All of lore.kernel.org
 help / color / mirror / Atom feed
* [PATCH V2 0/3] add support for cherryview gpio controller driver
@ 2021-08-31  5:07 Hongzhan Chen
  2021-08-31  5:07 ` [PATCH 1/3] drivers/gpio: core: Move out of OF config conditional compilation Hongzhan Chen
                   ` (3 more replies)
  0 siblings, 4 replies; 13+ messages in thread
From: Hongzhan Chen @ 2021-08-31  5:07 UTC (permalink / raw)
  To: xenomai

1. move out of OF config conditional compilation so that non-OF platform
  call same API to remove rtdm gpio chip device and remove _of for api
  name.
2. Introduce helper to find gpiochip as referring to pair of 
   rtdm_gpiochip_scan_of and rtdm_gpiochip_scan_array_of. 
3. Add Intel Cherryview pinctrl driver based on on 1 and 2.

I also did following tests with this patchset:
1. run /usr/lib/xenomai/testsuite/gpiobench -i 334 -i 335 -c INT33FF:02
   to validate patch 9afea5ff2d7ba97db96b22a005a9a7fcf5f2d892 when
   setting GPIO_RTIOC_TS
2. apply following patch, and rerun 1.

index f83d7689f..50afbd418 100644
--- a/testsuite/gpiobench/gpiobench.c
+++ b/testsuite/gpiobench/gpiobench.c
@@ -619,7 +619,7 @@ int main(int argc, char **argv)
                        goto out;
                }

-               ret = ioctl(ti.fd_dev_intr, GPIO_RTIOC_TS, &value);
+               ret = ioctl(ti.fd_dev_intr, GPIO_RTIOC_TS_MONO, &value);
                if (ret) {
                        printf("ioctl gpio port ts, failed\n");
                        goto out;

3. test result for inner related output:
   REALTIME clock:
	# Total: 000999999
	# Min Latencies: 00013
	# Avg Latencies: 16.879997
	# Max Latencies: 00043

   MONO clock:
	# Total: 000083988
	# Min Latencies: 00012
	# Avg Latencies: 2814377.000000
	# Max Latencies: 00041

Hardware env:
1. Rock PI X V1.4.
2. GPIO loopback connection between GPIO 334 and 335.

Hongzhan Chen (3):
  drivers/gpio: core: Move out of OF config conditional compilation
  drivers/gpio: core: Introduce helper to find gpiochip
  driver/gpio: Add Intel Cherryview pinctrl driver

 include/cobalt/kernel/rtdm/gpio.h     | 12 +++++--
 kernel/drivers/gpio/Kconfig           |  7 ++++
 kernel/drivers/gpio/Makefile          |  2 ++
 kernel/drivers/gpio/gpio-bcm2835.c    |  2 +-
 kernel/drivers/gpio/gpio-cherryview.c | 42 ++++++++++++++++++++++
 kernel/drivers/gpio/gpio-core.c       | 50 ++++++++++++++++++++++++---
 kernel/drivers/gpio/gpio-mxc.c        |  2 +-
 kernel/drivers/gpio/gpio-omap.c       |  2 +-
 kernel/drivers/gpio/gpio-sun8i-h3.c   |  2 +-
 kernel/drivers/gpio/gpio-xilinx.c     |  2 +-
 kernel/drivers/gpio/gpio-zynq7000.c   |  2 +-
 11 files changed, 112 insertions(+), 13 deletions(-)
 create mode 100644 kernel/drivers/gpio/gpio-cherryview.c

-- 
2.17.1



^ permalink raw reply related	[flat|nested] 13+ messages in thread
* [PATCH 0/3] add support for cherryview gpio controller driver
@ 2021-08-30  6:45 Hongzhan Chen
  2021-08-30  6:45 ` [PATCH 2/3] drivers/gpio: core: Introduce helper to find gpiochip Hongzhan Chen
  0 siblings, 1 reply; 13+ messages in thread
From: Hongzhan Chen @ 2021-08-30  6:45 UTC (permalink / raw)
  To: xenomai

1. move out of OF config conditional compilation so that non-OF platform
  call same API to remove rtdm gpio chip device.
2. Introduce helper to find gpiochip as referring to pair of 
   rtdm_gpiochip_scan_of and rtdm_gpiochip_scan_array_of. 
3. Add Intel Cherryview pinctrl driver based on on 1 and 2.

I also did following tests with this patchset:
1. run /usr/lib/xenomai/testsuite/gpiobench -i 334 -i 335 -c INT33FF:02
   to validate patch 9afea5ff2d7ba97db96b22a005a9a7fcf5f2d892 when
   setting GPIO_RTIOC_TS
2. apply following patch, and rerun 1.

index f83d7689f..50afbd418 100644
--- a/testsuite/gpiobench/gpiobench.c
+++ b/testsuite/gpiobench/gpiobench.c
@@ -619,7 +619,7 @@ int main(int argc, char **argv)
                        goto out;
                }

-               ret = ioctl(ti.fd_dev_intr, GPIO_RTIOC_TS, &value);
+               ret = ioctl(ti.fd_dev_intr, GPIO_RTIOC_TS_MONO, &value);
                if (ret) {
                        printf("ioctl gpio port ts, failed\n");
                        goto out;

Hardware env:
1. Rock PI X V1.4.
2. GPIO loopback connection between GPIO 334 and 335.
 

Hongzhan Chen (3):
  drivers/gpio: core: move out of OF config conditional compilation
  drivers/gpio: core: Introduce helper to find gpiochip
  driver/gpio: Add Intel Cherryview pinctrl driver

 include/cobalt/kernel/rtdm/gpio.h     | 10 ++++--
 kernel/drivers/gpio/Kconfig           |  7 ++++
 kernel/drivers/gpio/Makefile          |  2 ++
 kernel/drivers/gpio/gpio-cherryview.c | 42 ++++++++++++++++++++++++
 kernel/drivers/gpio/gpio-core.c       | 46 +++++++++++++++++++++++++--
 5 files changed, 103 insertions(+), 4 deletions(-)
 create mode 100644 kernel/drivers/gpio/gpio-cherryview.c

-- 
2.17.1



^ permalink raw reply related	[flat|nested] 13+ messages in thread
* [PATCH 1/3] rtdm/testing: latmus: introduce latmus driver
@ 2021-04-14  2:16 hongzha1
  2021-04-14  2:16 ` [PATCH 2/3] drivers/gpio: core: introduce helper to find gpiochip hongzha1
  0 siblings, 1 reply; 13+ messages in thread
From: hongzha1 @ 2021-04-14  2:16 UTC (permalink / raw)
  To: xenomai

