Re: 5.13-rt1 + KVM = WARNING: at fs/eventfd.c:74 eventfd_signal()

[Paolo Bonzini <pbonzini@redhat.com>]

On 14/07/21 12:41, Michael S. Tsirkin wrote:
>> --------------- 8< ---------------
>> From: Paolo Bonzini <pbonzini@redhat.com>
>> Subject: [PATCH] eventfd: protect eventfd_wake_count with a local_lock
>>
>> eventfd_signal assumes that spin_lock_irqsave/spin_unlock_irqrestore is
>> non-preemptable and therefore increments and decrements the percpu
>> variable inside the critical section.
>>
>> This obviously does not fly with PREEMPT_RT.  If eventfd_signal is
>> preempted and an unrelated thread calls eventfd_signal, the result is
>> a spurious WARN.  To avoid this, protect the percpu variable with a
>> local_lock.
>>
>> Reported-by: Daniel Bristot de Oliveira <bristot@redhat.com>
>> Fixes: b5e683d5cab8 ("eventfd: track eventfd_signal() recursion depth")
>> Cc: stable@vger.kernel.org
>> Cc: He Zhe <zhe.he@windriver.com>
>> Cc: Jens Axboe <axboe@kernel.dk>
>> Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
> 
> 
> Makes sense ...
> 
> Acked-by: Michael S. Tsirkin <mst@redhat.com>
> 
> want to send this to the windriver guys so they can test?
> Here's the list from that thread:

I included He Zhe, but it should be enough for Daniel to test it.  The 
bug is quite obvious once you wear your PREEMPT_RT goggles.

Paolo