Skip to contentSkip to conversation

cx88: High resolution timer for Remote Controls

PATCH In review Jean Delvare 1 participant 2 messages Jul 2, 2009 – Jul 3, 2009 Media (V4L/DVB)

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 cx88: High resolution timer for Remote Controls

From: Andrzej Hajda <[email protected]>

Patch solves problem of missed keystrokes on some remote controls, as reported on http://bugzilla.kernel.org/show_bug.cgi?id=9637 .

---
Resending because last attempt resulted in folded lines:
http://www.spinics.net/lists/linux-media/msg06884.html
Patch was already resent by Andrzej on June 4th but apparently it was overlooked.

Trent Piepho commented on the compatibility with kernels older than 2.6.20 being possibly broken:
http://www.spinics.net/lists/linux-media/msg06885.html
I don't think this is the case. The kernel version test was there because the workqueue API changed in 2.6.20, but the hrtimer API did not have such a change. This is why the version check has gone.

It is highly probable that the hrtimer API had its own incompatible changes since it was introduced in kernel 2.6.16. By looking at the code, I found the following ones:

* hrtimer_forward_now() was added with kernel 2.6.25 only:
http://git.kernel.org/?p=linux/kernel/git/torvalds/linux-2.6.git;a=commitdiff;h=5e05ad7d4e3b11f935998882b5d9c3b257137f1b
But this is an inline function, so I presume this shouldn't be too difficult to add to a compatibility header.

* Before 2.6.21, HRTIMER_MODE_REL was named HRTIMER_REL:
http://git.kernel.org/?p=linux/kernel/git/torvalds/linux-2.6.git;a=commitdiff;h=c9cb2e3d7c9178ab75d0942f96abb3abe0369906
This too should be solvable in a compatibility header.

The rest doesn't seem to cause compatibility issues, but only actual testing would confirm that.

This bug affects me, which is why I am motivated to get this fix upstream. Please let me know how I can help.

 linux/drivers/media/video/cx88/cx88-input.c |   37 ++++++++++++---------------
 1 file changed, 17 insertions(+), 20 deletions(-)
linux/drivers/media/video/cx88/cx88-input.c +17 -20
@@ -23,10 +23,10 @@
  */
 
 #include <linux/init.h>
-#include <linux/delay.h>
 #include <linux/input.h>
 #include <linux/pci.h>
 #include <linux/module.h>
+#include <linux/hrtimer.h>
 
 #include "compat.h"
 #include "cx88.h"
@@ -49,7 +49,7 @@ struct cx88_IR {
 
 	/* poll external decoder */
 	int polling;
-	struct delayed_work work;
+	struct hrtimer timer;
 	u32 gpio_addr;
 	u32 last_gpio;
 	u32 mask_keycode;
@@ -145,31 +145,28 @@ static void cx88_ir_handle_key(struct cx
 	}
 }
 
-#if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,20)
-static void cx88_ir_work(void *data)
-#else
-static void cx88_ir_work(struct work_struct *work)
-#endif
+enum hrtimer_restart cx88_ir_work(struct hrtimer *timer)
 {
-#if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,20)
-	struct cx88_IR *ir = data;
-#else
-	struct cx88_IR *ir = container_of(work, struct cx88_IR, work.work);
-#endif
+	unsigned long missed;
+	struct cx88_IR *ir = container_of(timer, struct cx88_IR, timer);
 
 	cx88_ir_handle_key(ir);
-	schedule_delayed_work(&ir->work, msecs_to_jiffies(ir->polling));
+	missed = hrtimer_forward_now(&ir->timer,
+				     ktime_set(0, ir->polling * 1000000));
+	if (missed > 1)
+		ir_dprintk("Missed ticks %ld\n", missed - 1);
+
+	return HRTIMER_RESTART;
 }
 
 void cx88_ir_start(struct cx88_core *core, struct cx88_IR *ir)
 {
 	if (ir->polling) {
-#if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,20)
-		INIT_DELAYED_WORK(&ir->work, cx88_ir_work, ir);
-#else
-		INIT_DELAYED_WORK(&ir->work, cx88_ir_work);
-#endif
-		schedule_delayed_work(&ir->work, 0);
+		hrtimer_init(&ir->timer, CLOCK_MONOTONIC, HRTIMER_MODE_REL);
+		ir->timer.function = cx88_ir_work;
+		hrtimer_start(&ir->timer,
+			      ktime_set(0, ir->polling * 1000000),
+			      HRTIMER_MODE_REL);
 	}
 	if (ir->sampling) {
 		core->pci_irqmask |= PCI_INT_IR_SMPINT;
@@ -186,7 +183,7 @@ void cx88_ir_stop(struct cx88_core *core
 	}
 
 	if (ir->polling)
-		cancel_delayed_work_sync(&ir->work);
+		hrtimer_cancel(&ir->timer);
 }
 
 /* ---------------------------------------------------------------------- */
Signature · 1 line
Jean Delvare
#1 lore raw
Quoted 29 lines · Jean Delvare

From: Andrzej Hajda <[email protected]>

Patch solves problem of missed keystrokes on some remote controls, as reported on http://bugzilla.kernel.org/show_bug.cgi?id=9637 .

---
Resending because last attempt resulted in folded lines:
http://www.spinics.net/lists/linux-media/msg06884.html
Patch was already resent by Andrzej on June 4th but apparently it was overlooked.

Trent Piepho commented on the compatibility with kernels older than 2.6.20 being possibly broken:
http://www.spinics.net/lists/linux-media/msg06885.html
I don't think this is the case. The kernel version test was there because the workqueue API changed in 2.6.20, but the hrtimer API did not have such a change. This is why the version check has gone.

It is highly probable that the hrtimer API had its own incompatible changes since it was introduced in kernel 2.6.16. By looking at the code, I found the following ones:

* hrtimer_forward_now() was added with kernel 2.6.25 only:
http://git.kernel.org/?p=linux/kernel/git/torvalds/linux-2.6.git;a=commitdiff;h=5e05ad7d4e3b11f935998882b5d9c3b257137f1b
But this is an inline function, so I presume this shouldn't be too difficult to add to a compatibility header.

Quoted 6 lines

* Before 2.6.21, HRTIMER_MODE_REL was named HRTIMER_REL:
http://git.kernel.org/?p=linux/kernel/git/torvalds/linux-2.6.git;a=commitdiff;h=c9cb2e3d7c9178ab75d0942f96abb3abe0369906
This too should be solvable in a compatibility header.

The rest doesn't seem to cause compatibility issues, but only actual testing would confirm that.

Actually there were more compatibility issues, the most important one being that not all functions of the hrtimer API are exported before 2.6.22. So unfortunately this bug fix means that the cx88 driver will no longer build on kernels < 2.6.22. I'll post a new patch with this change, and another one for the hrtimer compatibility code.

Signature · 1 line
Jean Delvare