Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 23.128.96.18 as permitted sender) client-ip=23.128.96.18;
Message-ID: <7c8594d5943f99bc647951269606ae24dd1b37de.camel@redhat.com>
Subject: Re: [PATCH 2/6] mm/swap: Introduce alternative per-cpu LRU cache
 locking
From:   nsaenzju@redhat.com
To:     Sebastian Andrzej Siewior <bigeasy@linutronix.de>
Cc:     Peter Zijlstra <peterz@infradead.org>, akpm@linux-foundation.org,
        frederic@kernel.org, linux-kernel@vger.kernel.org,
        linux-mm@kvack.org, tglx@linutronix.de, cl@linux.com,
        juri.lelli@redhat.com, mingo@redhat.com, mtosatti@redhat.com,
        nilal@redhat.com, mgorman@suse.de, ppandit@redhat.com,
        williams@redhat.com, anna-maria@linutronix.de,
        linux-rt-users@vger.kernel.org
Date:   Wed, 22 Sep 2021 11:50:07 +0200
In-Reply-To: <20210922092039.2j6efnkhmfxuzjnx@linutronix.de>
References: <20210921161323.607817-1-nsaenzju@redhat.com>
         <20210921161323.607817-3-nsaenzju@redhat.com>
         <20210921220358.GN4323@worktop.programming.kicks-ass.net>
         <0ec733daf2daaf8a6f5b1fbf56676b9892d5bf73.camel@redhat.com>
         <20210922092039.2j6efnkhmfxuzjnx@linutronix.de>
Content-Type: text/plain; charset="UTF-8"
User-Agent: Evolution 3.40.4 (3.40.4-1.fc34) 
MIME-Version: 1.0
Content-Transfer-Encoding: 8bit
Precedence: bulk

On Wed, 2021-09-22 at 11:20 +0200, Sebastian Andrzej Siewior wrote:
> On 2021-09-22 10:47:07 [+0200], nsaenzju@redhat.com wrote:
> > > *why* use migrate_disable(), that's horrible!
> > 
> > I was trying to be mindful of RT. They don't appreciate people taking spinlocks
> > just after having disabled preemption.
> > 
> > I think getting local_lock(&locks->local) is my only option then. But it adds
> > an extra redundant spinlock in the RT+NOHZ_FULL case.
> 
> spin_lock() does not disable preemption on PREEMPT_RT. You don't
> disables preemption on purpose or did I miss that?

Sorry my message wasn't clear. Adding code for context:

+ static inline void lru_cache_lock(struct lru_cache_locks *locks)
+ {
+ 	if (static_branch_unlikely(&remote_pcpu_cache_access)) {
+ 		/* Avoid migration between this_cpu_ptr() and spin_lock() */
+ 		migrate_disable();

IIUC PeterZ would've preferred that I disable preemption here. And what I meant
to explain is that, given the spinlock below, I choose migrate_disable() over
preempt_disable() to cater for RT.

+ 		spin_lock(this_cpu_ptr(&locks->spin));
+ 	} else {


So, to make both worlds happy, I think the only option left is using the
local_lock (at the expense of extra overhead in the RT+NOHZ_FULL case):

+ 	if (static_branch_unlikely(&remote_pcpu_cache_access)) {
+ 		/* Local lock avoids migration between this_cpu_ptr() and spin_lock() */
+		local_lock(&locks->local);
+ 		spin_lock(this_cpu_ptr(&locks->spin));
+	} else {

-- 
Nicolás Sáenz