* [PATCH v4 1/2] drivers/perf: hisi: Add support for HiSilicon NoC PMU
2025-07-17 7:41 [PATCH v4 0/2] drivers/perf: hisi: Add support for HiSilicon MN PMU driver Yicong Yang
@ 2025-07-17 7:41 ` Yicong Yang
2025-07-17 10:37 ` Jonathan Cameron
2025-07-17 7:41 ` [PATCH v4 2/2] drivers/perf: hisi: Add support for HiSilicon MN PMU driver Yicong Yang
1 sibling, 1 reply; 5+ messages in thread
From: Yicong Yang @ 2025-07-17 7:41 UTC (permalink / raw)
To: will, mark.rutland, linux-arm-kernel, linux-kernel
Cc: hejunhao3, jonathan.cameron, prime.zeng, linuxarm, yangyicong,
wangyushan12
From: Yicong Yang <yangyicong@hisilicon.com>
Adds the support for HiSilicon NoC (Network on Chip) PMU which
will be used to monitor the events on the system bus. The PMU
device will be named after the SCL ID (either Super CPU cluster
or Super IO cluster) and the index ID, just similar to other
HiSilicon Uncore PMUs. Below PMU formats are provided besides
the event:
- ch: the transaction channel (data, request, response, etc) which
can be used to filter the counting.
- tt_en: tracetag filtering enable. Just as other HiSilicon Uncore
PMUs the NoC PMU supports only counting the transactions with
tracetag.
The NoC PMU doesn't have an interrupt to indicate the overflow.
However we have a 64 bit counter which is large enough and it's
nearly impossible to overflow.
Signed-off-by: Yicong Yang <yangyicong@hisilicon.com>
---
Documentation/admin-guide/perf/hisi-pmu.rst | 11 +
drivers/perf/hisilicon/Makefile | 3 +-
drivers/perf/hisilicon/hisi_uncore_noc_pmu.c | 443 +++++++++++++++++++
3 files changed, 456 insertions(+), 1 deletion(-)
create mode 100644 drivers/perf/hisilicon/hisi_uncore_noc_pmu.c
diff --git a/Documentation/admin-guide/perf/hisi-pmu.rst b/Documentation/admin-guide/perf/hisi-pmu.rst
index 48992a0b8e94..6f0ea4f641cc 100644
--- a/Documentation/admin-guide/perf/hisi-pmu.rst
+++ b/Documentation/admin-guide/perf/hisi-pmu.rst
@@ -112,6 +112,17 @@ uring channel. It is 2 bits. Some important codes are as follows:
- 2'b00: default value, count the events which sent to the both uring and
uring_ext channel;
+6. ch: NoC PMU supports filtering the event counts of certain transaction
+channel with this option. The current supported channels are as follows:
+
+- 3'b010: Request channel
+- 3'b100: Snoop channel
+- 3'b110: Response channel
+- 3'b111: Data channel
+
+7. tt_en: NoC PMU supports counting only transactions that have tracetag set
+if this option is set. See the 2nd list for more information about tracetag.
+
Users could configure IDs to count data come from specific CCL/ICL, by setting
srcid_cmd & srcid_msk, and data desitined for specific CCL/ICL by setting
tgtid_cmd & tgtid_msk. A set bit in srcid_msk/tgtid_msk means the PMU will not
diff --git a/drivers/perf/hisilicon/Makefile b/drivers/perf/hisilicon/Makefile
index 48dcc8381ea7..dcec8f39719d 100644
--- a/drivers/perf/hisilicon/Makefile
+++ b/drivers/perf/hisilicon/Makefile
@@ -1,7 +1,8 @@
# SPDX-License-Identifier: GPL-2.0-only
obj-$(CONFIG_HISI_PMU) += hisi_uncore_pmu.o hisi_uncore_l3c_pmu.o \
hisi_uncore_hha_pmu.o hisi_uncore_ddrc_pmu.o hisi_uncore_sllc_pmu.o \
- hisi_uncore_pa_pmu.o hisi_uncore_cpa_pmu.o hisi_uncore_uc_pmu.o
+ hisi_uncore_pa_pmu.o hisi_uncore_cpa_pmu.o hisi_uncore_uc_pmu.o \
+ hisi_uncore_noc_pmu.o
obj-$(CONFIG_HISI_PCIE_PMU) += hisi_pcie_pmu.o
obj-$(CONFIG_HNS3_PMU) += hns3_pmu.o
diff --git a/drivers/perf/hisilicon/hisi_uncore_noc_pmu.c b/drivers/perf/hisilicon/hisi_uncore_noc_pmu.c
new file mode 100644
index 000000000000..2e3eb7881caa
--- /dev/null
+++ b/drivers/perf/hisilicon/hisi_uncore_noc_pmu.c
@@ -0,0 +1,443 @@
+// SPDX-License-Identifier: GPL-2.0
+/*
+ * Driver for HiSilicon Uncore NoC (Network on Chip) PMU device
+ *
+ * Copyright (c) 2025 HiSilicon Technologies Co., Ltd.
+ * Author: Yicong Yang <yangyicong@hisilicon.com>
+ */
+#include <linux/bitops.h>
+#include <linux/cpuhotplug.h>
+#include <linux/device.h>
+#include <linux/io.h>
+#include <linux/mod_devicetable.h>
+#include <linux/module.h>
+#include <linux/platform_device.h>
+#include <linux/property.h>
+#include <linux/sysfs.h>
+
+#include "hisi_uncore_pmu.h"
+
+#define NOC_PMU_VERSION 0x1e00
+#define NOC_PMU_GLOBAL_CTRL 0x1e04
+#define NOC_PMU_GLOBAL_CTRL_PMU_EN BIT(0)
+#define NOC_PMU_GLOBAL_CTRL_TT_EN BIT(1)
+#define NOC_PMU_CNT_INFO 0x1e08
+#define NOC_PMU_CNT_INFO_OVERFLOW(n) BIT(n)
+#define NOC_PMU_EVENT_CTRL0 0x1e20
+#define NOC_PMU_EVENT_CTRL_TYPE GENMASK(4, 0)
+/*
+ * Note channel of 0x0 will reset the counter value, so don't do it before
+ * we read out the counter.