Signed-off-by: hongzha1 <hongzhan.chen@intel.com>

diff --git a/include/rtdm/uapi/testing.h b/include/rtdm/uapi/testing.h
index f8207b8c7..b1723b9f8 100644
--- a/include/rtdm/uapi/testing.h
+++ b/include/rtdm/uapi/testing.h
@@ -118,6 +118,52 @@ struct rttst_heap_stathdr {
 	struct rttst_heap_stats *buf;
 };
 
+/* Latmus context types. */
+#define COBALT_LAT_IRQ   0
+#define COBALT_LAT_KERN  1
+#define COBALT_LAT_USER  2
+#define COBALT_LAT_SIRQ  3
+#define COBALT_LAT_LAST	COBALT_LAT_SIRQ
+
+struct latmus_setup {
+	__u32 type;
+	__u64 period;
+	__s32 priority;
+	__u32 cpu;
+	union {
+		struct {
+			__u32 verbosity;
+		} tune;
+		struct {
+			__u32 xfd;
+			__u32 hcells;
+		} measure;
+	} u;
+};
+
+/*
+ * The measurement record which the driver sends to userland each
+ * second through an xbuf channel.
+ */
+struct latmus_measurement {
+	__s64 sum_lat;
+	__s32 min_lat;
+	__s32 max_lat;
+	__u32 overruns;
+	__u32 samples;
+};
+
+struct latmus_measurement_result {
+	__u64 last_ptr;		/* (struct latmus_measurement __user *last) */
+	__u64 histogram_ptr;	/* (__s32 __user *histogram) */
+	__u32 len;
+};
+
+struct latmus_result {
+	__u64 data_ptr;		/* (void __user *data) */
+	__u32 len;
+};
+
 #define RTIOC_TYPE_TESTING		RTDM_CLASS_TESTING
 
 /*!
@@ -133,6 +179,8 @@ struct rttst_heap_stathdr {
 #define RTDM_SUBCLASS_RTDMTEST		3
 /** subclase name: "heapcheck" */
 #define RTDM_SUBCLASS_HEAPCHECK		4
+/** subclase name: "latmus" */
+#define RTDM_SUBCLASS_LATMUS		5
 /** @} */
 
 /*!
@@ -193,6 +241,21 @@ struct rttst_heap_stathdr {
 #define RTTST_RTIOC_HEAP_STAT_COLLECT \
 	_IOR(RTIOC_TYPE_TESTING, 0x45, int)
 
+#define RTTST_RTIOC_COBALT_LATIOC_TUNE \
+	_IOWR(RTIOC_TYPE_TESTING, 0x50, struct latmus_setup)
+
+#define RTTST_RTIOC_COBALT_LATIOC_MEASURE \
+	_IOWR(RTIOC_TYPE_TESTING, 0x51, struct latmus_setup)
+
+#define RTTST_RTIOC_COBALT_LATIOC_RUN \
+	_IOR(RTIOC_TYPE_TESTING, 0x52, struct latmus_result)
+
+#define RTTST_RTIOC_COBALT_LATIOC_PULSE \
+	_IOW(RTIOC_TYPE_TESTING, 0x53, __u64)
+
+#define RTTST_RTIOC_COBALT_LATIOC_RESET	\
+	_IO(RTIOC_TYPE_TESTING, 0x54)
+
 /** @} */
 
 #endif /* !_RTDM_UAPI_TESTING_H */
diff --git a/kernel/drivers/testing/Kconfig b/kernel/drivers/testing/Kconfig
index 88c043c82..5478dfe97 100644
--- a/kernel/drivers/testing/Kconfig
+++ b/kernel/drivers/testing/Kconfig
@@ -26,4 +26,14 @@ config XENO_DRIVERS_RTDMTEST
 	help
 	Kernel driver for performing RTDM unit tests.
 
+config COBALT_LATMUS
+	bool "Timer latency calibration and measurement"
+	depends on XENOMAI
+	default y
+	help
+	  This driver supports the latmus application for
+	  determining the best gravity values for the cobalt core
+	  clock, and measuring the response time to timer events.
+	  If in doubt, say 'Y'.
+
 endmenu
diff --git a/kernel/drivers/testing/Makefile b/kernel/drivers/testing/Makefile
index 09b076348..cad4ad50d 100644
--- a/kernel/drivers/testing/Makefile
+++ b/kernel/drivers/testing/Makefile
@@ -3,6 +3,7 @@ obj-$(CONFIG_XENO_DRIVERS_TIMERBENCH) += xeno_timerbench.o
 obj-$(CONFIG_XENO_DRIVERS_SWITCHTEST) += xeno_switchtest.o
 obj-$(CONFIG_XENO_DRIVERS_RTDMTEST)   += xeno_rtdmtest.o
 obj-$(CONFIG_XENO_DRIVERS_HEAPCHECK)   += xeno_heapcheck.o
+obj-$(CONFIG_COBALT_LATMUS)		+= latmus.o
 
