Add support for various PMU counters found on the Cavium ThunderX and
OcteonTx SoC.
The driver provides common "uncore" functions to avoid code duplication and
support adding more device PMUs (like L2 cache) in the future.
Patches are on top of 4.13-rc1.
Changes to v6:
- Make driver stand-alone again without hooking into EDAC
- depend on ARCH_THUNDER
Changes to v5:
- Only allow built-in CONFIG_CAVIUM_PMU
- Drop unneeded export of perf_event_update_userpage()
- Simplify callbacks in edac code, move CONFIG_CAVIUM_PMU check
to header file
- Fix some sparse static warnings
- Add documentation
- Fix OCX TLK event_valid check
- Add group validation in event_init
- Add a guard variable to prevent calling init twice
- Use kasprintf and fix pmu name allocation
- Remove unneeded check for embedded pmu
- Loop around local64_cmpxchg
- Simplify cvm_pmu_lmc_event_valid
- Fix list_add error case
Jan Glauber (3):
perf: cavium: Support memory controller PMU counters
perf: cavium: Support transmit-link PMU counters
perf: cavium: Add Documentation
Documentation/perf/cavium-pmu.txt | 74 +++++
drivers/perf/Kconfig | 8 +
drivers/perf/Makefile | 1 +
drivers/perf/cavium_pmu.c | 632 ++++++++++++++++++++++++++++++++++++++
include/linux/cpuhotplug.h | 1 +
5 files changed, 716 insertions(+)
create mode 100644 Documentation/perf/cavium-pmu.txt
create mode 100644 drivers/perf/cavium_pmu.c
--
2.9.0.rc0.21.g7777322
Add support for the PMU counters on Cavium SOC memory controllers.
This patch also adds generic functions to allow supporting more
devices with PMU counters.
Properties of the LMC PMU counters:
- not stoppable
- fixed purpose
- read-only
- one PCI device per memory controller
Signed-off-by: Jan Glauber <[email protected]>
---
drivers/perf/Kconfig | 8 +
drivers/perf/Makefile | 1 +
drivers/perf/cavium_pmu.c | 416 +++++++++++++++++++++++++++++++++++++++++++++
include/linux/cpuhotplug.h | 1 +
4 files changed, 426 insertions(+)
create mode 100644 drivers/perf/cavium_pmu.c
diff --git a/drivers/perf/Kconfig b/drivers/perf/Kconfig
index e5197ff..a46c3f0 100644
--- a/drivers/perf/Kconfig
+++ b/drivers/perf/Kconfig
@@ -43,4 +43,12 @@ config XGENE_PMU
help
Say y if you want to use APM X-Gene SoC performance monitors.
+config CAVIUM_PMU
+ bool "Cavium SOC PMU"
+ depends on ARCH_THUNDER
+ help
+ Provides access to various performance counters on Caviums
+ ARM64 SOCs. Adds support for memory controller (LMC) and
+ interconnect link (OCX TLK) counters.
+
endmenu
diff --git a/drivers/perf/Makefile b/drivers/perf/Makefile
index 6420bd4..b304646 100644
--- a/drivers/perf/Makefile
+++ b/drivers/perf/Makefile
@@ -3,3 +3,4 @@ obj-$(CONFIG_ARM_PMU_ACPI) += arm_pmu_acpi.o
obj-$(CONFIG_QCOM_L2_PMU) += qcom_l2_pmu.o
obj-$(CONFIG_QCOM_L3_PMU) += qcom_l3_pmu.o
obj-$(CONFIG_XGENE_PMU) += xgene_pmu.o
+obj-$(CONFIG_CAVIUM_PMU) += cavium_pmu.o
diff --git a/drivers/perf/cavium_pmu.c b/drivers/perf/cavium_pmu.c
new file mode 100644
index 0000000..7d22493
--- /dev/null
+++ b/drivers/perf/cavium_pmu.c
@@ -0,0 +1,416 @@
+/*
+ * Cavium ARM SOC "uncore" PMU counters
+ *
+ * This file is subject to the terms and conditions of the GNU General Public
+ * License. See the file "COPYING" in the main directory of this archive
+ * for more details.
+ *
+ * Copyright Cavium, Inc. 2017
+ * Author(s): Jan Glauber <[email protected]>
+ *
+ */
+#include <linux/cpumask.h>
+#include <linux/io.h>
+#include <linux/list.h>
+#include <linux/mutex.h>
+#include <linux/pci.h>
+#include <linux/perf_event.h>
+#include <linux/slab.h>
+
+enum cvm_pmu_type {
+ CVM_PMU_LMC,
+};
+
+/* maximum number of parallel hardware counters for all pmu types */
+#define CVM_PMU_MAX_COUNTERS 64
+
+/* generic struct to cover the different pmu types */
+struct cvm_pmu_dev {
+ struct pmu pmu;
+ const char *pmu_name;
+ bool (*event_valid)(u64);
+ void __iomem *map;
+ struct pci_dev *pdev;
+ int num_counters;
+ struct perf_event *events[CVM_PMU_MAX_COUNTERS];
+ struct list_head entry;
+ struct hlist_node cpuhp_node;
+ cpumask_t active_mask;
+};
+
+static struct list_head cvm_pmu_lmcs;
+
+/*
+ * Common Cavium PMU stuff
+ *
+ * Shared properties of the different PMU types:
+ * - all counters are 64 bit long
+ * - there are no overflow interrupts
+ * - all devices with PMU counters appear as PCI devices
+ *
+ * Counter control, access and device association depends on the
+ * PMU type.
+ */
+
+#define to_pmu_dev(x) container_of((x), struct cvm_pmu_dev, pmu)
+
+static int cvm_pmu_event_init(struct perf_event *event)
+{
+ struct hw_perf_event *hwc = &event->hw;
+ struct cvm_pmu_dev *pmu_dev;
+ struct perf_event *sibling;
+
+ if (event->attr.type != event->pmu->type)
+ return -ENOENT;
+
+ /* we do not support sampling */
+ if (is_sampling_event(event))
+ return -EINVAL;
+
+ /* PMU counters do not support any these bits */
+ if (event->attr.exclude_user ||
+ event->attr.exclude_kernel ||
+ event->attr.exclude_host ||
+ event->attr.exclude_guest ||
+ event->attr.exclude_hv ||
+ event->attr.exclude_idle)
+ return -EINVAL;
+
+ pmu_dev = to_pmu_dev(event->pmu);
+ if (!pmu_dev->event_valid(event->attr.config))
+ return -EINVAL;
+
+ /*
+ * Forbid groups containing mixed PMUs, software events are acceptable.
+ */
+ if (event->group_leader->pmu != event->pmu &&
+ !is_software_event(event->group_leader))
+ return -EINVAL;
+
+ list_for_each_entry(sibling, &event->group_leader->sibling_list,
+ group_entry)
+ if (sibling->pmu != event->pmu &&
+ !is_software_event(sibling))
+ return -EINVAL;
+
+ hwc->config = event->attr.config;
+ hwc->idx = -1;
+ return 0;
+}
+
+static void cvm_pmu_read(struct perf_event *event)
+{
+ struct cvm_pmu_dev *pmu_dev = to_pmu_dev(event->pmu);
+ struct hw_perf_event *hwc = &event->hw;
+ u64 prev, delta, new;
+
+again:
+ prev = local64_read(&hwc->prev_count);
+ new = readq(hwc->event_base + pmu_dev->map);
+
+ if (local64_cmpxchg(&hwc->prev_count, prev, new) != prev)
+ goto again;
+
+ delta = new - prev;
+ local64_add(delta, &event->count);
+}
+
+static void cvm_pmu_start(struct perf_event *event, int flags)
+{
+ struct cvm_pmu_dev *pmu_dev = to_pmu_dev(event->pmu);
+ struct hw_perf_event *hwc = &event->hw;
+ u64 new;
+
+ if (WARN_ON_ONCE(!(hwc->state & PERF_HES_STOPPED)))
+ return;
+
+ WARN_ON_ONCE(!(hwc->state & PERF_HES_UPTODATE));
+ hwc->state = 0;
+
+ /* update prev_count always in order support unstoppable counters */
+ new = readq(hwc->event_base + pmu_dev->map);
+ local64_set(&hwc->prev_count, new);
+
+ perf_event_update_userpage(event);
+}
+
+static void cvm_pmu_stop(struct perf_event *event, int flags)
+{
+ struct hw_perf_event *hwc = &event->hw;
+
+ WARN_ON_ONCE(hwc->state & PERF_HES_STOPPED);
+ hwc->state |= PERF_HES_STOPPED;
+
+ if ((flags & PERF_EF_UPDATE) && !(hwc->state & PERF_HES_UPTODATE)) {
+ cvm_pmu_read(event);
+ hwc->state |= PERF_HES_UPTODATE;
+ }
+}
+
+static int cvm_pmu_add(struct perf_event *event, int flags, u64 config_base,
+ u64 event_base)
+{
+ struct cvm_pmu_dev *pmu_dev = to_pmu_dev(event->pmu);
+ struct hw_perf_event *hwc = &event->hw;
+
+ if (!cmpxchg(&pmu_dev->events[hwc->config], NULL, event))
+ hwc->idx = hwc->config;
+
+ if (hwc->idx == -1)
+ return -EBUSY;
+
+ hwc->config_base = config_base;
+ hwc->event_base = event_base;
+ hwc->state = PERF_HES_UPTODATE | PERF_HES_STOPPED;
+
+ if (flags & PERF_EF_START)
+ pmu_dev->pmu.start(event, PERF_EF_RELOAD);
+
+ return 0;
+}
+
+static void cvm_pmu_del(struct perf_event *event, int flags)
+{
+ struct cvm_pmu_dev *pmu_dev = to_pmu_dev(event->pmu);
+ struct hw_perf_event *hwc = &event->hw;
+ int i;
+
+ event->pmu->stop(event, PERF_EF_UPDATE);
+
+ /*
+ * For programmable counters we need to check where we installed it.
+ * To keep this function generic always test the more complicated
+ * case (free running counters won't need the loop).
+ */
+ for (i = 0; i < pmu_dev->num_counters; i++)
+ if (cmpxchg(&pmu_dev->events[i], event, NULL) == event)
+ break;
+
+ perf_event_update_userpage(event);
+ hwc->idx = -1;
+}
+
+static ssize_t cvm_pmu_event_sysfs_show(struct device *dev,
+ struct device_attribute *attr,
+ char *page)
+{
+ struct perf_pmu_events_attr *pmu_attr =
+ container_of(attr, struct perf_pmu_events_attr, attr);
+
+ if (pmu_attr->event_str)
+ return sprintf(page, "%s", pmu_attr->event_str);
+
+ return 0;
+}
+
+/*
+ * The pmu events are independent from CPUs. Provide a cpumask
+ * nevertheless to prevent perf from adding the event per-cpu and just
+ * set the mask to one online CPU. Use the same cpumask for all "uncore"
+ * devices.
+ *
+ * There is a performance penalty for accessing a device from a CPU on
+ * another socket, but we do not care.