+ */
+#define NOC_PMU_EVENT_CTRL_CHANNEL GENMASK(10, 8)
+#define NOC_PMU_EVENT_CTRL_EN BIT(11)
+#define NOC_PMU_EVENT_COUNTER0 0x1e80
+
+#define NOC_PMU_NR_COUNTERS 4
+#define NOC_PMU_CH_DEFAULT 0x7
+
+#define NOC_PMU_EVENT_CTRLn(ctrl0, n) ((ctrl0) + 4 * (n))
+#define NOC_PMU_EVENT_CNTRn(cntr0, n) ((cntr0) + 8 * (n))
+
+HISI_PMU_EVENT_ATTR_EXTRACTOR(ch, config1, 2, 0);
+HISI_PMU_EVENT_ATTR_EXTRACTOR(tt_en, config1, 3, 3);
+
+/* Dynamic CPU hotplug state used by this PMU driver */
+static enum cpuhp_state hisi_noc_pmu_cpuhp_state;
+
+struct hisi_noc_pmu_regs {
+ u32 version;
+ u32 pmu_ctrl;
+ u32 event_ctrl0;
+ u32 event_cntr0;
+ u32 ovflow_status;
+};
+
+/*
+ * Tracetag filtering is not per event and all the events should keep
+ * the consistence. Return true if the new comer doesn't match the
+ * tracetag filtering configuration of the current scheduled events.
+ */
+static bool hisi_noc_pmu_check_global_filter(struct perf_event *curr,
+ struct perf_event *new)
+{
+ return hisi_get_tt_en(curr) == hisi_get_tt_en(new);
+}
+
+static void hisi_noc_pmu_write_evtype(struct hisi_pmu *noc_pmu, int idx, u32 type)
+{
+ struct hisi_noc_pmu_regs *reg_info = noc_pmu->dev_info->private;
+ u32 reg;
+
+ reg = readl(noc_pmu->base + NOC_PMU_EVENT_CTRLn(reg_info->event_ctrl0, idx));
+ reg &= ~NOC_PMU_EVENT_CTRL_TYPE;
+ reg |= FIELD_PREP(NOC_PMU_EVENT_CTRL_TYPE, type);
+ writel(reg, noc_pmu->base + NOC_PMU_EVENT_CTRLn(reg_info->event_ctrl0, idx));
+}
+
+static int hisi_noc_pmu_get_event_idx(struct perf_event *event)
+{
+ struct hisi_pmu *noc_pmu = to_hisi_pmu(event->pmu);
+ struct hisi_pmu_hwevents *pmu_events = &noc_pmu->pmu_events;
+ int cur_idx;
+
+ cur_idx = find_first_bit(pmu_events->used_mask, noc_pmu->num_counters);
+ if (cur_idx != noc_pmu->num_counters &&
+ !hisi_noc_pmu_check_global_filter(pmu_events->hw_events[cur_idx], event))
+ return -EAGAIN;
+
+ return hisi_uncore_pmu_get_event_idx(event);
+}
+
+static u64 hisi_noc_pmu_read_counter(struct hisi_pmu *noc_pmu,
+ struct hw_perf_event *hwc)
+{
+ struct hisi_noc_pmu_regs *reg_info = noc_pmu->dev_info->private;
+
+ return readq(noc_pmu->base + NOC_PMU_EVENT_CNTRn(reg_info->event_cntr0, hwc->idx));
+}
+
+static void hisi_noc_pmu_write_counter(struct hisi_pmu *noc_pmu,
+ struct hw_perf_event *hwc, u64 val)
+{
+ struct hisi_noc_pmu_regs *reg_info = noc_pmu->dev_info->private;
+
+ writeq(val, noc_pmu->base + NOC_PMU_EVENT_CNTRn(reg_info->event_cntr0, hwc->idx));
+}
+
+static void hisi_noc_pmu_enable_counter(struct hisi_pmu *noc_pmu,
+ struct hw_perf_event *hwc)
+{
+ struct hisi_noc_pmu_regs *reg_info = noc_pmu->dev_info->private;
+ u32 reg;
+
+ reg = readl(noc_pmu->base + NOC_PMU_EVENT_CTRLn(reg_info->event_ctrl0, hwc->idx));
+ reg |= NOC_PMU_EVENT_CTRL_EN;
+ writel(reg, noc_pmu->base + NOC_PMU_EVENT_CTRLn(reg_info->event_ctrl0, hwc->idx));
+}
+
+static void hisi_noc_pmu_disable_counter(struct hisi_pmu *noc_pmu,
+ struct hw_perf_event *hwc)
+{
+ struct hisi_noc_pmu_regs *reg_info = noc_pmu->dev_info->private;
+ u32 reg;
+
+ reg = readl(noc_pmu->base + NOC_PMU_EVENT_CTRLn(reg_info->event_ctrl0, hwc->idx));
+ reg &= ~NOC_PMU_EVENT_CTRL_EN;
+ writel(reg, noc_pmu->base + NOC_PMU_EVENT_CTRLn(reg_info->event_ctrl0, hwc->idx));
+}
+
+static void hisi_noc_pmu_enable_counter_int(struct hisi_pmu *noc_pmu,
+ struct hw_perf_event *hwc)
+{
+ /* We don't support interrupt, so a stub here. */
+}
+
+static void hisi_noc_pmu_disable_counter_int(struct hisi_pmu *noc_pmu,
+ struct hw_perf_event *hwc)
+{
+}
+
+static void hisi_noc_pmu_start_counters(struct hisi_pmu *noc_pmu)
+{
+ struct hisi_noc_pmu_regs *reg_info = noc_pmu->dev_info->private;
+ u32 reg;
+
+ reg = readl(noc_pmu->base + reg_info->pmu_ctrl);
+ reg |= NOC_PMU_GLOBAL_CTRL_PMU_EN;
