Re: [PATCH v3 05/14] locking/qspinlock: Remove unbounded cmpxchg loop from locking slowpath

[Waiman Long]

On 04/27/2018 06:16 AM, Will Deacon wrote:
> Hi Waiman,
>
> On Thu, Apr 26, 2018 at 04:16:30PM -0400, Waiman Long wrote:
>> On 04/26/2018 06:34 AM, Will Deacon wrote:
>>> diff --git a/kernel/locking/qspinlock_paravirt.h b/kernel/locking/qspinlock_paravirt.h
>>> index 2711940429f5..2dbad2f25480 100644
>>> --- a/kernel/locking/qspinlock_paravirt.h
>>> +++ b/kernel/locking/qspinlock_paravirt.h
>>> @@ -118,11 +118,6 @@ static __always_inline void set_pending(struct qspinlock *lock)
>>>  	WRITE_ONCE(lock->pending, 1);
>>>  }
>>>  
>>> -static __always_inline void clear_pending(struct qspinlock *lock)
>>> -{
>>> -	WRITE_ONCE(lock->pending, 0);
>>> -}
>>> -
>>>  /*
>>>   * The pending bit check in pv_queued_spin_steal_lock() isn't a memory
>>>   * barrier. Therefore, an atomic cmpxchg_acquire() is used to acquire the
>> There is another clear_pending() function after the "#else /*
>> _Q_PENDING_BITS == 8 */" line that need to be removed as well.
> Bugger, sorry I missed that one. Is the >= 16K CPUs case supported elsewhere
> in Linux? The x86 Kconfig appears to clamp NR_CPUS to 8192 iiuc.
>
> Anyway, additional patch below. Ingo -- please can you apply this on top?
>
I don't think we support >= 16k in any of the distros. However, this
will be a limit that we will reach eventually. That is why I said we can
wait.

Cheers,
Longman