Received: by 2002:a05:6a10:1287:0:0:0:0 with SMTP id d7csp2062415pxv; Sat, 24 Jul 2021 04:19:20 -0700 (PDT) X-Google-Smtp-Source: ABdhPJzFNVCvkGsTYke5XkB3580Pwn2Vjk1A3xYEKtAeWv3HmiBX3w2z9rQ1tzwu2wm+h2KNtGQa X-Received: by 2002:a05:6402:145a:: with SMTP id d26mr5214699edx.35.1627125559929; Sat, 24 Jul 2021 04:19:19 -0700 (PDT) ARC-Seal: i=1; a=rsa-sha256; t=1627125559; cv=none; d=google.com; s=arc-20160816; b=0sLW1IAcBhByxcMKUDRzTBda7UqBk255BJVCQOutXo2cm1tqtTLPftAg3/KaK13Npe y9EwDb/YKshM4DsEIL43zKzybxnrSbDhEp38J5IUXPe99NWFa2UM6BTLi+SXMyDsGyv9 WRcpeqR6oCnEZDSSJpnngqzaeGluQBNesHxsyjiGXqT74KE5bY0vYMLYvFYCdr6U2x/e rPTbO0o7pW/YqAiRX2+z6JRi+pEMhhVgmhMNhaoIUG2LV0V6P2kqbLAJs9WbzeF4VRaI cE38gcFDyoGnJTtSzRgfPAygDDkYAqxlDUH1sYdVjLyGJxSAQJbZESsrM1ey9bqbl4YI YSJA== ARC-Message-Signature: i=1; a=rsa-sha256; c=relaxed/relaxed; d=google.com; s=arc-20160816; h=list-id:precedence:content-transfer-encoding:mime-version :references:in-reply-to:message-id:cc:to:subject:from:date; bh=kKe5w2AujMf7VFMrXXseAx+YJiU3DL8t9FM97EAmvKk=; b=yhV0VeWJnogEbXGQLin28fIs54TIKESeJiw8+aWe0oQNxJqg0Uw7fwebaBADMgBSMP tJx3Vca6XKFBih7N2ihcOkoukvKUtgeHRGvIfLGDezibBykkuwq5pVXDF57Id2RgqfXu /3Zkv/kINPPfUyj9VlMstfjkL9pM+56yxqa+xXWKgEbErJ/3BT3m1lxMDb8r7oZN7GHP ndeKSKEumkiISHHMgKhrWGbvpCXKYbyGDGbo5g4HiRnzHbsrUsGyytbl0LmbP+LkKXK5 mxL8NaC8YFgfjW+F2lP3EWEdxelk9FthNLODgwKFePA0HfidzX8d6/5JxRKa2EtSqU1X 7zEQ== ARC-Authentication-Results: i=1; mx.google.com; spf=pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 23.128.96.18 as permitted sender) smtp.mailfrom=linux-kernel-owner@vger.kernel.org; dmarc=fail (p=NONE sp=NONE dis=NONE) header.from=crapouillou.net Return-Path: Received: from vger.kernel.org (vger.kernel.org. [23.128.96.18]) by mx.google.com with ESMTP id d2si11650060edy.190.2021.07.24.04.18.56; Sat, 24 Jul 2021 04:19:19 -0700 (PDT) Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 23.128.96.18 as permitted sender) client-ip=23.128.96.18; Authentication-Results: mx.google.com; spf=pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 23.128.96.18 as permitted sender) smtp.mailfrom=linux-kernel-owner@vger.kernel.org; dmarc=fail (p=NONE sp=NONE dis=NONE) header.from=crapouillou.net Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S234113AbhGXKgj convert rfc822-to-8bit (ORCPT + 99 others); Sat, 24 Jul 2021 06:36:39 -0400 Received: from aposti.net ([89.234.176.197]:51530 "EHLO aposti.net" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S232233AbhGXKgi (ORCPT ); Sat, 24 Jul 2021 06:36:38 -0400 Date: Sat, 24 Jul 2021 12:16:59 +0100 From: Paul Cercueil Subject: Re: [PATCH] cpuidle: JZ4780: Add Ingenic JZ4780 cpuidle driver. To: =?UTF-8?b?5ZGo55Cw5p2w?= Cc: rjw@rjwysocki.net, daniel.lezcano@linaro.org, linux-pm@vger.kernel.org, linux-mips@vger.kernel.org, linux-kernel@vger.kernel.org, dongsheng.qiu@ingenic.com, aric.pzqi@ingenic.com, rick.tyliu@ingenic.com, sihui.liu@ingenic.com, jun.jiang@ingenic.com, sernia.zhou@foxmail.com, Alex Smith Message-Id: In-Reply-To: <1627118399-125388-1-git-send-email-zhouyanjie@wanyeetech.com> References: <1627118399-125388-1-git-send-email-zhouyanjie@wanyeetech.com> MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8; format=flowed Content-Transfer-Encoding: 8BIT Precedence: bulk List-ID: X-Mailing-List: linux-kernel@vger.kernel.org Hi Zhou, Le sam., juil. 24 2021 at 17:19:59 +0800, 周琰杰 (Zhou Yanjie) a écrit : > The JZ4780 has a high overhead to executing a MIPS wait on SMP, as a > core must flush out dirty cache lines from its data cache before doing > so. This is because the core clock is gated during a wait and if the > other core tries to access a dirty line from the waiting core's cache, > it will lock up. > > To mitigate some of this impact, this driver provides a simple polling > top level idle state, to try to avoid the cache flushing overhead when > the wait will only be short. The second level state is implemented > with > the MIPS wait instruction. > > This patch first found in the github repository of CI20, the original > author is Alex Smith. Because there