+ writel(reg, noc_pmu->base + reg_info->pmu_ctrl);
+}
+
+static void hisi_noc_pmu_stop_counters(struct hisi_pmu *noc_pmu)
+{
+ struct hisi_noc_pmu_regs *reg_info = noc_pmu->dev_info->private;
+ u32 reg;
+
+ reg = readl(noc_pmu->base + reg_info->pmu_ctrl);
+ reg &= ~NOC_PMU_GLOBAL_CTRL_PMU_EN;
+ writel(reg, noc_pmu->base + reg_info->pmu_ctrl);
+}
+
+static u32 hisi_noc_pmu_get_int_status(struct hisi_pmu *noc_pmu)
+{
+ struct hisi_noc_pmu_regs *reg_info = noc_pmu->dev_info->private;
+
+ return readl(noc_pmu->base + reg_info->ovflow_status);
+}
+
+static void hisi_noc_pmu_clear_int_status(struct hisi_pmu *noc_pmu, int idx)
+{
+ struct hisi_noc_pmu_regs *reg_info = noc_pmu->dev_info->private;
+ u32 reg;
+
+ reg = readl(noc_pmu->base + reg_info->ovflow_status);
+ reg &= ~NOC_PMU_CNT_INFO_OVERFLOW(idx);
+ writel(reg, noc_pmu->base + reg_info->ovflow_status);
+}
+
+static void hisi_noc_pmu_enable_filter(struct perf_event *event)
+{
+ struct hisi_pmu *noc_pmu = to_hisi_pmu(event->pmu);
+ struct hisi_noc_pmu_regs *reg_info = noc_pmu->dev_info->private;
+ struct hw_perf_event *hwc = &event->hw;
+ u32 tt_en = hisi_get_tt_en(event);
+ u32 ch = hisi_get_ch(event);
+ u32 reg;
+
+ if (!ch)
+ ch = NOC_PMU_CH_DEFAULT;
+
+ reg = readl(noc_pmu->base + NOC_PMU_EVENT_CTRLn(reg_info->event_ctrl0, hwc->idx));
+ reg &= ~NOC_PMU_EVENT_CTRL_CHANNEL;
+ reg |= FIELD_PREP(NOC_PMU_EVENT_CTRL_CHANNEL, ch);
+ writel(reg, noc_pmu->base + NOC_PMU_EVENT_CTRLn(reg_info->event_ctrl0, hwc->idx));
+
+ /*
+ * Since tracetag filter applies to all the counters, don't touch it
+ * if user doesn't specify it explicitly.
+ */
+ if (tt_en) {
+ reg = readl(noc_pmu->base + reg_info->pmu_ctrl);
+ reg |= NOC_PMU_GLOBAL_CTRL_TT_EN;
+ writel(reg, noc_pmu->base + reg_info->pmu_ctrl);
+ }
+}
+
+static void hisi_noc_pmu_disable_filter(struct perf_event *event)
+{
+ struct hisi_pmu *noc_pmu = to_hisi_pmu(event->pmu);
+ struct hisi_noc_pmu_regs *reg_info = noc_pmu->dev_info->private;
+ u32 tt_en = hisi_get_tt_en(event);
+ u32 reg;
+
+ /*
+ * If we're not the last counter, don't touch the global tracetag
+ * configuration.
+ */
+ if (bitmap_weight(noc_pmu->pmu_events.used_mask, noc_pmu->num_counters) > 1)
+ return;
+
+ if (tt_en) {
+ reg = readl(noc_pmu->base + reg_info->pmu_ctrl);
+ reg &= ~NOC_PMU_GLOBAL_CTRL_TT_EN;
+ writel(reg, noc_pmu->base + reg_info->pmu_ctrl);
+ }
+}
+
+static const struct hisi_uncore_ops hisi_uncore_noc_ops = {
+ .write_evtype = hisi_noc_pmu_write_evtype,
+ .get_event_idx = hisi_noc_pmu_get_event_idx,
+ .read_counter = hisi_noc_pmu_read_counter,
+ .write_counter = hisi_noc_pmu_write_counter,
+ .enable_counter = hisi_noc_pmu_enable_counter,
+ .disable_counter = hisi_noc_pmu_disable_counter,
+ .enable_counter_int = hisi_noc_pmu_enable_counter_int,
+ .disable_counter_int = hisi_noc_pmu_disable_counter_int,
+ .start_counters = hisi_noc_pmu_start_counters,
+ .stop_counters = hisi_noc_pmu_stop_counters,
+ .get_int_status = hisi_noc_pmu_get_int_status,
+ .clear_int_status = hisi_noc_pmu_clear_int_status,
+ .enable_filter = hisi_noc_pmu_enable_filter,
+ .disable_filter = hisi_noc_pmu_disable_filter,
+};
+
+static struct attribute *hisi_noc_pmu_format_attrs[] = {
+ HISI_PMU_FORMAT_ATTR(event, "config:0-7"),
+ HISI_PMU_FORMAT_ATTR(ch, "config1:0-2"),
+ HISI_PMU_FORMAT_ATTR(tt_en, "config1:3"),
+ NULL
+};
+
+static const struct attribute_group hisi_noc_pmu_format_group = {
+ .name = "format",
+ .attrs = hisi_noc_pmu_format_attrs,
+};
+
+static struct attribute *hisi_noc_pmu_events_attrs[] = {
+ HISI_PMU_EVENT_ATTR(cycles, 0x0e),
+ /* Flux on/off the ring */
+ HISI_PMU_EVENT_ATTR(ingress_flow_sum, 0x1a),
+ HISI_PMU_EVENT_ATTR(egress_flow_sum, 0x17),
+ /* Buffer full duration on/off the ring */
+ HISI_PMU_EVENT_ATTR(ingress_buf_full, 0x19),