 xeno_timerbench-y := timerbench.o
 
diff --git a/kernel/drivers/testing/latmus.c b/kernel/drivers/testing/latmus.c
new file mode 100644
index 000000000..bef662260
--- /dev/null
+++ b/kernel/drivers/testing/latmus.c
@@ -0,0 +1,1237 @@
+/*
+ * SPDX-License-Identifier: GPL-2.0
+ *
+ * Derived from Xenomai Cobalt's autotune driver, https://xenomai.org/
+ * Copyright (C) 2014, 2018 Philippe Gerum  <rpm@xenomai.org>
+ */
+
+#include <linux/types.h>
+#include <linux/init.h>
+#include <linux/module.h>
+#include <linux/slab.h>
+#include <linux/kernel.h>
+#include <linux/sort.h>
+#include <linux/cdev.h>
+#include <linux/fs.h>
+#include <linux/fcntl.h>
+#include <cobalt/kernel/pipe.h>
+#include <cobalt/kernel/sched.h>
+#include <rtdm/ipc.h>
+#include <rtdm/testing.h>
+#include <rtdm/driver.h>
+#include <rtdm/compat.h>
+
+#define ONE_BILLION  1000000000
+#define TUNER_SAMPLING_TIME	500000000UL
+#define TUNER_WARMUP_STEPS	10
+#define TUNER_RESULT_STEPS	40
+
+#define progress(__runner, __fmt, __args...)				\
+	do {								\
+		if ((__runner)->verbosity > 1)				\
+			printk(XENO_INFO "latmus(%s) " __fmt "\n",	\
+			       (__runner)->name, ##__args);		\
+	} while (0)
+
+#define cobalt_init_xntimer_on_cpu(__timer, __cpu, __handler)		\
+		rtdm_timer_init_on_cpu(__timer, __handler,		\
+			#__handler, __cpu)
+
+struct latmus_message {
+	struct xnpipe_mh mh;
+	char data[];
+};
+
+struct tuning_score {
+	int pmean;
+	int stddev;
+	int minlat;
+	unsigned int step;
+	unsigned int gravity;
+};
+
+struct runner_state {
+	ktime_t ideal;
+	int offset;
+	int min_lat;
+	int max_lat;
+	int allmax_lat;
+	int prev_mean;
+	int prev_sqs;
+	int cur_sqs;
+	int sum;
+	unsigned int overruns;
+	unsigned int cur_samples;
+	unsigned int max_samples;
+};
+
+struct latmus_runner {
+	const char *name;
+	unsigned int (*get_gravity)(struct latmus_runner *runner);
+	void (*set_gravity)(struct latmus_runner *runner, unsigned int gravity);
+	unsigned int (*adjust_gravity)(struct latmus_runner *runner, int adjust);
+	int (*start)(struct latmus_runner *runner, ktime_t start_time);
+	void (*stop)(struct latmus_runner *runner);
+	void (*destroy)(struct latmus_runner *runner);
+	int (*add_sample)(struct latmus_runner *runner, ktime_t timestamp);
+	int (*run)(struct rtdm_fd *fd, struct latmus_runner *runner,
+			struct latmus_result *result);
+	void (*cleanup)(struct latmus_runner *runner);
+	struct runner_state state;
+	struct rtdm_event done;
+	int status;
+	int verbosity;
+	ktime_t period;
+	union {
+		struct {
+			struct tuning_score scores[TUNER_RESULT_STEPS];
+			int nscores;
+		};
+		struct {
+			unsigned int warmup_samples;
+			unsigned int warmup_limit;
+			int socketfd;
+			void *xbuf;
+			u32 hcells;
+			s32 *histogram;
+		};
+	};
+};
+
+struct irq_runner {
+	rtdm_timer_t timer;
+	struct latmus_runner runner;
+};
+
+struct kthread_runner {
+	rtdm_task_t kthread;
+	struct rtdm_event barrier;
+	ktime_t start_time;
+	struct latmus_runner runner;
+};
+
+struct uthread_runner {
+	rtdm_timer_t timer;
+	struct rtdm_event pulse;
+	struct latmus_runner runner;
+};
+
+struct sirq_runner {
+	int sirq;
+	struct sirq_runner * __percpu *sirq_percpu;
+	rtdm_timer_t timer;
+	struct latmus_runner runner;
+};
+
+struct latmus_state {
+	struct latmus_runner *runner;
+	rtdm_lock_t latmus_lock;
+};
+
+static inline void init_runner_base(struct latmus_runner *runner)
+{
+	rtdm_event_init(&runner->done, 0);
+	runner->status = 0;
+	runner->socketfd = -1;
+}
+
+static inline void destroy_runner_base(struct latmus_runner *runner)
+{
+	rtdm_event_destroy(&runner->done);
+	if (runner->cleanup)
+		runner->cleanup(runner);
+}
+
+static inline void done_sampling(struct latmus_runner *runner,
+				 int status)
+{
+	runner->status = status;
+	rtdm_event_signal(&runner->done);
+}
+
+static void send_measurement(struct latmus_runner *runner)
+{
+	struct runner_state *state = &runner->state;
+	struct latmus_measurement *meas;
+	struct latmus_message *mbuf;
+	int len;
+
+	len = sizeof(*meas) + sizeof(*mbuf);
+	mbuf = xnmalloc(len);
+	if (mbuf == NULL)
+		return;
+
+	meas = (struct latmus_measurement *)mbuf->data;
+	meas->min_lat = state->min_lat;
+	meas->max_lat = state->max_lat;
+	meas->sum_lat = state->sum;
+	meas->overruns = state->overruns;
+	meas->samples = state->cur_samples;
+
+	runner->xbuf = mbuf;
+	xnpipe_send(runner->socketfd, &mbuf->mh, len, XNPIPE_NORMAL);
+
+	/* Reset counters for next round. */
+	state->min_lat = INT_MAX;
+	state->max_lat = INT_MIN;
+	state->sum = 0;
+	state->overruns = 0;
+	state->cur_samples = 0;
+}
+
+static int add_measurement_sample(struct latmus_runner *runner,
+				  ktime_t timestamp)
+{
+	struct runner_state *state = &runner->state;
+	ktime_t period = runner->period;
+	int delta, cell, offset_delta;
+
+	/* Skip samples in warmup time. */
+	if (runner->warmup_samples < runner->warmup_limit) {
+		runner->warmup_samples++;
+		state->ideal = ktime_add(state->ideal, period);
+		return 0;
+	}
+
+	delta = (int)ktime_to_ns(ktime_sub(timestamp, state->ideal));
+	offset_delta = delta - state->offset;
+	if (offset_delta < state->min_lat)
+		state->min_lat = offset_delta;
+	if (offset_delta > state->max_lat)
+		state->max_lat = offset_delta;
+	if (offset_delta > state->allmax_lat) {
+		state->allmax_lat = offset_delta;
+		//trace_evl_latspot(offset_delta);
+		//trace_evl_trigger("latmus");
+	}
+
+	if (runner->histogram) {
+		cell = (offset_delta < 0 ? -offset_delta : offset_delta) / 1000; /* us */
+		if (cell >= runner->hcells)
+			cell = runner->hcells - 1;
+		runner->histogram[cell]++;
+	}
+
+	state->sum += offset_delta;
+	state->ideal = ktime_add(state->ideal, period);
+
+	while (delta > 0 &&
+		(unsigned int)delta > ktime_to_ns(period)) { /* period > 0 */
+		state->overruns++;
+		state->ideal = ktime_add(state->ideal, period);
+		delta -= ktime_to_ns(period);
+	}
+
+	if (++state->cur_samples >= state->max_samples)
+		send_measurement(runner);
+
+	return 0;	/* Always keep going. */
+}
+
+static int add_tuning_sample(struct latmus_runner *runner,
+			     ktime_t timestamp)
+{
+	struct runner_state *state = &runner->state;
+	int n, delta, cur_mean;
+
+	delta = (int)ktime_to_ns(ktime_sub(timestamp, state->ideal));