is a chance to cause kernel hang > scenarios which can occur within hours or even within days, so this > patch was abandoned, but now it is determined that this is not the > problem caused by this patch, but caused by the cache driver. With > the new Ingenic specific cache driver, it has been working properly > on CI20 v1 for more than one week. > > Tested-by: H. Nikolaus Schaller > Signed-off-by: Alex Smith > Signed-off-by: 周琰杰 (Zhou Yanjie) > --- > drivers/cpuidle/Kconfig.mips | 8 +++++ > drivers/cpuidle/Makefile | 1 + > drivers/cpuidle/cpuidle-jz4780.c | 74 > ++++++++++++++++++++++++++++++++++++++++ > 3 files changed, 83 insertions(+) > create mode 100644 drivers/cpuidle/cpuidle-jz4780.c > > diff --git a/drivers/cpuidle/Kconfig.mips > b/drivers/cpuidle/Kconfig.mips > index c3c011a..4a55d24 100644 > --- a/drivers/cpuidle/Kconfig.mips > +++ b/drivers/cpuidle/Kconfig.mips > @@ -16,3 +16,11 @@ config MIPS_CPS_CPUIDLE > Processing System (CPS) architecture. In order to make use of > the deepest idle states you will need to ensure that you are > also using the CONFIG_MIPS_CPS SMP implementation. > + > +config MIPS_JZ4780_CPUIDLE > + bool "CPU Idle driver for Ingenic JZ4780" > + depends on MACH_JZ4780 && SMP > + default y > + help > + Select this option to enable CPU idle state management through > + cpuidle for Ingenic JZ4780 platforms. > diff --git a/drivers/cpuidle/Makefile b/drivers/cpuidle/Makefile > index 26bbc5e..1dd372f 100644 > --- a/drivers/cpuidle/Makefile > +++ b/drivers/cpuidle/Makefile > @@ -29,6 +29,7 @@ obj-$(CONFIG_ARM_QCOM_SPM_CPUIDLE) += > cpuidle-qcom-spm.o > > ############################################################################### > # MIPS drivers > obj-$(CONFIG_MIPS_CPS_CPUIDLE) += cpuidle-cps.o > +obj-$(CONFIG_MIPS_JZ4780_CPUIDLE) += cpuidle-jz4780.o > > > ############################################################################### > # POWERPC drivers > diff --git a/drivers/cpuidle/cpuidle-jz4780.c > b/drivers/cpuidle/cpuidle-jz4780.c > new file mode 100644 > index 00000000..2025de4 > --- /dev/null > +++ b/drivers/cpuidle/cpuidle-jz4780.c > @@ -0,0 +1,74 @@ > +// SPDX-License-Identifier: GPL-2.0 > +/* > + * JZ4780 CPU idle driver > + * Copyright (C) 2015 Imagination Technologies > + * Author: Alex Smith > + * Copyright (c) 2020 周琰杰 (Zhou Yanjie) > > + */ > + > +#include > +#include > +#include > +#include > + > +#include > +#include > + > +/* > + * The JZ4780 has a high overhead to entering just the basic MIPS > wait on SMP, > + * due to the requirement to flush out dirty lines from the dcache > before > + * waiting. Therefore, we try to mitigate this overhead by using a > simple > + * polling loop for short waits. > + */ > +static int jz4780_cpuidle_poll_enter(struct cpuidle_device *dev, > + struct cpuidle_driver *drv, int index) > +{ > + if (!current_set_polling_and_test()) > + while (!need_resched() && !(read_c0_cause() & read_c0_status() & > CAUSEF_IP)) > + cpu_relax(); > + > + current_clr_polling(); > + local_irq_enable(); > + > + return index; > +} > + > +static struct cpuidle_driver jz4780_cpuidle_driver = { > + .name = "jz4780_cpuidle", > + .owner = THIS_MODULE, > + .states = { > + { > + .enter = jz4780_cpuidle_poll_enter, > + .exit_latency = 1, > + .target_residency = 1, > + .power_usage = UINT_MAX, > + .name = "poll", > + .desc = "polling loop", > + }, > + { > + .enter = mips_cpuidle_wait_enter, > + .exit_latency = 50, > + .target_residency = 300, > + .power_usage = UINT_MAX, > + .name = "wait", > + .desc = "MIPS wait", > + }, > + }, > + .state_count = 2, > +}; > + > +static int __init jz4780_cpuidle_init(void) > +{ > + int ret; > + > + ret = cpuidle_register(&jz4780_cpuidle_driver, NULL); You're missing something here - you never check that the kernel is actually running on a JZ4780. Cheers, -Paul > + if (ret) { > + pr_err("Failed to register JZ4780 idle driver: %d\n", ret); > + return ret; > + } > + > + pr_info("JZ4780 idle driver registered\n"); > + > + return 0; > +} > +device_initcall(jz4780_cpuidle_init); > -- > 2.7.4 >