Re: [PATCH v1 4/5] kvfree_rcu: Refactor kfree_rcu_monitor() function

["Paul E. McKenney" <paulmck@kernel.org>]

On Mon, May 10, 2021 at 12:09:01PM +0200, Uladzislau Rezki wrote:
> On Sun, May 09, 2021 at 04:59:54PM -0700, Andrew Morton wrote:
> > On Wed, 28 Apr 2021 15:44:21 +0200 "Uladzislau Rezki (Sony)" <urezki@gmail.com> wrote:
> > 
> > > Rearm the monitor work directly from its own function that
> > > is kfree_rcu_monitor(). So this patch puts the invocation
> > > timing control in one place.
> > >
> > > ...
> > >
> > > --- a/kernel/rcu/tree.c
> > > +++ b/kernel/rcu/tree.c
> > > @@ -3415,37 +3415,44 @@ static inline bool queue_kfree_rcu_work(struct kfree_rcu_cpu *krcp)
> > >  	return !repeat;
> > >  }
> > >  
> > > -static inline void kfree_rcu_drain_unlock(struct kfree_rcu_cpu *krcp,
> > > -					  unsigned long flags)
> > > +/*
> > > + * This function queues a new batch. If success or nothing to
> > > + * drain it returns 1. Otherwise 0 is returned indicating that
> > > + * a reclaim kthread has not processed a previous batch.
> > > + */
> > > +static inline int kfree_rcu_drain(struct kfree_rcu_cpu *krcp)
> > >  {
> > > +	unsigned long flags;
> > > +	int ret;
> > > +
> > > +	raw_spin_lock_irqsave(&krcp->lock, flags);
> > > +
> > >  	// Attempt to start a new batch.
> > > -	if (queue_kfree_rcu_work(krcp)) {
> > > +	ret = queue_kfree_rcu_work(krcp);
> > 
> > This code has changed slightly in mainline.  Can you please redo,
> > retest and resend?
> > 
> > > +	if (ret)
> > >  		// Success! Our job is done here.
> > >  		krcp->monitor_todo = false;
> > > -		raw_spin_unlock_irqrestore(&krcp->lock, flags);
> > > -		return;
> > > -	}
> > 
> > It's conventional to retain the braces here, otherwise the code looks
> > weird.  Unless you're a python programmer ;)
> > 
> > 
> Hello, Anrew.
> 
> This refactoring is not up to date and is obsolete, instead we have done 
> bigger rework of kfree_rcu_monitor(). It is located here:
> 
> https://kernel.googlesource.com/pub/scm/linux/kernel/git/paulmck/linux-rcu/+/2349a35d39e7af5eef9064cbd0e42309040551da%5E%21/#F0

If Andrew would like to start taking these kvfree_rcu() patches,
that would be all to the good.  For example, there is likely much
more opportunity for optimization by bringing them closer to the
sl*b allocators.  Yes, they will need some privileged access to RCU
internals, but not that much.  And at some point, they should move from
their current home in kernel/rcu/tree.c to somewhere in mm.

To that end, here is the list in -rcu against current mainline, from
youngest to oldest:

b5691dd1cd7a kvfree_rcu: Fix comments according to current code
2349a35d39e7 kvfree_rcu: Refactor kfree_rcu_monitor()
bfa15885893f kvfree_rcu: Release a page cache under memory pressure
de9d86c3b0b7 kvfree_rcu: Use [READ/WRITE]_ONCE() macros to access to nr_bkv_objs
54a0393340f7 kvfree_rcu: Add a bulk-list check when a scheduler is run
7490789de1ac kvfree_rcu: Update "monitor_todo" once a batch is started
28e690ce0347 kvfree_rcu: Use kfree_rcu_monitor() instead of open-coded variant

Please let me know how you would like to proceed.

							Thanx, Paul