Re: 5.13-rt1 + KVM = WARNING: at fs/eventfd.c:74 eventfd_signal()
From: Daniel Bristot de Oliveira
Date: Thu Jul 15 2021 - 08:34:34 EST
On 7/15/21 11:46 AM, Paolo Bonzini wrote:
> On 15/07/21 10:22, Daniel Bristot de Oliveira wrote:
>> On 7/14/21 12:35 PM, Paolo Bonzini wrote:
>>> On 14/07/21 11:23, Jason Wang wrote:
>>>> I think this can be probably fixed here:
>>>>
>>>> https://lore.kernel.org/lkml/20210618084412.18257-1-zhe.he@xxxxxxxxxxxxx/
>>>
>>> That seems wrong; in particular it wouldn't protect against AB/BA deadlocks.
>>> In fact, the bug is with the locking; the code 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; the right fix should be
>>> using a local_lock. Something like this (untested!!):
>>
>> the lock needs to be per-pcu...
>
> Great, thanks for testing and fixing the patch! Are you going to post
> it again once you've confirmed it works?
I can do that... unless you want to send it yourself...
-- Daniel
> Paolo