTIFY = 0,
> -	VFIO_GROUP_NOTIFY = 1,
>   };
>   
>   /* events for VFIO_IOMMU_NOTIFY */
>   #define VFIO_IOMMU_NOTIFY_DMA_UNMAP	BIT(0)
>   
> -/* events for VFIO_GROUP_NOTIFY */
> -#define VFIO_GROUP_NOTIFY_SET_KVM	BIT(0)
> -
>   extern int vfio_register_notifier(struct vfio_device *device,
>   				  enum vfio_notify_type type,
>   				  unsigned long *required_events,

Re: [PATCH v2 1/1] vfio: remove VFIO_GROUP_NOTIFY_SET_KVM

[Matthew Rosato]

On 5/19/22 12:23 PM, Tony Krowiak wrote:
> I made a few comments, but other than that this looks good to
> me:
> 
> Reviewed-by: Tony Krowiak <akrowiak@linux.ibm.com>
> 

...

> I'm not sure what version of the code on which the patch was rebased, 

Was on top of Jason's vfio_group_locking series, but now would apply on 
vfio-next since Alex pulled that series in to vfio-next.

> but in the
> latest master branch from our repository the kvm_get_kvm(kvm) function is
> called inside of the if block below. I'm fine with moving outside of the 
> block, but
> I don't see a corresponding removal of it from inside the block.

Yeah, I didn't notice those there.  v3 will simply remove my get/put 
additions and leave yours as-is.

...

>> vfio_ap_mdev_group_notifier;
>> -    events = VFIO_GROUP_NOTIFY_SET_KVM;
>> +    if (!vdev->kvm)
>> +        return -EPERM;
> 
> Perhaps -EINVAL or -EFAULT?
> 

Whichever you'd prefer?  If I don't hear back I'll just use -EINVAL in v3.

>> -    ret = vfio_register_notifier(vdev, VFIO_GROUP_NOTIFY, &events,
>> -                     &matrix_mdev->group_notifier);
>> +    ret = vfio_ap_mdev_set_kvm(matrix_mdev, vdev->kvm);
>>       if (ret)
>>           return ret;
>> @@ -1415,12 +1400,11 @@ static int vfio_ap_mdev_open_device(struct 
>> vfio_device *vdev)
>>       ret = vfio_register_notifier(vdev, VFIO_IOMMU_NOTIFY, &events,
>>                        &matrix_mdev->iommu_notifier);
>>       if (ret)
>> -        goto out_unregister_group;
>> +        goto err_kvm;
>>       return 0;
>> -out_unregister_group:
>> -    vfio_unregister_notifier(vdev, VFIO_GROUP_NOTIFY,
>> -                 &matrix_mdev->group_notifier);
>> +err_kvm:
>> +    vfio_ap_mdev_unset_kvm(matrix_mdev);
>>       return ret;
>>   }
>> @@ -1431,8 +1415,6 @@ static void vfio_ap_mdev_close_device(struct 
>> vfio_device *vdev)
>>       vfio_unregister_notifier(vdev, VFIO_IOMMU_NOTIFY,
>>                    &matrix_mdev->iommu_notifier);
>> -    vfio_unregister_notifier(vdev, VFIO_GROUP_NOTIFY,
>> -                 &matrix_mdev->group_notifier);
>>       vfio_ap_mdev_unset_kvm(matrix_mdev);
> 
> I'm not sure if this matters, but the vfio_ap_mdev_unset_kvm(matrix_mdev)
> function uses the KVM pointer stored in matrix_mdev->kvm. I can't imagine
> the KVM pointer stored in vdev->kvm being different than matrix_mdev->kvm,

With this patch matrix_mdev->kvm is set from the value in vdev->kvm 
during vfio_ap_mdev_set_kvm (basically, doing the work that the notifier 
was doing but instead of getting it from notifier data get it from the 
vfio_device)

> but thought I should point it out. Previously, this function was called 
> by the
> notifier handler which did not have access to the KVM pointer which is 
> why it
> was retrieved from matrix_mdev->kvm. Even if the vdev->kvm and
> matrix_mdev->kvm did not match, we should probably go ahead and call
> the unset function anyway to remove access to AP resources for the guest 
> and
> reset the queues.
>