+ HISI_PMU_EVENT_ATTR(egress_buf_full, 0x12),
+ /* Failure packets count on/off the ring */
+ HISI_PMU_EVENT_ATTR(cw_ingress_fail, 0x01),
+ HISI_PMU_EVENT_ATTR(cc_ingress_fail, 0x09),
+ HISI_PMU_EVENT_ATTR(cw_egress_fail, 0x03),
+ HISI_PMU_EVENT_ATTR(cc_egress_fail, 0x0b),
+ /* Flux of the ring */
+ HISI_PMU_EVENT_ATTR(cw_main_flow_sum, 0x05),
+ HISI_PMU_EVENT_ATTR(cc_main_flow_sum, 0x0d),
+ NULL
+};
+
+static const struct attribute_group hisi_noc_pmu_events_group = {
+ .name = "events",
+ .attrs = hisi_noc_pmu_events_attrs,
+};
+
+static const struct attribute_group *hisi_noc_pmu_attr_groups[] = {
+ &hisi_noc_pmu_format_group,
+ &hisi_noc_pmu_events_group,
+ &hisi_pmu_cpumask_attr_group,
+ &hisi_pmu_identifier_group,
+ NULL
+};
+
+static int hisi_noc_pmu_dev_init(struct platform_device *pdev, struct hisi_pmu *noc_pmu)
+{
+ struct hisi_noc_pmu_regs *reg_info;
+
+ hisi_uncore_pmu_init_topology(noc_pmu, &pdev->dev);
+
+ if (noc_pmu->topo.scl_id < 0)
+ return dev_err_probe(&pdev->dev, -EINVAL, "failed to get scl-id\n");
+
+ if (noc_pmu->topo.index_id < 0)
+ return dev_err_probe(&pdev->dev, -EINVAL, "failed to get idx-id\n");
+
+ if (noc_pmu->topo.sub_id < 0)
+ return dev_err_probe(&pdev->dev, -EINVAL, "failed to get sub-id\n");
+
+ noc_pmu->base = devm_platform_ioremap_resource(pdev, 0);
+ if (IS_ERR(noc_pmu->base))
+ return dev_err_probe(&pdev->dev, PTR_ERR(noc_pmu->base),
+ "fail to remap io memory\n");
+
+ noc_pmu->dev_info = device_get_match_data(&pdev->dev);
+ if (!noc_pmu->dev_info)
+ return -ENODEV;
+
+ noc_pmu->pmu_events.attr_groups = noc_pmu->dev_info->attr_groups;
+ noc_pmu->counter_bits = noc_pmu->dev_info->counter_bits;
+ noc_pmu->check_event = noc_pmu->dev_info->check_event;
+ noc_pmu->num_counters = NOC_PMU_NR_COUNTERS;
+ noc_pmu->ops = &hisi_uncore_noc_ops;
+ noc_pmu->dev = &pdev->dev;
+ noc_pmu->on_cpu = -1;
+
+ reg_info = noc_pmu->dev_info->private;
+ noc_pmu->identifier = readl(noc_pmu->base + reg_info->version);
+
+ return 0;
+}
+
+static void hisi_noc_pmu_remove_cpuhp_instance(void *hotplug_node)
+{
+ cpuhp_state_remove_instance_nocalls(hisi_noc_pmu_cpuhp_state, hotplug_node);
+}
+
+static void hisi_noc_pmu_unregister_pmu(void *pmu)
+{
+ perf_pmu_unregister(pmu);
+}
+
+static int hisi_noc_pmu_probe(struct platform_device *pdev)
+{
+ struct device *dev = &pdev->dev;
+ struct hisi_pmu *noc_pmu;
+ char *name;
+ int ret;
+
+ noc_pmu = devm_kzalloc(dev, sizeof(*noc_pmu), GFP_KERNEL);
+ if (!noc_pmu)
+ return -ENOMEM;
+
+ /*
+ * HiSilicon Uncore PMU framework needs to get common hisi_pmu device
+ * from device's drvdata.
+ */
+ platform_set_drvdata(pdev, noc_pmu);
+
+ ret = hisi_noc_pmu_dev_init(pdev, noc_pmu);
+ if (ret)
+ return ret;
+
+ ret = cpuhp_state_add_instance(hisi_noc_pmu_cpuhp_state, &noc_pmu->node);
+ if (ret)
+ return dev_err_probe(dev, ret, "Fail to register cpuhp instance\n");
+
+ ret = devm_add_action_or_reset(dev, hisi_noc_pmu_remove_cpuhp_instance,
+ &noc_pmu->node);
+ if (ret)
+ return ret;
+
+ hisi_pmu_init(noc_pmu, THIS_MODULE);
+
+ name = devm_kasprintf(dev, GFP_KERNEL, "hisi_scl%d_noc%d_%d",
+ noc_pmu->topo.scl_id, noc_pmu->topo.index_id,
+ noc_pmu->topo.sub_id);
+ if (!name)
+ return -ENOMEM;
+
+ ret = perf_pmu_register(&noc_pmu->pmu, name, -1);
+ if (ret)
+ return dev_err_probe(dev, ret, "Fail to register PMU\n");
+
+ return devm_add_action_or_reset(dev, hisi_noc_pmu_unregister_pmu,
+ &noc_pmu->pmu);
+}
+
+static struct hisi_noc_pmu_regs hisi_noc_v1_pmu_regs = {
+ .version = NOC_PMU_VERSION,
+ .pmu_ctrl = NOC_PMU_GLOBAL_CTRL,
+ .event_ctrl0 = NOC_PMU_EVENT_CTRL0,
+ .event_cntr0 = NOC_PMU_EVENT_COUNTER0,
+ .ovflow_status = NOC_PMU_CNT_INFO,
+};
+
+static const struct hisi_pmu_dev_info hisi_noc_v1 = {
+ .attr_groups = hisi_noc_pmu_attr_groups,
