Re: [PATCH v5 6/7] sched/deadline: Deferrable dl server

From: Daniel Bristot de Oliveira <bristot@kernel.org>
To: Peter Zijlstra <peterz@infradead.org>
Cc: Daniel Bristot de Oliveira <bristot@redhat.com>,
	Steven Rostedt <rostedt@goodmis.org>,
	Joel Fernandes <joel@joelfernandes.org>,
	Ingo Molnar <mingo@redhat.com>,
	Juri Lelli <juri.lelli@redhat.com>,
	Vincent Guittot <vincent.guittot@linaro.org>,
	Dietmar Eggemann <dietmar.eggemann@arm.com>,
	Ben Segall <bsegall@google.com>, Mel Gorman <mgorman@suse.de>,
	Valentin Schneider <vschneid@redhat.com>,
	linux-kernel@vger.kernel.org,
	Luca Abeni <luca.abeni@santannapisa.it>,
	Tommaso Cucinotta <tommaso.cucinotta@santannapisa.it>,
	Thomas Gleixner <tglx@linutronix.de>,
	Vineeth Pillai <vineeth@bitbyteword.org>,
	Shuah Khan <skhan@linuxfoundation.org>,
	Phil Auld <pauld@redhat.com>
Subject: Re: [PATCH v5 6/7] sched/deadline: Deferrable dl server
Date: Wed, 8 Nov 2023 14:58:12 +0100	[thread overview]
Message-ID: <be487754-5073-467b-8c73-6727e8dc3530@kernel.org> (raw)
In-Reply-To: <20231108124401.GQ8262@noisy.programming.kicks-ass.net>

On 11/8/23 13:44, Peter Zijlstra wrote:
> Should we rather not cap the runtime, something like so?
> 
> Because the above also causes period drift, which we do not want.

like in the example I showed before:

- 3/10 reservation (30%).
- w=waiting
- r=running
- s=sleeping
- T=throttled
- fair server dispatched at 0, starvation from RT.

|wwwwwwwwwwwwwwwwwwwwwwwwwwwwwwwwwwwwwwwwwwwwwwwwwwwwwwwwwwwwwwwwwwww|rrrrrrrrrrrrrrrrrrrrrrrrrrrrrr|TTTTTTTTTT[...]TTTTTTTTTTT|rrrrrrrrrrrrrrrrrrrrrrrrrrrrrrr|TTTTTTT
|___________________________period 1_________________________________|_________period 2________________________[...]___________|___period 3____________________|[.... internal-period
0---------1---------2---------3---------4---------5---------6--------7--------8---------9----------10.......11.[...]16.........17........18........19........20|[.... < Real-time
---------------------------------------------------------------------+---------------------------------------------------------|
                                                                     |                                                         +new period

From "real-world/wall clock" the internal period shift produces the
"zerolax" timeline. It runs 3 units of time before the 10's.

If one has a mix of DL and FIFO task, and want to enforce
a given response time to the fair server, they can reduce the
fair server period to achieve that.

-- Daniel