From patchwork Fri Aug 5 07:14:23 2022 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: Jiucheng Xu X-Patchwork-Id: 595893 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 48F08C00140 for ; Fri, 5 Aug 2022 07:14:47 +0000 (UTC) Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S235554AbiHEHOq (ORCPT ); Fri, 5 Aug 2022 03:14:46 -0400 Received: from lindbergh.monkeyblade.net ([23.128.96.19]:50030 "EHLO lindbergh.monkeyblade.net" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S234811AbiHEHOo (ORCPT ); Fri, 5 Aug 2022 03:14:44 -0400 Received: from mail-sh.amlogic.com (mail-sh.amlogic.com [58.32.228.43]) by lindbergh.monkeyblade.net (Postfix) with ESMTPS id EF4D4336; Fri, 5 Aug 2022 00:14:39 -0700 (PDT) Received: from droid01-xa.amlogic.com (10.88.11.200) by mail-sh.amlogic.com (10.18.11.5) with Microsoft SMTP Server id 15.1.2507.9; Fri, 5 Aug 2022 15:14:36 +0800 From: Jiucheng Xu To: , , , CC: Rob Herring , Krzysztof Kozlowski , Will Deacon , Mark Rutland , Neil Armstrong , Kevin Hilman , Jerome Brunet , Martin Blumenstingl , Chris Healy , Jiucheng Xu , kernel test robot Subject: [PATCH v4 1/4] perf/amlogic: Add support for Amlogic meson G12 SoC DDR PMU driver Date: Fri, 5 Aug 2022 15:14:23 +0800 Message-ID: <20220805071426.2598818-1-jiucheng.xu@amlogic.com> X-Mailer: git-send-email 2.25.1 MIME-Version: 1.0 X-Originating-IP: [10.88.11.200] Precedence: bulk List-ID: X-Mailing-List: devicetree@vger.kernel.org This patch adds support Amlogic meson G12 series SoC DDR bandwidth PMU driver framework and interfaces. The PMU not only can monitor the total DDR bandwidth, but also the bandwidth which is from individual IP module. Example usage: $ perf stat -a -e aml_ddr_bw/total_rw_bytes/ -I 1000 sleep 10 - or - $ perf stat -a -e \ aml_ddr_bw/total_rw_bytes/,\ aml_ddr_bw/chan_1_rw_bytes,arm=1/ -I 1000 \ sleep 10 g12 SoC support 4 channels to monitor DDR bandwidth simultaneously. Each channel can monitor up to 4 IP modules simultaneously. For Instance, If you want to get the sum of DDR bandwidth from CPU, GPU, USB3.0 and VDEC. You can use the following command parameters to display. $ perf stat -a -e \ aml_ddr_bw/chan_2_rw_bytes,arm=1,gpu=1,usb3_0=1,nna=1/ -I 1000 \ sleep 10 Other events are supported, and advertised via perf list. Signed-off-by: Jiucheng Xu Reported-by: kernel test robot --- Changes v3 -> v4: - No change Changes v2 -> v3: - Fix sh GCC 12.1.0 compiling warning - Rename prefix aml to meson for files and code Changes v1 -> v2: - Remove inline to let GCC make the decisions - Remove spinlock - Remove ddr_cnt_accumulate() - Remove the message which only indicate a bug - Remove all dev_warn() message - Use hweight64() helper instead of whole loop - Remove setting of hwc - Use for_each_set_bit() helper for bit loop - Use sysfs_emit() in sysfs show - Remove checking for bugs - Replace irq_set_affinity_hint() to irq_set_affinity() - Remove #ifdef CONFIG_OF - Use devm_platform_ioremap_resource() instead of platform_get_resource()&ioremap() - Use platform_get_irq() instead of platform_get_resource()&ioremap() - Replace IRQF_SHARED to IRQF_NOBALANCING - Remove meaningless log like "init ok" - Use compatible instead of creating new property to distinguish different platform. - Use the is_visible callback to avoid exposing unsupported fmt_attr - Use module_platform_driver_probe() instead of module_init/exit --- MAINTAINERS | 7 + drivers/perf/Kconfig | 2 + drivers/perf/Makefile | 1 + drivers/perf/amlogic/Kconfig | 10 + drivers/perf/amlogic/Makefile | 5 + drivers/perf/amlogic/meson_ddr_pmu_core.c | 544 ++++++++++++++++++++++ drivers/perf/amlogic/meson_g12_ddr_pmu.c | 388 +++++++++++++++ include/soc/amlogic/meson_ddr_pmu.h | 76 +++ 8 files changed, 1033 insertions(+) create mode 100644 drivers/perf/amlogic/Kconfig create mode 100644 drivers/perf/amlogic/Makefile create mode 100644 drivers/perf/amlogic/meson_ddr_pmu_core.c create mode 100644 drivers/perf/amlogic/meson_g12_ddr_pmu.c create mode 100644 include/soc/amlogic/meson_ddr_pmu.h diff --git a/MAINTAINERS b/MAINTAINERS index 1d99d54fa881..ac8a98dfbacc 100644 --- a/MAINTAINERS +++ b/MAINTAINERS @@ -1050,6 +1050,13 @@ S: Maintained F: Documentation/hid/amd-sfh* F: drivers/hid/amd-sfh-hid/ +AMLOGIC DDR PMU DRIVER +M: Jiucheng Xu +S: Supported +W: http://www.amlogic.com +F: drivers/perf/amlogic/ +F: include/soc/amlogic/ + AMPHION VPU CODEC V4L2 DRIVER M: Ming Qian M: Shijie Qin diff --git a/drivers/perf/Kconfig b/drivers/perf/Kconfig index 1e2d69453771..68200c798128 100644 --- a/drivers/perf/Kconfig +++ b/drivers/perf/Kconfig @@ -192,4 +192,6 @@ config MARVELL_CN10K_DDR_PMU Enable perf support for Marvell DDR Performance monitoring event on CN10K platform. +source "drivers/perf/amlogic/Kconfig" + endmenu diff --git a/drivers/perf/Makefile b/drivers/perf/Makefile index 57a279c61df5..4930482a15ff 100644 --- a/drivers/perf/Makefile +++ b/drivers/perf/Makefile @@ -20,3 +20,4 @@ obj-$(CONFIG_ARM_DMC620_PMU) += arm_dmc620_pmu.o obj-$(CONFIG_MARVELL_CN10K_TAD_PMU) += marvell_cn10k_tad_pmu.o obj-$(CONFIG_MARVELL_CN10K_DDR_PMU) += marvell_cn10k_ddr_pmu.o obj-$(CONFIG_APPLE_M1_CPU_PMU) += apple_m1_cpu_pmu.o +obj-$(CONFIG_MESON_DDR_PMU) += amlogic/ diff --git a/drivers/perf/amlogic/Kconfig b/drivers/perf/amlogic/Kconfig new file mode 100644 index 000000000000..f68db01a7f17 --- /dev/null +++ b/drivers/perf/amlogic/Kconfig @@ -0,0 +1,10 @@ +# SPDX-License-Identifier: GPL-2.0-only +config MESON_DDR_PMU + tristate "Amlogic DDR Bandwidth Performance Monitor" + depends on ARCH_MESON || COMPILE_TEST + help + Provides support for the DDR performance monitor + in Amlogic SoCs, which can give information about + memory throughput and other related events. It + supports multiple channels to monitor the memory + bandwidth simultaneously. diff --git