On Tue, Jul 14, 2020 at 09:02:53PM +0200, Sebastian Andrzej Siewior wrote: > On 2020-07-14 11:27:32 [-0700], Paul E. McKenney wrote: > > I believe that Ulad and Joel are working on an update. > > I expressed multiple times that I am unhappy with the raw_spinlock_t > which both want to keep. It is important to be future proof but at the > same time I am not sure if they know how many in-hardirq kmalloc() or > kfree() invocation we have as of today in PREEMPT_RT. > It is not about counting how many times kfree_rcu() gets called from the hard IRQ context for RT kernel. In fact it is about breaking of such possibility if a conversion to spinlock_t is done. Is not it? As a result the revert would change a behavior of the API function, that is used to accept such usage. That is a motivation. -- Vlad Rezki
