netdev - Re: [PATCH net-next] net/mlx5e: Reuse per-RQ XDP buffer to avoid stack zeroing overhead

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [day] [month] [year] [list]

Message-ID: <CA+h3auNLbmQFXrN1A5Ashek4UiMGa_j+EHaFFp-d74kGZvyjsA@mail.gmail.com>
Date: Wed, 21 May 2025 10:56:47 +0200
From: Samuel Dobron <sdobron@...hat.com>
To: Jesper Dangaard Brouer <hawk@...nel.org>
Cc: Tariq Toukan <ttoukan.linux@...il.com>, 
	Alexei Starovoitov <alexei.starovoitov@...il.com>, "David S. Miller" <davem@...emloft.net>, 
	Jakub Kicinski <kuba@...nel.org>, Paolo Abeni <pabeni@...hat.com>, Eric Dumazet <edumazet@...gle.com>, 
	Andrew Lunn <andrew+netdev@...n.ch>, Saeed Mahameed <saeedm@...dia.com>, 
	Leon Romanovsky <leon@...nel.org>, Alexei Starovoitov <ast@...nel.org>, Daniel Borkmann <daniel@...earbox.net>, 
	John Fastabend <john.fastabend@...il.com>, Network Development <netdev@...r.kernel.org>, 
	linux-rdma@...r.kernel.org, LKML <linux-kernel@...r.kernel.org>, 
	bpf <bpf@...r.kernel.org>, Moshe Shemesh <moshe@...dia.com>, Mark Bloch <mbloch@...dia.com>, 
	Gal Pressman <gal@...dia.com>, Carolina Jubran <cjubran@...dia.com>, 
	Sebastiano Miano <mianosebastiano@...il.com>, Benjamin Poirier <bpoirier@...hat.com>, 
	Toke Hoiland Jorgensen <toke@...hat.com>
Subject: Re: [PATCH net-next] net/mlx5e: Reuse per-RQ XDP buffer to avoid
 stack zeroing overhead

Hey,
I ran tests just on stack zeroing kernel.

> The XDP_TX number are actually lower than I expected.
> Hmm... I wonder if we regressed here(?)

The absolute numbers look more or less the same,
so I would say no. The first results we have for TX is from
6.13.0-0.rc1.20241202gite70140ba0d2b.14.eln144
comparing it to 6.15.0-0.rc5.250509g9c69f8884904.47.eln148
there is actually 1% improvement. But that might be a
random fluctuation (numbers are based on 1 iteration).
We don't have data for earlier kernels...

However, for TX I get better results:

XDP_TX: DPA, swap macs:
- baseline: 9.75 Mpps
- patched 10.78 Mpps (+10%)

Maybe just different test configuration? We use xdp-bench
in dpa mode+swapping macs.

XDP_DROP:
> >>> Stack zeroing enabled:
> >>> - XDP_DROP:
> >>>      * baseline:                     24.32 Mpps
> >>>      * baseline + per-RQ allocation: 32.27 Mpps (+32.7%)

Same results on my side:
- baseline 16.6 Mpps
- patched  24.6 Mpps (+32.5%)

Seems to be fixed :)

Sam.