KVM: irqfd: fix race between EPOLLHUP and irq_bypass_register_consumer

A comment warning against this bug is there, but the code is not doing what
the comment says.  Therefore it is possible that an EPOLLHUP races against
irq_bypass_register_consumer.  The EPOLLHUP handler schedules irqfd_shutdown,
and if that runs soon enough, you get a use-after-free.

Reported-by: syzbot <syzkaller@googlegroups.com>
Cc: stable@vger.kernel.org
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Reviewed-by: David Hildenbrand <david@redhat.com>

1 files changed, 6 insertions, 5 deletions

diff --git a/virt/kvm/eventfd.c b/virt/kvm/eventfd.c
index decefe944b0f..fe6eb0fe07f6 100644
--- a/virt/kvm/eventfd.c
+++ b/virt/kvm/eventfd.c
@@ -405,11 +405,6 @@ kvm_irqfd_assign(struct kvm *kvm, struct kvm_irqfd *args)
         if (events & EPOLLIN)
                 schedule_work(&irqfd->inject);
 
-        /*
-         * do not drop the file until the irqfd is fully initialized, otherwise
-         * we might race against the EPOLLHUP
-         */
-        fdput(f);
 #ifdef CONFIG_HAVE_KVM_IRQ_BYPASS
         if (kvm_arch_has_irq_bypass()) {
                 irqfd->consumer.token = (void *)irqfd->eventfd;
@@ -425,6 +420,12 @@ kvm_irqfd_assign(struct kvm *kvm, struct kvm_irqfd *args)
 #endif
 
         srcu_read_unlock(&kvm->irq_srcu, idx);
+
+        /*
+         * do not drop the file until the irqfd is fully initialized, otherwise
+         * we might race against the EPOLLHUP
+         */
+        fdput(f);
         return 0;
 
 fail: