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;
Message-ID: <f9ca2051-741b-e4bb-74c5-178778c788df@kernel.dk>
Date:   Sun, 8 Jan 2023 20:51:43 -0700
MIME-Version: 1.0
User-Agent: Mozilla/5.0 (X11; Linux aarch64; rv:102.0) Gecko/20100101
 Thunderbird/102.6.0
Subject: Re: [PATCHv2 00/12] iov_iter: replace import_single_range with ubuf
Content-Language: en-US
To:     Keith Busch <kbusch@meta.com>, linux-kernel@vger.kernel.org,
        Alexander Viro <viro@zeniv.linux.org.uk>
Cc:     linux-aio@kvack.org, linux-fsdevel@vger.kernel.org, hch@lst.de,
        io-uring@vger.kernel.org, Pavel Begunkov <asml.silence@gmail.com>,
        Eric Dumazet <edumazet@google.com>,
        Steven Rostedt <rostedt@goodmis.org>, davem@davemloft.net,
        Jakub Kicinski <kuba@kernel.org>,
        Paolo Abeni <pabeni@redhat.com>,
        David Howells <dhowells@redhat.com>,
        Jarkko Sakkinen <jarkko@kernel.org>,
        Paul Moore <paul@paul-moore.com>, keyrings@vger.kernel.org,
        linux-security-module@vger.kernel.org,
        linux-trace-kernel@vger.kernel.org, linux-block@vger.kernel.org,
        netdev@vger.kernel.org, Keith Busch <kbusch@kernel.org>
References: <20230105190741.2405013-1-kbusch@meta.com>
From:   Jens Axboe <axboe@kernel.dk>
In-Reply-To: <20230105190741.2405013-1-kbusch@meta.com>
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit
Precedence: bulk

On 1/5/23 12:07 PM, Keith Busch wrote:
> From: Keith Busch <kbusch@kernel.org>
> 
> ITER_UBUF is a more efficient representation when using single vector
> buffers, providing small optimizations over ITER_IOVEC. This series
> introduces a helper to set these up, and replaces all applicable users
> of import_single_range with the new helper. And since there are no
> single range users left after this change, the helper is no longer
> needed.
> 
> As noted in v1(*), there are some fundamental differences to how io_uring
> compares to read/write/readv/writev. There are only the two affected
> file_operations, and they already do not work with io_uring due to their
> diverging semantics for vectored vs non-vectored read/write. Therefore,
> this series having io_uring prefer ubuf iov_iter isn't introducing new
> breakage.

Pondering how to stage this, both for later upstream but also for
testing. Would probably make the best sense to stage 1-5 separately,
and then just punt the remaining ones to the appropriate subsystems.
And then 12/12 can go in when they have all been applied.

-- 
Jens Axboe