+ */
+static int cvm_pmu_offline_cpu(unsigned int old_cpu, struct hlist_node *node)
+{
+ struct cvm_pmu_dev *pmu_dev;
+ int new_cpu;
+
+ pmu_dev = hlist_entry_safe(node, struct cvm_pmu_dev, cpuhp_node);
+ if (!cpumask_test_and_clear_cpu(old_cpu, &pmu_dev->active_mask))
+ return 0;
+
+ new_cpu = cpumask_any_but(cpu_online_mask, old_cpu);
+ if (new_cpu >= nr_cpu_ids)
+ return 0;
+ perf_pmu_migrate_context(&pmu_dev->pmu, old_cpu, new_cpu);
+ cpumask_set_cpu(new_cpu, &pmu_dev->active_mask);
+ return 0;
+}
+
+static ssize_t cvm_pmu_attr_show_cpumask(struct device *dev,
+ struct device_attribute *attr,
+ char *buf)
+{
+ struct pmu *pmu = dev_get_drvdata(dev);
+ struct cvm_pmu_dev *pmu_dev = container_of(pmu, struct cvm_pmu_dev, pmu);
+
+ return cpumap_print_to_pagebuf(true, buf, &pmu_dev->active_mask);
+}
+
+static DEVICE_ATTR(cpumask, S_IRUGO, cvm_pmu_attr_show_cpumask, NULL);
+
+static struct attribute *cvm_pmu_attrs[] = {
+ &dev_attr_cpumask.attr,
+ NULL,
+};
+
+static struct attribute_group cvm_pmu_attr_group = {
+ .attrs = cvm_pmu_attrs,
+};
+
+/*
+ * LMC (memory controller) counters:
+ * - not stoppable, always on, read-only
+ * - one PCI device per memory controller
+ */
+#define LMC_CONFIG_OFFSET 0x188
+#define LMC_CONFIG_RESET_BIT BIT(17)
+
+/* LMC events */
+#define LMC_EVENT_IFB_CNT 0x1d0
+#define LMC_EVENT_OPS_CNT 0x1d8
+#define LMC_EVENT_DCLK_CNT 0x1e0
+#define LMC_EVENT_BANK_CONFLICT1 0x360
+#define LMC_EVENT_BANK_CONFLICT2 0x368
+
+#define CVM_PMU_LMC_EVENT_ATTR(_name, _id) \
+ &((struct perf_pmu_events_attr[]) { \
+ { \
+ __ATTR(_name, S_IRUGO, cvm_pmu_event_sysfs_show, NULL), \
+ _id, \
+ "lmc_event=" __stringify(_id), \
+ } \
+ })[0].attr.attr
+
+/* map counter numbers to register offsets */
+static int lmc_events[] = {
+ LMC_EVENT_IFB_CNT,
+ LMC_EVENT_OPS_CNT,
+ LMC_EVENT_DCLK_CNT,
+ LMC_EVENT_BANK_CONFLICT1,
+ LMC_EVENT_BANK_CONFLICT2,
+};
+
+static int cvm_pmu_lmc_add(struct perf_event *event, int flags)
+{
+ struct hw_perf_event *hwc = &event->hw;
+
+ return cvm_pmu_add(event, flags, LMC_CONFIG_OFFSET,
+ lmc_events[hwc->config]);
+}
+
+PMU_FORMAT_ATTR(lmc_event, "config:0-2");
+
+static struct attribute *cvm_pmu_lmc_format_attr[] = {
+ &format_attr_lmc_event.attr,
+ NULL,
+};
+
+static struct attribute_group cvm_pmu_lmc_format_group = {
+ .name = "format",
+ .attrs = cvm_pmu_lmc_format_attr,
+};
+
+static struct attribute *cvm_pmu_lmc_events_attr[] = {
+ CVM_PMU_LMC_EVENT_ATTR(ifb_cnt, 0),
+ CVM_PMU_LMC_EVENT_ATTR(ops_cnt, 1),
+ CVM_PMU_LMC_EVENT_ATTR(dclk_cnt, 2),
+ CVM_PMU_LMC_EVENT_ATTR(bank_conflict1, 3),
+ CVM_PMU_LMC_EVENT_ATTR(bank_conflict2, 4),
+ NULL,
+};
+
+static struct attribute_group cvm_pmu_lmc_events_group = {
+ .name = "events",
+ .attrs = cvm_pmu_lmc_events_attr,
+};
+
+static const struct attribute_group *cvm_pmu_lmc_attr_groups[] = {
+ &cvm_pmu_attr_group,
+ &cvm_pmu_lmc_format_group,
+ &cvm_pmu_lmc_events_group,
+ NULL,
+};
+
+static bool cvm_pmu_lmc_event_valid(u64 config)
+{
+ return (config < ARRAY_SIZE(lmc_events));
+}
+
+static int cvm_pmu_lmc_probe(struct pci_dev *pdev)
+{
+ struct cvm_pmu_dev *next, *lmc;
+ int nr = 0, ret = -ENOMEM;
+
+ lmc = kzalloc(sizeof(*lmc), GFP_KERNEL);
+ if (!lmc)
+ goto fail_nomem;
+
+ lmc->map = ioremap(pci_resource_start(pdev, 0),
+ pci_resource_len(pdev, 0));
+ list_for_each_entry(next, &cvm_pmu_lmcs, entry)
+ nr++;
+ lmc->pmu_name = kasprintf(GFP_KERNEL, "lmc%d", nr);
+
+ lmc->pdev = pdev;
+ lmc->num_counters = ARRAY_SIZE(lmc_events);
+ lmc->pmu = (struct pmu) {
+ .task_ctx_nr = perf_invalid_context,
+ .event_init = cvm_pmu_event_init,
+ .add = cvm_pmu_lmc_add,
+ .del = cvm_pmu_del,
+ .start = cvm_pmu_start,
+ .stop = cvm_pmu_stop,
+ .read = cvm_pmu_read,
+ .attr_groups = cvm_pmu_lmc_attr_groups,
+ };
+
+ cpuhp_state_add_instance_nocalls(CPUHP_AP_PERF_ARM_CVM_ONLINE,
+ &lmc->cpuhp_node);
+
+ /*
+ * perf PMU is CPU dependent so pick a random CPU and migrate away
+ * if it goes offline.
+ */
+ cpumask_set_cpu(smp_processor_id(), &lmc->active_mask);
+
+ ret = perf_pmu_register(&lmc->pmu, lmc->pmu_name, -1);
+ if (ret)
+ goto fail_hp;
+
+ list_add(&lmc->entry, &cvm_pmu_lmcs);
+
+ lmc->event_valid = cvm_pmu_lmc_event_valid;
+ dev_info(&pdev->dev, "Enabled %s PMU with %d counters\n",
+ lmc->pmu_name, lmc->num_counters);
+ return 0;
+
+fail_hp:
+ kfree(lmc->pmu_name);
+ cpuhp_state_remove_instance(CPUHP_AP_PERF_ARM_CVM_ONLINE,
+ &lmc->cpuhp_node);
+ iounmap(lmc->map);
+ kfree(lmc);
+fail_nomem:
+ return ret;
+}
+
+static int __init cvm_pmu_init(void)
+{
+ unsigned long implementor = read_cpuid_implementor();
+ unsigned int vendor_id = PCI_VENDOR_ID_CAVIUM;
+ struct pci_dev *pdev = NULL;
+ int rc;
+
+ if (implementor != ARM_CPU_IMP_CAVIUM)
+ return -ENODEV;
+
+ INIT_LIST_HEAD(&cvm_pmu_lmcs);
+
+ rc = cpuhp_setup_state_multi(CPUHP_AP_PERF_ARM_CVM_ONLINE,
+ "perf/arm/cvm:online", NULL,
+ cvm_pmu_offline_cpu);
+
+ /* detect LMC devices */
+ while ((pdev = pci_get_device(vendor_id, 0xa022, pdev))) {
+ if (!pdev)
+ break;
+ rc = cvm_pmu_lmc_probe(pdev);
+ if (rc)
+ return rc;
+ }
+ return 0;
+}
+late_initcall(cvm_pmu_init);
diff --git a/include/linux/cpuhotplug.h b/include/linux/cpuhotplug.h
index b56573b..78ac3d2 100644
--- a/include/linux/cpuhotplug.h
+++ b/include/linux/cpuhotplug.h
@@ -141,6 +141,7 @@ enum cpuhp_state {
CPUHP_AP_PERF_ARM_QCOM_L3_ONLINE,
CPUHP_AP_WORKQUEUE_ONLINE,
CPUHP_AP_RCUTREE_ONLINE,
+ CPUHP_AP_PERF_ARM_CVM_ONLINE,
CPUHP_AP_ONLINE_DYN,
CPUHP_AP_ONLINE_DYN_END = CPUHP_AP_ONLINE_DYN + 30,
CPUHP_AP_X86_HPET_ONLINE,
--
2.9.0.rc0.21.g7777322
Add support for the transmit-link (OCX TLK) PMU counters found
on Caviums SOCs with a processor interconnect.
