openwrt/target/linux/bcm27xx/patches-6.1/950-0841-gpio_fsm-Rework-the-atomic-vs-non-atomic-split.patch

From f0061ffc98c6e027c5774e2a24ceadcfee4167ea Mon Sep 17 00:00:00 2001
From: Phil Elwell <[email protected]>
Date: Tue, 5 Sep 2023 12:01:13 +0100
Subject: [PATCH] gpio_fsm: Rework the atomic-vs-non-atomic split

Partition the code to separate atomic and non-atomic methods so that
none of them have to handle both cases. The result avoids using deferred
work unless necessary, and should be easier to understand.

Signed-off-by: Phil Elwell <[email protected]>
---
 drivers/gpio/gpio-fsm.c | 84 ++++++++++++++++++++---------------------
 1 file changed, 41 insertions(+), 43 deletions(-)

--- a/drivers/gpio/gpio-fsm.c
+++ b/drivers/gpio/gpio-fsm.c
@@ -193,9 +193,6 @@ static void free_symbols(struct symtab_e
 	}
 }
 
-static void gpio_fsm_go_to_state(struct gpio_fsm *gf,
-				 struct fsm_state *new_state);
-
 static void gpio_fsm_set_soft(struct gpio_fsm *gf,
 			      unsigned int off, int val);
 
@@ -213,6 +210,7 @@ static void gpio_fsm_enter_state(struct
 	dev_dbg(gf->dev, "enter_state(%s)\n", state->name);
 
 	gf->current_state = state;
+	gf->delay_target_state = NULL;
 
 	// 1. Apply any listed signals
 	for (i = 0; i < state->num_signals; i++) {
@@ -271,7 +269,7 @@ static void gpio_fsm_enter_state(struct
 				dev_info(gf->dev,
 					 "GF_SOFT %d=%d -> %s\n", event->index,
 					 event->value, event->target->name);
-			gpio_fsm_go_to_state(gf, event->target);
+			gpio_fsm_enter_state(gf, event->target);
 			return;
 		}
 	}
@@ -284,7 +282,7 @@ static void gpio_fsm_enter_state(struct
 		inp_state->value = event->value;
 		inp_state->enabled = true;
 
-		value = gpiod_get_value(gf->input_gpios->desc[event->index]);
+		value = gpiod_get_value_cansleep(gf->input_gpios->desc[event->index]);
 
 		// Clear stale event state
 		disable_irq(inp_state->irq);
@@ -299,7 +297,7 @@ static void gpio_fsm_enter_state(struct
 				dev_info(gf->dev,
 					 "GF_IN %d=%d -> %s\n", event->index,
 					 event->value, event->target->name);
-			gpio_fsm_go_to_state(gf, event->target);
+			gpio_fsm_enter_state(gf, event->target);
 			return;
 		}
 	}
@@ -325,6 +323,33 @@ static void gpio_fsm_go_to_state(struct
 	dev_dbg(gf->dev, "go_to_state(%s)\n",
 		  new_state ? new_state->name : "<unset>");
 
+	state = gf->current_state;
+
+	/* Disable any enabled GPIO IRQs */
+	for (i = 0; i < state->num_gpio_events; i++) {
+		gp_ev = &state->gpio_events[i];
+		inp_state = &gf->input_gpio_states[gp_ev->index];
+		if (inp_state->enabled) {
+			inp_state->enabled = false;
+			irq_set_irq_type(inp_state->irq,
+					 IRQF_TRIGGER_NONE);
+		}
+	}
+
+	gpio_fsm_enter_state(gf, new_state);
+}
+
+static void gpio_fsm_go_to_state_deferred(struct gpio_fsm *gf,
+					  struct fsm_state *new_state)
+{
+	struct input_gpio_state *inp_state;
+	struct gpio_event *gp_ev;
+	struct fsm_state *state;
+	int i;
+
+	dev_dbg(gf->dev, "go_to_state_deferred(%s)\n",
+		  new_state ? new_state->name : "<unset>");
+
 	spin_lock(&gf->spinlock);
 
 	if (gf->next_state) {
@@ -335,57 +360,31 @@ static void gpio_fsm_go_to_state(struct
 
 	gf->next_state = new_state;
 	state = gf->current_state;
-	gf->delay_target_state = NULL;
 
