On 10.07.2017 20:53, Radim Krčmář wrote:
> kvm_vm_release() did not have slots_lock when calling
> kvm_io_bus_unregister_dev() and this went unnoticed until 4a12f9517728
> ("KVM: mark kvm->busses as rcu protected") added dynamic checks.
> Luckily, there should be no race at that point:
>
> =============================
> WARNING: suspicious RCU usage
> 4.12.0.kvm+ #0 Not tainted
> -----------------------------
> ./include/linux/kvm_host.h:479 suspicious rcu_dereference_check() usage!
>
> lockdep_rcu_suspicious+0xc5/0x100
> kvm_io_bus_unregister_dev+0x173/0x190 [kvm]
> kvm_free_pit+0x28/0x80 [kvm]
> kvm_arch_sync_events+0x2d/0x30 [kvm]
> kvm_put_kvm+0xa7/0x2a0 [kvm]
> kvm_vm_release+0x21/0x30 [kvm]
>
> Signed-off-by: Radim Krčmář <rkrcmar@xxxxxxxxxx>
> ---
> arch/x86/kvm/i8254.c | 2 ++
> 1 file changed, 2 insertions(+)
>
> diff --git a/arch/x86/kvm/i8254.c b/arch/x86/kvm/i8254.c
> index a78b445ce411..af192895b1fc 100644
> --- a/arch/x86/kvm/i8254.c
> +++ b/arch/x86/kvm/i8254.c
> @@ -724,8 +724,10 @@ void kvm_free_pit(struct kvm *kvm)
> struct kvm_pit *pit = kvm->arch.vpit;
>
> if (pit) {
> + mutex_lock(&kvm->slots_lock);
> kvm_io_bus_unregister_dev(kvm, KVM_PIO_BUS, &pit->dev);
> kvm_io_bus_unregister_dev(kvm, KVM_PIO_BUS, &pit->speaker_dev);
> + mutex_unlock(&kvm->slots_lock);
> kvm_pit_set_reinject(pit, false);
> hrtimer_cancel(&pit->pit_state.timer);
> kthread_destroy_worker(pit->worker);
>
I tried to verify that we don't have any hierarchical locking inversion
in the way here (kvm->lock, kvm->slots_lock, kvm->irq_lock).
Hopefully I did this carefully enough :)
Reviewed-by: David Hildenbrand <david@xxxxxxxxxx>
--
Thanks,
David