a/drivers/perf/amlogic/Makefile b/drivers/perf/amlogic/Makefile new file mode 100644 index 000000000000..8dd892a25efa --- /dev/null +++ b/drivers/perf/amlogic/Makefile @@ -0,0 +1,5 @@ +# SPDX-License-Identifier: GPL-2.0-only + +obj-$(CONFIG_MESON_DDR_PMU) += meson_ddr_pmu.o + +meson_ddr_pmu-y := meson_ddr_pmu_core.o meson_g12_ddr_pmu.o diff --git a/drivers/perf/amlogic/meson_ddr_pmu_core.c b/drivers/perf/amlogic/meson_ddr_pmu_core.c new file mode 100644 index 000000000000..7a51b445f6df --- /dev/null +++ b/drivers/perf/amlogic/meson_ddr_pmu_core.c @@ -0,0 +1,544 @@ +// SPDX-License-Identifier: GPL-2.0 +/* + * Copyright (c) 2022 Amlogic, Inc. All rights reserved. + */ +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include + +#include + +#define DDR_PERF_DEV_NAME "meson_ddr_bw" +#define MAX_AXI_PORTS_OF_CHANNEL 4 /* A DMC channel can monitor max 4 axi ports */ + +#define to_ddr_pmu(p) container_of(p, struct ddr_pmu, pmu) + +#define hw_info_to_pmu(p) container_of(p, struct ddr_pmu, info) + +static void dmc_pmu_enable(struct ddr_pmu *pmu) +{ + if (!pmu->pmu_enabled) + pmu->info.ops->enable(&pmu->info); + + pmu->pmu_enabled = true; +} + +static void dmc_pmu_disable(struct ddr_pmu *pmu) +{ + if (pmu->pmu_enabled) + pmu->info.ops->disable(&pmu->info); + + pmu->pmu_enabled = false; +} + +static void meson_ddr_config_axi_id(struct ddr_pmu *pmu, int axi_id, int chann) +{ + pmu->info.ops->config_axi_id(&pmu->info, axi_id, chann); +} + +static void meson_ddr_set_filter(struct perf_event *event, u8 axi_id) +{ + struct ddr_pmu *pmu = to_ddr_pmu(event->pmu); + int chann; + + if (event->attr.config > ALL_CHAN_COUNTER_ID && + event->attr.config < COUNTER_MAX_ID) { + chann = event->attr.config - CHAN1_COUNTER_ID; + + meson_ddr_config_axi_id(pmu, axi_id, chann); + } +} + +static void ddr_cnt_addition(struct dmc_counter *sum, + struct dmc_counter *add1, + struct dmc_counter *add2, + int chann_nr) +{ + int i; + u64 cnt1, cnt2; + + sum->all_cnt = add1->all_cnt + add2->all_cnt; + sum->all_req = add1->all_req + add2->all_req; + for (i = 0; i < chann_nr; i++) { + cnt1 = add1->channel_cnt[i]; + cnt2 = add2->channel_cnt[i]; + + sum->channel_cnt[i] = cnt1 + cnt2; + } +} + +static void meson_ddr_perf_event_update(struct perf_event *event) +{ + struct ddr_pmu *pmu = to_ddr_pmu(event->pmu); + u64 new_raw_count = 0; + struct dmc_counter dc = {0}, sum_dc = {0}; + int idx; + + /* get the remain counters in register. */ + pmu->info.ops->get_counters(&pmu->info, &dc); + + ddr_cnt_addition(&sum_dc, &pmu->counters, &dc, pmu->info.chann_nr); + + switch (event->attr.config) { + case ALL_CHAN_COUNTER_ID: + new_raw_count = sum_dc.all_cnt; + break; + case CHAN1_COUNTER_ID: + case CHAN2_COUNTER_ID: + case CHAN3_COUNTER_ID: + case CHAN4_COUNTER_ID: + case CHAN5_COUNTER_ID: + case CHAN6_COUNTER_ID: + case CHAN7_COUNTER_ID: + case CHAN8_COUNTER_ID: + idx = event->attr.config - CHAN1_COUNTER_ID; + new_raw_count = sum_dc.channel_cnt[idx]; + break; + } + + local64_set(&event->count, new_raw_count); +} + +static int meson_ddr_perf_event_init(struct perf_event *event) +{ + struct ddr_pmu *pmu = to_ddr_pmu(event->pmu); + u64 config1 = event->attr.config1; + + if (event->attr.type != event->pmu->type) + return -ENOENT; + + if (is_sampling_event(event) || event->attach_state & PERF_ATTACH_TASK) + return -EOPNOTSUPP; + + if (event->cpu < 0) + return -EOPNOTSUPP; + + /* check if the number of parameters is too much */ + if (event->attr.config != ALL_CHAN_COUNTER_ID && + hweight64(config1) > MAX_AXI_PORTS_OF_CHANNEL) + return -EOPNOTSUPP; + + event->cpu = pmu->cpu; + + return 0; +} + +static void meson_ddr_perf_event_start(struct perf_event *event, int flags) +{ + struct ddr_pmu *pmu = to_ddr_pmu(event->pmu); + + memset(&pmu->counters, 0, sizeof(pmu->counters)); + dmc_pmu_enable(pmu); +} + +static int meson_ddr_perf_event_add(struct perf_event *event, int flags) +{ + u64 config1 = event->attr.config1; + int i; + + for_each_set_bit(i, (const unsigned long *)&config1, sizeof(config1)) + meson_ddr_set_filter(event, i); + + if (flags & PERF_EF_START) + meson_ddr_perf_event_start(event, flags); + + return 0; +} + +static void meson_ddr_perf_event_stop(struct perf_event *event, int flags) +{ + struct ddr_pmu *pmu = to_ddr_pmu(event->pmu); + + if (flags & PERF_EF_UPDATE) + meson_ddr_perf_event_update(event); + + dmc_pmu_disable(pmu); +} + +static void meson_ddr_perf_event_del(struct perf_event *event, int flags) +{ + meson_ddr_perf_event_stop(event, PERF_EF_UPDATE); +} + +static ssize_t meson_ddr_perf_cpumask_show(struct device *dev, + struct device_attribute *attr, + char *buf) +{ + struct ddr_pmu *pmu = dev_get_drvdata(dev); + + return cpumap_print_to_pagebuf(true, buf, cpumask_of(pmu->cpu)); +} + +static struct device_attribute meson_ddr_perf_cpumask_attr = +__ATTR(cpumask, 0444, meson_ddr_perf_cpumask_show, NULL); + +static struct attribute *meson_ddr_perf_cpumask_attrs[] = { + &meson_ddr_perf_cpumask_attr.attr, + NULL, +}; + +static const struct attribute_group ddr_perf_cpumask_attr_group = { + .attrs = meson_ddr_perf_cpumask_attrs, +}; + +static ssize_t +pmu_event_show(struct device *dev, struct device_attribute *attr, + char *page) +{ + struct perf_pmu_events_attr *pmu_attr; + + pmu_attr = container_of(attr, struct perf_pmu_events_attr, attr); + return sysfs_emit(page, "event=0x%02llx\n", pmu_attr->id); +} + +static ssize_t +event_show_unit(struct device *dev, struct device_attribute *attr, + char *page) +{ + return sysfs_emit(page, "MB\n"); +} + +static ssize_t +event_show_scale(struct device *dev, struct device_attribute *attr, + char *page) +{ + /* one count = 16byte = 1.52587890625e-05 MB */ + return sysfs_emit(page, "1.52587890625e-05\n"); +} + +#define AML_DDR_PMU_EVENT_ATTR(_name, _id) \ +{ \ + .attr = __ATTR(_name, 