+	if (delta < state->min_lat)
+		state->min_lat = delta;
+	if (delta > state->max_lat)
+		state->max_lat = delta;
+	if (delta < 0)
+		delta = 0;
+
+	state->sum += delta;
+	state->ideal = ktime_add(state->ideal, runner->period);
+	n = ++state->cur_samples;
+
+	/* TAOCP (Vol 2), single-pass computation of variance. */
+	if (n == 1)
+		state->prev_mean = delta;
+	else {
+		cur_mean = state->prev_mean + (delta - state->prev_mean) / n;
+		state->cur_sqs = state->prev_sqs + (delta - state->prev_mean)
+			* (delta - cur_mean);
+		state->prev_mean = cur_mean;
+		state->prev_sqs = state->cur_sqs;
+	}
+
+	if (n >= state->max_samples) {
+		done_sampling(runner, 0);
+		return 1;	/* Finished. */
+	}
+
+	return 0;	/* Keep going. */
+}
+
+static void latmus_irq_handler(rtdm_timer_t *timer) /* hard irqs off */
+{
+	struct irq_runner *irq_runner;
+	ktime_t now;
+
+	irq_runner = container_of(timer, struct irq_runner, timer);
+	now = xnclock_read_raw(&nkclock);
+	if (irq_runner->runner.add_sample(&irq_runner->runner, now))
+		rtdm_timer_stop(timer);
+}
+
+static void destroy_irq_runner(struct latmus_runner *runner)
+{
+	struct irq_runner *irq_runner;
+
+	irq_runner = container_of(runner, struct irq_runner, runner);
+	rtdm_timer_destroy(&irq_runner->timer);
+	destroy_runner_base(runner);
+	kfree(irq_runner);
+}
+
+static unsigned int get_irq_gravity(struct latmus_runner *runner)
+{
+	return nkclock.gravity.irq;
+}
+
+static void set_irq_gravity(struct latmus_runner *runner, unsigned int gravity)
+{
+	nkclock.gravity.irq = gravity;
+}
+
+static unsigned int adjust_irq_gravity(struct latmus_runner *runner, int adjust)
+{
+	return nkclock.gravity.irq += adjust;
+}
+
+static int start_irq_runner(struct latmus_runner *runner,
+			    ktime_t start_time)
+{
+	struct irq_runner *irq_runner;
+
+	irq_runner = container_of(runner, struct irq_runner, runner);
+
+	rtdm_timer_start(&irq_runner->timer, start_time,
+				runner->period, RTDM_TIMERMODE_ABSOLUTE);
+
+	return 0;
+}
+
+static void stop_irq_runner(struct latmus_runner *runner)
+{
+	struct irq_runner *irq_runner;
+
+	irq_runner = container_of(runner, struct irq_runner, runner);
+
+	rtdm_timer_stop(&irq_runner->timer);
+}
+
+static struct latmus_runner *create_irq_runner(int cpu)
+{
+	struct irq_runner *irq_runner;
+
+	irq_runner = kzalloc(sizeof(*irq_runner), GFP_KERNEL);
+	if (irq_runner == NULL)
+		return NULL;
+
+	irq_runner->runner = (struct latmus_runner){
+		.name = "irqhand",
+		.destroy = destroy_irq_runner,
+		.get_gravity = get_irq_gravity,
+		.set_gravity = set_irq_gravity,
+		.adjust_gravity = adjust_irq_gravity,
+		.start = start_irq_runner,
+		.stop = stop_irq_runner,
+	};
+
+	init_runner_base(&irq_runner->runner);
+	cobalt_init_xntimer_on_cpu(&irq_runner->timer, cpu, latmus_irq_handler);
+
+	return &irq_runner->runner;
+}
+
+static irqreturn_t latmus_sirq_handler(int sirq, void *dev_id)
+{
+	struct sirq_runner * __percpu *self_percpu = dev_id;
+	struct sirq_runner *sirq_runner = *self_percpu;
+	ktime_t now;
+
+	now = xnclock_read_raw(&nkclock);
+	if (sirq_runner->runner.add_sample(&sirq_runner->runner, now))
+		rtdm_timer_stop(&sirq_runner->timer);
+
+	return IRQ_HANDLED;
+}
+
+static void latmus_sirq_timer_handler(rtdm_timer_t *timer) /* hard irqs off */
+{
+	struct sirq_runner *sirq_runner;
+	struct runner_state *state;
+	ktime_t now;
+
+	now = xnclock_read_raw(&nkclock);
+	sirq_runner = container_of(timer, struct sirq_runner, timer);
+	state = &sirq_runner->runner.state;
+	state->offset = (int)ktime_to_ns(ktime_sub(now, state->ideal));
+	irq_post_inband(sirq_runner->sirq);
+}
+
+static void destroy_sirq_runner(struct latmus_runner *runner)
+{
+	struct sirq_runner *sirq_runner;
+
+	sirq_runner = container_of(runner, struct sirq_runner, runner);
+	rtdm_timer_destroy(&sirq_runner->timer);
+	free_percpu_irq(sirq_runner->sirq, sirq_runner->sirq_percpu);
+	free_percpu(sirq_runner->sirq_percpu);
+	irq_dispose_mapping(sirq_runner->sirq);
+	destroy_runner_base(runner);
+	kfree(sirq_runner);
+}
+
+static int start_sirq_runner(struct latmus_runner *runner,
+			ktime_t start_time)
+{
+	struct sirq_runner *sirq_runner;
+
+	sirq_runner = container_of(runner, struct sirq_runner, runner);
+
+	rtdm_timer_start(&sirq_runner->timer, start_time,
+				runner->period, RTDM_TIMERMODE_ABSOLUTE);
+
+	return 0;
+}
+
+static void stop_sirq_runner(struct latmus_runner *runner)
+{
+	struct sirq_runner *sirq_runner;
+
+	sirq_runner = container_of(runner, struct sirq_runner, runner);
+
+	rtdm_timer_stop(&sirq_runner->timer);
+}
+
+static struct latmus_runner *create_sirq_runner(int cpu)
+{
+	struct sirq_runner * __percpu *sirq_percpu;
+	struct sirq_runner *sirq_runner;
+	int sirq, ret, _cpu;
+
+	sirq_percpu = alloc_percpu(struct sirq_runner *);
+	if (sirq_percpu == NULL)
+		return ERR_PTR(-ENOMEM);
+
+	sirq_runner = kzalloc(sizeof(*sirq_runner), GFP_KERNEL);
+	if (sirq_runner == NULL) {
+		free_percpu(sirq_percpu);
+		return NULL;
+	}
+
+	sirq = irq_create_direct_mapping(synthetic_irq_domain);
+	if (sirq == 0) {
+		free_percpu(sirq_percpu);
+		kfree(sirq_runner);
+		return ERR_PTR(-EAGAIN);
+	}
+
+	sirq_runner->runner = (struct latmus_runner){
+		.name = "sirqhand",
+		.destroy = destroy_sirq_runner,
+		.start = start_sirq_runner,
+		.stop = stop_sirq_runner,
+	};
+	sirq_runner->sirq = sirq;
+	sirq_runner->sirq_percpu = sirq_percpu;
+	init_runner_base(&sirq_runner->runner);
+	cobalt_init_xntimer_on_cpu(&sirq_runner->timer, cpu,
+			latmus_sirq_timer_handler);
+
+	for_each_possible_cpu(_cpu)
+		*per_cpu_ptr(sirq_percpu, _cpu) = sirq_runner;
+
+	ret = __request_percpu_irq(sirq, latmus_sirq_handler,
+				IRQF_NO_THREAD,
+				"latmus sirq",
+				sirq_percpu);
+	if (ret) {
+		rtdm_timer_destroy(&sirq_runner->timer);
+		irq_dispose_mapping(sirq);
+		free_percpu(sirq_percpu);
+		kfree(sirq_runner);
+		return ERR_PTR(ret);
+	}
+
+	return &sirq_runner->runner;
+}
+
+void kthread_handler(void *arg)
+{
+	struct kthread_runner *k_runner = arg;
+	ktime_t now;
+	int ret = 0;
+
+	for (;;) {
+		if (rtdm_task_should_stop())
+			break;
+
+		ret = rtdm_event_wait(&k_runner->barrier);
+		if (ret)
+			break;
+
+		ret = rtdm_task_set_period(&k_runner->kthread,
