From patchwork Sat Dec 18 13:00:12 2021 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: Daniel Lezcano X-Patchwork-Id: 12686047 X-Patchwork-Delegate: daniel.lezcano@linaro.org 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 39F38C433F5 for ; Sat, 18 Dec 2021 13:00:41 +0000 (UTC) Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S233264AbhLRNAk (ORCPT ); Sat, 18 Dec 2021 08:00:40 -0500 Received: from lindbergh.monkeyblade.net ([23.128.96.19]:33132 "EHLO lindbergh.monkeyblade.net" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S233256AbhLRNAj (ORCPT ); Sat, 18 Dec 2021 08:00:39 -0500 Received: from mail-wm1-x32d.google.com (mail-wm1-x32d.google.com [IPv6:2a00:1450:4864:20::32d]) by lindbergh.monkeyblade.net (Postfix) with ESMTPS id 56690C061574 for ; Sat, 18 Dec 2021 05:00:38 -0800 (PST) Received: by mail-wm1-x32d.google.com with SMTP id 203so344927wmb.5 for ; Sat, 18 Dec 2021 05:00:38 -0800 (PST) DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=linaro.org; s=google; h=from:to:cc:subject:date:message-id:in-reply-to:references :mime-version:content-transfer-encoding; bh=XRtSRhv3I6GjbIaBndvp8XiBC8v+f6H0Oym54m9OYJk=; b=Y2AOCv7Kfn4X9cGi87UWfX+lkA31SnWIHSQo7lkXKa/86ubsjsS80kg9ou94lsfeq6 Ed+4gRLsPEK42iatVPGqwkF4fpVahBaIGR2CQtkdtqKNkRjyrP5KP3uYoaNVOUWRexxE N7VqiOx+LIdoDB9UR/6DWGzwNEvDmdZq+lpoiaUAR57dnhll2jxpc2rYUgADUrl+JGh3 X6KmsR5MoRtWeNvN1qrZGqaZ4s7c8hTUL4nyA8PNEtaJnCoR8uBJTnv41oC8DADwvU+e 9o+YMgveHWaXhBVzwNYfzj2KKdV56g95TOG+VsOKF7mbvBimKjsNGEL2mKyCByqdeQPx wYyA== X-Google-DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=1e100.net; s=20210112; h=x-gm-message-state:from:to:cc:subject:date:message-id:in-reply-to :references:mime-version:content-transfer-encoding; bh=XRtSRhv3I6GjbIaBndvp8XiBC8v+f6H0Oym54m9OYJk=; b=AL9ildbGqibP+Vg2pdNUoP++HEzNae76ZpuI4aYBjEQTSdEYYTYHhE+4wpD+4ML86l JqzwqW+kSJqm70MDyn6WoRmno1AL2l7VOGxZCaCps1v+pWwAeq2+O3qraxW3iVtjst9x HOUg4EK6eb9FqM9SJf7Sv7IasBQEaR2MqnSFZtw16gtMeZcK5hEpd+ce+NYhfI4QJcxN bnmxZCuf+nxolnQ8lyGEv1sgUsWbhQnGDgU3pE0y3ZMUqdmh8xgzf9GFZ8jH/0e0JZLP vsm65KNRYQHkvOw3CqhjMjBcfHVS9m4YUVc3FkmiFVufflN4r2YlIyF7kA2utcU4p2wY 1izA== X-Gm-Message-State: AOAM5315aDMQdR3n/FA4LjmnneAvbehhQHOhRnASZS+qEr/thdJb5+xS EsJi8kjIOB32SGbKKu+FZXXPxA== X-Google-Smtp-Source: ABdhPJz+P72nJND4uskpnWqEqh60n4W45WpWU5viO93fwHtP1aExU91V4shWycbTXMQIn+kVGd+Csw== X-Received: by 2002:a05:600c:a4a:: with SMTP id c10mr5596846wmq.145.1639832436842; Sat, 18 Dec 2021 05:00:36 -0800 (PST) Received: from localhost.localdomain ([2a01:e34:ed2f:f020:1f0f:c9b8:ee5c:5c2f]) by smtp.gmail.com with ESMTPSA id j16sm1465785wms.12.2021.12.18.05.00.35 (version=TLS1_3 cipher=TLS_AES_256_GCM_SHA384 bits=256/256); Sat, 18 Dec 2021 05:00:36 -0800 (PST) From: Daniel Lezcano To: daniel.lezcano@linaro.org, rjw@rjwysocki.net Cc: lukasz.luba@arm.com, robh@kernel.org, heiko@sntech.de, arnd@linaro.org, linux-kernel@vger.kernel.org, linux-pm@vger.kernel.org, ulf.hansson@linaro.org, Chanwoo Choi , Kyungmin Park , MyungJoo Ham , "Rafael J. Wysocki" , Daniel Lezcano Subject: [PATCH v5 4/6] powercap/drivers/dtpm: Add dtpm devfreq with energy model support Date: Sat, 18 Dec 2021 14:00:12 +0100 Message-Id: <20211218130014.4037640-5-daniel.lezcano@linaro.org> X-Mailer: git-send-email 2.25.1 In-Reply-To: <20211218130014.4037640-1-daniel.lezcano@linaro.org> References: <20211218130014.4037640-1-daniel.lezcano@linaro.org> MIME-Version: 1.0 