0444, pmu_event_show, NULL), \ + .id = _id, \ +} + +#define AML_DDR_PMU_EVENT_UNIT_ATTR(_name) \ + __ATTR(_name.unit, 0444, event_show_unit, NULL) + +#define AML_DDR_PMU_EVENT_SCALE_ATTR(_name) \ + __ATTR(_name.scale, 0444, event_show_scale, NULL) + +static struct device_attribute event_unit_attrs[] = { + AML_DDR_PMU_EVENT_UNIT_ATTR(total_rw_bytes), + AML_DDR_PMU_EVENT_UNIT_ATTR(chan_1_rw_bytes), + AML_DDR_PMU_EVENT_UNIT_ATTR(chan_2_rw_bytes), + AML_DDR_PMU_EVENT_UNIT_ATTR(chan_3_rw_bytes), + AML_DDR_PMU_EVENT_UNIT_ATTR(chan_4_rw_bytes), + AML_DDR_PMU_EVENT_UNIT_ATTR(chan_5_rw_bytes), + AML_DDR_PMU_EVENT_UNIT_ATTR(chan_6_rw_bytes), + AML_DDR_PMU_EVENT_UNIT_ATTR(chan_7_rw_bytes), + AML_DDR_PMU_EVENT_UNIT_ATTR(chan_8_rw_bytes), +}; + +static struct device_attribute event_scale_attrs[] = { + AML_DDR_PMU_EVENT_SCALE_ATTR(total_rw_bytes), + AML_DDR_PMU_EVENT_SCALE_ATTR(chan_1_rw_bytes), + AML_DDR_PMU_EVENT_SCALE_ATTR(chan_2_rw_bytes), + AML_DDR_PMU_EVENT_SCALE_ATTR(chan_3_rw_bytes), + AML_DDR_PMU_EVENT_SCALE_ATTR(chan_4_rw_bytes), + AML_DDR_PMU_EVENT_SCALE_ATTR(chan_5_rw_bytes), + AML_DDR_PMU_EVENT_SCALE_ATTR(chan_6_rw_bytes), + AML_DDR_PMU_EVENT_SCALE_ATTR(chan_7_rw_bytes), + AML_DDR_PMU_EVENT_SCALE_ATTR(chan_8_rw_bytes), +}; + +static struct perf_pmu_events_attr event_attrs[] = { + AML_DDR_PMU_EVENT_ATTR(total_rw_bytes, ALL_CHAN_COUNTER_ID), + AML_DDR_PMU_EVENT_ATTR(chan_1_rw_bytes, CHAN1_COUNTER_ID), + AML_DDR_PMU_EVENT_ATTR(chan_2_rw_bytes, CHAN2_COUNTER_ID), + AML_DDR_PMU_EVENT_ATTR(chan_3_rw_bytes, CHAN3_COUNTER_ID), + AML_DDR_PMU_EVENT_ATTR(chan_4_rw_bytes, CHAN4_COUNTER_ID), + AML_DDR_PMU_EVENT_ATTR(chan_5_rw_bytes, CHAN5_COUNTER_ID), + AML_DDR_PMU_EVENT_ATTR(chan_6_rw_bytes, CHAN6_COUNTER_ID), + AML_DDR_PMU_EVENT_ATTR(chan_7_rw_bytes, CHAN7_COUNTER_ID), + AML_DDR_PMU_EVENT_ATTR(chan_8_rw_bytes, CHAN8_COUNTER_ID), +}; + +/* three attrs are combined an event */ +static struct attribute *ddr_perf_events_attrs[COUNTER_MAX_ID * 3]; + +static struct attribute_group ddr_perf_events_attr_group = { + .name = "events", + .attrs = ddr_perf_events_attrs, +}; + +/* an attr represents axi id, the max id is less than 256 */ +static struct attribute *ddr_perf_format_attrs[256]; + +static umode_t meson_ddr_perf_format_attr_visible(struct kobject *kobj, + struct attribute *attr, + int n) +{ + return attr->mode; +} + +static struct attribute_group ddr_perf_format_attr_group = { + .name = "format", + .attrs = ddr_perf_format_attrs, + .is_visible = meson_ddr_perf_format_attr_visible, +}; + +static ssize_t meson_ddr_perf_identifier_show(struct device *dev, + struct device_attribute *attr, + char *page) +{ + struct ddr_pmu *pmu = dev_get_drvdata(dev); + + return sysfs_emit(page, "%s\n", pmu->name); +} + +static struct device_attribute meson_ddr_perf_identifier_attr = +__ATTR(identifier, 0444, meson_ddr_perf_identifier_show, NULL); + +static struct attribute *meson_ddr_perf_identifier_attrs[] = { + &meson_ddr_perf_identifier_attr.attr, + NULL, +}; + +static const struct attribute_group ddr_perf_identifier_attr_group = { + .attrs = meson_ddr_perf_identifier_attrs, +}; + +static const struct attribute_group *attr_groups[] = { + &ddr_perf_events_attr_group, + &ddr_perf_format_attr_group, + &ddr_perf_cpumask_attr_group, + &ddr_perf_identifier_attr_group, + NULL, +}; + +static irqreturn_t dmc_irq_handler(int irq, void *dev_id) +{ + struct dmc_hw_info *info = dev_id; + struct ddr_pmu *pmu; + struct dmc_counter counters, *sum_cnter; + int i; + + pmu = hw_info_to_pmu(info); + + if (info->ops->irq_handler(info, &counters) != 0) + goto out; + + sum_cnter = &pmu->counters; + sum_cnter->all_cnt += counters.all_cnt; + sum_cnter->all_req += counters.all_req; + + for (i = 0; i < pmu->info.chann_nr; i++) + sum_cnter->channel_cnt[i] += counters.channel_cnt[i]; + + if (pmu->pmu_enabled) + /* + * the timer interrupt only supprt + * one shot mode, we have to re-enable + * it in ISR to support continue mode. + */ + info->ops->enable(info); + + dev_dbg(pmu->dev, "counts: %llu %llu %llu, %llu, %llu, %llu\t\t" + "sum: %llu %llu %llu, %llu, %llu, %llu\n", + counters.all_req, + counters.all_cnt, + counters.channel_cnt[0], + counters.channel_cnt[1], + counters.channel_cnt[2], + counters.channel_cnt[3], + + pmu->counters.all_req, + pmu->counters.all_cnt, + pmu->counters.channel_cnt[0], + pmu->counters.channel_cnt[1], + pmu->counters.channel_cnt[2], + pmu->counters.channel_cnt[3]); +out: + return IRQ_HANDLED; +} + +static int ddr_perf_offline_cpu(unsigned int cpu, struct hlist_node *node) +{ + struct ddr_pmu *pmu = hlist_entry_safe(node, struct ddr_pmu, node); + int target; + + if (cpu != pmu->cpu) + return 0; + + target = cpumask_any_but(cpu_online_mask, cpu); + if (target >= nr_cpu_ids) + return 0; + + perf_pmu_migrate_context(&pmu->pmu, cpu, target); + pmu->cpu = target; + + WARN_ON(irq_set_affinity(pmu->info.irq_num, cpumask_of(pmu->cpu))); + + return 0; +} + +static void fill_event_attr(struct ddr_pmu *pmu) +{ + int i, j, k; + struct attribute **dst = ddr_perf_events_attrs; + + j = 0; + k = 0; + + /* fill ALL_CHAN_COUNTER_ID event */ + dst[j++] = &event_attrs[k].attr.attr; + dst[j++] = &event_unit_attrs[k].attr; + dst[j++] = &event_scale_attrs[k].attr; + + k++; + + /* fill each channel event */ + for (i = 0; i < pmu->info.chann_nr; i++, k++) { + dst[j++] = &event_attrs[k].attr.attr; + dst[j++] = &event_unit_attrs[k].attr; + dst[j++] = &event_scale_attrs[k].attr; + } + + dst[j] = NULL; /* mark end */ +} + +static void fmt_attr_fill(struct attribute **fmt_attr) +{ + int i; + + for (i = 0; fmt_attr[i]; i++) + ddr_perf_format_attrs[i] = fmt_attr[i]; + + ddr_perf_format_attrs[i] = NULL; +} + +static int ddr_pmu_parse_dt(struct platform_device *pdev, struct dmc_hw_info *info) +{ + /*struct pinctrl *p;*/ + void __iomem *base; + int i, ret = -EINVAL; + + for (i = 0; i < info->dmc_nr; i++) { + /* resource 0 for ddr register base */ + base = devm_platform_ioremap_resource(pdev, i); + if (IS_ERR(base)) { + dev_err(&pdev->dev, "couldn't ioremap ddr reg %d\n", i); + return PTR_ERR(base); + } + info->ddr_reg[i] = base; + } + + /* resource i for pll register base */ + base = devm_platform_ioremap_resource(pdev, i); + if (IS_ERR(base)) { + dev_err(&pdev->dev, "couldn't ioremap for pll reg\n"); + return PTR_ERR(base); + } + info->pll_reg = base; + + ret = platform_get_irq(pdev, 0); + if (ret < 0) { + dev_err(&pdev->dev, "couldn't get irq\n"); + return ret; + } + info->irq_num = ret; + + ret = devm_request_irq(&pdev->dev, info->irq_num, dmc_irq_handler, + IRQF_NOBALANCING, dev_name(&pdev->dev), + (void *)info); + if (ret < 0) + dev_err(&pdev->dev, "ddr request irq failed\n"); + + return ret; +} + +int meson_ddr_pmu_create(struct platform_device *pdev, struct ddr_pmu *pmu) +{ + int ret; + char *name; + struct pmu tmp_pmu = { + .module = THIS_MODULE, + .capabilities = PERF_PMU_CAP_NO_EXCLUDE, + .task_ctx_nr = perf_invalid_context, + .attr_groups = attr_groups, + .event_init = meson_ddr_perf_event_init, + .add = meson_ddr_perf_event_add, + .del = meson_ddr_perf_event_del, + .start = meson_ddr_perf_event_start, + .stop = meson_ddr_perf_event_stop, + .read = meson_ddr_perf_event_update, + }; + + pmu->pmu = tmp_pmu; + + ret = ddr_pmu_parse_dt(pdev, &pmu->info); + if (ret < 0) + return ret; + + fmt_attr_fill(pmu->info.fmt_attr); + + pmu->cpu = raw_smp_processor_id(); + + name = devm_kasprintf(&pdev->dev, GFP_KERNEL, DDR_PERF_DEV_NAME); + if (!name) { + dev_err(&pdev->dev, "couldn't allocat name\n"); + return -ENOMEM; + } + + ret = cpuhp_setup_state_multi(CPUHP_AP_ONLINE_DYN, name, NULL, + ddr_perf_offline_cpu); + if (ret < 0) { + dev_err(&pdev->dev, "cpuhp_setup_state_multi failed\n"); + return ret; + } + pmu->cpuhp_state = ret; + + /* Register the pmu instance for cpu hotplug */ + ret = cpuhp_state_add_instance_nocalls(pmu->cpuhp_state, &pmu->node); + if (ret) { + dev_err(&pdev->dev, "Error %d registering hotplug\n", ret); + goto cpuhp_instance_err; + } + + fill_event_attr(pmu); + + ret = perf_pmu_register(&pmu->pmu, name, -1); + if (ret) { + dev_err(&pdev->dev, "perf pmu register failed\n"); + goto pmu_register_err; + } + + pmu->name = name; + pmu->dev = &pdev->dev; + pmu->pmu_enabled = false; + + platform_set_drvdata(pdev, pmu); + + return 0; + +pmu_register_err: + cpuhp_state_remove_instance_nocalls(pmu->cpuhp_state, &pmu->node); +cpuhp_instance_err: + cpuhp_remove_state(pmu->cpuhp_state); + + return ret; +} +EXPORT_SYMBOL(meson_ddr_pmu_create); + +int meson_ddr_pmu_remove(struct platform_device *pdev) +{ + struct ddr_pmu *pmu = platform_get_drvdata(pdev); + + perf_pmu_unregister(&pmu->pmu); + cpuhp_state_remove_instance_nocalls(pmu->cpuhp_state, &pmu->node); + cpuhp_remove_state(pmu->cpuhp_state); + + return 0; +} +EXPORT_SYMBOL(meson_ddr_pmu_remove); diff --git a/drivers/perf/amlogic/meson_g12_ddr_pmu.c b/drivers/perf/amlogic/meson_g12_ddr_pmu.c new file mode 100644 index 000000000000..ba1908733bee --- /dev/null +++ b/drivers/perf/amlogic/meson_g12_ddr_pmu.c @@ -0,0 +1,388 @@ +// SPDX-License-Identifier: GPL-2.0 +/* + * Copyright (c) 2022 Amlogic, Inc. All rights reserved. + */ +#include +#include +#include +#include +#include +#include +#include +#include +#include + +#include + +#define PORT_MAJOR 32 +#define DEFAULT_XTAL_FREQ 24000000UL + +#define DMC_QOS_IRQ BIT(30) + +/* DMC bandwidth monitor register address offset */ +#define DMC_MON_G12_CTRL0 (0x20 << 2) +#define DMC_MON_G12_CTRL1 (0x21 << 2) +#define DMC_MON_G12_CTRL2 (0x22 << 2) +#define DMC_MON_G12_CTRL3 (0x23 << 2) +#define DMC_MON_G12_CTRL4 (0x24 << 2) +#define DMC_MON_G12_CTRL5 (0x25 << 2) +#define DMC_MON_G12_CTRL6 (0x26 << 2) +#define DMC_MON_G12_CTRL7 (0x27 << 2) +#define DMC_MON_G12_CTRL8 (0x28 << 2) + +#define DMC_MON_G12_ALL_REQ_CNT (0x29 << 2) +#define DMC_MON_G12_ALL_GRANT_CNT (0x2a << 2) +#define DMC_MON_G12_ONE_GRANT_CNT (0x2b << 2) +#define DMC_MON_G12_SEC_GRANT_CNT (0x2c << 2) +#define DMC_MON_G12_THD_GRANT_CNT (0x2d << 2) +#define DMC_MON_G12_FOR_GRANT_CNT (0x2e << 2) +#define DMC_MON_G12_TIMER (0x2f << 2) + +/* Each bit represent a axi line */ +PMU_FORMAT_ATTR(event, "config:0-7"); +PMU_FORMAT_ATTR(arm, "config1:0"); +PMU_FORMAT_ATTR(gpu, "config1:1"); +PMU_FORMAT_ATTR(pcie, "config1:2"); +PMU_FORMAT_ATTR(hdcp, "config1:3"); +PMU_FORMAT_ATTR(hevc_front, "config1:4"); +PMU_FORMAT_ATTR(usb3_0, "config1:6"); +PMU_FORMAT_ATTR(device, "config1:7"); +PMU_FORMAT_ATTR(hevc_back, "config1:8"); +PMU_FORMAT_ATTR(h265enc, "config1:9"); +PMU_FORMAT_ATTR(vpu_read1, "config1:16"); +PMU_FORMAT_ATTR(vpu_read2, "config1:17"); +PMU_FORMAT_ATTR(vpu_read3, "config1:18"); +PMU_FORMAT_ATTR(vpu_write1, "config1:19"); +PMU_FORMAT_ATTR(vpu_write2, "config1:20"); +PMU_FORMAT_ATTR(vdec, "config1:21"); +PMU_FORMAT_ATTR(hcodec, "config1:22"); +PMU_FORMAT_ATTR(ge2d, "config1:23"); + +PMU_FORMAT_ATTR(spicc1, "config1:32"); +PMU_FORMAT_ATTR(usb0, "config1:33"); +PMU_FORMAT_ATTR(dma, "config1:34"); +PMU_FORMAT_ATTR(arb0, "config1:35"); +PMU_FORMAT_ATTR(sd_emmc_b, "config1:36"); +PMU_FORMAT_ATTR(usb1, "config1:37"); +PMU_FORMAT_ATTR(audio, "config1:38"); +PMU_FORMAT_ATTR(aififo, "config1:39"); +PMU_FORMAT_ATTR(parser, "config1:41"); +PMU_FORMAT_ATTR(ao_cpu, "config1:42"); +PMU_FORMAT_ATTR(sd_emmc_c, "config1:43"); +PMU_FORMAT_ATTR(spicc2, "config1:44"); +PMU_FORMAT_ATTR(ethernet, "config1:45"); +PMU_FORMAT_ATTR(sana, "config1:46"); + +/* for sm1 and g12b */ +PMU_FORMAT_ATTR(nna, "config1:10"); + +/* for g12b only */ +PMU_FORMAT_ATTR(gdc, "config1:11"); +PMU_FORMAT_ATTR(mipi_isp, "config1:12"); +PMU_FORMAT_ATTR(arm1, "config1:13"); +PMU_FORMAT_ATTR(sd_emmc_a, "config1:40"); + +static struct