+ .counter_bits = 64,
+ .check_event = NOC_PMU_EVENT_CTRL_TYPE,
+ .private = &hisi_noc_v1_pmu_regs,
+};
+
+static const struct acpi_device_id hisi_noc_pmu_ids[] = {
+ { "HISI04E0", (kernel_ulong_t) &hisi_noc_v1 },
+ { }
+};
+MODULE_DEVICE_TABLE(acpi, hisi_noc_pmu_ids);
+
+static struct platform_driver hisi_noc_pmu_driver = {
+ .driver = {
+ .name = "hisi_noc_pmu",
+ .acpi_match_table = hisi_noc_pmu_ids,
+ .suppress_bind_attrs = true,
+ },
+ .probe = hisi_noc_pmu_probe,
+};
+
+static int __init hisi_noc_pmu_module_init(void)
+{
+ int ret;
+
+ ret = cpuhp_setup_state_multi(CPUHP_AP_ONLINE_DYN, "perf/hisi/noc:online",
+ hisi_uncore_pmu_online_cpu,
+ hisi_uncore_pmu_offline_cpu);
+ if (ret < 0) {
+ pr_err("hisi_noc_pmu: Fail to setup cpuhp callbacks, ret = %d\n", ret);
+ return ret;
+ }
+ hisi_noc_pmu_cpuhp_state = ret;
+
+ ret = platform_driver_register(&hisi_noc_pmu_driver);
+ if (ret)
+ cpuhp_remove_multi_state(hisi_noc_pmu_cpuhp_state);
+
+ return ret;
+}
+module_init(hisi_noc_pmu_module_init);
+
+static void __exit hisi_noc_pmu_module_exit(void)
+{
+ platform_driver_unregister(&hisi_noc_pmu_driver);
+ cpuhp_remove_multi_state(hisi_noc_pmu_cpuhp_state);
+}
+module_exit(hisi_noc_pmu_module_exit);
+
+MODULE_IMPORT_NS("HISI_PMU");
+MODULE_DESCRIPTION("HiSilicon SoC Uncore NoC PMU driver");
+MODULE_LICENSE("GPL");
+MODULE_AUTHOR("Yicong Yang <yangyicong@hisilicon.com>");
--
2.24.0
^ permalink raw reply related [flat|nested] 5+ messages in thread* [PATCH v4 2/2] drivers/perf: hisi: Add support for HiSilicon MN PMU driver
2025-07-17 7:41 [PATCH v4 0/2] drivers/perf: hisi: Add support for HiSilicon MN PMU driver Yicong Yang
2025-07-17 7:41 ` [PATCH v4 1/2] drivers/perf: hisi: Add support for HiSilicon NoC PMU Yicong Yang
@ 2025-07-17 7:41 ` Yicong Yang
2025-07-17 10:39 ` Jonathan Cameron
1 sibling, 1 reply; 5+ messages in thread
From: Yicong Yang @ 2025-07-17 7:41 UTC (permalink / raw)
To: will, mark.rutland, linux-arm-kernel, linux-kernel
Cc: hejunhao3, jonathan.cameron, prime.zeng, linuxarm, yangyicong,
wangyushan12
From: Junhao He <hejunhao3@huawei.com>
MN (Miscellaneous Node) is a hybrid node in ARM CHI. It broadcasts the
following two types of requests: DVM operations and PCIe configuration.
MN PMU devices exist on both SCCL and SICL, so we named the MN pmu
driver after SCL (Super cluster) ID.
The MN PMU driver using the HiSilicon uncore PMU framework. And only
the event parameter is supported.
Signed-off-by: Junhao He <hejunhao3@huawei.com>
Signed-off-by: Yicong Yang <yangyicong@hisilicon.com>
---
drivers/perf/hisilicon/Makefile | 2 +-
drivers/perf/hisilicon/hisi_uncore_mn_pmu.c | 411 ++++++++++++++++++++
2 files changed, 412 insertions(+), 1 deletion(-)
create mode 100644 drivers/perf/hisilicon/hisi_uncore_mn_pmu.c
diff --git a/drivers/perf/hisilicon/Makefile b/drivers/perf/hisilicon/Makefile
index dcec8f39719d..186be3d02238 100644
--- a/drivers/perf/hisilicon/Makefile
+++ b/drivers/perf/hisilicon/Makefile
@@ -2,7 +2,7 @@
obj-$(CONFIG_HISI_PMU) += hisi_uncore_pmu.o hisi_uncore_l3c_pmu.o \
hisi_uncore_hha_pmu.o hisi_uncore_ddrc_pmu.o hisi_uncore_sllc_pmu.o \
hisi_uncore_pa_pmu.o hisi_uncore_cpa_pmu.o hisi_uncore_uc_pmu.o \
- hisi_uncore_noc_pmu.o
+ hisi_uncore_noc_pmu.o hisi_uncore_mn_pmu.o
obj-$(CONFIG_HISI_PCIE_PMU) += hisi_pcie_pmu.o
obj-$(CONFIG_HNS3_PMU) += hns3_pmu.o
diff --git a/drivers/perf/hisilicon/hisi_uncore_mn_pmu.c b/drivers/perf/hisilicon/hisi_uncore_mn_pmu.c
new file mode 100644
index 000000000000..4df4eebe243e
--- /dev/null
+++ b/drivers/perf/hisilicon/hisi_uncore_mn_pmu.c
@@ -0,0 +1,411 @@
+// SPDX-License-Identifier: GPL-2.0-only
+/*
+ * HiSilicon SoC MN uncore Hardware event counters support
+ *
+ * Copyright (c) 2025 HiSilicon Technologies Co., Ltd.
