[media] rc: refactor raw handler kthread

I think we can get rid of the spinlock protecting the kthread from being
interrupted by a wakeup in certain parts.
Even with the current implementation of the kthread the only lost wakeup
scenario could happen if the wakeup occurs between the kfifo_len check
and setting the state to TASK_INTERRUPTIBLE.

In the changed version we could lose a wakeup if it occurs between
processing the fifo content and setting the state to TASK_INTERRUPTIBLE.
This scenario is covered by an additional check for available events in
the fifo and setting the state to TASK_RUNNING in this case.

In addition the changed version flushes the kfifo before ending
when the kthread is stopped.

With this patch we gain:
- Get rid of the spinlock
- Simplify code
- Don't grep / release the mutex for each individual event but just once
  for the complete fifo content. This reduces overhead if a driver e.g.
  triggers processing after writing the content of a hw fifo to the kfifo.

Signed-off-by: Heiner Kallweit <hkallweit1@gmail.com>
Signed-off-by: Sean Young <sean@mess.org>
Signed-off-by: Mauro Carvalho Chehab <mchehab@s-opensource.com>

1 files changed, 0 insertions, 2 deletions

diff --git a/drivers/media/rc/rc-core-priv.h b/drivers/media/rc/rc-core-priv.h
index 585d5e52118d..577128ad0ae5 100644
--- a/drivers/media/rc/rc-core-priv.h
+++ b/drivers/media/rc/rc-core-priv.h
@@ -20,7 +20,6 @@
 #define	MAX_IR_EVENT_SIZE	512
 
 #include <linux/slab.h>
-#include <linux/spinlock.h>
 #include <media/rc-core.h>
 
 struct ir_raw_handler {
@@ -37,7 +36,6 @@ struct ir_raw_handler {
 struct ir_raw_event_ctrl {
 	struct list_head		list;		/* to keep track of raw clients */
 	struct task_struct		*thread;
-	spinlock_t			lock;
 	/* fifo for the pulse/space durations */
 	DECLARE_KFIFO(kfifo, struct ir_raw_event, MAX_IR_EVENT_SIZE);
 	ktime_t				last_event;	/* when last event occurred */