[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <20240703164818.13e1d33c@kernel.org>
Date: Wed, 3 Jul 2024 16:48:18 -0700
From: Jakub Kicinski <kuba@...nel.org>
To: Shengyu Qu <wiagn233@...look.com>
Cc: nbd@....name, sean.wang@...iatek.com, Mark-MC.Lee@...iatek.com,
lorenzo@...nel.org, davem@...emloft.net, edumazet@...gle.com,
pabeni@...hat.com, matthias.bgg@...il.com,
angelogioacchino.delregno@...labora.com, pablo@...filter.org,
netdev@...r.kernel.org, linux-kernel@...r.kernel.org,
linux-arm-kernel@...ts.infradead.org, linux-mediatek@...ts.infradead.org,
stable@...r.kernel.org, Elad Yifee <eladwf@...il.com>
Subject: Re: [PATCH v2] net: ethernet: mtk_ppe: Change PPE entries number to
16K
On Thu, 4 Jul 2024 01:38:50 +0800 Shengyu Qu wrote:
> BT download and PCDN would create tons of connections, and might be
> easily to reach the 8192 limit, one of my friend sees 50000+ links when
> hosting PCDN.
I don't know what PCDN is, but what we care about in Linux is whether
the change under Fixes introduced a regression. Optimizations, and
improvements no matter how trivial in terms of code are not fixes.
So did ba37b7caf1ed ("net: ethernet: mtk_eth_soc: add support for
initializing the PPE") make things worse. And if you're saying there
are 50k "links" in real world why is 16k a major win? it's 1/3rd of
the total.
Powered by blists - more mailing lists