Re: [PATCH 01/23] io_uring: shuffle rarely used ctx fields

From: Song Liu <songliubraving@fb.com>
To: Pavel Begunkov <asml.silence@gmail.com>
Cc: "io-uring@vger.kernel.org" <io-uring@vger.kernel.org>,
	Networking <netdev@vger.kernel.org>,
	"bpf@vger.kernel.org" <bpf@vger.kernel.org>,
	"linux-kernel@vger.kernel.org" <linux-kernel@vger.kernel.org>,
	Jens Axboe <axboe@kernel.dk>, Alexei Starovoitov <ast@kernel.org>,
	Daniel Borkmann <daniel@iogearbox.net>,
	Andrii Nakryiko <andrii@kernel.org>, Martin Lau <kafai@fb.com>,
	Yonghong Song <yhs@fb.com>,
	John Fastabend <john.fastabend@gmail.com>,
	KP Singh <kpsingh@kernel.org>,
	Horst Schirmeier <horst.schirmeier@tu-dortmund.de>,
	"Franz-B . Tuneke" <franz-bernhard.tuneke@tu-dortmund.de>,
	Christian Dietrich <stettberger@dokucode.de>
Subject: Re: [PATCH 01/23] io_uring: shuffle rarely used ctx fields
Date: Thu, 20 May 2021 21:46:47 +0000	[thread overview]
Message-ID: <0339DB35-27AC-4A50-B807-968C72ABB698@fb.com> (raw)
In-Reply-To: <485abb65cf032f4ddf13dcc0bd60e5475638efc2.1621424513.git.asml.silence@gmail.com>

> On May 19, 2021, at 7:13 AM, Pavel Begunkov <asml.silence@gmail.com> wrote:
> 
> There is a bunch of scattered around ctx fields that are almost never
> used, e.g. only on ring exit, plunge them to the end, better locality,
> better aesthetically.
> 
> Signed-off-by: Pavel Begunkov <asml.silence@gmail.com>
> ---
> fs/io_uring.c | 36 +++++++++++++++++-------------------
> 1 file changed, 17 insertions(+), 19 deletions(-)
> 
> diff --git a/fs/io_uring.c b/fs/io_uring.c
> index 9ac5e278a91e..7e3410ce100a 100644
> --- a/fs/io_uring.c
> +++ b/fs/io_uring.c
> @@ -367,9 +367,6 @@ struct io_ring_ctx {
> 		unsigned		cached_cq_overflow;
> 		unsigned long		sq_check_overflow;
> 
> -		/* hashed buffered write serialization */
> -		struct io_wq_hash	*hash_map;
> -
> 		struct list_head	defer_list;
> 		struct list_head	timeout_list;
> 		struct list_head	cq_overflow_list;
> @@ -386,9 +383,6 @@ struct io_ring_ctx {
> 
> 	struct io_rings	*rings;
> 
> -	/* Only used for accounting purposes */
> -	struct mm_struct	*mm_account;
> -
> 	const struct cred	*sq_creds;	/* cred used for __io_sq_thread() */
> 	struct io_sq_data	*sq_data;	/* if using sq thread polling */
> 
> @@ -409,14 +403,6 @@ struct io_ring_ctx {
> 	unsigned		nr_user_bufs;
> 	struct io_mapped_ubuf	**user_bufs;
> 
> -	struct user_struct	*user;
> -
> -	struct completion	ref_comp;
> -
> -#if defined(CONFIG_UNIX)
> -	struct socket		*ring_sock;
> -#endif
> -
> 	struct xarray		io_buffers;
> 
> 	struct xarray		personalities;
> @@ -460,12 +446,24 @@ struct io_ring_ctx {
> 
> 	struct io_restriction		restrictions;
> 
> -	/* exit task_work */
> -	struct callback_head		*exit_task_work;
> -
> 	/* Keep this last, we don't need it for the fast path */
> -	struct work_struct		exit_work;
> -	struct list_head		tctx_list;
> +	struct {

Why do we need an anonymous struct here? For cache line alignment?
Do we need ____cacheline_aligned_in_smp?

> +		#if defined(CONFIG_UNIX)
> +			struct socket		*ring_sock;
> +		#endif
> +		/* hashed buffered write serialization */
> +		struct io_wq_hash		*hash_map;
> +
> +		/* Only used for accounting purposes */
> +		struct user_struct		*user;
> +		struct mm_struct		*mm_account;
> +
> +		/* ctx exit and cancelation */
> +		struct callback_head		*exit_task_work;
> +		struct work_struct		exit_work;
> +		struct list_head		tctx_list;
> +		struct completion		ref_comp;
> +	};
> };
> 
> struct io_uring_task {
> -- 
> 2.31.1
> 