[<prev] [next>] [<thread-prev] [day] [month] [year] [list]
Message-ID: <44e83e1f-85e8-4485-bbe6-773214b9a264@intel.com>
Date: Thu, 13 Feb 2025 13:56:44 -0800
From: Jacob Keller <jacob.e.keller@...el.com>
To: Richard Cochran <richardcochran@...il.com>, Jakub Kicinski
<kuba@...nel.org>
CC: Jiawen Wu <jiawenwu@...stnetic.com>, <andrew+netdev@...n.ch>,
<davem@...emloft.net>, <edumazet@...gle.com>, <pabeni@...hat.com>,
<linux@...linux.org.uk>, <horms@...nel.org>, <netdev@...r.kernel.org>,
<vadim.fedorenko@...ux.dev>, <mengyuanlou@...-swift.com>
Subject: Re: [PATCH net-next v6 1/4] net: wangxun: Add support for PTP clock
On 2/13/2025 4:45 AM, Richard Cochran wrote:
> On Wed, Feb 12, 2025 at 06:17:44PM -0800, Jakub Kicinski wrote:
>
>> Give it a go, I think it will work better when machine is heavily
>> loaded and workqueues get blasted with other work items. But not
>> a hard requirement if it's difficult to get right.
>
> "work" items are essentially uncontrollable. They should be avoided
> if you need any kind of QoS for work being done.
>
> Thanks,
> Richard
+1.
I will note that simply using a kthread (like the PTP auxiliary work
thread) isn't a panacea, as this still requires configuration to ensure
that the thread has necessary priority to execute as necessary. We've
had a number of issues with deployments involving systems where the
solution requires carefully tuning the kthread to ensure it executes
reliably. Such tuning is possible with a kthread, while targeting
specific work items is not.
Thanks,
Jake
Powered by blists - more mailing lists