Precedence: bulk List-ID: X-Mailing-List: linux-pm@vger.kernel.org Currently the dtpm supports the CPUs via cpufreq and the energy model. This change provides the same for the device which supports devfreq. Each device supporting devfreq and having an energy model can be added to the hierarchy. The concept is the same as the cpufreq DTPM support: the QoS is used to aggregate the requests and the energy model gives the value of the instantaneous power consumption ponderated by the load of the device. Cc: Chanwoo Choi Cc: Lukasz Luba Cc: Kyungmin Park Cc: MyungJoo Ham Signed-off-by: Daniel Lezcano --- drivers/powercap/Kconfig | 7 ++ drivers/powercap/Makefile | 1 + drivers/powercap/dtpm_devfreq.c | 201 ++++++++++++++++++++++++++++++++ 3 files changed, 209 insertions(+) create mode 100644 drivers/powercap/dtpm_devfreq.c diff --git a/drivers/powercap/Kconfig b/drivers/powercap/Kconfig index b1ca339957e3..515e3ceb3393 100644 --- a/drivers/powercap/Kconfig +++ b/drivers/powercap/Kconfig @@ -57,4 +57,11 @@ config DTPM_CPU help This enables support for CPU power limitation based on energy model. + +config DTPM_DEVFREQ + bool "Add device power capping based on the energy model" + depends on DTPM && ENERGY_MODEL + help + This enables support for device power limitation based on + energy model. endif diff --git a/drivers/powercap/Makefile b/drivers/powercap/Makefile index fabcf388a8d3..494617cdad88 100644 --- a/drivers/powercap/Makefile +++ b/drivers/powercap/Makefile @@ -1,6 +1,7 @@ # SPDX-License-Identifier: GPL-2.0-only obj-$(CONFIG_DTPM) += dtpm.o obj-$(CONFIG_DTPM_CPU) += dtpm_cpu.o +obj-$(CONFIG_DTPM_DEVFREQ) += dtpm_devfreq.o obj-$(CONFIG_POWERCAP) += powercap_sys.o obj-$(CONFIG_INTEL_RAPL_CORE) += intel_rapl_common.o obj-$(CONFIG_INTEL_RAPL) += intel_rapl_msr.o diff --git a/drivers/powercap/dtpm_devfreq.c b/drivers/powercap/dtpm_devfreq.c new file mode 100644 index 000000000000..fd3817f71f44 --- /dev/null +++ b/drivers/powercap/dtpm_devfreq.c @@ -0,0 +1,201 @@ +// SPDX-License-Identifier: GPL-2.0-only +/* + * Copyright 2021 Linaro Limited + * + * Author: Daniel Lezcano + * + * The devfreq device combined with the energy model and the load can + * give an estimation of the power consumption as well as limiting the + * power. + * + */ +#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt + +#include +#include +#include +#include +#include +#include +#include +#include + +struct dtpm_devfreq { + struct dtpm dtpm; + struct dev_pm_qos_request qos_req; + struct devfreq *devfreq; +}; + +static struct dtpm_devfreq *to_dtpm_devfreq(struct dtpm *dtpm) +{ + return container_of(dtpm, struct dtpm_devfreq, dtpm); +} + +static int update_pd_power_uw(struct dtpm *dtpm) +{ + struct dtpm_devfreq *dtpm_devfreq = to_dtpm_devfreq(dtpm); + struct devfreq *devfreq = dtpm_devfreq->devfreq; + struct device *dev = devfreq->dev.parent; + struct em_perf_domain *pd = em_pd_get(dev); + + dtpm->power_min = pd->table[0].power; + dtpm->power_min *= MICROWATT_PER_MILLIWATT; + + dtpm->power_max = pd->table[pd->nr_perf_states - 1].power; + dtpm->power_max *= MICROWATT_PER_MILLIWATT; + + return 0; +} + +static u64 set_pd_power_limit(struct dtpm *dtpm, u64 power_limit) +{ + struct dtpm_devfreq *dtpm_devfreq = to_dtpm_devfreq(dtpm); + struct devfreq *devfreq = dtpm_devfreq->devfreq; + struct device *dev = devfreq->dev.parent; + struct em_perf_domain *pd = em_pd_get(dev); + unsigned long freq; + u64 power; + int i; + + for (i = 0; i < pd->nr_perf_states; i++) { + + power = pd->table[i].power * MICROWATT_PER_MILLIWATT; + if (power > power_limit) + break; + } + + freq = pd->table[i - 1].frequency; + + dev_pm_qos_update_request(&dtpm_devfreq->qos_req, freq); + + power_limit = pd->table[i - 1].power * MICROWATT_PER_MILLIWATT; + + return power_limit; +} + +static void _normalize_load(struct devfreq_dev_status *status) +{ + if (status->total_time > 0xfffff) { + status->total_time >>= 10; + status->busy_time >>= 10; + } + + status->busy_time <<= 10; + status->busy_time /= status->total_time ? : 1; + + status->busy_time = status->busy_time ? : 1; + status->total_time = 1024; +} + +static u64 get_pd_power_uw(struct dtpm *dtpm) +{ + struct dtpm_devfreq *dtpm_devfreq = to_dtpm_devfreq(dtpm); + struct devfreq *devfreq = dtpm_devfreq->devfreq; + struct device *dev = devfreq->dev.parent; + struct em_perf_domain *pd = em_pd_get(dev); + struct devfreq_dev_status status; + unsigned long freq; + u64 power; + int i; + + mutex_lock(&devfreq->lock); + status = devfreq->last_status; + mutex_unlock(&devfreq->lock); + + freq = DIV_ROUND_UP(status.current_frequency, HZ_PER_KHZ); + _normalize_load(&status); + + for (i = 0; i < pd->nr_perf_states; i++) { + + if (pd->table[i].frequency < freq) + continue; + + power = pd->table[i].power * MICROWATT_PER_MILLIWATT; + power *= status.busy_time; + power >>= 10; + + return power; + } + + return 0; +} + +static void pd_release(struct dtpm *dtpm) +{ + struct dtpm_devfreq *dtpm_devfreq = to_dtpm_devfreq(dtpm); + + if (dev_pm_qos_request_active(&dtpm_devfreq->qos_req)) + dev_pm_qos_remove_request(&dtpm_devfreq->qos_req); + + kfree(dtpm_devfreq); +} + +static struct dtpm_ops dtpm_ops = { + .set_power_uw = set_pd_power_limit, + .get_power_uw = get_pd_power_uw, + .update_power_uw = update_pd_power_uw, + .release = pd_release, +}; + +static int __dtpm_devfreq_setup(struct devfreq *devfreq, struct dtpm *parent) +{ + struct device *dev = devfreq->dev.parent; + struct dtpm_devfreq *dtpm_devfreq; + struct em_perf_domain *pd; + int ret = -ENOMEM; + + pd = em_pd_get(dev); + if (!pd) { + ret = dev_pm_opp_of_register_em(dev, NULL); + if (ret) { + pr_err("No energy model available for '%s'\n", dev_name(dev)); + return -EINVAL; + } + } + + dtpm_devfreq = kzalloc(sizeof(*dtpm_devfreq), GFP_KERNEL); + if (!dtpm_devfreq) + return -ENOMEM; + + dtpm_init(&dtpm_devfreq->dtpm, &dtpm_ops); + + dtpm_devfreq->devfreq = devfreq; + + ret = dtpm_register(dev_name(dev), &dtpm_devfreq->dtpm, parent); + if (ret) { + pr_err("Failed to register '%s': %d\n", dev_name(dev), ret); + goto out_dtpm_devfreq; + } + + ret = dev_pm_qos_add_request(dev, &dtpm_devfreq->qos_req, + DEV_PM_QOS_MAX_FREQUENCY, + PM_QOS_MAX_FREQUENCY_DEFAULT_VALUE); + if (ret) { + pr_err("Failed to add QoS request: %d\n", ret); + goto out_dtpm_unregister; + } + + dtpm_update_power(&dtpm_devfreq->dtpm); + + return 0; + +out_dtpm_unregister: + dtpm_unregister(&dtpm_devfreq->dtpm); +out_dtpm_devfreq: + kfree(dtpm_devfreq); + + return ret; +} + +static int dtpm_devfreq_setup(struct dtpm *dtpm, struct device_node *np) +{ + struct devfreq *devfreq; + + devfreq = devfreq_get_devfreq_by_node(np); + if (IS_ERR(devfreq)) + return 0; + + return __dtpm_devfreq_setup(devfreq, dtpm); +} + +DTPM_DECLARE(dtpm_dev, NULL, dtpm_devfreq_setup);