attribute *g12_pmu_format_attrs[] = { + &format_attr_event.attr, + &format_attr_arm.attr, + &format_attr_gpu.attr, + &format_attr_nna.attr, + &format_attr_gdc.attr, + &format_attr_arm1.attr, + &format_attr_mipi_isp.attr, + &format_attr_sd_emmc_a.attr, + &format_attr_pcie.attr, + &format_attr_hdcp.attr, + &format_attr_hevc_front.attr, + &format_attr_usb3_0.attr, + &format_attr_device.attr, + &format_attr_hevc_back.attr, + &format_attr_h265enc.attr, + &format_attr_vpu_read1.attr, + &format_attr_vpu_read2.attr, + &format_attr_vpu_read3.attr, + &format_attr_vpu_write1.attr, + &format_attr_vpu_write2.attr, + &format_attr_vdec.attr, + &format_attr_hcodec.attr, + &format_attr_ge2d.attr, + &format_attr_spicc1.attr, + &format_attr_usb0.attr, + &format_attr_dma.attr, + &format_attr_arb0.attr, + &format_attr_sd_emmc_b.attr, + &format_attr_usb1.attr, + &format_attr_audio.attr, + &format_attr_aififo.attr, + &format_attr_parser.attr, + &format_attr_ao_cpu.attr, + &format_attr_sd_emmc_c.attr, + &format_attr_spicc2.attr, + &format_attr_ethernet.attr, + &format_attr_sana.attr, + NULL, +}; + +/* calculate ddr clock */ +static unsigned long dmc_g12_get_freq_quick(struct dmc_hw_info *info) +{ + unsigned int val; + unsigned int n, m, od1; + unsigned int od_div = 0xfff; + unsigned long freq = 0; + + val = readl(info->pll_reg); + val = val & 0xfffff; + switch ((val >> 16) & 7) { + case 0: + od_div = 2; + break; + + case 1: + od_div = 3; + break; + + case 2: + od_div = 4; + break; + + case 3: + od_div = 6; + break; + + case 4: + od_div = 8; + break; + + default: + break; + } + + m = val & 0x1ff; + n = ((val >> 10) & 0x1f); + od1 = (((val >> 19) & 0x1)) == 1 ? 2 : 1; + freq = DEFAULT_XTAL_FREQ / 1000; /* avoid overflow */ + if (n) + freq = ((((freq * m) / n) >> od1) / od_div) * 1000; + + return freq; +} + +#ifdef DEBUG +static void g12_dump_reg(struct dmc_hw_info *db) +{ + int s = 0, i; + unsigned int r; + + for (i = 0; i < 9; i++) { + r = readl(db->ddr_reg[0] + (DMC_MON_G12_CTRL0 + (i << 2))); + pr_notice("DMC_MON_CTRL%d: %08x\n", i, r); + } + r = readl(db->ddr_reg[0] + DMC_MON_G12_ALL_REQ_CNT); + pr_notice("DMC_MON_ALL_REQ_CNT: %08x\n", r); + r = readl(db->ddr_reg[0] + DMC_MON_G12_ALL_GRANT_CNT); + pr_notice("DMC_MON_ALL_GRANT_CNT:%08x\n", r); + r = readl(db->ddr_reg[0] + DMC_MON_G12_ONE_GRANT_CNT); + pr_notice("DMC_MON_ONE_GRANT_CNT:%08x\n", r); + r = readl(db->ddr_reg[0] + DMC_MON_G12_SEC_GRANT_CNT); + pr_notice("DMC_MON_SEC_GRANT_CNT:%08x\n", r); + r = readl(db->ddr_reg[0] + DMC_MON_G12_THD_GRANT_CNT); + pr_notice("DMC_MON_THD_GRANT_CNT:%08x\n", r); + r = readl(db->ddr_reg[0] + DMC_MON_G12_FOR_GRANT_CNT); + pr_notice("DMC_MON_FOR_GRANT_CNT:%08x\n", r); + r = readl(db->ddr_reg[0] + DMC_MON_G12_TIMER); + pr_notice("DMC_MON_TIMER: %08x\n", r); +} +#endif + +static void dmc_g12_counter_enable(struct dmc_hw_info *info) +{ + unsigned int val; + unsigned long clock_count = dmc_g12_get_freq_quick(info) / 10; /* 100ms */ + + writel(clock_count, info->ddr_reg[0] + DMC_MON_G12_TIMER); + + val = readl(info->ddr_reg[0] + DMC_MON_G12_CTRL0); + + /* enable all channel */ + val = BIT(31) | /* enable bit */ + BIT(20) | /* use timer */ + 0x0f; /* 4 channels */ + + writel(val, info->ddr_reg[0] + DMC_MON_G12_CTRL0); + +#ifdef DEBUG + g12_dump_reg(info); +#endif +} + +static void dmc_g12_config_fiter(struct dmc_hw_info *info, + int port, int channel) +{ + u32 val; + u32 rp[MAX_CHANNEL_NUM] = {DMC_MON_G12_CTRL1, DMC_MON_G12_CTRL3, + DMC_MON_G12_CTRL5, DMC_MON_G12_CTRL7}; + u32 rs[MAX_CHANNEL_NUM] = {DMC_MON_G12_CTRL2, DMC_MON_G12_CTRL4, + DMC_MON_G12_CTRL6, DMC_MON_G12_CTRL8}; + int subport = -1; + + /* clear all port mask */ + if (port < 0) { + writel(0, info->ddr_reg[0] + rp[channel]); + writel(0, info->ddr_reg[0] + rs[channel]); + return; + } + + if (port >= PORT_MAJOR) + subport = port - PORT_MAJOR; + + if (subport < 0) { + val = readl(info->ddr_reg[0] + rp[channel]); + val |= (1 << port); + writel(val, info->ddr_reg[0] + rp[channel]); + val = 0xffff; + writel(val, info->ddr_reg[0] + rs[channel]); + } else { + val = BIT(23); /* select device */ + writel(val, info->ddr_reg[0] + rp[channel]); + val = readl(info->ddr_reg[0] + rs[channel]); + val |= (1 << subport); + writel(val, info->ddr_reg[0] + rs[channel]); + } +} + +static void dmc_g12_config_axi_id(struct dmc_hw_info *info, int axi_id, int channel) +{ + if (channel > info->chann_nr) + return; + + dmc_g12_config_fiter(info, axi_id, channel); +} + +static void dmc_g12_counter_disable(struct dmc_hw_info *info) +{ + int i; + + /* clear timer */ + writel(0, info->ddr_reg[0] + DMC_MON_G12_CTRL0); + writel(0, info->ddr_reg[0] + DMC_MON_G12_TIMER); + + writel(0, info->ddr_reg[0] + DMC_MON_G12_ALL_REQ_CNT); + writel(0, info->ddr_reg[0] + DMC_MON_G12_ALL_GRANT_CNT); + writel(0, info->ddr_reg[0] + DMC_MON_G12_ONE_GRANT_CNT); + writel(0, info->ddr_reg[0] + DMC_MON_G12_SEC_GRANT_CNT); + writel(0, info->ddr_reg[0] + DMC_MON_G12_THD_GRANT_CNT); + writel(0, info->ddr_reg[0] + DMC_MON_G12_FOR_GRANT_CNT); + + /* clear port channel mapping */ + for (i = 0; i < info->chann_nr; i++) + dmc_g12_config_fiter(info, -1, i); +} + +static void dmc_g12_get_counters(struct dmc_hw_info *info, + struct dmc_counter *counter) +{ + int i; + unsigned int reg; + + counter->all_cnt = readl(info->ddr_reg[0] + DMC_MON_G12_ALL_GRANT_CNT); + counter->all_req = readl(info->ddr_reg[0] + DMC_MON_G12_ALL_REQ_CNT); + + for (i = 0; i < info->chann_nr; i++) { + reg = DMC_MON_G12_ONE_GRANT_CNT + (i << 2); + counter->channel_cnt[i] = readl(info->ddr_reg[0] + reg); + } +} + +static int dmc_g12_irq_handler(struct dmc_hw_info *info, + struct dmc_counter *counter) +{ + unsigned int val; + int ret = -EINVAL; + + val = readl(info->ddr_reg[0] + DMC_MON_G12_CTRL0); + if (val & DMC_QOS_IRQ) { + dmc_g12_get_counters(info, counter); + /* clear