Re: [PATCH 11/12] evtchn: convert vIRQ lock to an r/w one

[Jan Beulich <jbeulich@xxxxxxxx>]

On 30.09.2020 09:58, Paul Durrant wrote:
>> From: Xen-devel <xen-devel-bounces@xxxxxxxxxxxxxxxxxxxx> On Behalf Of Jan 
>> Beulich
>> Sent: 28 September 2020 12:02
>>
>> @@ -334,6 +334,12 @@ void _spin_unlock_recursive(spinlock_t *
>>      }
>>  }
>>
>> +void _rw_barrier(rwlock_t *lock)
>> +{
>> +    check_barrier(&lock->lock.debug);
>> +    do { smp_mb(); } while ( _rw_is_locked(lock) );
>> +}
> 
> Should you not have a cpu_relax() somewhere in here?
> 
> TBH though, the fact this lock is never taken as a writer makes me
> wonder whether there needs to be a lock at all.

For both of these - see the discussion Julien and I also had. The
construct will now move to the subsequent patch anyway, and change
as per Julien's request.

Jan