Properties of the OCX TLK counters:
- per-unit control
- fixed purpose
- writable
- one PCI device with multiple TLK units
Signed-off-by: Jan Glauber <[email protected]>
---
drivers/perf/cavium_pmu.c | 216 ++++++++++++++++++++++++++++++++++++++++++++++
1 file changed, 216 insertions(+)
diff --git a/drivers/perf/cavium_pmu.c b/drivers/perf/cavium_pmu.c
index 7d22493..e14a93d 100644
--- a/drivers/perf/cavium_pmu.c
+++ b/drivers/perf/cavium_pmu.c
@@ -19,6 +19,7 @@
enum cvm_pmu_type {
CVM_PMU_LMC,
+ CVM_PMU_TLK,
};
/* maximum number of parallel hardware counters for all pmu types */
@@ -39,6 +40,7 @@ struct cvm_pmu_dev {
};
static struct list_head cvm_pmu_lmcs;
+static struct list_head cvm_pmu_tlks;
/*
* Common Cavium PMU stuff
@@ -387,6 +389,210 @@ static int cvm_pmu_lmc_probe(struct pci_dev *pdev)
return ret;
}
+/*
+ * CCPI interface controller (OCX) Transmit link (TLK) counters:
+ * - per-unit control
+ * - writable
+ * - one PCI device with multiple TLK units
+ */
+
+#define TLK_NR_UNITS 3
+#define TLK_UNIT_OFFSET 0x2000
+#define TLK_UNIT_LEN 0x7ff
+#define TLK_START_ADDR 0x10000
+#define TLK_STAT_CTL_OFFSET 0x40
+#define TLK_STAT_OFFSET 0x400
+
+#define TLK_STAT_ENABLE_BIT BIT(0)
+#define TLK_STAT_RESET_BIT BIT(1)
+
+#define CVM_PMU_TLK_EVENT_ATTR(_name, _id) \
+ &((struct perf_pmu_events_attr[]) { \
+ { \
+ __ATTR(_name, S_IRUGO, cvm_pmu_event_sysfs_show, NULL), \
+ _id, \
+ "tlk_event=" __stringify(_id), \
+ } \
+ })[0].attr.attr
+
+static void cvm_pmu_tlk_enable_pmu(struct pmu *pmu)
+{
+ struct cvm_pmu_dev *pmu_dev = container_of(pmu, struct cvm_pmu_dev, pmu);
+
+ /* enable all counters */
+ writeb(TLK_STAT_ENABLE_BIT, pmu_dev->map + TLK_STAT_CTL_OFFSET);
+}
+
+static void cvm_pmu_tlk_disable_pmu(struct pmu *pmu)
+{
+ struct cvm_pmu_dev *pmu_dev = container_of(pmu, struct cvm_pmu_dev, pmu);
+
+ /* disable all counters */
+ writeb(0, pmu_dev->map + TLK_STAT_CTL_OFFSET);
+}
+
+static int cvm_pmu_tlk_add(struct perf_event *event, int flags)
+{
+ struct hw_perf_event *hwc = &event->hw;
+
+ return cvm_pmu_add(event, flags, TLK_STAT_CTL_OFFSET,
+ TLK_STAT_OFFSET + hwc->config * 8);
+}
+
+PMU_FORMAT_ATTR(tlk_event, "config:0-5");
+
+static struct attribute *cvm_pmu_tlk_format_attr[] = {
+ &format_attr_tlk_event.attr,
+ NULL,
+};
+
+static struct attribute_group cvm_pmu_tlk_format_group = {
+ .name = "format",
+ .attrs = cvm_pmu_tlk_format_attr,
+};
+
+static struct attribute *cvm_pmu_tlk_events_attr[] = {
+ CVM_PMU_TLK_EVENT_ATTR(idle_cnt, 0x00),
+ CVM_PMU_TLK_EVENT_ATTR(data_cnt, 0x01),
+ CVM_PMU_TLK_EVENT_ATTR(sync_cnt, 0x02),
+ CVM_PMU_TLK_EVENT_ATTR(retry_cnt, 0x03),
+ CVM_PMU_TLK_EVENT_ATTR(err_cnt, 0x04),
+ CVM_PMU_TLK_EVENT_ATTR(mat0_cnt, 0x08),
+ CVM_PMU_TLK_EVENT_ATTR(mat1_cnt, 0x09),
+ CVM_PMU_TLK_EVENT_ATTR(mat2_cnt, 0x0a),
+ CVM_PMU_TLK_EVENT_ATTR(mat3_cnt, 0x0b),
+ CVM_PMU_TLK_EVENT_ATTR(vc0_cmd, 0x10),
+ CVM_PMU_TLK_EVENT_ATTR(vc1_cmd, 0x11),
+ CVM_PMU_TLK_EVENT_ATTR(vc2_cmd, 0x12),
+ CVM_PMU_TLK_EVENT_ATTR(vc3_cmd, 0x13),
+ CVM_PMU_TLK_EVENT_ATTR(vc4_cmd, 0x14),
+ CVM_PMU_TLK_EVENT_ATTR(vc5_cmd, 0x15),
+ CVM_PMU_TLK_EVENT_ATTR(vc0_pkt, 0x20),
+ CVM_PMU_TLK_EVENT_ATTR(vc1_pkt, 0x21),
+ CVM_PMU_TLK_EVENT_ATTR(vc2_pkt, 0x22),
+ CVM_PMU_TLK_EVENT_ATTR(vc3_pkt, 0x23),
+ CVM_PMU_TLK_EVENT_ATTR(vc4_pkt, 0x24),
+ CVM_PMU_TLK_EVENT_ATTR(vc5_pkt, 0x25),
+ CVM_PMU_TLK_EVENT_ATTR(vc6_pkt, 0x26),
+ CVM_PMU_TLK_EVENT_ATTR(vc7_pkt, 0x27),
+ CVM_PMU_TLK_EVENT_ATTR(vc8_pkt, 0x28),
+ CVM_PMU_TLK_EVENT_ATTR(vc9_pkt, 0x29),
+ CVM_PMU_TLK_EVENT_ATTR(vc10_pkt, 0x2a),
+ CVM_PMU_TLK_EVENT_ATTR(vc11_pkt, 0x2b),
+ CVM_PMU_TLK_EVENT_ATTR(vc12_pkt, 0x2c),
+ CVM_PMU_TLK_EVENT_ATTR(vc13_pkt, 0x2d),
+ CVM_PMU_TLK_EVENT_ATTR(vc0_con, 0x30),
+ CVM_PMU_TLK_EVENT_ATTR(vc1_con, 0x31),
+ CVM_PMU_TLK_EVENT_ATTR(vc2_con, 0x32),
+ CVM_PMU_TLK_EVENT_ATTR(vc3_con, 0x33),
+ CVM_PMU_TLK_EVENT_ATTR(vc4_con, 0x34),
+ CVM_PMU_TLK_EVENT_ATTR(vc5_con, 0x35),
+ CVM_PMU_TLK_EVENT_ATTR(vc6_con, 0x36),
+ CVM_PMU_TLK_EVENT_ATTR(vc7_con, 0x37),
+ CVM_PMU_TLK_EVENT_ATTR(vc8_con, 0x38),
+ CVM_PMU_TLK_EVENT_ATTR(vc9_con, 0x39),
+ CVM_PMU_TLK_EVENT_ATTR(vc10_con, 0x3a),
+ CVM_PMU_TLK_EVENT_ATTR(vc11_con, 0x3b),
+ CVM_PMU_TLK_EVENT_ATTR(vc12_con, 0x3c),
+ CVM_PMU_TLK_EVENT_ATTR(vc13_con, 0x3d),
+ NULL,
+};
+
+static struct attribute_group cvm_pmu_tlk_events_group = {
+ .name = "events",
+ .attrs = cvm_pmu_tlk_events_attr,
+};
+
+static const struct attribute_group *cvm_pmu_tlk_attr_groups[] = {
+ &cvm_pmu_attr_group,
+ &cvm_pmu_tlk_format_group,
+ &cvm_pmu_tlk_events_group,
+ NULL,
+};
+
+static bool cvm_pmu_tlk_event_valid(u64 config)
+{
+ struct perf_pmu_events_attr *attr;
+ int i;
+
+ for (i = 0; i < ARRAY_SIZE(cvm_pmu_tlk_events_attr) - 1; i++) {
+ attr = (struct perf_pmu_events_attr *)cvm_pmu_tlk_events_attr[i];
+ if (attr->id == config)
+ return true;
+ }
+ return false;
+}
+
+static int cvm_pmu_tlk_probe_unit(struct pci_dev *pdev, int nr)
+{
+ struct cvm_pmu_dev *tlk;
+ int ret = -ENOMEM;
+
+ tlk = kzalloc(sizeof(*tlk), GFP_KERNEL);
+ if (!tlk)
+ goto fail_nomem;
+
+ tlk->map = ioremap(pci_resource_start(pdev, 0) + TLK_START_ADDR +
+ nr * TLK_UNIT_OFFSET, TLK_UNIT_LEN);
+ tlk->pmu_name = kasprintf(GFP_KERNEL, "ocx_tlk%d", nr);
+
+ tlk->pdev = pdev;
+ tlk->num_counters = ARRAY_SIZE(cvm_pmu_tlk_events_attr) - 1;
+ tlk->pmu = (struct pmu) {
+ .task_ctx_nr = perf_invalid_context,
+ .pmu_enable = cvm_pmu_tlk_enable_pmu,
+ .pmu_disable = cvm_pmu_tlk_disable_pmu,
+ .event_init = cvm_pmu_event_init,
+ .add = cvm_pmu_tlk_add,
+ .del = cvm_pmu_del,
+ .start = cvm_pmu_start,
+ .stop = cvm_pmu_stop,
+ .read = cvm_pmu_read,
+ .attr_groups = cvm_pmu_tlk_attr_groups,
+ };
+
+ cpuhp_state_add_instance_nocalls(CPUHP_AP_PERF_ARM_CVM_ONLINE,
+ &tlk->cpuhp_node);
+
+ /*
+ * perf PMU is CPU dependent so pick a random CPU and migrate away
+ * if it goes offline.
+ */
+ cpumask_set_cpu(smp_processor_id(), &tlk->active_mask);
+
+ ret = perf_pmu_register(&tlk->pmu, tlk->pmu_name, -1);
+ if (ret)
+ goto fail_hp;
+
+ list_add(&tlk->entry, &cvm_pmu_tlks);
+
+ tlk->event_valid = cvm_pmu_tlk_event_valid;
+ dev_info(&pdev->dev, "Enabled %s PMU with %d counters\n",
+ tlk->pmu_name, tlk->num_counters);
+ return 0;
+
+fail_hp:
+ kfree(tlk->pmu_name);
+ cpuhp_state_remove_instance(CPUHP_AP_PERF_ARM_CVM_ONLINE,
+ &tlk->cpuhp_node);
+ iounmap(tlk->map);
+ kfree(tlk);
+fail_nomem:
+ return ret;
+}
+
+static int cvm_pmu_tlk_probe(struct pci_dev *pdev)
+{
+ int rc, i;
+
+ for (i = 0; i < TLK_NR_UNITS; i++) {
+ rc = cvm_pmu_tlk_probe_unit(pdev, i);
+ if (rc)
+ return rc;
+ }
+ return 0;
+}
+
static int __init cvm_pmu_init(void)
{
unsigned long implementor = read_cpuid_implementor();
@@ -398,6 +604,7 @@ static int __init cvm_pmu_init(void)
return -ENODEV;
INIT_LIST_HEAD(&cvm_pmu_lmcs);
+ INIT_LIST_HEAD(&cvm_pmu_tlks);
rc = cpuhp_setup_state_multi(CPUHP_AP_PERF_ARM_CVM_ONLINE,
"perf/arm/cvm:online", NULL,
@@ -411,6 +618,15 @@ static int __init cvm_pmu_init(void)
if (rc)
return rc;
}
+
+ /* detect OCX TLK devices */
+ while ((pdev = pci_get_device(vendor_id, 0xa013, pdev))) {
+ if (!pdev)
+ break;
+ rc = cvm_pmu_tlk_probe(pdev);
+ if (rc)
+ return rc;
+ }
return 0;
}
late_initcall(cvm_pmu_init);
--
2.9.0.rc0.21.g7777322
Document Cavium SoC PMUs.
Signed-off-by: Jan Glauber <[email protected]>
---
Documentation/perf/cavium-pmu.txt | 74 +++++++++++++++++++++++++++++++++++++++
1 file changed, 74 insertions(+)
create mode 100644 Documentation/perf/cavium-pmu.txt
diff --git a/Documentation/perf/cavium-pmu.txt b/Documentation/perf/cavium-pmu.txt
new file mode 100644
index 0000000..37669b9
--- /dev/null
+++ b/Documentation/perf/cavium-pmu.txt
@@ -0,0 +1,74 @@
+Cavium ThunderX and OcteonTx Performance Monitoring Unit (PMU)
+==============================================================
+
+Cavium SoCs contain various system devices such as L2 caches, processor
+interconnect and memory controllers. Unfortunately the PMU counters
+are not following a common design so each device has a slightly different
+approach how to control and use the PMU counters.
+
+Common properties of all devices carrying PMU counters:
+- The devices are PCI devices and the counters are embedded somewhere
+ in the PCI register space.
+- All counters are 64 bit wide.
+- There are no overflow interrupts (unnecessary because of the 64 bit wide counters).
+
+Properties depending on the device type:
+- How to start/stop the counters
+- Programmable vs. fixed purpose counters
+- Stoppable vs. always running counters
+- Independent vs. grouped counters
+- Read-only vs. writable counters
+- PCI device to PMU group relationship
+
+
+Devices with PMU counters
+-------------------------
+
+Memory controller (LMC):
+- one PCI device per LMC
+- fixed-purpose counters
+- always running counters without start/stop/reset control
+- read-only counters
+
+CCPI interface controller (OCX) Transmit link (TLK) counters:
+- writable counters
+- only one PCI device exposes multiple TLK units (3 units on T88)
+- start/stop control per unit
+- only present on multi-socket systems
+
+PMU (perf) driver
+-----------------
+
+The cavium-pmu driver registers several perf PMU drivers. Each of the perf
+driver provides description of its available events and configuration options
+in sysfs, see /sys/devices/<lmcX/ocx_tlkX>/.
+
+The "format" directory describes format of the config (event ID),
+The "events" directory shows the names of the events and provides configuration
+templates for all supported event types that can be used with perf tool. For
+example, "lmc0/dclk_cnt/" is an equivalent of "lmc0/config=2/".
+
+Each perf driver also provides a "cpumask" sysfs attribute, which contains a
+single CPU ID of the processor which will be used to handle all the PMU events.
+
+Example for perf tool use:
+
+ / # perf list | grep -e lmc
+ lmc0/bank_conflict1/ [Kernel PMU event]
+ lmc0/bank_conflict2/ [Kernel PMU event]
+ lmc0/dclk_cnt/ [Kernel PMU event]
+ lmc0/ifb_cnt/ [Kernel PMU event]
+ lmc0/ops_cnt/ [Kernel PMU event]
+
+ / # perf stat -a -e lmc0/ops_cnt/,lmc0/dclk_cnt/ -- sleep 1
+
+ Performance counter stats for 'system wide':
+
+ 176,133 lmc0/ops_cnt/
+ 670,243,653 lmc0/dclk_cnt/
+
+ 1.005479295 seconds time elapsed
+
+The driver does not support sampling, therefore "perf record" will
+not work. System wide mode ("-a") must be used as per-task (without "-a")
+perf sessions are not supported.
--
2.9.0.rc0.21.g7777322
On Wed, 19 Jul 2017 14:08:45 +0200
Jan Glauber <[email protected]> wrote:
> Add support for the PMU counters on Cavium SOC memory controllers.
>
> This patch also adds generic functions to allow supporting more
> devices with PMU counters.
>
> Properties of the LMC PMU counters:
> - not stoppable
> - fixed purpose
> - read-only
> - one PCI device per memory controller
>
> Signed-off-by: Jan Glauber <[email protected]>
Hi Jan,
A few little things I noticed whilst taking a quick look.
Jonathan
> ---
> drivers/perf/Kconfig | 8 +
> drivers/perf/Makefile | 1 +
> drivers/perf/cavium_pmu.c | 416 +++++++++++++++++++++++++++++++++++++++++++++
> include/linux/cpuhotplug.h | 1 +
> 4 files changed, 426 insertions(+)
> create mode 100644 drivers/perf/cavium_pmu.c
<snip>
> +
> +/*
> + * The pmu events are independent from CPUs. Provide a cpumask
> + * nevertheless to prevent perf from adding the event per-cpu and just
> + * set the mask to one online CPU. Use the same cpumask for all "uncore"
> + * devices.
> + *
> + * There is a performance penalty for accessing a device from a CPU on
> + * another socket, but we do not care.
> + */
> +static int cvm_pmu_offline_cpu(unsigned int old_cpu, struct hlist_node *node)
> +{
> + struct cvm_pmu_dev *pmu_dev;
> + int new_cpu;
> +
> + pmu_dev = hlist_entry_safe(node, struct cvm_pmu_dev, cpuhp_node);
> + if (!cpumask_test_and_clear_cpu(old_cpu, &pmu_dev->active_mask))
> + return 0;
> +
> + new_cpu = cpumask_any_but(cpu_online_mask, old_cpu);
> + if (new_cpu >= nr_cpu_ids)
> + return 0;
Blank line.
> + perf_pmu_migrate_context(&pmu_dev->pmu, old_cpu, new_cpu);
> + cpumask_set_cpu(new_cpu, &pmu_dev->active_mask);
nitpick : blank line here would help readability.
> + return 0;
> +}
> +
> +static ssize_t cvm_pmu_attr_show_cpumask(struct device *dev,
> + struct device_attribute *attr,
> + char *buf)
> +{
> + struct pmu *pmu = dev_get_drvdata(dev);
> + struct cvm_pmu_dev *pmu_dev = container_of(pmu, struct cvm_pmu_dev, pmu);
> +
> + return cpumap_print_to_pagebuf(true, buf, &pmu_dev->active_mask);
> +}
> +
> +static DEVICE_ATTR(cpumask, S_IRUGO, cvm_pmu_attr_show_cpumask, NULL);
> +
> +static struct attribute *cvm_pmu_attrs[] = {
> + &dev_attr_cpumask.attr,
> + NULL,
> +};
> +
> +static struct attribute_group cvm_pmu_attr_group = {
> + .attrs = cvm_pmu_attrs,
> +};
> +
> +/*
> + * LMC (memory controller) counters:
> + * - not stoppable, always on, read-only
> + * - one PCI device per memory controller
> + */
> +#define LMC_CONFIG_OFFSET 0x188
> +#define LMC_CONFIG_RESET_BIT BIT(17)
> +
> +/* LMC events */
> +#define LMC_EVENT_IFB_CNT 0x1d0
> +#define LMC_EVENT_OPS_CNT 0x1d8
> +#define LMC_EVENT_DCLK_CNT 0x1e0
> +#define LMC_EVENT_BANK_CONFLICT1 0x360
> +#define LMC_EVENT_BANK_CONFLICT2 0x368
> +
> +#define CVM_PMU_LMC_EVENT_ATTR(_name, _id) \
> + &((struct perf_pmu_events_attr[]) { \
> + { \
> + __ATTR(_name, S_IRUGO, cvm_pmu_event_sysfs_show, NULL), \
> + _id, \
> + "lmc_event=" __stringify(_id), \
> + } \
> + })[0].attr.attr
> +
> +/* map counter numbers to register offsets */
> +static int lmc_events[] = {
> + LMC_EVENT_IFB_CNT,
> + LMC_EVENT_OPS_CNT,
> + LMC_EVENT_DCLK_CNT,
> + LMC_EVENT_BANK_CONFLICT1,
> + LMC_EVENT_BANK_CONFLICT2,
> +};
> +
> +static int cvm_pmu_lmc_add(struct perf_event *event, int flags)
> +{
> + struct hw_perf_event *hwc = &event->hw;
> +
> + return cvm_pmu_add(event, flags, LMC_CONFIG_OFFSET,
> + lmc_events[hwc->config]);
> +}
> +
> +PMU_FORMAT_ATTR(lmc_event, "config:0-2");
> +
> +static struct attribute *cvm_pmu_lmc_format_attr[] = {
> + &format_attr_lmc_event.attr,
> + NULL,
> +};
> +
> +static struct attribute_group cvm_pmu_lmc_format_group = {
> + .name = "format",
> + .attrs = cvm_pmu_lmc_format_attr,
> +};
> +
> +static struct attribute *cvm_pmu_lmc_events_attr[] = {
> + CVM_PMU_LMC_EVENT_ATTR(ifb_cnt, 0),
> + CVM_PMU_LMC_EVENT_ATTR(ops_cnt, 1),
> + CVM_PMU_LMC_EVENT_ATTR(dclk_cnt, 2),
> + CVM_PMU_LMC_EVENT_ATTR(bank_conflict1, 3),
> + CVM_PMU_LMC_EVENT_ATTR(bank_conflict2, 4),
> + NULL,
> +};
> +
> +static struct attribute_group cvm_pmu_lmc_events_group = {
> + .name = "events",
> + .attrs = cvm_pmu_lmc_events_attr,
> +};
> +
> +static const struct attribute_group *cvm_pmu_lmc_attr_groups[] = {
> + &cvm_pmu_attr_group,
> + &cvm_pmu_lmc_format_group,
> + &cvm_pmu_lmc_events_group,
> + NULL,
> +};
> +
> +static bool cvm_pmu_lmc_event_valid(u64 config)
> +{
> + return (config < ARRAY_SIZE(lmc_events));
> +}
> +
> +static int cvm_pmu_lmc_probe(struct pci_dev *pdev)
> +{
> + struct cvm_pmu_dev *next, *lmc;
> + int nr = 0, ret = -ENOMEM;
> +
> + lmc = kzalloc(sizeof(*lmc), GFP_KERNEL);
> + if (!lmc)
> + goto fail_nomem;
cleaner to do a direct return here as nothing to unwind.
return -ENOMEM;
> +
> + lmc->map = ioremap(pci_resource_start(pdev, 0),
> + pci_resource_len(pdev, 0));
Might be overly paranoid, but ioremap can return NULL,
so should really be checked.
> + list_for_each_entry(next, &cvm_pmu_lmcs, entry)
> + nr++;
> + lmc->pmu_name = kasprintf(GFP_KERNEL, "lmc%d", nr);
Check for error on the allocation.
> +
> + lmc->pdev = pdev;
> + lmc->num_counters = ARRAY_SIZE(lmc_events);
> + lmc->pmu = (struct pmu) {
> + .task_ctx_nr = perf_invalid_context,
> + .event_init = cvm_pmu_event_init,
> + .add = cvm_pmu_lmc_add,
> + .del = cvm_pmu_del,
> + .start = cvm_pmu_start,
> + .stop = cvm_pmu_stop,
> + .read = cvm_pmu_read,
> + .attr_groups = cvm_pmu_lmc_attr_groups,
> + };
> +
> + cpuhp_state_add_instance_nocalls(CPUHP_AP_PERF_ARM_CVM_ONLINE,
> + &lmc->cpuhp_node);
This function can fail by the look of it. Should be checking it's
return value really.
> +
> + /*
> + * perf PMU is CPU dependent so pick a random CPU and migrate away
> + * if it goes offline.
> + */
> + cpumask_set_cpu(smp_processor_id(), &lmc->active_mask);
> +
> + ret = perf_pmu_register(&lmc->pmu, lmc->pmu_name, -1);
> + if (ret)
> + goto fail_hp;
> +
> + list_add(&lmc->entry, &cvm_pmu_lmcs);
> +
> + lmc->event_valid = cvm_pmu_lmc_event_valid;
These last two elements rather look like they might cause a
potential race as the perf_pmu_register has made the counters
available.
