Date:   Mon, 6 Feb 2023 19:19:06 +0000
From:   Matthew Wilcox <willy@infradead.org>
To:     Marcelo Tosatti <mtosatti@redhat.com>
Cc:     Christoph Lameter <cl@linux.com>,
        Aaron Tomlin <atomlin@atomlin.com>,
        Frederic Weisbecker <frederic@kernel.org>,
        Andrew Morton <akpm@linux-foundation.org>,
        linux-kernel@vger.kernel.org, linux-mm@kvack.org
Subject: Re: [PATCH 3/5] mm/vmstat: use cmpxchg loop in cpu_vm_stats_fold
Message-ID: <Y+FSqkDYfhKuBZIe@casper.infradead.org>
References: <20230201195013.881721887@redhat.com>
 <20230201195104.460373427@redhat.com>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <20230201195104.460373427@redhat.com>
Precedence: bulk

On Wed, Feb 01, 2023 at 04:50:16PM -0300, Marcelo Tosatti wrote:
> In preparation to switch vmstat shepherd to flush
> per-CPU counters remotely, use a cmpxchg loop 
> instead of a pair of read/write instructions.

FYI, try_cmpxchg() is preferred to plain cmpxchg() these days.
Apparently it generates better code on x86.

> -				v = pzstats->vm_stat_diff[i];
> -				pzstats->vm_stat_diff[i] = 0;
> +				do {
> +					v = pzstats->vm_stat_diff[i];
> +				} while (cmpxchg(&pzstats->vm_stat_diff[i], v, 0) != v);

I think this would be:

				do {
					v = pzstats->vm_stat_diff[i];
				} while (!try_cmpxchg(&pzstats->vm_stat_diff[i], v, 0));