Re: [PATCH 3/3] kvfree_rcu: use migrate_disable/enable()

From: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
To: "Uladzislau Rezki (Sony)" <urezki@gmail.com>
Cc: LKML <linux-kernel@vger.kernel.org>, RCU <rcu@vger.kernel.org>,
	"Paul E . McKenney" <paulmck@kernel.org>,
	Michael Ellerman <mpe@ellerman.id.au>,
	Andrew Morton <akpm@linux-foundation.org>,
	Daniel Axtens <dja@axtens.net>,
	Frederic Weisbecker <frederic@kernel.org>,
	Neeraj Upadhyay <neeraju@codeaurora.org>,
	Joel Fernandes <joel@joelfernandes.org>,
	Peter Zijlstra <peterz@infradead.org>,
	Michal Hocko <mhocko@suse.com>,
	Thomas Gleixner <tglx@linutronix.de>,
	"Theodore Y . Ts'o" <tytso@mit.edu>,
	Oleksiy Avramchenko <oleksiy.avramchenko@sonymobile.com>
Subject: Re: [PATCH 3/3] kvfree_rcu: use migrate_disable/enable()
Date: Wed, 20 Jan 2021 20:45:54 +0100	[thread overview]
Message-ID: <20210120194554.34dby47iipmyplts@linutronix.de> (raw)
In-Reply-To: <20210120162148.1973-3-urezki@gmail.com>

On 2021-01-20 17:21:48 [+0100], Uladzislau Rezki (Sony) wrote:
> --- a/kernel/rcu/tree.c
> +++ b/kernel/rcu/tree.c
> @@ -3489,10 +3489,12 @@ add_ptr_to_bulk_krc_lock(struct kfree_rcu_cpu **krcp,
>  			(*krcp)->bkvhead[idx]->nr_records == KVFREE_BULK_MAX_ENTR) {
>  		bnode = get_cached_bnode(*krcp);
>  		if (!bnode && can_alloc) {
> +			migrate_disable();
>  			krc_this_cpu_unlock(*krcp, *flags);

Why is krc_this_cpu_unlock() defined as
| static inline void
| krc_this_cpu_unlock(struct kfree_rcu_cpu *krcp, unsigned long flags)
| {
|         raw_spin_unlock(&krcp->lock);
|         local_irq_restore(flags);
| }

instead of raw_spin_unlock_irqrestore()?
Should something with the locked section trigger a scheduling event by
setting TIF_NEED_RESCHED then there will be no scheduling event on
unlock. It will be delayed until a later "random" preempt_enable().

raw_spin_unlock_irqrestore() will reschedule if the flag is set,
local_irq_restore() will not.

Sebastian