Received-SPF: softfail (google.com: domain of transitioning linux-kernel-owner@vger.kernel.org does not designate 23.128.96.19 as permitted sender) client-ip=23.128.96.19;
Message-ID: <0a6ed6b9-0917-0d83-5c45-70ff58fad429@kernel.dk>
Date:   Fri, 20 May 2022 09:44:25 -0600
MIME-Version: 1.0
User-Agent: Mozilla/5.0 (X11; Linux aarch64; rv:91.0) Gecko/20100101
 Thunderbird/91.8.1
Subject: Re: [PATCH v4 0/3] random: convert to using iters, for Al Viro
Content-Language: en-US
To:     "Jason A. Donenfeld" <Jason@zx2c4.com>
Cc:     Theodore Ts'o <tytso@mit.edu>, Christoph Hellwig <hch@lst.de>,
        LKML <linux-kernel@vger.kernel.org>,
        Al Viro <viro@zeniv.linux.org.uk>
References: <20220520094459.116240-1-Jason@zx2c4.com>
 <Yoey+FOYO69lS5qP@zx2c4.com> <bbdeeca7-9c99-e661-6596-f04ea783a30b@kernel.dk>
 <Yoe2PPA/pzqhf+GZ@zx2c4.com>
From:   Jens Axboe <axboe@kernel.dk>
In-Reply-To: <Yoe2PPA/pzqhf+GZ@zx2c4.com>
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 7bit
Precedence: bulk

On 5/20/22 9:39 AM, Jason A. Donenfeld wrote:
> Hi Jens,
> 
> On Fri, May 20, 2022 at 09:34:46AM -0600, Jens Axboe wrote:
>> On 5/20/22 9:25 AM, Jason A. Donenfeld wrote:
>>> Hi Jens,
>>>
>>> On Fri, May 20, 2022 at 11:44:56AM +0200, Jason A. Donenfeld wrote:
>>>> Jens Axboe (3):
>>>>   random: convert to using fops->read_iter()
>>>>   random: convert to using fops->write_iter()
>>>>   random: wire up fops->splice_{read,write}_iter()
>>>
>>> FYI, this series makes reads from /dev/urandom slower, from around 616
>>> MiB/s to 598 MiB/s on my system. That seems rather unfortunate.
>>
>> How reproducible is that? That seems like a huge difference for the
>> change. How big are the reads?
> 
> Fairly reproducible. Actually, if anything, it reproduces consistently
> with worst results; I chose the most favorable ones for the new code.
> This isn't any fancy `perf` profiling, but just running:
> 
> $ pv /dev/urandom > /dev/null
> 
> From looking at strace, the read size appears to be 131072.

Ran 32, 1k, 4k here and it does seem to be down aboout 3%. Which is
definitely bigger than I expected, particularly for larger reads. If
anything, the 32b read seems comparably better than eg 1k or 4k, which
is also unexpected. Let me do a bit of profiling to see what is up.

If you're worried about it, I'd just keep the read/write and add the
iter variants on the side.

-- 
Jens Axboe