+					k_runner->start_time,
+					k_runner->runner.period);
+		if (ret)
+			break;
+
+		for (;;) {
+			ret = rtdm_task_wait_period(NULL);
+			if (ret && ret != -ETIMEDOUT)
+				goto out;
+
+			now = xnclock_read_raw(&nkclock);
+			if (k_runner->runner.add_sample(&k_runner->runner, now)) {
+				rtdm_task_set_period(&k_runner->kthread, 0, 0);
+				break;
+			}
+		}
+	}
+out:
+	done_sampling(&k_runner->runner, ret);
+	rtdm_task_destroy(&k_runner->kthread);
+}
+
+static void destroy_kthread_runner(struct latmus_runner *runner)
+{
+	struct kthread_runner *k_runner;
+
+	k_runner = container_of(runner, struct kthread_runner, runner);
+	rtdm_task_destroy(&k_runner->kthread);
+	rtdm_event_destroy(&k_runner->barrier);
+	destroy_runner_base(runner);
+	kfree(k_runner);
+}
+
+static unsigned int get_kthread_gravity(struct latmus_runner *runner)
+{
+	return nkclock.gravity.kernel;
+}
+
+static void
+set_kthread_gravity(struct latmus_runner *runner, unsigned int gravity)
+{
+	nkclock.gravity.kernel = gravity;
+}
+
+static unsigned int
+adjust_kthread_gravity(struct latmus_runner *runner, int adjust)
+{
+	return nkclock.gravity.kernel += adjust;
+}
+
+static int start_kthread_runner(struct latmus_runner *runner,
+				ktime_t start_time)
+{
+	struct kthread_runner *k_runner;
+
+	k_runner = container_of(runner, struct kthread_runner, runner);
+
+	k_runner->start_time = start_time;
+	rtdm_event_signal(&k_runner->barrier);
+
+	return 0;
+}
+
+static struct latmus_runner *
+create_kthread_runner(int priority, int cpu)
+{
+	struct kthread_runner *k_runner;
+	int ret;
+	char *taskname;
+
+	k_runner = kzalloc(sizeof(*k_runner), GFP_KERNEL);
+	if (k_runner == NULL)
+		return NULL;
+
+	k_runner->runner = (struct latmus_runner){
+		.name = "kthread",
+		.destroy = destroy_kthread_runner,
+		.get_gravity = get_kthread_gravity,
+		.set_gravity = set_kthread_gravity,
+		.adjust_gravity = adjust_kthread_gravity,
+		.start = start_kthread_runner,
+	};
+
+	init_runner_base(&k_runner->runner);
+	rtdm_event_init(&k_runner->barrier, 0);
+
+	taskname = kasprintf(GFP_KERNEL, "latmus-klat:%d",
+			task_pid_nr(current));
+	ret = rtdm_task_init_on_cpu(&k_runner->kthread, cpu,
+				taskname,
+				kthread_handler,
+				k_runner,
+				priority,
+				0);
+	kfree(taskname);
+	if (ret) {
+		kfree(k_runner);
+		return ERR_PTR(ret);
+	}
+
+	return &k_runner->runner;
+}
+
+static void latmus_pulse_handler(rtdm_timer_t *timer)
+{
+	struct uthread_runner *u_runner;
+
+	u_runner = container_of(timer, struct uthread_runner, timer);
+	rtdm_event_signal(&u_runner->pulse);
+}
+
+static void destroy_uthread_runner(struct latmus_runner *runner)
+{
+	struct uthread_runner *u_runner;
+
+	u_runner = container_of(runner, struct uthread_runner, runner);
+	rtdm_timer_destroy(&u_runner->timer);
+	rtdm_event_destroy(&u_runner->pulse);
+	destroy_runner_base(runner);
+	kfree(u_runner);
+}
+
+static unsigned int get_uthread_gravity(struct latmus_runner *runner)
+{
+	return nkclock.gravity.user;
+}
+
+static void set_uthread_gravity(struct latmus_runner *runner,
+				unsigned int gravity)
+{
+	nkclock.gravity.user = gravity;
+}
+
+static unsigned int
+adjust_uthread_gravity(struct latmus_runner *runner, int adjust)
+{
+	return nkclock.gravity.user += adjust;
+}
+
+static int start_uthread_runner(struct latmus_runner *runner,
+				ktime_t start_time)
+{
+	struct uthread_runner *u_runner;
+
+	u_runner = container_of(runner, struct uthread_runner, runner);
+
+	rtdm_timer_start(&u_runner->timer, start_time,
+				runner->period, RTDM_TIMERMODE_ABSOLUTE);
+
+	return 0;
+}
+
+static void stop_uthread_runner(struct latmus_runner *runner)
+{
+	struct uthread_runner *u_runner;
+
+	u_runner = container_of(runner, struct uthread_runner, runner);
+
+	rtdm_timer_stop(&u_runner->timer);
+}
+
+static int add_uthread_sample(struct latmus_runner *runner,
+			      ktime_t user_timestamp)
+{
+	struct uthread_runner *u_runner;
+	int ret;
+
+	u_runner = container_of(runner, struct uthread_runner, runner);
+
+	if (user_timestamp &&
+	    u_runner->runner.add_sample(runner, user_timestamp)) {
+		rtdm_timer_stop(&u_runner->timer);
+		/* Tell the caller to park until next round. */
+		ret = -EPIPE;
+	} else {
+		ret = rtdm_event_wait(&u_runner->pulse);
+	}
+
+	return ret;
+}
+
+static struct latmus_runner *create_uthread_runner(int cpu)
+{
+	struct uthread_runner *u_runner;
+
+	u_runner = kzalloc(sizeof(*u_runner), GFP_KERNEL);
+	if (u_runner == NULL)
+		return NULL;
+
+	u_runner->runner = (struct latmus_runner){
+		.name = "uthread",
+		.destroy = destroy_uthread_runner,
+		.get_gravity = get_uthread_gravity,
+		.set_gravity = set_uthread_gravity,
+		.adjust_gravity = adjust_uthread_gravity,
+		.start = start_uthread_runner,
+		.stop = stop_uthread_runner,
+	};
+
+	init_runner_base(&u_runner->runner);
+	cobalt_init_xntimer_on_cpu(&u_runner->timer, cpu, latmus_pulse_handler);
+	xntimer_set_gravity(&u_runner->timer, XNTIMER_UGRAVITY);
+	rtdm_event_init(&u_runner->pulse, 0);
+
+	return &u_runner->runner;
+}
+
+static inline void build_score(struct latmus_runner *runner, int step)
+{
+	struct runner_state *state = &runner->state;
+	unsigned int variance, n;
+
+	n = state->cur_samples;
+	runner->scores[step].pmean = state->sum / n;
+	variance = n > 1 ? state->cur_sqs / (n - 1) : 0;
+	runner->scores[step].stddev = int_sqrt(variance);
+	runner->scores[step].minlat = state->min_lat;
+	runner->scores[step].gravity = runner->get_gravity(runner);
+	runner->scores[step].step = step;
+	runner->nscores++;
+}
+
+static int cmp_score_mean(const void *c, const void *r)
+{
+	const struct tuning_score *sc = c, *sr = r;
+
+	return sc->pmean - sr->pmean;
+}
+
+static int cmp_score_stddev(const void *c, const void *r)
+{
+	const struct tuning_score *sc = c, *sr = r;
+
+	return sc->stddev - sr->stddev;
+}
+
+static int cmp_score_minlat(const void *c, const void *r)
+{
+	const struct tuning_score *sc = c, *sr = r;
+
+	return sc->minlat - sr->minlat;
+}
+
+static int cmp_score_gravity(const void *c, const void *r)
+{
+	const struct tuning_score *sc = c, *sr = r;
+
+	return sc->gravity - sr->gravity;
+}
+
+static int filter_mean(struct latmus_runner *runner)
+{
+	sort(runner->scores, runner->nscores,
