Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::1:20 as permitted sender) client-ip=2620:137:e000::1:20;
Date:   Tue, 10 May 2022 14:16:07 -0400
From:   Johannes Weiner <hannes@cmpxchg.org>
To:     Minchan Kim <minchan@kernel.org>
Cc:     Andrew Morton <akpm@linux-foundation.org>,
        LKML <linux-kernel@vger.kernel.org>,
        linux-mm <linux-mm@kvack.org>,
        Suren Baghdasaryan <surenb@google.com>,
        Michal Hocko <mhocko@suse.com>,
        John Dias <joaodias@google.com>,
        Tim Murray <timmurray@google.com>,
        Matthew Wilcox <willy@infradead.org>,
        Vladimir Davydov <vdavydov.dev@gmail.com>,
        Martin Liu <liumartin@google.com>
Subject: Re: [PATCH v3] mm: don't be stuck to rmap lock on reclaim path
Message-ID: <Ynqr57CVLPOqwhZa@cmpxchg.org>
References: <20220510171100.48500-1-minchan@kernel.org>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <20220510171100.48500-1-minchan@kernel.org>
Precedence: bulk

Hi Minchan,

The patch looks reasonable to me, but one thing stands out:

On Tue, May 10, 2022 at 10:11:00AM -0700, Minchan Kim wrote:
> @@ -1391,6 +1391,10 @@ static enum page_references folio_check_references(struct folio *folio,
>  	if (vm_flags & VM_LOCKED)
>  		return PAGEREF_ACTIVATE;
>  
> +	/* page_referenced didn't work due to lock contention */
> +	if (referenced_ptes == -1)
> +		return PAGEREF_KEEP;
> +
>  	if (referenced_ptes) {
>  		/*
>  		 * All mapped folios start out with page table

This means contended inactive pages get rotated.

> @@ -2492,7 +2496,7 @@ static void shrink_active_list(unsigned long nr_to_scan,
>  		}
>  
>  		if (folio_referenced(folio, 0, sc->target_mem_cgroup,
> -				     &vm_flags)) {
> +				     &vm_flags) > 0) {
>  			/*
>  			 * Identify referenced, file-backed active pages and
>  			 * give them one more trip around the active list. So

This means contended active pages do NOT get rotated.

It's a bit of an arbitrary choice what to do with reclaim candidates
for which you can get no reference information, but staying consistent
is likely a good idea. My preference would be to just rotate contended
pages throughout rather than risk dropping the workingset on the floor.

		/* Referenced or rmap lock contention: rotate */
		if (folio_referenced() != 0) {
			...

What do you think?