-	if (state) {
-		/* Disarm any GPIO IRQs */
-		for (i = 0; i < state->num_gpio_events; i++) {
-			gp_ev = &state->gpio_events[i];
-			inp_state = &gf->input_gpio_states[gp_ev->index];
-			inp_state->target = NULL;
-		}
+	/* Disarm any GPIO IRQs */
+	for (i = 0; i < state->num_gpio_events; i++) {
+		gp_ev = &state->gpio_events[i];
+		inp_state = &gf->input_gpio_states[gp_ev->index];
+		inp_state->target = NULL;
 	}
 
 	spin_unlock(&gf->spinlock);
 
-	if (new_state)
-		schedule_work(&gf->work);
+	schedule_work(&gf->work);
 }
 
 static void gpio_fsm_work(struct work_struct *work)
 {
-	struct input_gpio_state *inp_state;
 	struct fsm_state *new_state;
-	struct fsm_state *state;
-	struct gpio_event *gp_ev;
 	struct gpio_fsm *gf;
-	int i;
 
 	gf = container_of(work, struct gpio_fsm, work);
 	spin_lock(&gf->spinlock);
-	state = gf->current_state;
 	new_state = gf->next_state;
-	if (!new_state)
-		new_state = gf->delay_target_state;
 	gf->next_state = NULL;
-	gf->delay_target_state = NULL;
 	spin_unlock(&gf->spinlock);
 
-	if (state) {
-		/* Disable any enabled GPIO IRQs */
-		for (i = 0; i < state->num_gpio_events; i++) {
-			gp_ev = &state->gpio_events[i];
-			inp_state = &gf->input_gpio_states[gp_ev->index];
-			if (inp_state->enabled) {
-				inp_state->enabled = false;
-				irq_set_irq_type(inp_state->irq,
-						 IRQF_TRIGGER_NONE);
-			}
-		}
-	}
-
-	if (new_state)
-		gpio_fsm_enter_state(gf, new_state);
+	gpio_fsm_go_to_state(gf, new_state);
 }
 
 static irqreturn_t gpio_fsm_gpio_irq_handler(int irq, void *dev_id)
@@ -404,7 +403,7 @@ static irqreturn_t gpio_fsm_gpio_irq_han
 	if (gf->debug)
 		dev_info(gf->dev, "GF_IN %d->%d -> %s\n",
 			 inp_state->index, inp_state->value, target->name);
-	gpio_fsm_go_to_state(gf, target);
+	gpio_fsm_go_to_state_deferred(gf, target);
 	return IRQ_HANDLED;
 }
 
@@ -416,12 +415,11 @@ static void gpio_fsm_timer(struct timer_
 	target = gf->delay_target_state;
 	if (!target)
 		return;
-
 	if (gf->debug)
 		dev_info(gf->dev, "GF_DELAY %d -> %s\n", gf->delay_ms,
 			 target->name);
 
-	gpio_fsm_go_to_state(gf, target);
+	gpio_fsm_go_to_state_deferred(gf, target);
 }
 
 int gpio_fsm_parse_signals(struct gpio_fsm *gf, struct fsm_state *state,
@@ -1119,7 +1117,7 @@ static int gpio_fsm_probe(struct platfor
 	if (gf->debug)
 		dev_info(gf->dev, "Start -> %s\n", gf->start_state->name);
 
-	gpio_fsm_go_to_state(gf, gf->start_state);
+	gpio_fsm_enter_state(gf, gf->start_state);
 
 	return devm_gpiochip_add_data(dev, &gf->gc, gf);
 }