+	     sizeof(struct tuning_score),
+	     cmp_score_mean, NULL);
+
+	/* Top half of the best pondered means. */
+
+	return (runner->nscores + 1) / 2;
+}
+
+static int filter_stddev(struct latmus_runner *runner)
+{
+	sort(runner->scores, runner->nscores,
+	     sizeof(struct tuning_score),
+	     cmp_score_stddev, NULL);
+
+	/* Top half of the best standard deviations. */
+
+	return (runner->nscores + 1) / 2;
+}
+
+static int filter_minlat(struct latmus_runner *runner)
+{
+	sort(runner->scores, runner->nscores,
+	     sizeof(struct tuning_score),
+	     cmp_score_minlat, NULL);
+
+	/* Top half of the minimum latencies. */
+
+	return (runner->nscores + 1) / 2;
+}
+
+static int filter_gravity(struct latmus_runner *runner)
+{
+	sort(runner->scores, runner->nscores,
+	     sizeof(struct tuning_score),
+	     cmp_score_gravity, NULL);
+
+	/* Smallest gravity required among the shortest latencies. */
+
+	return runner->nscores;
+}
+
+static void dump_scores(struct latmus_runner *runner)
+{
+	int n;
+
+	if (runner->verbosity < 2)
+		return;
+
+	for (n = 0; n < runner->nscores; n++)
+		printk(XENO_INFO
+		       ".. S%.2d pmean=%d stddev=%d minlat=%d gravity=%u\n",
+		       runner->scores[n].step,
+		       runner->scores[n].pmean,
+		       runner->scores[n].stddev,
+		       runner->scores[n].minlat,
+		       runner->scores[n].gravity);
+}
+
+static inline void filter_score(struct latmus_runner *runner,
+				int (*filter)(struct latmus_runner *runner))
+{
+	runner->nscores = filter(runner);
+	dump_scores(runner);
+}
+
+static void __latmus_free_handler(void *buf, void *skarg) /* nklock free */
+{
+	struct latmus_runner *runner = skarg;
+
+	xnfree(runner->xbuf);
+}
+
+static int measure_continously(struct latmus_runner *runner)
+{
+	struct runner_state *state = &runner->state;
+	ktime_t period = runner->period;
+	int ret;
+
+	state->max_samples = ONE_BILLION / (int)ktime_to_ns(period);
+	runner->add_sample = add_measurement_sample;
+	state->min_lat = INT_MAX;
+	state->max_lat = INT_MIN;
+	state->allmax_lat = INT_MIN;
+	state->sum = 0;
+	state->overruns = 0;
+	state->cur_samples = 0;
+	state->offset = 0;	/* for SIRQ latency only. */
+	state->ideal = ktime_add(xnclock_read_raw(&nkclock), period);
+
+	ret = runner->start(runner, state->ideal);
+	if (ret)
+		goto out;
+
+	ret = rtdm_event_wait(&runner->done) ?: runner->status;
+
+	if (runner->stop)
+		runner->stop(runner);
+out:
+	return ret;
+}
+
+static int tune_gravity(struct latmus_runner *runner)
+{
+	struct runner_state *state = &runner->state;
+	int ret, step, gravity_limit, adjust;
+	ktime_t period = runner->period;
+	unsigned int orig_gravity;
+
+	state->max_samples = TUNER_SAMPLING_TIME / (int)ktime_to_ns(period);
+	orig_gravity = runner->get_gravity(runner);
+	runner->add_sample = add_tuning_sample;
+	runner->set_gravity(runner, 0);
+	runner->nscores = 0;
+	adjust = 500; /* Gravity adjustment step */
+	gravity_limit = 0;
+	progress(runner, "warming up...");
+
+	for (step = 0; step < TUNER_WARMUP_STEPS + TUNER_RESULT_STEPS; step++) {
+		state->ideal = ktime_add_ns(xnclock_read_raw(&nkclock),
+			    ktime_to_ns(period) * TUNER_WARMUP_STEPS);
+		state->min_lat = INT_MAX;
+		state->max_lat = INT_MIN;
+		state->prev_mean = 0;
+		state->prev_sqs = 0;
+		state->cur_sqs = 0;
+		state->sum = 0;
+		state->cur_samples = 0;
+
+		ret = runner->start(runner, state->ideal);
+		if (ret)
+			goto fail;
+
+		/* Runner stops when posting. */
+		ret = rtdm_event_wait(&runner->done);
+		if (ret)
+			goto fail;
+
+		ret = runner->status;
+		if (ret)
+			goto fail;
+
+		if (step < TUNER_WARMUP_STEPS) {
+			if (state->min_lat > gravity_limit) {
+				gravity_limit = state->min_lat;
+				progress(runner, "gravity limit set to %u ns (%d)",
+					 gravity_limit, state->min_lat);
+			}
+			continue;
+		}
+
+		if (state->min_lat < 0) {
+			if (runner->get_gravity(runner) < -state->min_lat) {
+				printk(XENO_WARNING
+				       "latmus(%s) failed with early shot (%d ns)\n",
+				       runner->name,
+				       -(runner->get_gravity(runner) + state->min_lat));
+				ret = -EAGAIN;
+				goto fail;
+			}
+			break;
+		}
+
+		if (((step - TUNER_WARMUP_STEPS) % 5) == 0)
+			progress(runner, "calibrating... (slice %d)",
+				 (step - TUNER_WARMUP_STEPS) / 5 + 1);
+
+		build_score(runner, step - TUNER_WARMUP_STEPS);
+
+		/*
+		 * Anticipating by more than the minimum latency
+		 * detected at warmup would make no sense: cap the
+		 * gravity we may try.
+		 */
+		if (runner->adjust_gravity(runner, adjust) > gravity_limit) {
+			progress(runner, "beyond gravity limit at %u ns",
+				 runner->get_gravity(runner));
+			break;
+		}
+	}
+
+	progress(runner, "calibration scores");
+	dump_scores(runner);
+	progress(runner, "pondered mean filter");
+	filter_score(runner, filter_mean);
+	progress(runner, "standard deviation filter");
+	filter_score(runner, filter_stddev);
+	progress(runner, "minimum latency filter");
+	filter_score(runner, filter_minlat);
+	progress(runner, "gravity filter");
+	filter_score(runner, filter_gravity);
+	runner->set_gravity(runner, runner->scores[0].gravity);
+
+	return 0;
+fail:
+	runner->set_gravity(runner, orig_gravity);
+
+	return ret;
+}
+
+static int setup_tuning(struct latmus_runner *runner,
+			struct latmus_setup *setup)
+{
+	runner->verbosity = setup->u.tune.verbosity;
+	runner->period = setup->period;
+
+	return 0;
+}
+
+static int run_tuning(struct rtdm_fd *fd,
+		struct latmus_runner *runner,
+		struct latmus_result *result)
+{
+	__u32 gravity;
+	int ret;
+
+	ret = tune_gravity(runner);
+	if (ret)
+		return ret;
+
+	gravity = runner->get_gravity(runner);
+
+	if (rtdm_safe_copy_to_user(fd, (void *)(long)result->data_ptr,
+				&gravity, sizeof(gravity)))
+		return -EFAULT;
+
+	return 0;
+}
+
+static int setup_measurement(struct latmus_runner *runner,
+			     struct latmus_setup *setup)
+{
+	struct xnpipe_operations ops;
+	int xnpipefd;
+
+	memset(&ops, 0, sizeof(ops));
+	ops.free_obuf = &__latmus_free_handler;
+	xnpipefd = xnpipe_connect(-1, &ops, runner);
+	if (xnpipefd < 0)
+		return xnpipefd;
+	runner->socketfd = xnpipefd;
+	runner->period = setup->period;
