diff mbox

[V5,21/26] coresight: etm-perf: new PMU driver for ETM tracers

Message ID 1448849687-5724-22-git-send-email-mathieu.poirier@linaro.org
State New
Headers show

Commit Message

Mathieu Poirier Nov. 30, 2015, 2:14 a.m. UTC
Perf is a well known and used tool for performance monitoring
and much more. A such it is an ideal candidate for integration
with coresight based HW tracing.

This patch introduces a PMU that represent a coresight tracer to
the Perf core.

Signed-off-by: Mathieu Poirier <mathieu.poirier@linaro.org>

---
 drivers/hwtracing/coresight/Makefile             |   3 +-
 drivers/hwtracing/coresight/coresight-etm-perf.c | 465 +++++++++++++++++++++++
 drivers/hwtracing/coresight/coresight-etm-perf.h |  32 ++
 drivers/hwtracing/coresight/coresight-etm3x.c    |   7 +
 include/linux/coresight-pmu.h                    |  27 ++
 5 files changed, 533 insertions(+), 1 deletion(-)
 create mode 100644 drivers/hwtracing/coresight/coresight-etm-perf.c
 create mode 100644 drivers/hwtracing/coresight/coresight-etm-perf.h
 create mode 100644 include/linux/coresight-pmu.h

-- 
2.1.4

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/

Comments

Mathieu Poirier Dec. 1, 2015, 5:25 p.m. UTC | #1
On 30 November 2015 at 16:23, Alexander Shishkin
<alexander.shishkin@linux.intel.com> wrote:
> Mathieu Poirier <mathieu.poirier@linaro.org> writes:

>

>> +static void etm_event_destroy(struct perf_event *event) {}

>> +

>> +static int etm_event_init(struct perf_event *event)

>> +{

>> +     if (event->attr.type != etm_pmu.type)

>> +             return -ENOENT;

>> +

>> +     if (event->cpu >= nr_cpu_ids)

>> +             return -EINVAL;

>> +

>> +     event->destroy = etm_event_destroy;

>

> You don't have to do this if it's a nop, event::destroy can be NULL.


ACK

>

>> +

>> +     return 0;

>> +}

>

>

>> +static void *alloc_event_data(int cpu)

>> +{

>> +     int lcpu, size;

>> +     cpumask_t *mask;

>> +     struct etm_cpu_data *cpu_data;

>> +     struct etm_event_data *event_data;

>> +

>> +     /* First get memory for the session's data */

>> +     event_data = kzalloc(sizeof(struct etm_event_data), GFP_KERNEL);

>> +      if (!event_data)

>

> Looks like a whitespace mixup.


ACK

>

>> +             return NULL;

>> +

>> +     /* Make sure nothing disappears under us */

>> +     get_online_cpus();

>> +     size = num_online_cpus();

>> +

>> +     mask = &event_data->mask;

>> +     if (cpu != -1)

>> +             cpumask_set_cpu(cpu, mask);

>> +     else

>> +             cpumask_copy(mask, cpu_online_mask);

>

> It would be nice to have a comment somewhere here explaining that you

> have to set up tracer on each cpu in case of per-thread counter and

> why. We must have discussed this, but I forgot already.


That's a very good idea and I'm not entirely sure I've explained it
plainly before either.  Coresight has several types of tracers and
each have their little differences with configuration registers
changing often from one version to another.  It is also possible to
have different types of tracers on one SoC (ex. big.LITTLE platforms).
As such the global configuration from Perf can be interpreted
differently depending on the implemented tracer version.  Sorting out
tracer configuration on each CPU before a run is started is much more
efficient than parsing the Perf configuration each time an event is
scheduled on a CPU.

Last but not least Coresight tracers have many configuration options
that I haven't exposed to Perf yet.  One such option is address range
filtering.  Processing those each time an event is about to be
scheduled would be highly inefficient.

>

> Btw, do you want to also set 'size' to 1 for cpu != -1 case?


I thought long and hard about that one.  Per CPU tracing is really a
corner case of the general scenario where all CPUs are part of an
event.  The 'size' variable is to allocate an array of 'struct
etm_cpu_data' pointers, where the index of the array is the CPU the
data pertains to.  Whether one or all CPUs are involved in a trace
session, I decided to allocate that array the same way in order to 1)
make access to CPU data generic and 2) make trace configuration
retrieval for a CPU very fast by using that CPU number as an index.

We could have an an array with only the CPUs that were configured but
that would also mean that we'd loose the quick access CPU indexing
provides.  In my opinion that was much worse than the extra memory
needed for corner cases.

>

>> +     put_online_cpus();

>> +

>> +     /* Allocate an array of cpu_data to work with */

>> +     event_data->cpu_data = kcalloc(size,

>> +                                    sizeof(struct etm_cpu_data *),

>> +                                    GFP_KERNEL);

>> +     if (!event_data->cpu_data)

>> +             goto free_event_data;

>> +

>> +     /* Allocate a cpu_data for each CPU this event is dealing with */

>> +     for_each_cpu(lcpu, mask) {

>> +             cpu_data = kzalloc(sizeof(struct etm_cpu_data), GFP_KERNEL);

>> +             if (!cpu_data)

>> +                     goto free_event_data;

>> +

>> +             event_data->cpu_data[lcpu] = cpu_data;

>> +     }

>

> Wouldn't it be easier to allocate the whole thing with one

>

> event_data->cpu_data = kcalloc(size, sizeof(struct etm_cpu_data), GFP_KERNEL);

>

> ?


Right, that would work if 'cpu_data[]' wasn't used as an index table.
As I said above, I thought it was better to loose a few bytes of
memory to quicken access to trace configuration when events are
scheduled in.

Special thanks for taking the time to review this.
Mathieu

>

> Regards,

> --

> Alex

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/
diff mbox

Patch

diff --git a/drivers/hwtracing/coresight/Makefile b/drivers/hwtracing/coresight/Makefile
index 233d66cf22d3..cf8c6d689747 100644
--- a/drivers/hwtracing/coresight/Makefile
+++ b/drivers/hwtracing/coresight/Makefile
@@ -9,6 +9,7 @@  obj-$(CONFIG_CORESIGHT_SINK_ETBV10) += coresight-etb10.o
 obj-$(CONFIG_CORESIGHT_LINKS_AND_SINKS) += coresight-funnel.o \
 					   coresight-replicator.o
 obj-$(CONFIG_CORESIGHT_SOURCE_ETM3X) += coresight-etm3x.o coresight-etm-cp14.o \
-					coresight-etm3x-sysfs.o
+					coresight-etm3x-sysfs.o \
+					coresight-etm-perf.o
 obj-$(CONFIG_CORESIGHT_SOURCE_ETM4X) += coresight-etm4x.o
 obj-$(CONFIG_CORESIGHT_QCOM_REPLICATOR) += coresight-replicator-qcom.o
diff --git a/drivers/hwtracing/coresight/coresight-etm-perf.c b/drivers/hwtracing/coresight/coresight-etm-perf.c
new file mode 100644
index 000000000000..cc033980f822
--- /dev/null
+++ b/drivers/hwtracing/coresight/coresight-etm-perf.c
@@ -0,0 +1,465 @@ 
+/*
+ * Copyright(C) 2015 Linaro Limited. All rights reserved.
+ * Author: Mathieu Poirier <mathieu.poirier@linaro.org>
+ *
+ * This program is free software; you can redistribute it and/or modify it
+ * under the terms of the GNU General Public License version 2 as published by
+ * the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License for
+ * more details.
+ *
+ * You should have received a copy of the GNU General Public License along with
+ * this program.  If not, see <http://www.gnu.org/licenses/>.
+ */
+
+#include <linux/coresight.h>
+#include <linux/coresight-pmu.h>
+#include <linux/cpumask.h>
+#include <linux/device.h>
+#include <linux/list.h>
+#include <linux/mm.h>
+#include <linux/module.h>
+#include <linux/perf_event.h>
+#include <linux/slab.h>
+#include <linux/types.h>
+
+#include "coresight-priv.h"
+
+static struct pmu etm_pmu;
+static bool etm_perf_up;
+
+/**
+ * struct etm_cpu_data - Coresight specifics accociated to a single CPU
+ * @src_config:		The tracer configuration.
+ * @snk_config:		The sink configuration.
+ * @patch:		The path from source to sink.
+ */
+struct etm_cpu_data {
+	void *src_config;
+	void *snk_config;
+	struct list_head *path;
+};
+
+/**
+ * struct etm_event_data - Coresight specifics associated to an event
+ * @mask:		Hold the CPU(s) this event was set for.
+ * @cpu_data:		An array of cpu data, each slot for one CPU.
+ */
+struct etm_event_data {
+	cpumask_t mask;
+	struct etm_cpu_data **cpu_data;
+};
+
+static DEFINE_PER_CPU(struct perf_output_handle, ctx_handle);
+static DEFINE_PER_CPU(struct coresight_device *, csdev_src);
+
+/* ETMv3.5/PTM's ETMCR is 'config' */
+PMU_FORMAT_ATTR(cycacc,		"config:" __stringify(ETM_OPT_CYCACC));
+PMU_FORMAT_ATTR(timestamp,	"config:" __stringify(ETM_OPT_TS));
+
+static struct attribute *etm_config_formats_attr[] = {
+	&format_attr_cycacc.attr,
+	&format_attr_timestamp.attr,
+	NULL,
+};
+
+static struct attribute_group etm_pmu_format_group = {
+	.name   = "format",
+	.attrs  = etm_config_formats_attr,
+};
+
+static const struct attribute_group *etm_pmu_attr_groups[] = {
+	&etm_pmu_format_group,
+	NULL,
+};
+
+static void etm_event_read(struct perf_event *event) {}
+static void etm_event_destroy(struct perf_event *event) {}
+
+static int etm_event_init(struct perf_event *event)
+{
+	if (event->attr.type != etm_pmu.type)
+		return -ENOENT;
+
+	if (event->cpu >= nr_cpu_ids)
+		return -EINVAL;
+
+	event->destroy = etm_event_destroy;
+
+	return 0;
+}
+
+static void free_cpu_data(int cpu, struct etm_cpu_data *cpu_data)
+{
+	struct coresight_device *sink, *csdev;
+
+	csdev = per_cpu(csdev_src, cpu);
+	if (!csdev)
+		return;
+
+	if (source_ops(csdev)->put_config)
+		source_ops(csdev)->put_config(cpu_data->src_config);
+
+	/* No need to continue if there isn't a path to work with */
+	if (!cpu_data->path)
+		return;
+
+	sink = coresight_get_sink(cpu_data->path);
+	if (sink_ops(sink)->put_config)
+		sink_ops(sink)->put_config(cpu_data->snk_config);
+
+	coresight_release_path(cpu_data->path);
+}
+
+static void free_event_data(struct etm_event_data *event_data)
+{
+	int cpu;
+	cpumask_t *mask = &event_data->mask;
+
+	for_each_cpu(cpu, mask) {
+		if (event_data->cpu_data[cpu])
+			free_cpu_data(cpu, event_data->cpu_data[cpu]);
+		kfree(event_data->cpu_data[cpu]);
+	}
+
+	kfree(event_data->cpu_data);
+	kfree(event_data);
+}
+
+static void *alloc_event_data(int cpu)
+{
+	int lcpu, size;
+	cpumask_t *mask;
+	struct etm_cpu_data *cpu_data;
+	struct etm_event_data *event_data;
+
+	/* First get memory for the session's data */
+	event_data = kzalloc(sizeof(struct etm_event_data), GFP_KERNEL);
+	 if (!event_data)
+		return NULL;
+
+	/* Make sure nothing disappears under us */
+	get_online_cpus();
+	size = num_online_cpus();
+
+	mask = &event_data->mask;
+	if (cpu != -1)
+		cpumask_set_cpu(cpu, mask);
+	else
+		cpumask_copy(mask, cpu_online_mask);
+	put_online_cpus();
+
+	/* Allocate an array of cpu_data to work with */
+	event_data->cpu_data = kcalloc(size,
+				       sizeof(struct etm_cpu_data *),
+				       GFP_KERNEL);
+	if (!event_data->cpu_data)
+		goto free_event_data;
+
+	/* Allocate a cpu_data for each CPU this event is dealing with */
+	for_each_cpu(lcpu, mask) {
+		cpu_data = kzalloc(sizeof(struct etm_cpu_data), GFP_KERNEL);
+		if (!cpu_data)
+			goto free_event_data;
+
+		event_data->cpu_data[lcpu] = cpu_data;
+	}
+
+out:
+	return event_data;
+
+free_event_data:
+	free_event_data(event_data);
+	event_data = NULL;
+
+	goto out;
+}
+
+static void *etm_setup_aux(struct perf_event *event, void **pages,
+			   int nr_pages, bool overwrite)
+{
+	int cpu;
+	cpumask_t *mask;
+	struct etm_event_data *event_data = NULL;
+	struct coresight_device *csdev;
+
+	event_data = alloc_event_data(event->cpu);
+	if (!event_data)
+		return NULL;
+
+	mask = &event_data->mask;
+
+	for_each_cpu(cpu, mask) {
+		struct etm_cpu_data *cpu_data;
+		struct coresight_device *sink;
+
+		csdev = per_cpu(csdev_src, cpu);
+		if (!csdev)
+			goto err;
+
+		cpu_data = event_data->cpu_data[cpu];
+
+		/* Get the tracer's config from perf */
+		if (!source_ops(csdev)->get_config)
+			goto err;
+
+		cpu_data->src_config =
+				source_ops(csdev)->get_config(csdev, event);
+		if (!cpu_data->src_config)
+			goto err;
+
+		/*
+		 * Building a path doesn't enable it, it simply builds a
+		 * list of devices from source to sink that can be
+		 * referenced later when the path is actually needed.
+		 */
+		cpu_data->path = coresight_build_path(csdev);
+		if (!cpu_data->path)
+			goto err;
+
+		/* Grab the sink at the end of the path */
+		sink = coresight_get_sink(cpu_data->path);
+		if (!sink)
+			goto err;
+
+		if (!sink_ops(sink)->get_config)
+			goto err;
+
+		/* Finally get the AUX specific data from the sink buffer */
+		cpu_data->snk_config =
+				sink_ops(sink)->get_config(sink, cpu, pages,
+							   nr_pages, overwrite);
+		if (!cpu_data->snk_config)
+			goto err;
+
+	}
+
+out:
+	return event_data;
+
+err:
+	free_event_data(event_data);
+	event_data = NULL;
+	goto out;
+
+	return NULL;
+}
+
+static void etm_free_aux(void *data)
+{
+	free_event_data(data);
+}
+
+static void etm_event_stop(struct perf_event *event, int mode)
+{
+	int cpu = smp_processor_id();
+	struct coresight_device *csdev = per_cpu(csdev_src, cpu);
+
+	if (event->hw.state == PERF_HES_STOPPED)
+		return;
+
+	if (!csdev)
+		return;
+
+	/* stop tracer */
+	source_ops(csdev)->disable(csdev);
+
+	/* tell the core */
+	event->hw.state = PERF_HES_STOPPED;
+
+
+	if (mode & PERF_EF_UPDATE) {
+		struct coresight_device *sink;
+		struct perf_output_handle *handle = this_cpu_ptr(&ctx_handle);
+		struct etm_event_data *event_data = perf_get_aux(handle);
+		struct etm_cpu_data *cpu_data = event_data->cpu_data[cpu];
+
+		if (WARN_ON_ONCE(handle->event != event))
+			return;
+
+		if (WARN_ON_ONCE(!cpu_data))
+			return;
+
+		sink = coresight_get_sink(cpu_data->path);
+		if (WARN_ON_ONCE(!sink))
+			return;
+
+		/* update trace information */
+		if (!sink_ops(sink)->update_buffer)
+			return;
+
+		sink_ops(sink)->update_buffer(sink, handle,
+					      cpu_data->snk_config);
+	}
+}
+
+static void etm_event_start(struct perf_event *event, int flags)
+{
+	int cpu = smp_processor_id();
+	struct coresight_device *csdev = per_cpu(csdev_src, cpu);
+
+	if (!csdev)
+		goto fail;
+
+	/* tell the perf core the event is alive */
+	event->hw.state = 0;
+
+	if (source_ops(csdev)->enable(csdev, CS_MODE_PERF))
+		goto fail;
+
+	return;
+
+fail:
+	event->hw.state = PERF_HES_STOPPED;
+}
+
+static void etm_event_del(struct perf_event *event, int mode)
+{
+	bool lost = true;
+	int cpu = smp_processor_id();
+	unsigned long size;
+	struct coresight_device *sink;
+	struct perf_output_handle *handle = this_cpu_ptr(&ctx_handle);
+	struct etm_event_data *event_data = perf_get_aux(handle);
+	struct etm_cpu_data *cpu_data = event_data->cpu_data[cpu];
+
+	if (WARN_ON_ONCE(!cpu_data))
+		return;
+
+	etm_event_stop(event, PERF_EF_UPDATE);
+
+	sink = coresight_get_sink(cpu_data->path);
+	if (!sink)
+		return;
+
+	if (!sink_ops(sink)->reset_buffer)
+		return;
+
+	size = sink_ops(sink)->reset_buffer(sink, handle,
+					    cpu_data->snk_config,
+					    &lost);
+
+	perf_aux_output_end(handle, size, lost);
+
+	coresight_disable_path(cpu_data->path);
+}
+
+static int etm_event_add(struct perf_event *event, int mode)
+{
+	int ret = -EBUSY, cpu = smp_processor_id();
+	struct etm_event_data *event_data;
+	struct etm_cpu_data *cpu_data;
+	struct perf_output_handle *handle = this_cpu_ptr(&ctx_handle);
+	struct hw_perf_event *hwc = &event->hw;
+	struct coresight_device *csdev = per_cpu(csdev_src, cpu);
+	struct coresight_device *sink;
+
+	if (handle->event)
+		goto out;
+
+	event_data = perf_aux_output_begin(handle, event);
+	ret = -EINVAL;
+	if (WARN_ON_ONCE(!event_data))
+		goto fail_stop;
+
+	cpu_data = event_data->cpu_data[cpu];
+
+	sink = coresight_get_sink(cpu_data->path);
+	if (!sink)
+		goto fail_end_stop;
+
+	if (!sink_ops(sink)->set_buffer)
+		goto fail_end_stop;
+
+	ret = sink_ops(sink)->set_buffer(sink, handle,
+					 cpu_data->snk_config);
+
+	if (ret)
+		goto fail_end_stop;
+
+	if (!source_ops(csdev)->set_config) {
+		ret = -EINVAL;
+		goto fail_end_stop;
+	}
+
+	source_ops(csdev)->set_config(csdev, cpu_data->src_config);
+
+	ret = coresight_enable_path(cpu_data->path, CS_MODE_PERF);
+	if (ret)
+		goto fail_end_stop;
+
+	if (mode & PERF_EF_START) {
+		etm_event_start(event, 0);
+		if (hwc->state & PERF_HES_STOPPED) {
+			etm_event_del(event, 0);
+			ret = -EBUSY;
+			goto fail_path_end_stop;
+		}
+	}
+
+out:
+	return ret;
+
+fail_path_end_stop:
+	coresight_disable_path(cpu_data->path);
+fail_end_stop:
+	perf_aux_output_end(handle, 0, true);
+fail_stop:
+	hwc->state = PERF_HES_STOPPED;
+	goto out;
+
+	return 0;
+}
+
+int etm_perf_symlink(struct coresight_device *csdev, bool link)
+{
+	char entry[sizeof("cpu9999999")];
+	int ret = 0, cpu = source_ops(csdev)->cpu_id(csdev);
+	struct device *pmu_dev = etm_pmu.dev;
+	struct device *cs_dev = &csdev->dev;
+
+	sprintf(entry, "cpu%d", cpu);
+
+	if (!etm_perf_up)
+		return -EPROBE_DEFER;
+
+	if (link) {
+		ret = sysfs_create_link(&pmu_dev->kobj, &cs_dev->kobj, entry);
+		if (ret)
+			return ret;
+		per_cpu(csdev_src, cpu) = csdev;
+	} else {
+		sysfs_remove_link(&pmu_dev->kobj, entry);
+		per_cpu(csdev_src, cpu) = NULL;
+	}
+
+	return 0;
+}
+
+static int __init etm_perf_init(void)
+{
+	int ret;
+
+	etm_pmu.capabilities	= PERF_PMU_CAP_EXCLUSIVE;
+
+	etm_pmu.attr_groups	= etm_pmu_attr_groups;
+	etm_pmu.task_ctx_nr	= perf_sw_context;
+	etm_pmu.read		= etm_event_read;
+	etm_pmu.event_init	= etm_event_init;
+	etm_pmu.setup_aux	= etm_setup_aux;
+	etm_pmu.free_aux	= etm_free_aux;
+	etm_pmu.stop		= etm_event_stop;
+	etm_pmu.start		= etm_event_start;
+	etm_pmu.del		= etm_event_del;
+	etm_pmu.add		= etm_event_add;
+
+	ret = perf_pmu_register(&etm_pmu, CORESIGHT_ETM_PMU_NAME, -1);
+	if (ret == 0)
+		etm_perf_up = true;
+
+	return ret;
+}
+module_init(etm_perf_init);
diff --git a/drivers/hwtracing/coresight/coresight-etm-perf.h b/drivers/hwtracing/coresight/coresight-etm-perf.h
new file mode 100644
index 000000000000..87f5a134eb6f
--- /dev/null
+++ b/drivers/hwtracing/coresight/coresight-etm-perf.h
@@ -0,0 +1,32 @@ 
+/*
+ * Copyright(C) 2015 Linaro Limited. All rights reserved.
+ * Author: Mathieu Poirier <mathieu.poirier@linaro.org>
+ *
+ * This program is free software; you can redistribute it and/or modify it
+ * under the terms of the GNU General Public License version 2 as published by
+ * the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License for
+ * more details.
+ *
+ * You should have received a copy of the GNU General Public License along with
+ * this program.  If not, see <http://www.gnu.org/licenses/>.
+ */
+
+#ifndef _CORESIGHT_ETM_PERF_H
+#define _CORESIGHT_ETM_PERF_H
+
+struct coresight_device;
+
+#ifdef CONFIG_CORESIGHT
+int etm_perf_symlink(struct coresight_device *csdev, bool link);
+
+#else
+static inline int etm_perf_symlink(struct coresight_device *csdev, bool link)
+{ return -EINVAL; }
+
+#endif /* CONFIG_CORESIGHT */
+
+#endif
diff --git a/drivers/hwtracing/coresight/coresight-etm3x.c b/drivers/hwtracing/coresight/coresight-etm3x.c
index fbd04979e21c..6170df1b9e59 100644
--- a/drivers/hwtracing/coresight/coresight-etm3x.c
+++ b/drivers/hwtracing/coresight/coresight-etm3x.c
@@ -35,6 +35,7 @@ 
 #include <asm/sections.h>
 
