[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <20231120171959.GC245676@kernel.org>
Date: Mon, 20 Nov 2023 17:19:59 +0000
From: Simon Horman <horms@...nel.org>
To: Coco Li <lixiaoyan@...gle.com>
Cc: Jakub Kicinski <kuba@...nel.org>, Eric Dumazet <edumazet@...gle.com>,
Neal Cardwell <ncardwell@...gle.com>,
Mubashir Adnan Qureshi <mubashirq@...gle.com>,
Paolo Abeni <pabeni@...hat.com>, Andrew Lunn <andrew@...n.ch>,
Jonathan Corbet <corbet@....net>, David Ahern <dsahern@...nel.org>,
Daniel Borkmann <daniel@...earbox.net>, netdev@...r.kernel.org,
Chao Wu <wwchao@...gle.com>, Wei Wang <weiwan@...gle.com>,
Pradeep Nemavat <pnemavat@...gle.com>
Subject: Re: [PATCH v7 net-next 1/5] Documentations: Analyze heavily used
Networking related structs
On Mon, Nov 13, 2023 at 11:32:57PM +0000, Coco Li wrote:
> Analyzed a few structs in the networking stack by looking at variables
> within them that are used in the TCP/IP fast path.
>
> Fast path is defined as TCP path where data is transferred from sender to
> receiver unidirectionally. It doesn't include phases other than
> TCP_ESTABLISHED, nor does it look at error paths.
>
> We hope to re-organizing variables that span many cachelines whose fast
> path variables are also spread out, and this document can help future
> developers keep networking fast path cachelines small.
>
> Optimized_cacheline field is computed as
> (Fastpath_Bytes/L3_cacheline_size_x86), and not the actual organized
> results (see patches to come for these).
>
> Investigation is done on 6.5
>
> Name Struct_Cachelines Cur_fastpath_cache Fastpath_Bytes Optimized_cacheline
> tcp_sock 42 (2664 Bytes) 12 396 8
> net_device 39 (2240 bytes) 12 234 4
> inet_sock 15 (960 bytes) 14 922 14
> Inet_connection_sock 22 (1368 bytes) 18 1166 18
> Netns_ipv4 (sysctls) 12 (768 bytes) 4 77 2
> linux_mib 16 (1060) 6 104 2
>
> Note how there isn't much improvement space for inet_sock and
> Inet_connection_sock because sk and icsk_inet respectively takes up so
> much of the struct that rest of the variables become a small portion of
> the struct size.
>
> So, we decided to reorganize tcp_sock, net_device, Netns_ipv4, linux_mib
>
> Signed-off-by: Coco Li <lixiaoyan@...gle.com>
> Suggested-by: Eric Dumazet <edumazet@...gle.com>
...
> diff --git a/Documentation/networking/net_cachelines/index.rst b/Documentation/networking/net_cachelines/index.rst
> new file mode 100644
> index 0000000000000..92a6fbe93af35
> --- /dev/null
> +++ b/Documentation/networking/net_cachelines/index.rst
> @@ -0,0 +1,13 @@
Hi Coco,
A minor nit from my side: an SPDX header is probably appropriate at the top
of each .rst file
> +===================================
> +Common Networking Struct Cachelines
> +===================================
> +
> +.. toctree::
> + :maxdepth: 1
> +
> + inet_connection_sock
> + inet_sock
> + net_device
> + netns_ipv4_sysctl
> + snmp
> + tcp_sock
...
Powered by blists - more mailing lists