+	runner->warmup_limit = ONE_BILLION / (int)ktime_to_ns(setup->period); /* 1s warmup */
+	runner->histogram = NULL;
+	runner->hcells = setup->u.measure.hcells;
+	if (runner->hcells == 0)
+		return 0;
+
+	if (runner->hcells > 1000) /* LART */
+		return -EINVAL;
+
+	runner->histogram = kzalloc(runner->hcells * sizeof(s32),
+				    GFP_KERNEL);
+
+	return runner->histogram ? 0 : -ENOMEM;
+}
+
+static int run_measurement(struct rtdm_fd *fd,
+			struct latmus_runner *runner,
+			struct latmus_result *result)
+{
+	struct runner_state *state = &runner->state;
+	struct latmus_measurement_result mr;
+	struct latmus_measurement last;
+	size_t len;
+	int ret;
+
+	if (result->len != sizeof(mr))
+		return -EINVAL;
+
+	if (rtdm_safe_copy_from_user(fd, &mr, (void *)(long)result->data_ptr,
+				sizeof(mr)))
+		return -EFAULT;
+
+	ret = measure_continously(runner);
+	if (ret != -EINTR)
+		return ret;
+
+	/*
+	 * Copy the last bulk of consolidated measurements and the
+	 * histogram distribution data back to userland.
+	 */
+	last.min_lat = state->min_lat;
+	last.max_lat = state->max_lat;
+	last.sum_lat = state->sum;
+	last.overruns = state->overruns;
+	last.samples = state->cur_samples;
+	if (rtdm_safe_copy_to_user(fd, (void *)(long)mr.last_ptr,
+				&last, sizeof(last)))
+		return -EFAULT;
+
+	if (runner->histogram) {
+		len = runner->hcells * sizeof(s32);
+		if (len > mr.len)
+			len = result->len;
+		if (len > 0 &&
+		    rtdm_safe_copy_to_user(fd, (void *)(long)mr.histogram_ptr,
+					   runner->histogram, len))
+			return -EFAULT;
+	}
+
+	return 0;
+}
+
+static void cleanup_measurement(struct latmus_runner *runner)
+{
+	if (runner->socketfd >= 0) {
+		xnpipe_disconnect(runner->socketfd);
+		runner->socketfd = -1;
+	}
+
+	if (runner->histogram)
+		kfree(runner->histogram);
+}
+
+static int latmus_ioctl(struct rtdm_fd *fd, unsigned int cmd, void *arg)
+{
+	struct latmus_state *ls = rtdm_fd_to_private(fd);
+	int (*setup)(struct latmus_runner *runner,
+		     struct latmus_setup *setup_data);
+	int (*run)(struct rtdm_fd *fd,
+		struct latmus_runner *runner,
+		struct latmus_result *result);
+	void (*cleanup)(struct latmus_runner *runner);
+	struct latmus_setup setup_data;
+	struct latmus_runner *runner;
+	int ret;
+
+	if (cmd == RTTST_RTIOC_COBALT_LATIOC_RESET) {
+		xnclock_reset_gravity(&nkclock);
+		return 0;
+	}
+
+	/* All other cmds require a setup struct to be passed. */
+
+	if (rtdm_copy_from_user(fd, &setup_data, arg, sizeof(setup_data)))
+		return -EFAULT;
+
+	if (setup_data.type == COBALT_LAT_SIRQ &&
+			cmd != RTTST_RTIOC_COBALT_LATIOC_MEASURE)
+		return -EINVAL;
+
+	switch (cmd) {
+	case RTTST_RTIOC_COBALT_LATIOC_TUNE:
+		setup = setup_tuning;
+		run = run_tuning;
+		cleanup = NULL;
+		break;
+	case RTTST_RTIOC_COBALT_LATIOC_MEASURE:
+		setup = setup_measurement;
+		run = run_measurement;
+		cleanup = cleanup_measurement;
+		break;
+	default:
+		return -ENOSYS;
+	}
+
+	if (setup_data.period <= 0 ||
+	    setup_data.period > ONE_BILLION)
+		return -EINVAL;
+
+	if (setup_data.priority < 1 ||
+	    setup_data.priority > XNSCHED_FIFO_MAX_PRIO)
+		return -EINVAL;
+
+	if (setup_data.cpu >= num_possible_cpus() ||
+		!xnsched_supported_cpu(setup_data.cpu))
+		return -EINVAL;
+
+	/* Clear previous runner. */
+	runner = ls->runner;
+	if (runner) {
+		runner->destroy(runner);
+		ls->runner = NULL;
+	}
+
+	switch (setup_data.type) {
+	case COBALT_LAT_IRQ:
+		runner = create_irq_runner(setup_data.cpu);
+		break;
+	case COBALT_LAT_KERN:
+		runner = create_kthread_runner(setup_data.priority,
+					       setup_data.cpu);
+		break;
+	case COBALT_LAT_USER:
+		runner = create_uthread_runner(setup_data.cpu);
+		break;
+	case COBALT_LAT_SIRQ:
+		runner = create_sirq_runner(setup_data.cpu);
+		break;
+	default:
+		return -EINVAL;
+	}
+
+	if (IS_ERR(runner))
+		return PTR_ERR(runner);
+
+	ret = setup(runner, &setup_data);
+	if (ret) {
+		runner->destroy(runner);
+		return ret;
+	}
+
+	runner->run = run;
+	runner->cleanup = cleanup;
+	ls->runner = runner;
+
+	return ((cmd == RTTST_RTIOC_COBALT_LATIOC_MEASURE) ?
+			runner->socketfd : 0);
+}
+
+static int latmus_oob_ioctl(struct rtdm_fd *fd, unsigned int cmd, void *arg)
+{
+	struct latmus_state *ls = rtdm_fd_to_private(fd);
+	struct latmus_runner *runner;
+	struct latmus_result result;
+	__u64 timestamp;
+	int ret;
+
+	switch (cmd) {
+	case RTTST_RTIOC_COBALT_LATIOC_RUN:
+		runner = ls->runner;
+		if (runner == NULL)
+			return -EAGAIN;
+		ret = rtdm_safe_copy_from_user(fd, &result,
+				arg, sizeof(result));
+
+		if (ret)
+			return -EFAULT;
+		ret = runner->run(fd, runner, &result);
+		break;
+	case RTTST_RTIOC_COBALT_LATIOC_PULSE:
+		runner = ls->runner;
+		if (runner == NULL)
+			return -EAGAIN;
+		if (runner->start != start_uthread_runner)
+			return -EINVAL;
+		ret = rtdm_safe_copy_from_user(fd,  &timestamp,
+						arg, sizeof(timestamp));
+		if (ret)
+			return -EFAULT;
+		ret = add_uthread_sample(runner, ns_to_ktime(timestamp));
+		break;
+	default:
+		ret = -ENOSYS;
+	}
+
+	return ret;
+}
+
+static int latmus_open(struct rtdm_fd *fd, int oflags)
+{
+	struct latmus_state *ls;
+
+	ls = rtdm_fd_to_private(fd);
+	ls->runner = NULL;
+	rtdm_lock_init(&ls->latmus_lock);
+
+	return 0;
+}
+
+static void latmus_close(struct rtdm_fd *fd)
+{
+	struct latmus_state *ls = rtdm_fd_to_private(fd);
+	struct latmus_runner *runner;
+
+	runner = ls->runner;
+	if (runner)
+		runner->destroy(runner);
+
+}
+
+static struct rtdm_driver latmus_driver = {
+	.profile_info		= RTDM_PROFILE_INFO(latmus,
+						    RTDM_CLASS_TESTING,
+						    RTDM_SUBCLASS_LATMUS,
+						    RTTST_PROFILE_VER),
+	.device_flags		= RTDM_NAMED_DEVICE | RTDM_EXCLUSIVE,
+	.device_count		= 1,
+	.context_size		= sizeof(struct latmus_state),
+	.ops = {
+		.open		= latmus_open,
+		.close		= latmus_close,
+		.ioctl_rt	= latmus_oob_ioctl,
+		.ioctl_nrt	= latmus_ioctl,
+	},
+};
+
+static struct rtdm_device device = {
+	.driver = &latmus_driver,
+	.label = "latmus",
+};
+
+static int __init __latmus_init(void)
+{
+	return rtdm_dev_register(&device);
+}
+
+static void __latmus_exit(void)
+{
+	rtdm_dev_unregister(&device);
+}
+
+module_init(__latmus_init);
+module_exit(__latmus_exit);
+
+MODULE_LICENSE("GPL");
-- 
2.17.1



^ permalink raw reply related	[flat|nested] 13+ messages in thread
end of thread, other threads:[~2021-09-01  1:16 UTC | newest]

Thread overview: 13+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2021-08-31  5:07 [PATCH V2 0/3] add support for cherryview gpio controller driver Hongzhan Chen
2021-08-31  5:07 ` [PATCH 1/3] drivers/gpio: core: Move out of OF config conditional compilation Hongzhan Chen
2021-08-31 13:05   ` Jan Kiszka
2021-09-01  1:16     ` Chen, Hongzhan
2021-08-31  5:07 ` [PATCH 2/3] drivers/gpio: core: Introduce helper to find gpiochip Hongzhan Chen
2021-08-31  6:47   ` Jan Kiszka
2021-08-31  7:07     ` Chen, Hongzhan
2021-08-31  7:14       ` Jan Kiszka
2021-08-31  5:07 ` [PATCH 3/3] driver/gpio: Add Intel Cherryview pinctrl driver Hongzhan Chen
2021-08-31  6:51   ` Jan Kiszka
2021-08-31  7:15 ` [PATCH V2 0/3] add support for cherryview gpio controller driver Jan Kiszka
  -- strict thread matches above, loose matches on Subject: below --
2021-08-30  6:45 [PATCH " Hongzhan Chen
2021-08-30  6:45 ` [PATCH 2/3] drivers/gpio: core: Introduce helper to find gpiochip Hongzhan Chen
2021-04-14  2:16 [PATCH 1/3] rtdm/testing: latmus: introduce latmus driver hongzha1
2021-04-14  2:16 ` [PATCH 2/3] drivers/gpio: core: introduce helper to find gpiochip hongzha1

This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.