+ */
+#include <linux/cpuhotplug.h>
+#include <linux/interrupt.h>
+#include <linux/iopoll.h>
+#include <linux/irq.h>
+#include <linux/list.h>
+#include <linux/mod_devicetable.h>
+#include <linux/property.h>
+
+#include "hisi_uncore_pmu.h"
+
+/* Dynamic CPU hotplug state used by MN PMU */
+static enum cpuhp_state hisi_mn_pmu_online;
+
+/* MN register definition */
+#define HISI_MN_DYNAMIC_CTRL_REG 0x400
+#define HISI_MN_DYNAMIC_CTRL_EN BIT(0)
+#define HISI_MN_PERF_CTRL_REG 0x408
+#define HISI_MN_PERF_CTRL_EN BIT(6)
+#define HISI_MN_INT_MASK_REG 0x800
+#define HISI_MN_INT_STATUS_REG 0x808
+#define HISI_MN_INT_CLEAR_REG 0x80C
+#define HISI_MN_EVENT_CTRL_REG 0x1C00
+#define HISI_MN_VERSION_REG 0x1C04
+#define HISI_MN_EVTYPE0_REG 0x1d00
+#define HISI_MN_EVTYPE_MASK GENMASK(7, 0)
+#define HISI_MN_CNTR0_REG 0x1e00
+#define HISI_MN_EVTYPE_REGn(evtype0, n) ((evtype0) + (n) * 4)
+#define HISI_MN_CNTR_REGn(cntr0, n) ((cntr0) + (n) * 8)
+
+#define HISI_MN_NR_COUNTERS 4
+#define HISI_MN_TIMEOUT_US 500U
+
+struct hisi_mn_pmu_regs {
+ u32 version;
+ u32 dyn_ctrl;
+ u32 perf_ctrl;
+ u32 int_mask;
+ u32 int_clear;
+ u32 int_status;
+ u32 event_ctrl;
+ u32 event_type0;
+ u32 event_cntr0;
+};
+
+/*
+ * Each event request takes a certain amount of time to complete. If
+ * we counting the latency related event, we need to wait for the all
+ * requests complete. Otherwise, the value of counter is slightly larger.
+ */
+static void hisi_mn_pmu_counter_flush(struct hisi_pmu *mn_pmu)
+{
+ struct hisi_mn_pmu_regs *reg_info = mn_pmu->dev_info->private;
+ int ret;
+ u32 val;
+
+ val = readl(mn_pmu->base + reg_info->dyn_ctrl);
+ val |= HISI_MN_DYNAMIC_CTRL_EN;
+ writel(val, mn_pmu->base + reg_info->dyn_ctrl);
+
+ ret = readl_poll_timeout_atomic(mn_pmu->base + reg_info->dyn_ctrl,
+ val, !(val & HISI_MN_DYNAMIC_CTRL_EN),
+ 1, HISI_MN_TIMEOUT_US);
+ if (ret)
+ dev_warn(mn_pmu->dev, "Counter flush timeout\n");
+}
+
+static u64 hisi_mn_pmu_read_counter(struct hisi_pmu *mn_pmu,
+ struct hw_perf_event *hwc)
+{
+ struct hisi_mn_pmu_regs *reg_info = mn_pmu->dev_info->private;
+
+ return readq(mn_pmu->base + HISI_MN_CNTR_REGn(reg_info->event_cntr0, hwc->idx));
+}
+
+static void hisi_mn_pmu_write_counter(struct hisi_pmu *mn_pmu,
+ struct hw_perf_event *hwc, u64 val)
+{
+ struct hisi_mn_pmu_regs *reg_info = mn_pmu->dev_info->private;
+
+ writeq(val, mn_pmu->base + HISI_MN_CNTR_REGn(reg_info->event_cntr0, hwc->idx));
+}
+
+static void hisi_mn_pmu_write_evtype(struct hisi_pmu *mn_pmu, int idx, u32 type)
+{
+ struct hisi_mn_pmu_regs *reg_info = mn_pmu->dev_info->private;
+ u32 val;
+
+ /*
+ * Select the appropriate event select register.
+ * There are 2 32-bit event select registers for the
+ * 8 hardware counters, each event code is 8-bit wide.
+ */
+ val = readl(mn_pmu->base + HISI_MN_EVTYPE_REGn(reg_info->event_type0, idx / 4));
+ val &= ~(HISI_MN_EVTYPE_MASK << HISI_PMU_EVTYPE_SHIFT(idx));
+ val |= (type << HISI_PMU_EVTYPE_SHIFT(idx));
+ writel(val, mn_pmu->base + HISI_MN_EVTYPE_REGn(reg_info->event_type0, idx / 4));
+}
+
+static void hisi_mn_pmu_start_counters(struct hisi_pmu *mn_pmu)
+{
+ struct hisi_mn_pmu_regs *reg_info = mn_pmu->dev_info->private;
+ u32 val;
+
+ val = readl(mn_pmu->base + reg_info->perf_ctrl);
+ val |= HISI_MN_PERF_CTRL_EN;
+ writel(val, mn_pmu->base + reg_info->perf_ctrl);
+}
+
+static void hisi_mn_pmu_stop_counters(struct hisi_pmu *mn_pmu)
+{
+ struct hisi_mn_pmu_regs *reg_info = mn_pmu->dev_info->private;
+ u32 val;
+
+ val = readl(mn_pmu->base + reg_info->perf_ctrl);
+ val &= ~HISI_MN_PERF_CTRL_EN;
+ writel(val, mn_pmu->base + reg_info->perf_ctrl);
+
+ hisi_mn_pmu_counter_flush(mn_pmu);
+}
+
+static void hisi_mn_pmu_enable_counter(struct hisi_pmu *mn_pmu,
+ struct hw_perf_event *hwc)
+{
+ struct hisi_mn_pmu_regs *reg_info = mn_pmu->dev_info->private;
+ u32 val;
+
+ val = readl(mn_pmu->base + reg_info->event_ctrl);
+ val |= BIT(hwc->idx);
+ writel(val, mn_pmu->base + reg_info->event_ctrl);
+}
+
+static void hisi_mn_pmu_disable_counter(struct hisi_pmu *mn_pmu,
+ struct hw_perf_event *hwc)
+{
