[<prev] [next>] [<thread-prev] [day] [month] [year] [list]
Message-ID: <ZtVUVfWYJtl_3hX9@slm.duckdns.org>
Date: Sun, 1 Sep 2024 19:59:49 -1000
From: Tejun Heo <tj@...nel.org>
To: Changwoo Min <multics69@...il.com>
Cc: Andrea Righi <andrea.righi@...ux.dev>, void@...ifault.com,
kernel-team@...a.com, linux-kernel@...r.kernel.org,
Daniel Hodges <hodges.daniel.scott@...il.com>,
Dan Schatzberg <schatzberg.dan@...il.com>,
"kernel-dev@...lia.com" <kernel-dev@...lia.com>
Subject: Re: [PATCH 10/11] sched_ext: Implement
scx_bpf_dispatch[_vtime]_from_dsq()
Hello,
On Mon, Sep 02, 2024 at 10:53:33AM +0900, Changwoo Min wrote:
..
> > IMHO we can simply drop them, introducing a separate struct makes the
> > API a bit inconsistent with scx_bpf_dispatch() (and I don't think we
> > want to change also scx_bpf_dispatch() for that).
>
> Dropping @slice and @vtime would be cleaner in terms of the API
> interface. Some use cases simply move a task from one DSQ to
> another (e.g., from a shared DSQ to a per-domain DSQ).
Yeap, I sent an updated version where both parameters are dropped. There are
separate kfunc calls that can precede the dispatch call to override
slice/vtime if necessary.
Thanks.
--
tejun
Powered by blists - more mailing lists