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;
From:   SeongJae Park <sj@kernel.org>
To:     Matthew Wilcox <willy@infradead.org>
Cc:     SeongJae Park <sj@kernel.org>, akpm@linux-foundation.org,
        vbabka@suse.cz, paulmck@kernel.org, linux-mm@kvack.org,
        linux-kernel@vger.kernel.org
Subject: Re: [PATCH v2 2/2] mm/slab: break up RCU readers on SLAB_TYPESAFE_BY_RCU example code
Date:   Sat, 15 Apr 2023 16:27:39 +0000
Message-Id: <20230415162739.5898-1-sj@kernel.org>
In-Reply-To: <ZDoe2K7mMu31/6Lb@casper.infradead.org>
References: 
MIME-Version: 1.0
Content-Transfer-Encoding: 8bit
Precedence: bulk

On Sat, 15 Apr 2023 04:49:44 +0100 Matthew Wilcox <willy@infradead.org> wrote:

> On Sat, Apr 15, 2023 at 03:31:59AM +0000, SeongJae Park wrote:
> > The SLAB_TYPESAFE_BY_RCU example code snippet is having not tiny RCU
> > read-side critical section.  'Documentation/RCU/rculist_nulls.rst' has
> > similar example code snippet, and commit da82af04352b ("doc: Update and
> > wordsmith rculist_nulls.rst") has broken it.  Apply the change to
> > SLAB_TYPESAFE_BY_RCU example code snippet, too.
> 
> so the page cache (eg find_get_entry()) does not follow this "split
> the RCU critical section" pattern.  Should it?  What's the benefit?

The benefit would be shorter RCU grace period that allows lower memory
footprint, iiuc.

Whether it should split the section or not would depend on the lookup speed and
number of retries, I think.  If the total lookup takes a time that long enough
to make the grace period too long and therefore the amount of RCU-protected
objects that cannot freed due to the grace priod is huge, I think it would
better to follow the pattern.


Thanks,
SJ