[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <da3eefaf-a73a-4d7b-8bee-5a6c874cd071@linux.dev>
Date: Tue, 20 Feb 2024 22:54:36 +0800
From: Zhu Yanjun <yanjun.zhu@...ux.dev>
To: Yevgeny Kliteynik <kliteyn@...dia.com>, Arnd Bergmann <arnd@...nel.org>,
Saeed Mahameed <saeedm@...dia.com>, Leon Romanovsky <leon@...nel.org>
Cc: Arnd Bergmann <arnd@...db.de>, "David S. Miller" <davem@...emloft.net>,
Eric Dumazet <edumazet@...gle.com>, Jakub Kicinski <kuba@...nel.org>,
Paolo Abeni <pabeni@...hat.com>, Alex Vesker <valex@...dia.com>,
Erez Shitrit <erezsh@...dia.com>, Hamdan Igbaria <hamdani@...dia.com>,
netdev@...r.kernel.org, linux-rdma@...r.kernel.org,
linux-kernel@...r.kernel.org
Subject: Re: [PATCH 1/2] net/mlx5: pre-initialize sprintf buffers
在 2024/2/20 14:57, Yevgeny Kliteynik 写道:
> On 20-Feb-24 07:50, Zhu Yanjun wrote:
>> External email: Use caution opening links or attachments
>>
>>
>> 在 2024/2/19 18:04, Arnd Bergmann 写道:
>>> From: Arnd Bergmann <arnd@...db.de>
>>>
>>> The debugfs files always in this driver all use an extra round-trip
>>> through an snprintf() before getting put into a mlx5dr_dbg_dump_buff()
>>> rather than the normal seq_printf().
>>>
>>> Zhu Yanjun noticed that the buffers are not initialized before being
>>> filled or reused and requested them to always be zeroed as a
>>> preparation for having more reused between the buffers.
>>
>> I think that you are the first to find this.
>
> The buffers are not initialized intentionally.
> The content is overwritten from the buffer's beginning.
> Initializing is not needed as it will only cause perf degradation.
If the mentioned functions are not in the critical path, this
initialization will not make too much difference on the performance.
But if this function knows how many bytes are written and read, it is
not necessary to initialize the buffer. Or else we should initialize the
buffer before it is used again.
Zhu Yanjun
>
> -- YK
>
>
Powered by blists - more mailing lists