[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <aYJXJp-hFOyG1tBz@horms.kernel.org>
Date: Tue, 3 Feb 2026 20:14:30 +0000
From: Simon Horman <horms@...nel.org>
To: Eric Dumazet <edumazet@...gle.com>
Cc: "David S. Miller" <davem@...emloft.net>,
Jakub Kicinski <kuba@...nel.org>, Paolo Abeni <pabeni@...hat.com>,
Willem de Bruijn <willemb@...gle.com>, netdev@...r.kernel.org,
eric.dumazet@...il.com
Subject: Re: [PATCH net-next] net: add vlan_get_protocol_offset_inline()
helper
On Mon, Feb 02, 2026 at 08:22:00AM +0000, Eric Dumazet wrote:
> skb_protocol() is bloated, and forces slow stack canaries in many
> fast paths.
>
> Add vlan_get_protocol_offset_inline() which deals with the non-vlan
> common cases.
>
> __vlan_get_protocol_offset() is now out of line.
>
> It returns a type_depth struct to avoid stack canaries in callers.
>
> struct type_depth {
> __be16 type;
> u16 depth;
> };
>
> $ scripts/bloat-o-meter -t vmlinux.0 vmlinux.1
> add/remove: 0/2 grow/shrink: 0/22 up/down: 0/-6320 (-6320)
> Function old new delta
> vlan_get_protocol_dgram 61 59 -2
> __pfx_skb_protocol 16 - -16
> __vlan_get_protocol_offset 307 273 -34
> tap_get_user 1374 1207 -167
> ip_md_tunnel_xmit 1625 1452 -173
> tap_sendmsg 940 753 -187
> netif_skb_features 1079 866 -213
> netem_enqueue 3017 2800 -217
> vlan_parse_protocol 271 50 -221
> tso_start 567 344 -223
> fq_dequeue 1908 1685 -223
> skb_network_protocol 434 205 -229
> ip6_tnl_xmit 2639 2409 -230
> br_dev_queue_push_xmit 474 236 -238
> skb_protocol 258 - -258
> packet_parse_headers 621 357 -264
> __ip6_tnl_rcv 1306 1039 -267
> skb_csum_hwoffload_help 515 224 -291
> ip_tunnel_xmit 2635 2339 -296
> sch_frag_xmit_hook 1582 1233 -349
> bpf_skb_ecn_set_ce 868 457 -411
> IP6_ECN_decapsulate 1297 768 -529
> ip_tunnel_rcv 2121 1489 -632
> ipip6_rcv 2572 1922 -650
> Total: Before=24892803, After=24886483, chg -0.03%
>
> Signed-off-by: Eric Dumazet <edumazet@...gle.com>
Reviewed-by: Simon Horman <horms@...nel.org>
Powered by blists - more mailing lists