Re: [PATCH] Increase default MLOCK_LIMIT to 8 MiB

From: Vito Caputo <vcaputo@pengaru.com>
To: Drew DeVault <sir@cmpwn.com>
Cc: Jens Axboe <axboe@kernel.dk>,
	Matthew Wilcox <willy@infradead.org>,
	Andrew Morton <akpm@linux-foundation.org>,
	Ammar Faizi <ammarfaizi2@gnuweeb.org>,
	linux-kernel@vger.kernel.org, linux-api@vger.kernel.org,
	io_uring Mailing List <io-uring@vger.kernel.org>,
	Pavel Begunkov <asml.silence@gmail.com>,
	linux-mm@kvack.org
Subject: Re: [PATCH] Increase default MLOCK_LIMIT to 8 MiB
Date: Tue, 16 Nov 2021 11:46:44 -0800	[thread overview]
Message-ID: <20211116194644.uyvfz4zzzjlbfqbm@shells.gnugeneration.com> (raw)
In-Reply-To: <CFRG8CM6QUPN.1Z75SA6XN02W1@taiga>

On Tue, Nov 16, 2021 at 08:25:33PM +0100, Drew DeVault wrote:
> On Tue Nov 16, 2021 at 8:21 PM CET, Vito Caputo wrote:
> > Considering a single fullscreen 32bpp 4K-resolution framebuffer is
> > ~32MiB, I'm not convinced this is really correct in nearly 2022.
> 
> Can you name a practical use-case where you'll be doing I/O with
> uncompressed 4K framebuffers? The kind of I/O which is supported by
> io_uring, to be specific, not, say, handing it off to libdrm.

Obviously video/image editing software tends to operate on raw frames,
and sometimes even persists them via filesystems.

I haven't given it a lot of thought, but a framebuffer is a commonly
used unit of memory allocation in code run on the CPU I've written
over the years.  Being able to pin those for something like io_uring
(or some other DMA-possible interface) seems like an obvious
memory-hungry thing to consider here if we're talking default upper
bounds.

Regards,
Vito Caputo