 #include "coresight-etm.h"
+#include "coresight-etm-perf.h"
 
 static int boot_enable;
 module_param_named(boot_enable, boot_enable, int, S_IRUGO);
@@ -879,6 +880,12 @@  static int etm_probe(struct amba_device *adev, const struct amba_id *id)
 		goto err_arch_supported;
 	}
 
+	ret = etm_perf_symlink(drvdata->csdev, true);
+	if (ret) {
+		coresight_unregister(drvdata->csdev);
+		goto err_arch_supported;
+	}
+
 	pm_runtime_put(&adev->dev);
 	dev_info(dev, "%s initialized\n", (char *)id->data);
 
diff --git a/include/linux/coresight-pmu.h b/include/linux/coresight-pmu.h
new file mode 100644
index 000000000000..6c5386b23b10
--- /dev/null
+++ b/include/linux/coresight-pmu.h
@@ -0,0 +1,27 @@ 
+/*
+ * Copyright(C) 2015 Linaro Limited. All rights reserved.
+ * Author: Mathieu Poirier <mathieu.poirier@linaro.org>
+ *
+ * This program is free software; you can redistribute it and/or modify it
+ * under the terms of the GNU General Public License version 2 as published by
+ * the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License for
+ * more details.
+ *
+ * You should have received a copy of the GNU General Public License along with
+ * this program.  If not, see <http://www.gnu.org/licenses/>.
+ */
+
+#ifndef _LINUX_CORESIGHT_PMU_H
+#define _LINUX_CORESIGHT_PMU_H
+
+#define CORESIGHT_ETM_PMU_NAME "cs_etm"
+
+/* ETMv3.5/PTM's ETMCR config bit */
+#define ETM_OPT_CYCACC  12
+#define ETM_OPT_TS      28
+
+#endif