irq flags */ + writel(val, info->ddr_reg[0] + DMC_MON_G12_CTRL0); + ret = 0; + } + return ret; +} + +static struct dmc_pmu_hw_ops g12_ops = { + .enable = dmc_g12_counter_enable, + .disable = dmc_g12_counter_disable, + .irq_handler = dmc_g12_irq_handler, + .get_counters = dmc_g12_get_counters, + .config_axi_id = dmc_g12_config_axi_id, +}; + +static int __init g12_ddr_pmu_probe(struct platform_device *pdev) +{ + struct ddr_pmu *pmu; + + if (of_device_is_compatible(pdev->dev.of_node, + "amlogic,g12a-ddr-pmu")) { + format_attr_nna.attr.mode = 0; + format_attr_gdc.attr.mode = 0; + format_attr_arm1.attr.mode = 0; + format_attr_mipi_isp.attr.mode = 0; + } else if (of_device_is_compatible(pdev->dev.of_node, + "amlogic,sm1-ddr-pmu")) { + format_attr_gdc.attr.mode = 0; + format_attr_arm1.attr.mode = 0; + format_attr_mipi_isp.attr.mode = 0; + } + + pmu = devm_kzalloc(&pdev->dev, sizeof(struct ddr_pmu), GFP_KERNEL); + if (!pmu) + return -ENOMEM; + + /* + * G12 series Soc have single dmc controller and + * 4x ddr bandwidth monitor channels + */ + pmu->info.dmc_nr = 1; + pmu->info.chann_nr = 4; + pmu->info.ops = &g12_ops; + pmu->info.fmt_attr = g12_pmu_format_attrs; + + return meson_ddr_pmu_create(pdev, pmu); +} + +static int __exit g12_ddr_pmu_remove(struct platform_device *pdev) +{ + meson_ddr_pmu_remove(pdev); + + return 0; +} + +static const struct of_device_id meson_ddr_pmu_dt_match[] = { + { + .compatible = "amlogic,g12-ddr-pmu", + }, + { + .compatible = "amlogic,g12a-ddr-pmu", + }, + { + .compatible = "amlogic,g12b-ddr-pmu", + }, + { + .compatible = "amlogic,sm1-ddr-pmu", + }, + {} +}; + +static struct platform_driver g12_ddr_pmu_driver = { + .driver = { + .name = "amlogic,ddr-pmu", + .of_match_table = meson_ddr_pmu_dt_match, + }, + .remove = __exit_p(g12_ddr_pmu_remove), +}; + +module_platform_driver_probe(g12_ddr_pmu_driver, g12_ddr_pmu_probe); +MODULE_AUTHOR("Jiucheng Xu"); +MODULE_LICENSE("GPL"); +MODULE_DESCRIPTION("Amlogic G12 series SoC DDR PMU"); diff --git a/include/soc/amlogic/meson_ddr_pmu.h b/include/soc/amlogic/meson_ddr_pmu.h new file mode 100644 index 000000000000..882efe3c2f58 --- /dev/null +++ b/include/soc/amlogic/meson_ddr_pmu.h @@ -0,0 +1,76 @@ +/* SPDX-License-Identifier: GPL-2.0 */ +/* + * Copyright (c) 2022 Amlogic, Inc. All rights reserved. + */ + +#ifndef __MESON_DDR_PMU_H__ +#define __MESON_DDR_PMU_H__ + +#define MAX_CHANNEL_NUM 8 + +enum { + ALL_CHAN_COUNTER_ID, + CHAN1_COUNTER_ID, + CHAN2_COUNTER_ID, + CHAN3_COUNTER_ID, + CHAN4_COUNTER_ID, + CHAN5_COUNTER_ID, + CHAN6_COUNTER_ID, + CHAN7_COUNTER_ID, + CHAN8_COUNTER_ID, + COUNTER_MAX_ID, +}; + +struct dmc_hw_info; + +struct dmc_counter { + u64 all_cnt; /* The count of all requests come in/out ddr controller */ + union { + u64 all_req; + struct { + u64 all_idle_cnt; + u64 all_16bit_cnt; + }; + }; + u64 channel_cnt[MAX_CHANNEL_NUM]; /* To save a DMC bandwidth-monitor channel counter */ +}; + +struct dmc_pmu_hw_ops { + void (*enable)(struct dmc_hw_info *info); + void (*disable)(struct dmc_hw_info *info); + /* Bind an axi line to a bandwidth-monitor channel */ + void (*config_axi_id)(struct dmc_hw_info *info, int axi_id, int chann); + int (*irq_handler)(struct dmc_hw_info *info, + struct dmc_counter *counter); + void (*get_counters)(struct dmc_hw_info *info, + struct dmc_counter *counter); +}; + +struct dmc_hw_info { + struct dmc_pmu_hw_ops *ops; + void __iomem *ddr_reg[4]; + unsigned long timer_value; /* Timer value in TIMER register */ + void __iomem *pll_reg; + int irq_num; /* irq vector number */ + int dmc_nr; /* The number of dmc controller */ + int chann_nr; /* The number of dmc bandwidth monitor channels */ + int id; /* The number of supported channels */ + struct attribute **fmt_attr; +}; + +struct ddr_pmu { + struct pmu pmu; + struct dmc_hw_info info; + struct dmc_counter counters; /* save counters from hw */ + bool pmu_enabled; + struct device *dev; + char *name; + struct hlist_node node; + enum cpuhp_state cpuhp_state; + int cpu; /* for cpu hotplug */ +}; + +int meson_ddr_pmu_create(struct platform_device *pdev, struct ddr_pmu *pmu); +int meson_ddr_pmu_remove(struct platform_device *pdev); + +#endif /* __MESON_DDR_PMU_H__ */ From patchwork Fri Aug 5 07:14:24 2022 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: Jiucheng Xu X-Patchwork-Id: 595638 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 00FBEC19F2D for ; Fri, 5 Aug 2022 07:14:49 +0000 (UTC) Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S240230AbiHEHOr (ORCPT ); Fri, 5 Aug 2022 03:14:47 -0400 Received: from lindbergh.monkeyblade.net ([23.128.96.19]:50046 "EHLO lindbergh.monkeyblade.net" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S240195AbiHEHOq (ORCPT ); Fri, 5 Aug 2022 03:14:46 -0400 Received: from mail-sh.amlogic.com (mail-sh.amlogic.com [58.32.228.43]) by lindbergh.monkeyblade.net (Postfix) with ESMTPS id 53203D40; Fri, 5 Aug 2022 00:14:45 -0700 (PDT) Received: from droid01-xa.amlogic.com (10.88.11.200) by mail-sh.amlogic.com (10.18.11.5) with Microsoft SMTP Server id 15.1.2507.9; Fri, 5 Aug 2022 15:14:37 +0800 From: Jiucheng Xu To: , , , CC: Rob Herring , Krzysztof Kozlowski , Will Deacon , Mark Rutland , Neil Armstrong , Kevin Hilman , Jerome Brunet , Martin Blumenstingl , Chris Healy , Jiucheng Xu Subject: [PATCH v4 2/4] docs/perf: Add documentation for the Amlogic G12 DDR PMU Date: Fri, 5 Aug 2022 15:14:24 +0800 Message-ID: <20220805071426.2598818-2-jiucheng.xu@amlogic.com> X-Mailer: git-send-email 2.25.1 In-Reply-To: <20220805071426.2598818-1-jiucheng.xu@amlogic.com> References: <20220805071426.2598818-1-jiucheng.xu@amlogic.com> MIME-Version: 1.0 X-Originating-IP: [10.88.11.200] Precedence: bulk List-ID: X-Mailing-List: devicetree@vger.kernel.org Add a user guide to show how to use DDR PMU to monitor