KVM/Eventfd: Avoid crash when assign and deassign specific eventfd in parallel.

Syzbot reports crashes in kvm_irqfd_assign(), caused by use-after-free
when kvm_irqfd_assign() and kvm_irqfd_deassign() run in parallel
for one specific eventfd. When the assign path hasn't finished but irqfd
has been added to kvm->irqfds.items list, another thead may deassign the
eventfd and free struct kvm_kernel_irqfd(). The assign path then uses
the struct kvm_kernel_irqfd that has been freed by deassign path. To avoid
such issue, keep irqfd under kvm->irq_srcu protection after the irqfd
has been added to kvm->irqfds.items list, and call synchronize_srcu()
in irq_shutdown() to make sure that irqfd has been fully initialized in
the assign path.

Reported-by: Dmitry Vyukov <dvyukov@google.com>
Cc: Paolo Bonzini <pbonzini@redhat.com>
Cc: Radim Krčmář <rkrcmar@redhat.com>
Cc: Dmitry Vyukov <dvyukov@google.com>
Signed-off-by: Tianyu Lan <tianyu.lan@intel.com>
Cc: stable@vger.kernel.org
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

1 files changed, 5 insertions, 1 deletions

diff --git a/virt/kvm/eventfd.c b/virt/kvm/eventfd.c
index 6e865e8b5b10..decefe944b0f 100644
--- a/virt/kvm/eventfd.c
+++ b/virt/kvm/eventfd.c
@@ -119,8 +119,12 @@ irqfd_shutdown(struct work_struct *work)
 {
         struct kvm_kernel_irqfd *irqfd =
                 container_of(work, struct kvm_kernel_irqfd, shutdown);
+        struct kvm *kvm = irqfd->kvm;
         u64 cnt;
 
+        /* Make sure irqfd has been initalized in assign path. */
+        synchronize_srcu(&kvm->irq_srcu);
+
         /*
          * Synchronize with the wait-queue and unhook ourselves to prevent
          * further events.
@@ -387,7 +391,6 @@ kvm_irqfd_assign(struct kvm *kvm, struct kvm_irqfd *args)
 
         idx = srcu_read_lock(&kvm->irq_srcu);
         irqfd_update(kvm, irqfd);
-        srcu_read_unlock(&kvm->irq_srcu, idx);
 
         list_add_tail(&irqfd->list, &kvm->irqfds.items);
 
@@ -421,6 +424,7 @@ kvm_irqfd_assign(struct kvm *kvm, struct kvm_irqfd *args)
         }
 #endif
 
+        srcu_read_unlock(&kvm->irq_srcu, idx);
         return 0;
 
 fail: