[RFC,5/5] pwm: pwm-omap-dmtimer: Add capture functionality

Message ID 20180108154336.GE4077@lenoch
State New
Headers show
Series
  • Add capture functionality to OMAP pwm driver
Related show

Commit Message

Ladislav Michl Jan. 8, 2018, 3:43 p.m.
Call for help:
For testing purposes pwm_evt pin is driven by a signal generator,
in this example at 200us period and 80% duty cycle.
Repeated testing gives:
$ cat /sys/class/pwm/pwmchip4/pwm0/capture
200000 160000
or:
$ cat /sys/class/pwm/pwmchip4/pwm0/capture
200000 40000
Here it seems hardware can capture both edges, but I do not see a way
how to tell it I want start from either low to high or high to low
transition. Clues?

Also there is probably bug in the code as for shorter periods (here 10us)
I'm getting:
$ cat /sys/class/pwm/pwmchip4/pwm0/capture
omap_dm_timer_read_status: timer not available or enabled.
10000 2000
Unfortunately I do not see how it can happen in the code.

Any hints greatly appreciated, thank you...

Signed-off-by: Ladislav Michl <ladis@linux-mips.org>

Comments

Tony Lindgren Jan. 8, 2018, 9:59 p.m. | #1
* Ladislav Michl <ladis@linux-mips.org> [180108 15:46]:
> Here it seems hardware can capture both edges, but I do not see a way
> how to tell it I want start from either low to high or high to low
> transition. Clues?

At least dm3730 TRM documents TCM bits [9:8] for TCLR, but you
probably know that already..

If you're having hard time getting things starting, maybe something
like this helps:

stop timer in TCLR register
configure timer in TCLR
write some value to TLDR, maybe 0?
set ST bit in TCLR to start

Regards,

Tony
--
To unsubscribe from this list: send the line "unsubscribe linux-pwm" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Ladislav Michl Jan. 8, 2018, 10:06 p.m. | #2
On Mon, Jan 08, 2018 at 01:59:31PM -0800, Tony Lindgren wrote:
> * Ladislav Michl <ladis@linux-mips.org> [180108 15:46]:
> > Here it seems hardware can capture both edges, but I do not see a way
> > how to tell it I want start from either low to high or high to low
> > transition. Clues?
> 
> At least dm3730 TRM documents TCM bits [9:8] for TCLR, but you
> probably know that already..
> 
> If you're having hard time getting things starting, maybe something
> like this helps:
> 
> stop timer in TCLR register
> configure timer in TCLR
> write some value to TLDR, maybe 0?
> set ST bit in TCLR to start

Let me clarify it a bit more. I have no problem starting timer and capture
events. I just didn't find a way how to tell hardware I want to start
with for example rising edge, so rising edge goes to TCAR1 and failing edge
to TCAR2. Substracting those gives pulse width.

	ladis
--
To unsubscribe from this list: send the line "unsubscribe linux-pwm" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Tony Lindgren Jan. 8, 2018, 10:13 p.m. | #3
* Ladislav Michl <ladis@linux-mips.org> [180108 22:09]:
> On Mon, Jan 08, 2018 at 01:59:31PM -0800, Tony Lindgren wrote:
> > * Ladislav Michl <ladis@linux-mips.org> [180108 15:46]:
> > > Here it seems hardware can capture both edges, but I do not see a way
> > > how to tell it I want start from either low to high or high to low
> > > transition. Clues?
> > 
> > At least dm3730 TRM documents TCM bits [9:8] for TCLR, but you
> > probably know that already..
> > 
> > If you're having hard time getting things starting, maybe something
> > like this helps:
> > 
> > stop timer in TCLR register
> > configure timer in TCLR
> > write some value to TLDR, maybe 0?
> > set ST bit in TCLR to start
> 
> Let me clarify it a bit more. I have no problem starting timer and capture
> events. I just didn't find a way how to tell hardware I want to start
> with for example rising edge, so rising edge goes to TCAR1 and failing edge
> to TCAR2. Substracting those gives pulse width.

Oh I see, yeah that would be cool :) Maybe you can first configure
an interrupt to trigger on rising edge, then configure things for
falling edge, then subtract..

Regards,

Tony
--
To unsubscribe from this list: send the line "unsubscribe linux-pwm" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Ladislav Michl Jan. 8, 2018, 10:26 p.m. | #4
On Mon, Jan 08, 2018 at 02:13:27PM -0800, Tony Lindgren wrote:
> * Ladislav Michl <ladis@linux-mips.org> [180108 22:09]:
> > On Mon, Jan 08, 2018 at 01:59:31PM -0800, Tony Lindgren wrote:
> > > * Ladislav Michl <ladis@linux-mips.org> [180108 15:46]:
> > > > Here it seems hardware can capture both edges, but I do not see a way
> > > > how to tell it I want start from either low to high or high to low
> > > > transition. Clues?
> > > 
> > > At least dm3730 TRM documents TCM bits [9:8] for TCLR, but you
> > > probably know that already..
> > > 
> > > If you're having hard time getting things starting, maybe something
> > > like this helps:
> > > 
> > > stop timer in TCLR register
> > > configure timer in TCLR
> > > write some value to TLDR, maybe 0?
> > > set ST bit in TCLR to start
> > 
> > Let me clarify it a bit more. I have no problem starting timer and capture
> > events. I just didn't find a way how to tell hardware I want to start
> > with for example rising edge, so rising edge goes to TCAR1 and failing edge
> > to TCAR2. Substracting those gives pulse width.
> 
> Oh I see, yeah that would be cool :) Maybe you can first configure
> an interrupt to trigger on rising edge, then configure things for
> falling edge, then subtract..

That will work only for very long periods. I did test where value captured
into TCAR was compared to value read from TCRR. Sometimes those differ by
few hundreds microseconds...

I started this work to overcome GPIO edge interrupts latency, but so far
I'm a bit dissapointed.

(also please read the code, reconfiguration is done to get period and
duty cycle - but in case of duty cycle it is not clear whenever we are
measuring pulse or space length)

	ladis
--
To unsubscribe from this list: send the line "unsubscribe linux-pwm" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Patch

diff --git a/drivers/pwm/pwm-omap-dmtimer.c b/drivers/pwm/pwm-omap-dmtimer.c
index ee1cd92b1744..90d07df0f69d 100644
--- a/drivers/pwm/pwm-omap-dmtimer.c
+++ b/drivers/pwm/pwm-omap-dmtimer.c
@@ -17,7 +17,9 @@ 
  */
 
 #include <linux/clk.h>
+#include <clocksource/dmtimer.h>
 #include <linux/err.h>
+#include <linux/interrupt.h>
 #include <linux/kernel.h>
 #include <linux/module.h>
 #include <linux/mutex.h>
@@ -30,17 +32,21 @@ 
 #include <linux/pwm.h>
 #include <linux/slab.h>
 #include <linux/time.h>
+#include <linux/wait.h>
 
 #define DM_TIMER_LOAD_MIN 0xfffffffe
 #define DM_TIMER_MAX      0xffffffff
 
 struct pwm_omap_dmtimer_chip {
 	struct pwm_chip chip;
+	spinlock_t lock;
 	struct mutex mutex;
+	wait_queue_head_t wait;
 	pwm_omap_dmtimer *dm_timer;
 	struct omap_dm_timer_ops *pdata;
 	struct platform_device *dm_timer_pdev;
 	unsigned long freq;
+	unsigned int ev_cnt, overflow, width;
 };
 
 static inline struct pwm_omap_dmtimer_chip *
@@ -55,6 +61,22 @@  pwm_omap_dmtimer_get_clock_cycles(struct pwm_omap_dmtimer_chip *omap, int ns)
 	return DIV_ROUND_CLOSEST_ULL((u64)omap->freq * ns, NSEC_PER_SEC);
 }
 
+static inline unsigned int
+pwm_omap_dmtimer_ticks_to_ns(struct pwm_omap_dmtimer_chip *omap,
+			     unsigned int ticks)
+{
+	return DIV_ROUND_CLOSEST_ULL((u64)ticks * NSEC_PER_SEC, omap->freq);
+}
+
+static unsigned int pwm_omap_dmtimer_get_width(struct pwm_omap_dmtimer_chip *omap,
+					       unsigned int c1, unsigned int c2)
+{
+	if (c1 <= c2)
+		return c2 - c1;
+
+	return (c2 - omap->overflow) + (DM_TIMER_MAX - c1);
+}
+
 static void pwm_omap_dmtimer_start(struct pwm_omap_dmtimer_chip *omap)
 {
 	/*
@@ -218,7 +240,101 @@  static int pwm_omap_dmtimer_set_polarity(struct pwm_chip *chip,
 	return 0;
 }
 
+static int pwm_omap_dmtimer_capture(struct pwm_chip *chip,
+				    struct pwm_device *pwm,
+				    struct pwm_capture *result,
+				    unsigned long timeout)
+
+{
+	int res;
+	unsigned long flags;
+	struct pwm_omap_dmtimer_chip *omap = to_pwm_omap_dmtimer_chip(chip);
+
+	mutex_lock(&omap->mutex);
+
+	if (pm_runtime_active(&omap->dm_timer_pdev->dev))
+		omap->pdata->stop(omap->dm_timer);
+
+	/* Let timer overflow every 1 second */
+	omap->overflow = DM_TIMER_MAX - (1 * omap->freq);
+	timeout = msecs_to_jiffies(timeout);
+	result->period = result->duty_cycle = 0;
+
+	spin_lock_irqsave(&omap->lock, flags);
+	omap->pdata->set_int_enable(omap->dm_timer, OMAP_TIMER_INT_CAPTURE |
+				    OMAP_TIMER_INT_OVERFLOW);
+	omap->pdata->set_capture(omap->dm_timer, 1, 1);
+	omap->pdata->set_load_start(omap->dm_timer, true, omap->overflow);
+	omap->ev_cnt = omap->width = 0;
+	spin_unlock_irqrestore(&omap->lock, flags);
+
+	res = wait_event_interruptible_timeout(omap->wait,
+					       omap->width > 0, timeout);
+	if (res > 0) {
+		spin_lock_irqsave(&omap->lock, flags);
+		result->period =
+			pwm_omap_dmtimer_ticks_to_ns(omap, omap->width);
+		omap->pdata->stop(omap->dm_timer);
+		omap->pdata->set_capture(omap->dm_timer, 1, 3);
+		omap->pdata->start(omap->dm_timer);
+		omap->ev_cnt = omap->width = 0;
+		spin_unlock_irqrestore(&omap->lock, flags);
+
+		res = wait_event_interruptible_timeout(omap->wait,
+						       omap->width > 0,
+						       timeout);
+	}
+
+	spin_lock_irqsave(&omap->lock, flags);
+	omap->pdata->stop(omap->dm_timer);
+	omap->pdata->set_int_disable(omap->dm_timer, OMAP_TIMER_INT_CAPTURE |
+				     OMAP_TIMER_INT_OVERFLOW);
+	spin_unlock_irqrestore(&omap->lock, flags);
+
+	if (res == 0) {
+		res = -ETIMEDOUT;
+	} else if (res > 0) {
+		result->duty_cycle =
+			pwm_omap_dmtimer_ticks_to_ns(omap, omap->width);
+		res = 0;
+	}
+
+	mutex_unlock(&omap->mutex);
+
+	return res;
+}
+
+static irqreturn_t pwm_omap_dmtimer_irq(int irq, void *dev_id)
+{
+	u32 l;
+	unsigned int c1, c2;
+	unsigned long flags;
+        struct pwm_omap_dmtimer_chip *omap = dev_id;
+
+	spin_lock_irqsave(&omap->lock, flags);
+
+	l = omap->pdata->read_status(omap->dm_timer);
+	if (l & OMAP_TIMER_INT_CAPTURE) {
+		if (!omap->width && omap->ev_cnt == 1) {
+			omap->pdata->read_capture(omap->dm_timer, &c1, &c2);
+			omap->width = pwm_omap_dmtimer_get_width(omap, c1, c2);
+			wake_up(&omap->wait);
+		}
+		omap->ev_cnt++;
+	}
+	/* TODO: handle overflow as well
+	if (l & OMAP_TIMER_INT_OVERFLOW)
+		overflow++;
+	*/
+	omap->pdata->write_status(omap->dm_timer, l);
+
+	spin_unlock_irqrestore(&omap->lock, flags);
+
+	return IRQ_HANDLED;
+}
+
 static const struct pwm_ops pwm_omap_dmtimer_ops = {
+	.capture = pwm_omap_dmtimer_capture,
 	.enable	= pwm_omap_dmtimer_enable,
 	.disable = pwm_omap_dmtimer_disable,
 	.config	= pwm_omap_dmtimer_config,
@@ -237,7 +353,7 @@  static int pwm_omap_dmtimer_probe(struct platform_device *pdev)
 	pwm_omap_dmtimer *dm_timer;
 	struct clk *fclk;
 	u32 v;
-	int status;
+	int irq, status;
 
 	timer = of_parse_phandle(np, "ti,timers", 0);
 	if (!timer)
@@ -338,7 +454,18 @@  static int pwm_omap_dmtimer_probe(struct platform_device *pdev)
 	omap->chip.of_xlate = of_pwm_xlate_with_flags;
 	omap->chip.of_pwm_n_cells = 3;
 
+	spin_lock_init(&omap->lock);
 	mutex_init(&omap->mutex);
+	init_waitqueue_head(&omap->wait);
+
+	irq = omap->pdata->get_irq(omap->dm_timer);
+	if (irq < 0)
+		return irq;
+
+	status = devm_request_irq(&pdev->dev, irq, pwm_omap_dmtimer_irq, 0,
+				  "event_capture", omap);
+	if (status)
+		return status;
 
 	status = pwmchip_add(&omap->chip);
 	if (status < 0) {