DDR bandwidth on Amlogic G12 SoC Signed-off-by: Jiucheng Xu --- Changes v3 -> v4: - No change Changes v2 -> v3: - Rename doc name from aml-ddr-pmu.rst to meson-ddr-pmu.rst Changes v1 -> v2: - Nothing was changed --- .../admin-guide/perf/meson-ddr-pmu.rst | 68 +++++++++++++++++++ MAINTAINERS | 1 + 2 files changed, 69 insertions(+) create mode 100644 Documentation/admin-guide/perf/meson-ddr-pmu.rst diff --git a/Documentation/admin-guide/perf/meson-ddr-pmu.rst b/Documentation/admin-guide/perf/meson-ddr-pmu.rst new file mode 100644 index 000000000000..f8601f3b103e --- /dev/null +++ b/Documentation/admin-guide/perf/meson-ddr-pmu.rst @@ -0,0 +1,68 @@ +.. SPDX-License-Identifier: GPL-2.0 +=========================================================== +Amlogic SoC DDR Bandwidth Performance Monitoring Unit (PMU) +=========================================================== + +There is a bandwidth monitor inside the DRAM contorller. The monitor include +4 channels which can count the read/write request of accessing DRAM individually. +It can be helpful to show if the performance bottleneck is on DDR bandwidth. + +Currently, this driver supports the following 5 Perf events: + +aml_ddr_bw/total_rw_bytes/ +aml_ddr_bw/chan_1_rw_bytes/ +aml_ddr_bw/chan_2_rw_bytes/ +aml_ddr_bw/chan_3_rw_bytes/ +aml_ddr_bw/chan_4_rw_bytes/ + +aml_ddr_bw/chan_{1,2,3,4}_rw_bytes/ events are the channel related events. +Each channel support using keywords as filter, which can let the channel +to monitor the individual IP module in SoC. + +The following keywords are the filter: + +arm - DDR access request from CPU +vpu_read1 - DDR access request from OSD + VPP read +gpu - DDR access request from 3D GPU +pcie - DDR access request from PCIe controller +hdcp - DDR access request from HDCP controller +hevc_front - DDR access request from HEVC codec front end +usb3_0 - DDR access request from USB3.0 controller +hevc_back - DDR access request from HEVC codec back end +h265enc - DDR access request from HEVC encoder +vpu_read2 - DDR access request from DI read +vpu_write1 - DDR access request from VDIN write +vpu_write2 - DDR access request from di write +vdec - DDR access request from legacy codec video decoder +hcodec - DDR access request from H264 encoder +ge2d - DDR access request from ge2d +spicc1 - DDR access request from SPI controller 1 +usb0 - DDR access request from USB2.0 controller 0 +dma - DDR access request from system DMA controller 1 +arb0 - DDR access request from arb0 +sd_emmc_b - DDR access request from SD eMMC b controller +usb1 - DDR access request from USB2.0 controller 1 +audio - DDR access request from Audio module +sd_emmc_c - DDR access request from SD eMMC c controller +spicc2 - DDR access request from SPI controller 2 +ethernet - DDR access request from Ethernet controller + +The following command is to show the total DDR bandwidth: + + .. code-block::bash + + perf stat -a -e aml_ddr_bw/total_rw_bytes/ -I 1000 sleep 10 + +This command will print the total DDR bandwidth per second. + +The following commands are to show how to use filter parameters: + + .. code-block::bash + + perf stat -a -e aml_ddr_bw/chan_1_rw_bytes,arm=1/ -I 1000 sleep 10 + perf stat -a -e aml_ddr_bw/chan_2_rw_bytes,gpu=1/ -I 1000 sleep 10 + perf stat -a -e aml_ddr_bw/chan_3_rw_bytes,arm=1,gpu=1/ -I 1000 sleep 10 + +The 1st command show how to use channel 1 to monitor the DDR bandwidth from ARM. +The 2nd command show using channel 2 to get the DDR bandwidth of GPU. +The 3rd command show using channel 3 to monitor the sum of ARM and GPU. diff --git a/MAINTAINERS b/MAINTAINERS index ac8a98dfbacc..8ee68e699e6d 100644 --- a/MAINTAINERS +++ b/MAINTAINERS @@ -1054,6 +1054,7 @@ AMLOGIC DDR PMU DRIVER M: Jiucheng Xu S: Supported W: http://www.amlogic.com +F: Documentation/admin-guide/perf/meson-ddr-pmu.rst F: drivers/perf/amlogic/ F: include/soc/amlogic/ From patchwork Fri Aug 5 07:14:25 2022 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: Jiucheng Xu X-Patchwork-Id: 595892 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 E66CBC00140 for ; Fri, 5 Aug 2022 07:14:57 +0000 (UTC) Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S240287AbiHEHOz (ORCPT ); Fri, 5 Aug 2022 03:14:55 -0400 Received: from lindbergh.monkeyblade.net ([23.128.96.19]:50114 "EHLO lindbergh.monkeyblade.net" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S240243AbiHEHOu (ORCPT ); Fri, 5 Aug 2022 03:14:50 -0400 Received: from mail-sh.amlogic.com (mail-sh.amlogic.com [58.32.228.43]) by lindbergh.monkeyblade.net (Postfix) with ESMTPS id E69FB336; Fri, 5 Aug 2022 00:14:46 -0700 (PDT) Received: from droid01-xa.amlogic.com (10.88.11.200) by mail-sh.amlogic.com (10.18.11.5) with Microsoft SMTP Server id 15.1.2507.9; Fri, 5 Aug 2022 15:14:38 +0800 From: Jiucheng Xu To: , , , CC: Rob Herring , Krzysztof Kozlowski , Will Deacon , Mark Rutland , Neil Armstrong , Kevin Hilman , Jerome Brunet , Martin Blumenstingl , Chris Healy , Jiucheng Xu Subject: [PATCH v4 3/4] arm64: dts: meson: Add DDR PMU node Date: Fri, 5 Aug 2022 15:14:25 +0800 Message-ID: <20220805071426.2598818-3-jiucheng.xu@amlogic.com> X-Mailer: git-send-email 2.25.1 In-Reply-To: <20220805071426.2598818-1-jiucheng.xu@amlogic.com> References: <20220805071426.2598818-1-jiucheng.xu@amlogic.com> MIME-Version: 1.0 X-Originating-IP: [10.88.11.200] Precedence: bulk List-ID: X-Mailing-List: devicetree@vger.kernel.org Add DDR PMU device node for G12 series SoC Signed-off-by: Jiucheng Xu --- Changes v3 -> v4: - No change Changes v2 -> v3: - No change Changes v1 -> v2: - Remove model, dmc_nr, chann_nr properties - Add g12a-ddr-pmu, g12b-ddr-pmu, sm1-ddr-pmu compatibles as identifier --- arch/arm64/boot/dts/amlogic/meson-g12-common.dtsi | 7 +++++++ arch/arm64/boot/dts/amlogic/meson-g12a.dtsi | 4 ++++ arch/arm64/boot/dts/amlogic/meson-g12b.dtsi | 4 ++++ arch/arm64/boot/dts/amlogic/meson-sm1.dtsi | 4 ++++ 4 files changed, 19 insertions(+) diff --git a/arch/arm64/boot/dts/amlogic/meson-g12-common.dtsi b/arch/arm64/boot/dts/amlogic/meson-g12-common.dtsi index 45947c1031c4..7e556fe575be 100644 --- a/arch/arm64/boot/dts/amlogic/meson-g12-common.dtsi +++ b/arch/arm64/boot/dts/amlogic/meson-g12-common.dtsi @@ -2146,6 +2146,13 @@ hdmi_tx_out: endpoint { }; }; + ddr_pmu: ddr_pmu@ff638000 { + compatible = "amlogic,g12-ddr-pmu"; + reg = <0x0 0xff638000 0x0 0x100 + 0x0 0xff638c00 0x0 0x100>; + interrupts = ; + }; + gic: interrupt-controller@ffc01000 { compatible = "arm,gic-400"; reg = <0x0 0xffc01000 0 0x1000>, diff --git a/arch/arm64/boot/dts/amlogic/meson-g12a.dtsi b/arch/arm64/boot/dts/amlogic/meson-g12a.dtsi index fb0ab27d1f64..4a32e081e70e 100644 --- a/arch/arm64/boot/dts/amlogic/meson-g12a.dtsi +++ b/arch/arm64/boot/dts/amlogic/meson-g12a.dtsi @@ -133,3 +133,7 @@ map1 { }; }; }; + +&ddr_pmu { + compatible = "amlogic,g12a-ddr-pmu"; +}; diff --git a/arch/arm64/boot/dts/amlogic/meson-g12b.dtsi b/arch/arm64/boot/dts/amlogic/meson-g12b.dtsi index ee8fcae9f9f0..d91eca5a9afc 100644 --- a/arch/arm64/boot/dts/amlogic/meson-g12b.dtsi +++ b/arch/arm64/boot/dts/amlogic/meson-g12b.dtsi @@ -139,3 +139,7 @@ map1 { &mali { dma-coherent; }; + +&ddr_pmu { + compatible = "amlogic,g12b-ddr-pmu"; +}; diff --git a/arch/arm64/boot/dts/amlogic/meson-sm1.dtsi b/arch/arm64/boot/dts/amlogic/meson-sm1.dtsi index 80737731af3f..7d62c661fde5 100644 --- a/arch/arm64/boot/dts/amlogic/meson-sm1.dtsi +++ b/arch/arm64/boot/dts/amlogic/meson-sm1.dtsi @@ -543,3 +543,7 @@ &vpu { &usb { power-domains = <&pwrc PWRC_SM1_USB_ID>; }; + +&ddr_pmu { + compatible = "amlogic,sm1-ddr-pmu"; +}; From patchwork Fri Aug 5 07:14:26 2022 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: Jiucheng Xu X-Patchwork-Id: 595637 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 7418AC00140 for ; Fri, 5 Aug 2022 07:15:02 +0000 (UTC) Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S240125AbiHEHPA (ORCPT ); Fri, 5 Aug 2022 03:15:00 -0400 Received: from lindbergh.monkeyblade.net ([23.128.96.19]:50264 "EHLO lindbergh.monkeyblade.net" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S240283AbiHEHOz (ORCPT ); Fri, 5 Aug 2022 03:14:55 -0400 Received: from mail-sh.amlogic.com (mail-sh.amlogic.com [58.32.228.43]) by lindbergh.monkeyblade.net (Postfix) with ESMTPS id B11451A383; Fri, 5 Aug 2022 00:14:50 -0700 (PDT) Received: from droid01-xa.amlogic.com (10.88.11.200) by mail-sh.amlogic.com (10.18.11.5) with Microsoft SMTP Server id 15.1.2507.9; Fri, 5 Aug 2022 15:14:38 +0800 From: Jiucheng Xu To: , , , CC: Rob Herring , Krzysztof Kozlowski , Will Deacon , Mark Rutland , Neil Armstrong , Kevin Hilman , Jerome Brunet , Martin Blumenstingl , Chris Healy , Jiucheng Xu Subject: [PATCH v4 4/4] dt-binding: perf: Add Amlogic DDR PMU Date: Fri, 5 Aug 2022 15:14:26 +0800 Message-ID: <20220805071426.2598818-4-jiucheng.xu@amlogic.com> X-Mailer: git-send-email 2.25.1 In-Reply-To: <20220805071426.2598818-1-jiucheng.xu@amlogic.com> References: <20220805071426.2598818-1-jiucheng.xu@amlogic.com> MIME-Version: 1.0 X-Originating-IP: [10.88.11.200] Precedence: bulk List-ID: X-Mailing-List: devicetree@vger.kernel.org Add binding documentation for the Amlogic G12 series DDR performance monitor unit. Signed-off-by: Jiucheng Xu --- Changes v3 -> v4: - Fix "$id: relative path/filename doesn't match actual path or filename" warning Changes v2 -> v3: - Remove oneOf - Add descriptions - Fix compiling warning Changes v1 -> v2: - Rename file, from aml_ddr_pmu.yaml to amlogic,g12_ddr_pmu.yaml - Delete "model", "dmc_nr", "chann_nr" new properties - Fix compiling error --- .../bindings/perf/amlogic,g12-ddr-pmu.yaml | 52 +++++++++++++++++++ MAINTAINERS | 1 + 2 files changed, 53 insertions(+) create mode 100644 Documentation/devicetree/bindings/perf/amlogic,g12-ddr-pmu.yaml diff --git a/Documentation/devicetree/bindings/perf/amlogic,g12-ddr-pmu.yaml b/Documentation/devicetree/bindings/perf/amlogic,g12-ddr-pmu.yaml new file mode 100644 index 000000000000..4d018574ffd1 --- /dev/null +++ b/Documentation/devicetree/bindings/perf/amlogic,g12-ddr-pmu.yaml @@ -0,0 +1,52 @@ +# SPDX-License-Identifier: (GPL-2.0-only OR BSD-2-Clause) +%YAML 1.2 +--- +$id: http://devicetree.org/schemas/perf/amlogic,g12-ddr-pmu.yaml# +$schema: http://devicetree.org/meta-schemas/core.yaml# + +title: Amlogic G12 DDR performance monitor + +maintainers: + - Jiucheng Xu + +description: | + Amlogic G12 series SoC integrate DDR bandwidth monitor. + A timer is inside and can generate interrupt when timeout. + The bandwidth is counted in the timer ISR. Different platform + has different subset of event format attribute. + +properties: + compatible: + items: + - enum: + - amlogic,g12b-ddr-pmu + - amlogic,g12a-ddr-pmu + - amlogic,sm1-ddr-pmu + + reg: + items: + - description: Physical address of DMC bandwidth register + and size of the configuration address space. + - description: Physical address of DMC PLL register and + size of the configuration address space. + + interrupts: + items: + - description: The IRQ of the inside timer timeout. + +required: + - compatible + - reg + - interrupts + +additionalProperties: false + +examples: + - | + #include + ddr_pmu@ff638000 { + compatible = "amlogic,g12a-ddr-pmu"; + reg = <0xff638000 0x100 + 0xff638c00 0x100>; + interrupts = ; + }; diff --git a/MAINTAINERS b/MAINTAINERS index 8ee68e699e6d..67c2c9e8c4ea 100644 --- a/MAINTAINERS +++ b/MAINTAINERS @@ -1055,6 +1055,7 @@ M: Jiucheng Xu S: Supported W: http://www.amlogic.com F: Documentation/admin-guide/perf/meson-ddr-pmu.rst +F: Documentation/devicetree/bindings/perf/amlogic,g12-ddr-pmu.yaml F: drivers/perf/amlogic/ F: include/soc/amlogic/