V4L/DVB: IR: replace workqueue with kthread

It is perfectly possible to have ir_raw_event_work
running concurently on two cpus, thus we must protect
it from that situation.

This stems from the fact that if hardware sends short packets of samples
we might end up queueing the work item more times that nessesary.

Such job isn't well suited for a workqueue, so use a kernel thread.

Signed-off-by: Maxim Levitsky <maximlevitsky@gmail.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab@redhat.com>

2 files changed, 32 insertions, 12 deletions

diff --git a/drivers/media/IR/ir-core-priv.h b/drivers/media/IR/ir-core-priv.h
index dc26e2beeefb..84c7a9a5cad4 100644
--- a/drivers/media/IR/ir-core-priv.h
+++ b/drivers/media/IR/ir-core-priv.h
@@ -32,7 +32,7 @@ struct ir_raw_handler {
 
 struct ir_raw_event_ctrl {
 	struct list_head		list;		/* to keep track of raw clients */
-	struct work_struct		rx_work;	/* for the rx decoding workqueue */
+	struct task_struct		*thread;
 	struct kfifo			kfifo;		/* fifo for the pulse/space durations */
 	ktime_t				last_event;	/* when last event occurred */
 	enum raw_event_type		last_type;	/* last event type */
diff --git a/drivers/media/IR/ir-raw-event.c b/drivers/media/IR/ir-raw-event.c
index 9d5c029cfc0c..d0c18db4c0d3 100644
--- a/drivers/media/IR/ir-raw-event.c
+++ b/drivers/media/IR/ir-raw-event.c
@@ -12,9 +12,10 @@
  *  GNU General Public License for more details.
  */
 
-#include <linux/workqueue.h>
+#include <linux/kthread.h>
 #include <linux/mutex.h>
 #include <linux/sched.h>
+#include <linux/freezer.h>
 #include "ir-core-priv.h"
 
 /* Define the max number of pulse/space transitions to buffer */
@@ -33,20 +34,30 @@ static u64 available_protocols;
 static struct work_struct wq_load;
 #endif
 
-static void ir_raw_event_work(struct work_struct *work)
+static int ir_raw_event_thread(void *data)
 {
 	struct ir_raw_event ev;
 	struct ir_raw_handler *handler;
-	struct ir_raw_event_ctrl *raw =
-		container_of(work, struct ir_raw_event_ctrl, rx_work);
+	struct ir_raw_event_ctrl *raw = (struct ir_raw_event_ctrl *)data;
+
+	while (!kthread_should_stop()) {
+		try_to_freeze();
 
-	while (kfifo_out(&raw->kfifo, &ev, sizeof(ev)) == sizeof(ev)) {
 		mutex_lock(&ir_raw_handler_lock);
-		list_for_each_entry(handler, &ir_raw_handler_list, list)
-			handler->decode(raw->input_dev, ev);
+
+		while (kfifo_out(&raw->kfifo, &ev, sizeof(ev)) == sizeof(ev)) {
+			list_for_each_entry(handler, &ir_raw_handler_list, list)
+				handler->decode(raw->input_dev, ev);
+			raw->prev_ev = ev;
+		}
+
 		mutex_unlock(&ir_raw_handler_lock);
-		raw->prev_ev = ev;
+
+		set_current_state(TASK_INTERRUPTIBLE);
+		schedule();
 	}
+
+	return 0;
 }
 
 /**
@@ -141,7 +152,7 @@ void ir_raw_event_handle(struct input_dev *input_dev)
 	if (!ir->raw)
 		return;
 
-	schedule_work(&ir->raw->rx_work);
+	wake_up_process(ir->raw->thread);
 }
 EXPORT_SYMBOL_GPL(ir_raw_event_handle);
 
@@ -170,7 +181,7 @@ int ir_raw_event_register(struct input_dev *input_dev)
 		return -ENOMEM;
 
 	ir->raw->input_dev = input_dev;
-	INIT_WORK(&ir->raw->rx_work, ir_raw_event_work);
+
 	ir->raw->enabled_protocols = ~0;
 	rc = kfifo_alloc(&ir->raw->kfifo, sizeof(s64) * MAX_IR_EVENT_SIZE,
 			 GFP_KERNEL);
@@ -180,6 +191,15 @@ int ir_raw_event_register(struct input_dev *input_dev)
 		return rc;
 	}
 
+	ir->raw->thread = kthread_run(ir_raw_event_thread, ir->raw,
+			"rc%u",  (unsigned int)ir->devno);
+
+	if (IS_ERR(ir->raw->thread)) {
+		kfree(ir->raw);
+		ir->raw = NULL;
+		return PTR_ERR(ir->raw->thread);
+	}
+
 	mutex_lock(&ir_raw_handler_lock);
 	list_add_tail(&ir->raw->list, &ir_raw_client_list);
 	list_for_each_entry(handler, &ir_raw_handler_list, list)
@@ -198,7 +218,7 @@ void ir_raw_event_unregister(struct input_dev *input_dev)
 	if (!ir->raw)
 		return;
 
-	cancel_work_sync(&ir->raw->rx_work);
+	kthread_stop(ir->raw->thread);
 
 	mutex_lock(&ir_raw_handler_lock);
 	list_del(&ir->raw->list);