> + dev_info(&pdev->dev, "Enabled %s PMU with %d counters\n",
> + lmc->pmu_name, lmc->num_counters);
> + return 0;
> +
> +fail_hp:
> + kfree(lmc->pmu_name);
Convention is to unwind in the opposite order to the way things
were created / set up. Becomes more relevant once you are
checking more error codes.
> + cpuhp_state_remove_instance(CPUHP_AP_PERF_ARM_CVM_ONLINE,
> + &lmc->cpuhp_node);
> + iounmap(lmc->map);
> + kfree(lmc);
> +fail_nomem:
> + return ret;
> +}
> +
> +static int __init cvm_pmu_init(void)
> +{
> + unsigned long implementor = read_cpuid_implementor();
> + unsigned int vendor_id = PCI_VENDOR_ID_CAVIUM;
> + struct pci_dev *pdev = NULL;
> + int rc;
> +
> + if (implementor != ARM_CPU_IMP_CAVIUM)
> + return -ENODEV;
> +
> + INIT_LIST_HEAD(&cvm_pmu_lmcs);
> +
> + rc = cpuhp_setup_state_multi(CPUHP_AP_PERF_ARM_CVM_ONLINE,
> + "perf/arm/cvm:online", NULL,
> + cvm_pmu_offline_cpu);
> +
> + /* detect LMC devices */
> + while ((pdev = pci_get_device(vendor_id, 0xa022, pdev))) {
> + if (!pdev)
> + break;
> + rc = cvm_pmu_lmc_probe(pdev);
> + if (rc)
> + return rc;
> + }
> + return 0;
> +}
A comment perhaps explaining why this is a late_initcall? What
dependency are we forcing to be true?
> +late_initcall(cvm_pmu_init);
> diff --git a/include/linux/cpuhotplug.h b/include/linux/cpuhotplug.h
> index b56573b..78ac3d2 100644
> --- a/include/linux/cpuhotplug.h
> +++ b/include/linux/cpuhotplug.h
> @@ -141,6 +141,7 @@ enum cpuhp_state {
> CPUHP_AP_PERF_ARM_QCOM_L3_ONLINE,
> CPUHP_AP_WORKQUEUE_ONLINE,
> CPUHP_AP_RCUTREE_ONLINE,
> + CPUHP_AP_PERF_ARM_CVM_ONLINE,
> CPUHP_AP_ONLINE_DYN,
> CPUHP_AP_ONLINE_DYN_END = CPUHP_AP_ONLINE_DYN + 30,
> CPUHP_AP_X86_HPET_ONLINE,
Hi Jonathan,
On Wed, Jul 19, 2017 at 09:31:01PM +0800, Jonathan Cameron wrote:
> On Wed, 19 Jul 2017 14:08:45 +0200
> Jan Glauber <[email protected]> wrote:
>
> > Add support for the PMU counters on Cavium SOC memory controllers.
> >
> > This patch also adds generic functions to allow supporting more
> > devices with PMU counters.
> >
> > Properties of the LMC PMU counters:
> > - not stoppable
> > - fixed purpose
> > - read-only
> > - one PCI device per memory controller
> >
> > Signed-off-by: Jan Glauber <[email protected]>
> Hi Jan,
>
> A few little things I noticed whilst taking a quick look.
>
> Jonathan
> > ---
> > drivers/perf/Kconfig | 8 +
> > drivers/perf/Makefile | 1 +
> > drivers/perf/cavium_pmu.c | 416 +++++++++++++++++++++++++++++++++++++++++++++
> > include/linux/cpuhotplug.h | 1 +
> > 4 files changed, 426 insertions(+)
> > create mode 100644 drivers/perf/cavium_pmu.c
> <snip>
> > +
> > +/*
> > + * The pmu events are independent from CPUs. Provide a cpumask
> > + * nevertheless to prevent perf from adding the event per-cpu and just
> > + * set the mask to one online CPU. Use the same cpumask for all "uncore"
> > + * devices.
> > + *
> > + * There is a performance penalty for accessing a device from a CPU on
> > + * another socket, but we do not care.
> > + */
> > +static int cvm_pmu_offline_cpu(unsigned int old_cpu, struct hlist_node *node)
> > +{
> > + struct cvm_pmu_dev *pmu_dev;
> > + int new_cpu;
> > +
> > + pmu_dev = hlist_entry_safe(node, struct cvm_pmu_dev, cpuhp_node);
> > + if (!cpumask_test_and_clear_cpu(old_cpu, &pmu_dev->active_mask))
> > + return 0;
> > +
> > + new_cpu = cpumask_any_but(cpu_online_mask, old_cpu);
> > + if (new_cpu >= nr_cpu_ids)
> > + return 0;
> Blank line.
OK
> > + perf_pmu_migrate_context(&pmu_dev->pmu, old_cpu, new_cpu);
> > + cpumask_set_cpu(new_cpu, &pmu_dev->active_mask);
> nitpick : blank line here would help readability.
OK
> > + return 0;
> > +}
> > +
> > +static ssize_t cvm_pmu_attr_show_cpumask(struct device *dev,
> > + struct device_attribute *attr,
> > + char *buf)
> > +{
> > + struct pmu *pmu = dev_get_drvdata(dev);
> > + struct cvm_pmu_dev *pmu_dev = container_of(pmu, struct cvm_pmu_dev, pmu);
> > +
> > + return cpumap_print_to_pagebuf(true, buf, &pmu_dev->active_mask);
> > +}
> > +
> > +static DEVICE_ATTR(cpumask, S_IRUGO, cvm_pmu_attr_show_cpumask, NULL);
> > +
> > +static struct attribute *cvm_pmu_attrs[] = {
> > + &dev_attr_cpumask.attr,
> > + NULL,
> > +};
> > +
> > +static struct attribute_group cvm_pmu_attr_group = {
> > + .attrs = cvm_pmu_attrs,
> > +};
> > +
> > +/*
> > + * LMC (memory controller) counters:
> > + * - not stoppable, always on, read-only
> > + * - one PCI device per memory controller
> > + */
> > +#define LMC_CONFIG_OFFSET 0x188
> > +#define LMC_CONFIG_RESET_BIT BIT(17)
> > +
> > +/* LMC events */
> > +#define LMC_EVENT_IFB_CNT 0x1d0
> > +#define LMC_EVENT_OPS_CNT 0x1d8
> > +#define LMC_EVENT_DCLK_CNT 0x1e0
> > +#define LMC_EVENT_BANK_CONFLICT1 0x360
> > +#define LMC_EVENT_BANK_CONFLICT2 0x368
> > +
> > +#define CVM_PMU_LMC_EVENT_ATTR(_name, _id) \
> > + &((struct perf_pmu_events_attr[]) { \
> > + { \
> > + __ATTR(_name, S_IRUGO, cvm_pmu_event_sysfs_show, NULL), \
> > + _id, \
> > + "lmc_event=" __stringify(_id), \
> > + } \
> > + })[0].attr.attr
> > +
> > +/* map counter numbers to register offsets */
> > +static int lmc_events[] = {
> > + LMC_EVENT_IFB_CNT,
> > + LMC_EVENT_OPS_CNT,
> > + LMC_EVENT_DCLK_CNT,
> > + LMC_EVENT_BANK_CONFLICT1,
> > + LMC_EVENT_BANK_CONFLICT2,
> > +};
> > +
> > +static int cvm_pmu_lmc_add(struct perf_event *event, int flags)
> > +{
> > + struct hw_perf_event *hwc = &event->hw;
> > +
> > + return cvm_pmu_add(event, flags, LMC_CONFIG_OFFSET,
> > + lmc_events[hwc->config]);
> > +}
> > +
> > +PMU_FORMAT_ATTR(lmc_event, "config:0-2");
> > +
> > +static struct attribute *cvm_pmu_lmc_format_attr[] = {
> > + &format_attr_lmc_event.attr,
> > + NULL,
> > +};
> > +
> > +static struct attribute_group cvm_pmu_lmc_format_group = {
> > + .name = "format",
> > + .attrs = cvm_pmu_lmc_format_attr,
> > +};
> > +
> > +static struct attribute *cvm_pmu_lmc_events_attr[] = {
> > + CVM_PMU_LMC_EVENT_ATTR(ifb_cnt, 0),
> > + CVM_PMU_LMC_EVENT_ATTR(ops_cnt, 1),
> > + CVM_PMU_LMC_EVENT_ATTR(dclk_cnt, 2),
> > + CVM_PMU_LMC_EVENT_ATTR(bank_conflict1, 3),
> > + CVM_PMU_LMC_EVENT_ATTR(bank_conflict2, 4),
> > + NULL,
> > +};
> > +
> > +static struct attribute_group cvm_pmu_lmc_events_group = {
> > + .name = "events",
> > + .attrs = cvm_pmu_lmc_events_attr,
> > +};
> > +
> > +static const struct attribute_group *cvm_pmu_lmc_attr_groups[] = {
> > + &cvm_pmu_attr_group,
> > + &cvm_pmu_lmc_format_group,
> > + &cvm_pmu_lmc_events_group,
> > + NULL,
> > +};
> > +
> > +static bool cvm_pmu_lmc_event_valid(u64 config)
> > +{
> > + return (config < ARRAY_SIZE(lmc_events));
> > +}
> > +
> > +static int cvm_pmu_lmc_probe(struct pci_dev *pdev)
> > +{
> > + struct cvm_pmu_dev *next, *lmc;
> > + int nr = 0, ret = -ENOMEM;
> > +
> > + lmc = kzalloc(sizeof(*lmc), GFP_KERNEL);
> > + if (!lmc)
> > + goto fail_nomem;
> cleaner to do a direct return here as nothing to unwind.
> return -ENOMEM;
OK.
> > +
> > + lmc->map = ioremap(pci_resource_start(pdev, 0),
> > + pci_resource_len(pdev, 0));
> Might be overly paranoid, but ioremap can return NULL,
> so should really be checked.
>
I'll add the check.
> > + list_for_each_entry(next, &cvm_pmu_lmcs, entry)
> > + nr++;
> > + lmc->pmu_name = kasprintf(GFP_KERNEL, "lmc%d", nr);
> Check for error on the allocation.
OK, missed that.
> > +
> > + lmc->pdev = pdev;
> > + lmc->num_counters = ARRAY_SIZE(lmc_events);
> > + lmc->pmu = (struct pmu) {
> > + .task_ctx_nr = perf_invalid_context,
> > + .event_init = cvm_pmu_event_init,
> > + .add = cvm_pmu_lmc_add,
> > + .del = cvm_pmu_del,
> > + .start = cvm_pmu_start,
> > + .stop = cvm_pmu_stop,
> > + .read = cvm_pmu_read,
> > + .attr_groups = cvm_pmu_lmc_attr_groups,
> > + };
> > +
> > + cpuhp_state_add_instance_nocalls(CPUHP_AP_PERF_ARM_CVM_ONLINE,
> > + &lmc->cpuhp_node);
> This function can fail by the look of it. Should be checking it's
> return value really.
>
I'm ignoring the return value on purpose as I don't want probe() to fail
even if this would return an error.