+ struct hisi_mn_pmu_regs *reg_info = mn_pmu->dev_info->private;
+ u32 val;
+
+ val = readl(mn_pmu->base + reg_info->event_ctrl);
+ val &= ~BIT(hwc->idx);
+ writel(val, mn_pmu->base + reg_info->event_ctrl);
+}
+
+static void hisi_mn_pmu_enable_counter_int(struct hisi_pmu *mn_pmu,
+ struct hw_perf_event *hwc)
+{
+ struct hisi_mn_pmu_regs *reg_info = mn_pmu->dev_info->private;
+ u32 val;
+
+ val = readl(mn_pmu->base + reg_info->int_mask);
+ val &= ~BIT(hwc->idx);
+ writel(val, mn_pmu->base + reg_info->int_mask);
+}
+
+static void hisi_mn_pmu_disable_counter_int(struct hisi_pmu *mn_pmu,
+ struct hw_perf_event *hwc)
+{
+ struct hisi_mn_pmu_regs *reg_info = mn_pmu->dev_info->private;
+ u32 val;
+
+ val = readl(mn_pmu->base + reg_info->int_mask);
+ val |= BIT(hwc->idx);
+ writel(val, mn_pmu->base + reg_info->int_mask);
+}
+
+static u32 hisi_mn_pmu_get_int_status(struct hisi_pmu *mn_pmu)
+{
+ struct hisi_mn_pmu_regs *reg_info = mn_pmu->dev_info->private;
+
+ return readl(mn_pmu->base + reg_info->int_status);
+}
+
+static void hisi_mn_pmu_clear_int_status(struct hisi_pmu *mn_pmu, int idx)
+{
+ struct hisi_mn_pmu_regs *reg_info = mn_pmu->dev_info->private;
+
+ writel(BIT(idx), mn_pmu->base + reg_info->int_clear);
+}
+
+static struct attribute *hisi_mn_pmu_format_attr[] = {
+ HISI_PMU_FORMAT_ATTR(event, "config:0-7"),
+ NULL
+};
+
+static const struct attribute_group hisi_mn_pmu_format_group = {
+ .name = "format",
+ .attrs = hisi_mn_pmu_format_attr,
+};
+
+static struct attribute *hisi_mn_pmu_events_attr[] = {
+ HISI_PMU_EVENT_ATTR(req_eobarrier_num, 0x00),
+ HISI_PMU_EVENT_ATTR(req_ecbarrier_num, 0x01),
+ HISI_PMU_EVENT_ATTR(req_dvmop_num, 0x02),
+ HISI_PMU_EVENT_ATTR(req_dvmsync_num, 0x03),
+ HISI_PMU_EVENT_ATTR(req_retry_num, 0x04),
+ HISI_PMU_EVENT_ATTR(req_writenosnp_num, 0x05),
+ HISI_PMU_EVENT_ATTR(req_readnosnp_num, 0x06),
+ HISI_PMU_EVENT_ATTR(snp_dvm_num, 0x07),
+ HISI_PMU_EVENT_ATTR(snp_dvmsync_num, 0x08),
+ HISI_PMU_EVENT_ATTR(l3t_req_dvm_num, 0x09),
+ HISI_PMU_EVENT_ATTR(l3t_req_dvmsync_num, 0x0A),
+ HISI_PMU_EVENT_ATTR(mn_req_dvm_num, 0x0B),
+ HISI_PMU_EVENT_ATTR(mn_req_dvmsync_num, 0x0C),
+ HISI_PMU_EVENT_ATTR(pa_req_dvm_num, 0x0D),
+ HISI_PMU_EVENT_ATTR(pa_req_dvmsync_num, 0x0E),
+ HISI_PMU_EVENT_ATTR(snp_dvm_latency, 0x80),
+ HISI_PMU_EVENT_ATTR(snp_dvmsync_latency, 0x81),
+ HISI_PMU_EVENT_ATTR(l3t_req_dvm_latency, 0x82),
+ HISI_PMU_EVENT_ATTR(l3t_req_dvmsync_latency, 0x83),
+ HISI_PMU_EVENT_ATTR(mn_req_dvm_latency, 0x84),
+ HISI_PMU_EVENT_ATTR(mn_req_dvmsync_latency, 0x85),
+ HISI_PMU_EVENT_ATTR(pa_req_dvm_latency, 0x86),
+ HISI_PMU_EVENT_ATTR(pa_req_dvmsync_latency, 0x87),
+ NULL
+};
+
+static const struct attribute_group hisi_mn_pmu_events_group = {
+ .name = "events",
+ .attrs = hisi_mn_pmu_events_attr,
+};
+
+static const struct attribute_group *hisi_mn_pmu_attr_groups[] = {
+ &hisi_mn_pmu_format_group,
+ &hisi_mn_pmu_events_group,
+ &hisi_pmu_cpumask_attr_group,
+ &hisi_pmu_identifier_group,
+ NULL
+};
+
+static const struct hisi_uncore_ops hisi_uncore_mn_ops = {
+ .write_evtype = hisi_mn_pmu_write_evtype,
+ .get_event_idx = hisi_uncore_pmu_get_event_idx,
+ .start_counters = hisi_mn_pmu_start_counters,
+ .stop_counters = hisi_mn_pmu_stop_counters,
+ .enable_counter = hisi_mn_pmu_enable_counter,
+ .disable_counter = hisi_mn_pmu_disable_counter,
+ .enable_counter_int = hisi_mn_pmu_enable_counter_int,
+ .disable_counter_int = hisi_mn_pmu_disable_counter_int,
+ .write_counter = hisi_mn_pmu_write_counter,
+ .read_counter = hisi_mn_pmu_read_counter,
+ .get_int_status = hisi_mn_pmu_get_int_status,
+ .clear_int_status = hisi_mn_pmu_clear_int_status,
+};
+
+static int hisi_mn_pmu_dev_init(struct platform_device *pdev,
+ struct hisi_pmu *mn_pmu)
+{
+ struct hisi_mn_pmu_regs *reg_info;
+ int ret;
+
+ hisi_uncore_pmu_init_topology(mn_pmu, &pdev->dev);
+
+ if (mn_pmu->topo.scl_id < 0)
+ return dev_err_probe(&pdev->dev, -EINVAL,
+ "Failed to read MN scl id\n");
+
+ if (mn_pmu->topo.index_id < 0)
+ return dev_err_probe(&pdev->dev, -EINVAL,
+ "Failed to read MN index id\n");
+
+ mn_pmu->base = devm_platform_ioremap_resource(pdev, 0);
