[PATCH v3 01/19] KVM: Treat the device list as an rculist

[Oliver Upton <oliver.upton@linux.dev>]

A subsequent change to KVM/arm64 will necessitate walking the device
list outside of the kvm->lock. Prepare by converting to an rculist. This
has zero effect on the VM destruction path, as it is expected every
reader is backed by a reference on the kvm struct.

On the other hand, ensure a given device is completely destroyed before
dropping the kvm->lock in the release() path, as certain devices expect
to be a singleton (e.g. the vfio-kvm device).

Cc: Paolo Bonzini <pbonzini@redhat.com>
Cc: Sean Christopherson <seanjc@google.com>
Signed-off-by: Oliver Upton <oliver.upton@linux.dev>
---
 virt/kvm/kvm_main.c | 14 +++++++++++---
 virt/kvm/vfio.c     |  2 ++
 2 files changed, 13 insertions(+), 3 deletions(-)

diff --git a/virt/kvm/kvm_main.c b/virt/kvm/kvm_main.c
index fb49c2a60200..6c09fe40948f 100644
--- a/virt/kvm/kvm_main.c
+++ b/virt/kvm/kvm_main.c
@@ -1329,6 +1329,12 @@ static void kvm_destroy_devices(struct kvm *kvm)
 	 * We do not need to take the kvm->lock here, because nobody else
 	 * has a reference to the struct kvm at this point and therefore
 	 * cannot access the devices list anyhow.
+	 *
+	 * The device list is generally managed as an rculist, but list_del()
+	 * is used intentionally here. If a bug in KVM introduced a reader that
+	 * was not backed by a reference on the kvm struct, the hope is that
+	 * it'd consume the poisoned forward pointer instead of suffering a
+	 * use-after-free, even though this cannot be guaranteed.
 	 */
 	list_for_each_entry_safe(dev, tmp, &kvm->devices, vm_node) {
 		list_del(&dev->vm_node);
@@ -4725,7 +4731,8 @@ static int kvm_device_release(struct inode *inode, struct file *filp)
 
 	if (dev->ops->release) {
 		mutex_lock(&kvm->lock);
-		list_del(&dev->vm_node);
+		list_del_rcu(&dev->vm_node);
+		synchronize_rcu();
 		dev->ops->release(dev);
 		mutex_unlock(&kvm->lock);
 	}
@@ -4808,7 +4815,7 @@ static int kvm_ioctl_create_device(struct kvm *kvm,
 		kfree(dev);
 		return ret;
 	}
-	list_add(&dev->vm_node, &kvm->devices);
+	list_add_rcu(&dev->vm_node, &kvm->devices);
 	mutex_unlock(&kvm->lock);
 
 	if (ops->init)
@@ -4819,7 +4826,8 @@ static int kvm_ioctl_create_device(struct kvm *kvm,
 	if (ret < 0) {
 		kvm_put_kvm_no_destroy(kvm);
 		mutex_lock(&kvm->lock);
-		list_del(&dev->vm_node);
+		list_del_rcu(&dev->vm_node);
+		synchronize_rcu();
 		if (ops->release)
 			ops->release(dev);
 		mutex_unlock(&kvm->lock);
diff --git a/virt/kvm/vfio.c b/virt/kvm/vfio.c
index ca24ce120906..76b7f6085dcd 100644
--- a/virt/kvm/vfio.c
+++ b/virt/kvm/vfio.c
@@ -366,6 +366,8 @@ static int kvm_vfio_create(struct kvm_device *dev, u32 type)
 	struct kvm_device *tmp;
 	struct kvm_vfio *kv;
 
+	lockdep_assert_held(&dev->kvm->lock);
+
 	/* Only one VFIO "device" per VM */
 	list_for_each_entry(tmp, &dev->kvm->devices, vm_node)
 		if (tmp->ops == &kvm_vfio_ops)
-- 
2.44.0.769.g3c40516874-goog