> > +
> > + /*
> > + * perf PMU is CPU dependent so pick a random CPU and migrate away
> > + * if it goes offline.
> > + */
> > + cpumask_set_cpu(smp_processor_id(), &lmc->active_mask);
> > +
> > + ret = perf_pmu_register(&lmc->pmu, lmc->pmu_name, -1);
> > + if (ret)
> > + goto fail_hp;
> > +
> > + list_add(&lmc->entry, &cvm_pmu_lmcs);
> > +
> > + lmc->event_valid = cvm_pmu_lmc_event_valid;
> These last two elements rather look like they might cause a
> potential race as the perf_pmu_register has made the counters
> available.
Yes, they should happen before perf_pmu_register.
> > + dev_info(&pdev->dev, "Enabled %s PMU with %d counters\n",
> > + lmc->pmu_name, lmc->num_counters);
> > + return 0;
> > +
> > +fail_hp:
> > + kfree(lmc->pmu_name);
> Convention is to unwind in the opposite order to the way things
> were created / set up. Becomes more relevant once you are
> checking more error codes.
Agreed.
> > + cpuhp_state_remove_instance(CPUHP_AP_PERF_ARM_CVM_ONLINE,
> > + &lmc->cpuhp_node);
> > + iounmap(lmc->map);
> > + kfree(lmc);
> > +fail_nomem:
> > + return ret;
> > +}
> > +
> > +static int __init cvm_pmu_init(void)
> > +{
> > + unsigned long implementor = read_cpuid_implementor();
> > + unsigned int vendor_id = PCI_VENDOR_ID_CAVIUM;
> > + struct pci_dev *pdev = NULL;
> > + int rc;
> > +
> > + if (implementor != ARM_CPU_IMP_CAVIUM)
> > + return -ENODEV;
> > +
> > + INIT_LIST_HEAD(&cvm_pmu_lmcs);
> > +
> > + rc = cpuhp_setup_state_multi(CPUHP_AP_PERF_ARM_CVM_ONLINE,
> > + "perf/arm/cvm:online", NULL,
> > + cvm_pmu_offline_cpu);
> > +
> > + /* detect LMC devices */
> > + while ((pdev = pci_get_device(vendor_id, 0xa022, pdev))) {
> > + if (!pdev)
> > + break;
> > + rc = cvm_pmu_lmc_probe(pdev);
> > + if (rc)
> > + return rc;
> > + }
> > + return 0;
> > +}
> A comment perhaps explaining why this is a late_initcall? What
> dependency are we forcing to be true?
PCI probing should be finished before, I'll add a comment.
thanks for the review,
Jan
> > +late_initcall(cvm_pmu_init);
> > diff --git a/include/linux/cpuhotplug.h b/include/linux/cpuhotplug.h
> > index b56573b..78ac3d2 100644
> > --- a/include/linux/cpuhotplug.h
> > +++ b/include/linux/cpuhotplug.h
> > @@ -141,6 +141,7 @@ enum cpuhp_state {
> > CPUHP_AP_PERF_ARM_QCOM_L3_ONLINE,
> > CPUHP_AP_WORKQUEUE_ONLINE,
> > CPUHP_AP_RCUTREE_ONLINE,
> > + CPUHP_AP_PERF_ARM_CVM_ONLINE,
> > CPUHP_AP_ONLINE_DYN,
> > CPUHP_AP_ONLINE_DYN_END = CPUHP_AP_ONLINE_DYN + 30,
> > CPUHP_AP_X86_HPET_ONLINE,
>
On Mon, 24 Jul 2017 10:57:18 +0200
Jan Glauber <[email protected]> wrote:
> Hi Jonathan,
>
> On Wed, Jul 19, 2017 at 09:31:01PM +0800, Jonathan Cameron wrote:
> > On Wed, 19 Jul 2017 14:08:45 +0200
> > Jan Glauber <[email protected]> wrote:
> >
> > > Add support for the PMU counters on Cavium SOC memory controllers.
> > >
> > > This patch also adds generic functions to allow supporting more
> > > devices with PMU counters.
> > >
> > > Properties of the LMC PMU counters:
> > > - not stoppable
> > > - fixed purpose
> > > - read-only
> > > - one PCI device per memory controller
> > >
> > > Signed-off-by: Jan Glauber <[email protected]>
> > Hi Jan,
> >
> > A few little things I noticed whilst taking a quick look.
> >
> > Jonathan
> > > ---
> > > drivers/perf/Kconfig | 8 +
> > > drivers/perf/Makefile | 1 +
> > > drivers/perf/cavium_pmu.c | 416 +++++++++++++++++++++++++++++++++++++++++++++
> > > include/linux/cpuhotplug.h | 1 +
> > > 4 files changed, 426 insertions(+)
> > > create mode 100644 drivers/perf/cavium_pmu.c
> > <snip>
> > > +
> > > +/*
> > > + * The pmu events are independent from CPUs. Provide a cpumask
> > > + * nevertheless to prevent perf from adding the event per-cpu and just
> > > + * set the mask to one online CPU. Use the same cpumask for all "uncore"
> > > + * devices.
> > > + *
> > > + * There is a performance penalty for accessing a device from a CPU on
> > > + * another socket, but we do not care.
> > > + */
> > > +static int cvm_pmu_offline_cpu(unsigned int old_cpu, struct hlist_node *node)
> > > +{
> > > + struct cvm_pmu_dev *pmu_dev;
> > > + int new_cpu;
> > > +
> > > + pmu_dev = hlist_entry_safe(node, struct cvm_pmu_dev, cpuhp_node);
> > > + if (!cpumask_test_and_clear_cpu(old_cpu, &pmu_dev->active_mask))
> > > + return 0;
> > > +
> > > + new_cpu = cpumask_any_but(cpu_online_mask, old_cpu);
> > > + if (new_cpu >= nr_cpu_ids)
> > > + return 0;
> > Blank line.
>
> OK
>
> > > + perf_pmu_migrate_context(&pmu_dev->pmu, old_cpu, new_cpu);
> > > + cpumask_set_cpu(new_cpu, &pmu_dev->active_mask);
> > nitpick : blank line here would help readability.
>
> OK
>
> > > + return 0;
> > > +}
> > > +
> > > +static ssize_t cvm_pmu_attr_show_cpumask(struct device *dev,
> > > + struct device_attribute *attr,
> > > + char *buf)
> > > +{
> > > + struct pmu *pmu = dev_get_drvdata(dev);
> > > + struct cvm_pmu_dev *pmu_dev = container_of(pmu, struct cvm_pmu_dev, pmu);
> > > +
> > > + return cpumap_print_to_pagebuf(true, buf, &pmu_dev->active_mask);
> > > +}
> > > +
> > > +static DEVICE_ATTR(cpumask, S_IRUGO, cvm_pmu_attr_show_cpumask, NULL);
> > > +
> > > +static struct attribute *cvm_pmu_attrs[] = {
> > > + &dev_attr_cpumask.attr,
> > > + NULL,
> > > +};
> > > +
> > > +static struct attribute_group cvm_pmu_attr_group = {
> > > + .attrs = cvm_pmu_attrs,
> > > +};
> > > +
> > > +/*
> > > + * LMC (memory controller) counters:
> > > + * - not stoppable, always on, read-only
> > > + * - one PCI device per memory controller
> > > + */
> > > +#define LMC_CONFIG_OFFSET 0x188
> > > +#define LMC_CONFIG_RESET_BIT BIT(17)
> > > +
> > > +/* LMC events */
> > > +#define LMC_EVENT_IFB_CNT 0x1d0
> > > +#define LMC_EVENT_OPS_CNT 0x1d8
> > > +#define LMC_EVENT_DCLK_CNT 0x1e0
> > > +#define LMC_EVENT_BANK_CONFLICT1 0x360
> > > +#define LMC_EVENT_BANK_CONFLICT2 0x368
> > > +
> > > +#define CVM_PMU_LMC_EVENT_ATTR(_name, _id) \
> > > + &((struct perf_pmu_events_attr[]) { \
> > > + { \
> > > + __ATTR(_name, S_IRUGO, cvm_pmu_event_sysfs_show, NULL), \
> > > + _id, \
> > > + "lmc_event=" __stringify(_id), \
> > > + } \
> > > + })[0].attr.attr
> > > +
> > > +/* map counter numbers to register offsets */
> > > +static int lmc_events[] = {
> > > + LMC_EVENT_IFB_CNT,
> > > + LMC_EVENT_OPS_CNT,
> > > + LMC_EVENT_DCLK_CNT,
> > > + LMC_EVENT_BANK_CONFLICT1,
> > > + LMC_EVENT_BANK_CONFLICT2,
> > > +};
> > > +
> > > +static int cvm_pmu_lmc_add(struct perf_event *event, int flags)
> > > +{
> > > + struct hw_perf_event *hwc = &event->hw;
> > > +
> > > + return cvm_pmu_add(event, flags, LMC_CONFIG_OFFSET,
> > > + lmc_events[hwc->config]);
> > > +}
> > > +
> > > +PMU_FORMAT_ATTR(lmc_event, "config:0-2");
> > > +
> > > +static struct attribute *cvm_pmu_lmc_format_attr[] = {
> > > + &format_attr_lmc_event.attr,
> > > + NULL,
> > > +};
> > > +
> > > +static struct attribute_group cvm_pmu_lmc_format_group = {
> > > + .name = "format",
> > > + .attrs = cvm_pmu_lmc_format_attr,
> > > +};
> > > +
> > > +static struct attribute *cvm_pmu_lmc_events_attr[] = {
> > > + CVM_PMU_LMC_EVENT_ATTR(ifb_cnt, 0),
> > > + CVM_PMU_LMC_EVENT_ATTR(ops_cnt, 1),
> > > + CVM_PMU_LMC_EVENT_ATTR(dclk_cnt, 2),
> > > + CVM_PMU_LMC_EVENT_ATTR(bank_conflict1, 3),
> > > + CVM_PMU_LMC_EVENT_ATTR(bank_conflict2, 4),
> > > + NULL,
> > > +};
> > > +
> > > +static struct attribute_group cvm_pmu_lmc_events_group = {
> > > + .name = "events",
> > > + .attrs = cvm_pmu_lmc_events_attr,
> > > +};
> > > +
> > > +static const struct attribute_group *cvm_pmu_lmc_attr_groups[] = {
> > > + &cvm_pmu_attr_group,
> > > + &cvm_pmu_lmc_format_group,
> > > + &cvm_pmu_lmc_events_group,
> > > + NULL,
> > > +};
> > > +
> > > +static bool cvm_pmu_lmc_event_valid(u64 config)
> > > +{
> > > + return (config < ARRAY_SIZE(lmc_events));
> > > +}
> > > +
> > > +static int cvm_pmu_lmc_probe(struct pci_dev *pdev)
> > > +{
> > > + struct cvm_pmu_dev *next, *lmc;
> > > + int nr = 0, ret = -ENOMEM;
> > > +
> > > + lmc = kzalloc(sizeof(*lmc), GFP_KERNEL);
> > > + if (!lmc)
> > > + goto fail_nomem;
> > cleaner to do a direct return here as nothing to unwind.