+ if (IS_ERR(mn_pmu->base))
+ return dev_err_probe(&pdev->dev, PTR_ERR(mn_pmu->base),
+ "Failed to ioremap resource\n");
+
+ ret = hisi_uncore_pmu_init_irq(mn_pmu, pdev);
+ if (ret)
+ return ret;
+
+ mn_pmu->dev_info = device_get_match_data(&pdev->dev);
+ if (!mn_pmu->dev_info)
+ return -ENODEV;
+
+ mn_pmu->pmu_events.attr_groups = mn_pmu->dev_info->attr_groups;
+ mn_pmu->counter_bits = mn_pmu->dev_info->counter_bits;
+ mn_pmu->check_event = mn_pmu->dev_info->check_event;
+ mn_pmu->num_counters = HISI_MN_NR_COUNTERS;
+ mn_pmu->ops = &hisi_uncore_mn_ops;
+ mn_pmu->dev = &pdev->dev;
+ mn_pmu->on_cpu = -1;
+
+ reg_info = mn_pmu->dev_info->private;
+ mn_pmu->identifier = readl(mn_pmu->base + reg_info->version);
+
+ return 0;
+}
+
+static void hisi_mn_pmu_remove_cpuhp(void *hotplug_node)
+{
+ cpuhp_state_remove_instance_nocalls(hisi_mn_pmu_online, hotplug_node);
+}
+
+static void hisi_mn_pmu_unregister(void *pmu)
+{
+ perf_pmu_unregister(pmu);
+}
+
+static int hisi_mn_pmu_probe(struct platform_device *pdev)
+{
+ struct hisi_pmu *mn_pmu;
+ char *name;
+ int ret;
+
+ mn_pmu = devm_kzalloc(&pdev->dev, sizeof(*mn_pmu), GFP_KERNEL);
+ if (!mn_pmu)
+ return -ENOMEM;
+
+ platform_set_drvdata(pdev, mn_pmu);
+
+ ret = hisi_mn_pmu_dev_init(pdev, mn_pmu);
+ if (ret)
+ return ret;
+
+ name = devm_kasprintf(&pdev->dev, GFP_KERNEL, "hisi_scl%d_mn%d",
+ mn_pmu->topo.scl_id, mn_pmu->topo.index_id);
+ if (!name)
+ return -ENOMEM;
+
+ ret = cpuhp_state_add_instance(hisi_mn_pmu_online, &mn_pmu->node);
+ if (ret)
+ return dev_err_probe(&pdev->dev, ret, "Failed to register cpu hotplug\n");
+
+ ret = devm_add_action_or_reset(&pdev->dev, hisi_mn_pmu_remove_cpuhp, &mn_pmu->node);
+ if (ret)
+ return ret;
+
+ hisi_pmu_init(mn_pmu, THIS_MODULE);
+
+ ret = perf_pmu_register(&mn_pmu->pmu, name, -1);
+ if (ret)
+ return dev_err_probe(mn_pmu->dev, ret, "Failed to register MN PMU\n");
+
+ return devm_add_action_or_reset(&pdev->dev, hisi_mn_pmu_unregister, &mn_pmu->pmu);
+}
+
+static struct hisi_mn_pmu_regs hisi_mn_v1_pmu_regs = {
+ .version = HISI_MN_VERSION_REG,
+ .dyn_ctrl = HISI_MN_DYNAMIC_CTRL_REG,
+ .perf_ctrl = HISI_MN_PERF_CTRL_REG,
+ .int_mask = HISI_MN_INT_MASK_REG,
+ .int_clear = HISI_MN_INT_CLEAR_REG,
+ .int_status = HISI_MN_INT_STATUS_REG,
+ .event_ctrl = HISI_MN_EVENT_CTRL_REG,
+ .event_type0 = HISI_MN_EVTYPE0_REG,
+ .event_cntr0 = HISI_MN_CNTR0_REG,
+};
+
+static const struct hisi_pmu_dev_info hisi_mn_v1 = {
+ .attr_groups = hisi_mn_pmu_attr_groups,
+ .counter_bits = 48,
+ .check_event = HISI_MN_EVTYPE_MASK,
+ .private = &hisi_mn_v1_pmu_regs,
+};
+
+static const struct acpi_device_id hisi_mn_pmu_acpi_match[] = {
+ { "HISI0222", (kernel_ulong_t) &hisi_mn_v1 },
+ { }
+};
+MODULE_DEVICE_TABLE(acpi, hisi_mn_pmu_acpi_match);
+
+static struct platform_driver hisi_mn_pmu_driver = {
+ .driver = {
+ .name = "hisi_mn_pmu",
+ .acpi_match_table = hisi_mn_pmu_acpi_match,
+ /*
+ * We have not worked out a safe bind/unbind process,
+ * Forcefully unbinding during sampling will lead to a
+ * kernel panic, so this is not supported yet.
+ */
+ .suppress_bind_attrs = true,
+ },
+ .probe = hisi_mn_pmu_probe,
+};
+
+static int __init hisi_mn_pmu_module_init(void)
+{
+ int ret;
+
+ ret = cpuhp_setup_state_multi(CPUHP_AP_ONLINE_DYN, "perf/hisi/mn:online",
+ hisi_uncore_pmu_online_cpu,
+ hisi_uncore_pmu_offline_cpu);
+ if (ret < 0) {
+ pr_err("hisi_mn_pmu: Failed to setup MN PMU hotplug: %d\n", ret);
+ return ret;
+ }
+ hisi_mn_pmu_online = ret;
+
+ ret = platform_driver_register(&hisi_mn_pmu_driver);
+ if (ret)
+ cpuhp_remove_multi_state(hisi_mn_pmu_online);
+
+ return ret;
+}
+module_init(hisi_mn_pmu_module_init);
+
+static void __exit hisi_mn_pmu_module_exit(void)
+{
+ platform_driver_unregister(&hisi_mn_pmu_driver);
+ cpuhp_remove_multi_state(hisi_mn_pmu_online);
+}
+module_exit(hisi_mn_pmu_module_exit);
+
+MODULE_IMPORT_NS("HISI_PMU");
+MODULE_DESCRIPTION("HiSilicon SoC MN uncore PMU driver");
+MODULE_LICENSE("GPL");
+MODULE_AUTHOR("Junhao He <hejunhao3@huawei.com>");
--
2.24.0
^ permalink raw reply related [flat|nested] 5+ messages in thread