From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: X-Spam-Checker-Version: SpamAssassin 3.4.0 (2014-02-07) on aws-us-west-2-korg-lkml-1.web.codeaurora.org Received: from vger.kernel.org (vger.kernel.org [23.128.96.18]) by smtp.lore.kernel.org (Postfix) with ESMTP id 25AC6C433FE for ; Mon, 10 Oct 2022 15:42:32 +0000 (UTC) Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S229790AbiJJPm3 (ORCPT ); Mon, 10 Oct 2022 11:42:29 -0400 Received: from lindbergh.monkeyblade.net ([23.128.96.19]:35554 "EHLO lindbergh.monkeyblade.net" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S229436AbiJJPmU (ORCPT ); Mon, 10 Oct 2022 11:42:20 -0400 Received: from foss.arm.com (foss.arm.com [217.140.110.172]) by lindbergh.monkeyblade.net (Postfix) with ESMTP id 9817B733C1 for ; Mon, 10 Oct 2022 08:42:19 -0700 (PDT) Received: from usa-sjc-imap-foss1.foss.arm.com (unknown [10.121.207.14]) by usa-sjc-mx-foss1.foss.arm.com (Postfix) with ESMTP id 6F03312FC; Mon, 10 Oct 2022 08:42:25 -0700 (PDT) Received: from wubuntu (unknown [10.57.35.2]) by usa-sjc-imap-foss1.foss.arm.com (Postfix) with ESMTPSA id 391FD3F67D; Mon, 10 Oct 2022 08:42:18 -0700 (PDT) Date: Mon, 10 Oct 2022 16:42:16 +0100 From: Qais Yousef To: Hillf Danton Cc: John Stultz , LKML , linux-mm@kvack.org, Connor O'Brien , Peter Zijlstra , Steven Rostedt Subject: Re: [RFC PATCH v4 2/3] sched: Avoid placing RT threads on cores handling long softirqs Message-ID: <20221010154216.6mw7fszdaoajurvm@wubuntu> References: <20221003232033.3404802-3-jstultz@google.com> <20221004013611.1822-1-hdanton@sina.com> <20221005002149.1876-1-hdanton@sina.com> <20221005060155.1571-1-hdanton@sina.com> MIME-Version: 1.0 Content-Type: text/plain; charset=utf-8 Content-Disposition: inline In-Reply-To: <20221005060155.1571-1-hdanton@sina.com> Precedence: bulk List-ID: X-Mailing-List: linux-kernel@vger.kernel.org On 10/05/22 14:01, Hillf Danton wrote: > On 4 Oct 2022 18:13:52 -0700 John Stultz > > On Tue, Oct 4, 2022 at 5:22 PM Hillf Danton wrote: > > > On 3 Oct 2022 19:29:36 -0700 John Stultz > > > > > > > > Why would ksoftirqd preempt the rt task? > > > > > > > For example the kthread becomes sensitive to latency. > > > > Is it the case where > > the ksoftirqd thread is configured to run at higher rtprio? > > > Yes, you are right. I don't see a problem here. If a sys-admin configures their ksoftirqds to be a higher priority RT tasks than the audio threads, then they better know what they're doing :-) The issue at hand here is that the softirqs boundedness is hard to control. And the scheduling delays ensued are hard to deal with by any sys-admin. Networking has actually introduced some knobs to help control that - but the tricky bit of still being able to deliver high throughput networking while keeping the softirq bounded to minimize scheduling delays/latencies. I think even for PREEMPT_RT, high performance networking could be impacted to achieve the required low latency. See this paper which explores this duality: https://citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.702.7571&rep=rep1&type=pdf With WiFi 6 and 5G mobile networks, phones are actually expected to deliver multi-gigabit network throughputs. Cheers -- Qais Yousef