> > return -ENOMEM;
>
> OK.
>
> > > +
> > > + lmc->map = ioremap(pci_resource_start(pdev, 0),
> > > + pci_resource_len(pdev, 0));
> > Might be overly paranoid, but ioremap can return NULL,
> > so should really be checked.
> >
>
> I'll add the check.
>
> > > + list_for_each_entry(next, &cvm_pmu_lmcs, entry)
> > > + nr++;
> > > + lmc->pmu_name = kasprintf(GFP_KERNEL, "lmc%d", nr);
> > Check for error on the allocation.
>
> OK, missed that.
>
> > > +
> > > + lmc->pdev = pdev;
> > > + lmc->num_counters = ARRAY_SIZE(lmc_events);
> > > + lmc->pmu = (struct pmu) {
> > > + .task_ctx_nr = perf_invalid_context,
> > > + .event_init = cvm_pmu_event_init,
> > > + .add = cvm_pmu_lmc_add,
> > > + .del = cvm_pmu_del,
> > > + .start = cvm_pmu_start,
> > > + .stop = cvm_pmu_stop,
> > > + .read = cvm_pmu_read,
> > > + .attr_groups = cvm_pmu_lmc_attr_groups,
> > > + };
> > > +
> > > + cpuhp_state_add_instance_nocalls(CPUHP_AP_PERF_ARM_CVM_ONLINE,
> > > + &lmc->cpuhp_node);
> > This function can fail by the look of it. Should be checking it's
> > return value really.
> >
>
> I'm ignoring the return value on purpose as I don't want probe() to fail
> even if this would return an error.
A comment perhaps to explain this.
>
> > > +
> > > + /*
> > > + * perf PMU is CPU dependent so pick a random CPU and migrate away
> > > + * if it goes offline.
> > > + */
> > > + cpumask_set_cpu(smp_processor_id(), &lmc->active_mask);
> > > +
> > > + ret = perf_pmu_register(&lmc->pmu, lmc->pmu_name, -1);
> > > + if (ret)
> > > + goto fail_hp;
> > > +
> > > + list_add(&lmc->entry, &cvm_pmu_lmcs);
> > > +
> > > + lmc->event_valid = cvm_pmu_lmc_event_valid;
> > These last two elements rather look like they might cause a
> > potential race as the perf_pmu_register has made the counters
> > available.
>
> Yes, they should happen before perf_pmu_register.
>
> > > + dev_info(&pdev->dev, "Enabled %s PMU with %d counters\n",
> > > + lmc->pmu_name, lmc->num_counters);
> > > + return 0;
> > > +
> > > +fail_hp:
> > > + kfree(lmc->pmu_name);
> > Convention is to unwind in the opposite order to the way things
> > were created / set up. Becomes more relevant once you are
> > checking more error codes.
>
> Agreed.
>
> > > + cpuhp_state_remove_instance(CPUHP_AP_PERF_ARM_CVM_ONLINE,
> > > + &lmc->cpuhp_node);
> > > + iounmap(lmc->map);
> > > + kfree(lmc);
> > > +fail_nomem:
> > > + return ret;
> > > +}
> > > +
> > > +static int __init cvm_pmu_init(void)
> > > +{
> > > + unsigned long implementor = read_cpuid_implementor();
> > > + unsigned int vendor_id = PCI_VENDOR_ID_CAVIUM;
> > > + struct pci_dev *pdev = NULL;
> > > + int rc;
> > > +
> > > + if (implementor != ARM_CPU_IMP_CAVIUM)
> > > + return -ENODEV;
> > > +
> > > + INIT_LIST_HEAD(&cvm_pmu_lmcs);
> > > +
> > > + rc = cpuhp_setup_state_multi(CPUHP_AP_PERF_ARM_CVM_ONLINE,
> > > + "perf/arm/cvm:online", NULL,
> > > + cvm_pmu_offline_cpu);
> > > +
> > > + /* detect LMC devices */
> > > + while ((pdev = pci_get_device(vendor_id, 0xa022, pdev))) {
> > > + if (!pdev)
> > > + break;
> > > + rc = cvm_pmu_lmc_probe(pdev);
> > > + if (rc)
> > > + return rc;
> > > + }
> > > + return 0;
> > > +}
> > A comment perhaps explaining why this is a late_initcall? What
> > dependency are we forcing to be true?
>
> PCI probing should be finished before, I'll add a comment.
>
> thanks for the review,
You are welcome.
> Jan
>
> > > +late_initcall(cvm_pmu_init);
> > > diff --git a/include/linux/cpuhotplug.h b/include/linux/cpuhotplug.h
> > > index b56573b..78ac3d2 100644
> > > --- a/include/linux/cpuhotplug.h
> > > +++ b/include/linux/cpuhotplug.h
> > > @@ -141,6 +141,7 @@ enum cpuhp_state {
> > > CPUHP_AP_PERF_ARM_QCOM_L3_ONLINE,
> > > CPUHP_AP_WORKQUEUE_ONLINE,
> > > CPUHP_AP_RCUTREE_ONLINE,
> > > + CPUHP_AP_PERF_ARM_CVM_ONLINE,
> > > CPUHP_AP_ONLINE_DYN,
> > > CPUHP_AP_ONLINE_DYN_END = CPUHP_AP_ONLINE_DYN + 30,
> > > CPUHP_AP_X86_HPET_ONLINE,
> >
On Wed, 19 Jul 2017 14:08:46 +0200
Jan Glauber <[email protected]> wrote:
> Add support for the transmit-link (OCX TLK) PMU counters found
> on Caviums SOCs with a processor interconnect.
>
> Properties of the OCX TLK counters:
> - per-unit control
> - fixed purpose
> - writable
> - one PCI device with multiple TLK units
>
> Signed-off-by: Jan Glauber <[email protected]>
A few really trivial bits in here.
Jonathan
> ---
> drivers/perf/cavium_pmu.c | 216 ++++++++++++++++++++++++++++++++++++++++++++++
> 1 file changed, 216 insertions(+)
>
> diff --git a/drivers/perf/cavium_pmu.c b/drivers/perf/cavium_pmu.c
> index 7d22493..e14a93d 100644
> --- a/drivers/perf/cavium_pmu.c
> +++ b/drivers/perf/cavium_pmu.c
> @@ -19,6 +19,7 @@
>
> enum cvm_pmu_type {
> CVM_PMU_LMC,
> + CVM_PMU_TLK,
> };
>
> /* maximum number of parallel hardware counters for all pmu types */
> @@ -39,6 +40,7 @@ struct cvm_pmu_dev {
> };
>
> static struct list_head cvm_pmu_lmcs;
> +static struct list_head cvm_pmu_tlks;
>
> /*
> * Common Cavium PMU stuff
> @@ -387,6 +389,210 @@ static int cvm_pmu_lmc_probe(struct pci_dev *pdev)
> return ret;
> }
>
> +/*
> + * CCPI interface controller (OCX) Transmit link (TLK) counters:
> + * - per-unit control
> + * - writable
> + * - one PCI device with multiple TLK units
> + */
> +
> +#define TLK_NR_UNITS 3
> +#define TLK_UNIT_OFFSET 0x2000
> +#define TLK_UNIT_LEN 0x7ff
> +#define TLK_START_ADDR 0x10000
> +#define TLK_STAT_CTL_OFFSET 0x40
> +#define TLK_STAT_OFFSET 0x400
> +
> +#define TLK_STAT_ENABLE_BIT BIT(0)
> +#define TLK_STAT_RESET_BIT BIT(1)
> +
> +#define CVM_PMU_TLK_EVENT_ATTR(_name, _id) \
> + &((struct perf_pmu_events_attr[]) { \
> + { \
> + __ATTR(_name, S_IRUGO, cvm_pmu_event_sysfs_show, NULL), \
> + _id, \
> + "tlk_event=" __stringify(_id), \
> + } \
> + })[0].attr.attr
> +
> +static void cvm_pmu_tlk_enable_pmu(struct pmu *pmu)
> +{
> + struct cvm_pmu_dev *pmu_dev = container_of(pmu, struct cvm_pmu_dev, pmu);
> +
> + /* enable all counters */
> + writeb(TLK_STAT_ENABLE_BIT, pmu_dev->map + TLK_STAT_CTL_OFFSET);
> +}
> +
> +static void cvm_pmu_tlk_disable_pmu(struct pmu *pmu)
> +{
> + struct cvm_pmu_dev *pmu_dev = container_of(pmu, struct cvm_pmu_dev, pmu);
> +
> + /* disable all counters */
> + writeb(0, pmu_dev->map + TLK_STAT_CTL_OFFSET);
> +}
> +
> +static int cvm_pmu_tlk_add(struct perf_event *event, int flags)
> +{
> + struct hw_perf_event *hwc = &event->hw;
> +
> + return cvm_pmu_add(event, flags, TLK_STAT_CTL_OFFSET,
> + TLK_STAT_OFFSET + hwc->config * 8);
> +}
> +
> +PMU_FORMAT_ATTR(tlk_event, "config:0-5");
> +
> +static struct attribute *cvm_pmu_tlk_format_attr[] = {
> + &format_attr_tlk_event.attr,
> + NULL,
> +};
> +
> +static struct attribute_group cvm_pmu_tlk_format_group = {
> + .name = "format",
> + .attrs = cvm_pmu_tlk_format_attr,
> +};
> +
> +static struct attribute *cvm_pmu_tlk_events_attr[] = {
> + CVM_PMU_TLK_EVENT_ATTR(idle_cnt, 0x00),
> + CVM_PMU_TLK_EVENT_ATTR(data_cnt, 0x01),
> + CVM_PMU_TLK_EVENT_ATTR(sync_cnt, 0x02),
> + CVM_PMU_TLK_EVENT_ATTR(retry_cnt, 0x03),
> + CVM_PMU_TLK_EVENT_ATTR(err_cnt, 0x04),
> + CVM_PMU_TLK_EVENT_ATTR(mat0_cnt, 0x08),
> + CVM_PMU_TLK_EVENT_ATTR(mat1_cnt, 0x09),
> + CVM_PMU_TLK_EVENT_ATTR(mat2_cnt, 0x0a),
> + CVM_PMU_TLK_EVENT_ATTR(mat3_cnt, 0x0b),
> + CVM_PMU_TLK_EVENT_ATTR(vc0_cmd, 0x10),
> + CVM_PMU_TLK_EVENT_ATTR(vc1_cmd, 0x11),
> + CVM_PMU_TLK_EVENT_ATTR(vc2_cmd, 0x12),
> + CVM_PMU_TLK_EVENT_ATTR(vc3_cmd, 0x13),
> + CVM_PMU_TLK_EVENT_ATTR(vc4_cmd, 0x14),
> + CVM_PMU_TLK_EVENT_ATTR(vc5_cmd, 0x15),
> + CVM_PMU_TLK_EVENT_ATTR(vc0_pkt, 0x20),
> + CVM_PMU_TLK_EVENT_ATTR(vc1_pkt, 0x21),
> + CVM_PMU_TLK_EVENT_ATTR(vc2_pkt, 0x22),
> + CVM_PMU_TLK_EVENT_ATTR(vc3_pkt, 0x23),
> + CVM_PMU_TLK_EVENT_ATTR(vc4_pkt, 0x24),
> + CVM_PMU_TLK_EVENT_ATTR(vc5_pkt, 0x25),
> + CVM_PMU_TLK_EVENT_ATTR(vc6_pkt, 0x26),
> + CVM_PMU_TLK_EVENT_ATTR(vc7_pkt, 0x27),
> + CVM_PMU_TLK_EVENT_ATTR(vc8_pkt, 0x28),
> + CVM_PMU_TLK_EVENT_ATTR(vc9_pkt, 0x29),
> + CVM_PMU_TLK_EVENT_ATTR(vc10_pkt, 0x2a),
> + CVM_PMU_TLK_EVENT_ATTR(vc11_pkt, 0x2b),
> + CVM_PMU_TLK_EVENT_ATTR(vc12_pkt, 0x2c),
> + CVM_PMU_TLK_EVENT_ATTR(vc13_pkt, 0x2d),
> + CVM_PMU_TLK_EVENT_ATTR(vc0_con, 0x30),
> + CVM_PMU_TLK_EVENT_ATTR(vc1_con, 0x31),
> + CVM_PMU_TLK_EVENT_ATTR(vc2_con, 0x32),
> + CVM_PMU_TLK_EVENT_ATTR(vc3_con, 0x33),
> + CVM_PMU_TLK_EVENT_ATTR(vc4_con, 0x34),
> + CVM_PMU_TLK_EVENT_ATTR(vc5_con, 0x35),
> + CVM_PMU_TLK_EVENT_ATTR(vc6_con, 0x36),
> + CVM_PMU_TLK_EVENT_ATTR(vc7_con, 0x37),
> + CVM_PMU_TLK_EVENT_ATTR(vc8_con, 0x38),
> + CVM_PMU_TLK_EVENT_ATTR(vc9_con, 0x39),
> + CVM_PMU_TLK_EVENT_ATTR(vc10_con, 0x3a),
> + CVM_PMU_TLK_EVENT_ATTR(vc11_con, 0x3b),
> + CVM_PMU_TLK_EVENT_ATTR(vc12_con, 0x3c),
> + CVM_PMU_TLK_EVENT_ATTR(vc13_con, 0x3d),
> + NULL,
> +};
> +
> +static struct attribute_group cvm_pmu_tlk_events_group = {
> + .name = "events",
> + .attrs = cvm_pmu_tlk_events_attr,
> +};
> +
> +static const struct attribute_group *cvm_pmu_tlk_attr_groups[] = {
> + &cvm_pmu_attr_group,
> + &cvm_pmu_tlk_format_group,
> + &cvm_pmu_tlk_events_group,
> + NULL,
> +};
> +
> +static bool cvm_pmu_tlk_event_valid(u64 config)
> +{
> + struct perf_pmu_events_attr *attr;
> + int i;
> +
> + for (i = 0; i < ARRAY_SIZE(cvm_pmu_tlk_events_attr) - 1; i++) {
> + attr = (struct perf_pmu_events_attr *)cvm_pmu_tlk_events_attr[i];
> + if (attr->id == config)
> + return true;
> + }
> + return false;
> +}
> +
> +static int cvm_pmu_tlk_probe_unit(struct pci_dev *pdev, int nr)
> +{
> + struct cvm_pmu_dev *tlk;
> + int ret = -ENOMEM;
> +
> + tlk = kzalloc(sizeof(*tlk), GFP_KERNEL);
> + if (!tlk)
Direct return generally preferred when there is no cleanup to do.
> + goto fail_nomem;
> +
> + tlk->map = ioremap(pci_resource_start(pdev, 0) + TLK_START_ADDR +
> + nr * TLK_UNIT_OFFSET, TLK_UNIT_LEN);
> + tlk->pmu_name = kasprintf(GFP_KERNEL, "ocx_tlk%d", nr);
Check allocation succeeded.
> +
> + tlk->pdev = pdev;
> + tlk->num_counters = ARRAY_SIZE(cvm_pmu_tlk_events_attr) - 1;
> + tlk->pmu = (struct pmu) {
> + .task_ctx_nr = perf_invalid_context,
> + .pmu_enable = cvm_pmu_tlk_enable_pmu,
> + .pmu_disable = cvm_pmu_tlk_disable_pmu,
> + .event_init = cvm_pmu_event_init,
> + .add = cvm_pmu_tlk_add,
> + .del = cvm_pmu_del,
> + .start = cvm_pmu_start,
> + .stop = cvm_pmu_stop,
> + .read = cvm_pmu_read,
> + .attr_groups = cvm_pmu_tlk_attr_groups,
> + };
> +
> + cpuhp_state_add_instance_nocalls(CPUHP_AP_PERF_ARM_CVM_ONLINE,
> + &tlk->cpuhp_node);
> +
> + /*
> + * perf PMU is CPU dependent so pick a random CPU and migrate away
> + * if it goes offline.
> + */
> + cpumask_set_cpu(smp_processor_id(), &tlk->active_mask);
> +
> + ret = perf_pmu_register(&tlk->pmu, tlk->pmu_name, -1);
> + if (ret)
> + goto fail_hp;
> +
> + list_add(&tlk->entry, &cvm_pmu_tlks);
> +
> + tlk->event_valid = cvm_pmu_tlk_event_valid;
> + dev_info(&pdev->dev, "Enabled %s PMU with %d counters\n",
> + tlk->pmu_name, tlk->num_counters);
> + return 0;
> +
> +fail_hp:
> + kfree(tlk->pmu_name);
> + cpuhp_state_remove_instance(CPUHP_AP_PERF_ARM_CVM_ONLINE,
> + &tlk->cpuhp_node);
> + iounmap(tlk->map);
> + kfree(tlk);
> +fail_nomem:
> + return ret;
> +}
> +
> +static int cvm_pmu_tlk_probe(struct pci_dev *pdev)
> +{
> + int rc, i;
> +
> + for (i = 0; i < TLK_NR_UNITS; i++) {
> + rc = cvm_pmu_tlk_probe_unit(pdev, i);
> + if (rc)
> + return rc;
> + }
> + return 0;
> +}
> +
> static int __init cvm_pmu_init(void)
> {
> unsigned long implementor = read_cpuid_implementor();
> @@ -398,6 +604,7 @@ static int __init cvm_pmu_init(void)
> return -ENODEV;
>
> INIT_LIST_HEAD(&cvm_pmu_lmcs);
> + INIT_LIST_HEAD(&cvm_pmu_tlks);
>
> rc = cpuhp_setup_state_multi(CPUHP_AP_PERF_ARM_CVM_ONLINE,
> "perf/arm/cvm:online", NULL,
> @@ -411,6 +618,15 @@ static int __init cvm_pmu_init(void)
> if (rc)
> return rc;
> }
> +
> + /* detect OCX TLK devices */
> + while ((pdev = pci_get_device(vendor_id, 0xa013, pdev))) {
> + if (!pdev)
> + break;
> + rc = cvm_pmu_tlk_probe(pdev);
> + if (rc)
> + return rc;
> + }
> return 0;
> }
> late_initcall(cvm_pmu_init);
On Wed, 19 Jul 2017 14:08:47 +0200
Jan Glauber <[email protected]> wrote:
> Document Cavium SoC PMUs.
>
> Signed-off-by: Jan Glauber <[email protected]>
And just to finish - the utterly trivial ;)
> ---
> Documentation/perf/cavium-pmu.txt | 74 +++++++++++++++++++++++++++++++++++++++
> 1 file changed, 74 insertions(+)
> create mode 100644 Documentation/perf/cavium-pmu.txt
>
> diff --git a/Documentation/perf/cavium-pmu.txt b/Documentation/perf/cavium-pmu.txt
> new file mode 100644
> index 0000000..37669b9
> --- /dev/null
> +++ b/Documentation/perf/cavium-pmu.txt
> @@ -0,0 +1,74 @@
> +Cavium ThunderX and OcteonTx Performance Monitoring Unit (PMU)
> +==============================================================
> +
> +Cavium SoCs contain various system devices such as L2 caches, processor
> +interconnect and memory controllers. Unfortunately the PMU counters
> +are not following a common design so each device has a slightly different
> +approach how to control and use the PMU counters.
> +
> +Common properties of all devices carrying PMU counters:
> +- The devices are PCI devices and the counters are embedded somewhere
> + in the PCI register space.
> +- All counters are 64 bit wide.
> +- There are no overflow interrupts (unnecessary because of the 64 bit wide counters).
Too long a line > 80 chars.
> +
> +Properties depending on the device type:
> +- How to start/stop the counters
> +- Programmable vs. fixed purpose counters
> +- Stoppable vs. always running counters
> +- Independent vs. grouped counters
> +- Read-only vs. writable counters
> +- PCI device to PMU group relationship
> +
> +
> +Devices with PMU counters
> +-------------------------
> +
> +Memory controller (LMC):
> +- one PCI device per LMC
> +- fixed-purpose counters
> +- always running counters without start/stop/reset control
> +- read-only counters
> +
> +CCPI interface controller (OCX) Transmit link (TLK) counters:
> +- writable counters
> +- only one PCI device exposes multiple TLK units (3 units on T88)
> +- start/stop control per unit
> +- only present on multi-socket systems
> +
> +PMU (perf) driver
> +-----------------
> +
> +The cavium-pmu driver registers several perf PMU drivers. Each of the perf
> +driver provides description of its available events and configuration options
> +in sysfs, see /sys/devices/<lmcX/ocx_tlkX>/.
> +
> +The "format" directory describes format of the config (event ID),
> +The "events" directory shows the names of the events and provides configuration
> +templates for all supported event types that can be used with perf tool. For
> +example, "lmc0/dclk_cnt/" is an equivalent of "lmc0/config=2/".
> +
> +Each perf driver also provides a "cpumask" sysfs attribute, which contains a
> +single CPU ID of the processor which will be used to handle all the PMU events.
> +
> +Example for perf tool use:
> +
> + / # perf list | grep -e lmc
> + lmc0/bank_conflict1/ [Kernel PMU event]
> + lmc0/bank_conflict2/ [Kernel PMU event]
> + lmc0/dclk_cnt/ [Kernel PMU event]
> + lmc0/ifb_cnt/ [Kernel PMU event]
> + lmc0/ops_cnt/ [Kernel PMU event]
> +
> + / # perf stat -a -e lmc0/ops_cnt/,lmc0/dclk_cnt/ -- sleep 1
> +
> + Performance counter stats for 'system wide':
> +
> + 176,133 lmc0/ops_cnt/
> + 670,243,653 lmc0/dclk_cnt/
> +
> + 1.005479295 seconds time elapsed
> +
> +The driver does not support sampling, therefore "perf record" will
> +not work. System wide mode ("-a") must be used as per-task (without "-a")
> +perf sessions are not supported.